SPATIAL INTERACTION MODELING OF ...

23 downloads 0 Views 1MB Size Report
Aug 8, 1989 - 191 Guns, howitzers, mortars, related equipment, or parts, 30 mm ...... private truck and for-hire truck; truck and air; truck and water; truck and ...
SPATIAL INTERACTION MODELING OF INTERREGIONAL COMMODITY FLOWS

DISSERTATION

Presented in Partial Fulfillment of the Requirements for the Degree Doctor of Philosophy in the Graduate School of The Ohio State University

By Huseyin Murat Celik, B.C.P., M.U.P. *****

The Ohio State University 2001

Dissertation Committee: Professor Jean-Michel Guldmann, Adviser Professor Burkhard von Rabenau Professor Phillip Viton

Approved by _________________________________ Adviser Department of City and Regional Planning

ABSTRACT

In contrast to passenger flows, there has been relatively little empirical research on interregional commodity flows, primarily because of limited available data. However, a better understanding of the determinants of interregional commodity flows may be critical for both transportation infrastructure planning and regional development policies. This research attempts

to expand past spatial interaction models of commodity flows by

incorporating new variables into the model, using a flexible Box-Cox functional form, and applying the analysis to all manufacturing commodities . The recently released 1993 and 1997 U.S. Commodity Flows Surveys provide the empirical basis for estimating state-tostate flow models for 16 manufactured commodity groups over the 48 continental U.S. states. Overall, the results show that the selected variables and functional form are very successful in explaining flow variations. First, having a common physical border increases trade between states for all commodity groups. As other destinations are physically closer to a specific destination, the amount of commodity shipped to that destination decreases because of demand competition effects. Similar effects are uncovered around the origin. Having a state involved in foreign exports and/or imports through a custom district may affect in- or out-shipments of commodities. Transportation costs, as proxied by distance, have a negative effect on out-shipments for all commodities. Population and personal percapita income at the origins have significant negative effects, suggesting that increasing local

demand

at

the

origin

decreases

the

amount

of

commodities

available

for

interregional shipment. The usual origin supply variables, sectoral employment and sectoral value-added, are significant and positive for most commodity groups. Wholesale employment at both the origins and destinations are important facilitators of the flows of many commodities. As a proxy for destinations

intermediate demand, manufacturing employment at

has significant and positive effects on the flows of wood, chemicals, mineral

products, and primary metals. The average size of

industrial establishments

at the

origins has a positive effect in a few cases, implying economies of scale, but mostly 2

negative effects,

pointing to diversification effects. Finally, the population and personal

income per-capita at the destinations, taken as proxies for final demand at destinations, have generally positive and significant effects on flows. In all cases, the Box-Cox specification obtained through maximum likelihood estimation is statistically superior to the log-linear form.

3

To my father, Fikri Celik, who taught me hardworking and honor

4

ACKNOWLEDGMENTS I would like to extend my cordial gratitude to my adviser, Dr. Jean-Michel Guldmann, for not only using his scholarship to guide me through this research, but also for being a patient counselor, a sincere friend, and a perfect draft proofreader, although he often strongly objected to this task. He has been a “father figure”, and, without his personal support, this dissertation would never have been completed. His smile and friendship will be highly missed. I also would like to thank my Ph.D. Committee Members, Dr. Philip Viton and Dr. Burkhard von Rabenau. Their timely advice and criticism contributed to the completion of this dissertation. I would like to thank the Faculty of City and Regional Planning in the Knowlton School of Architecture for supporting my research by awarding me the 2001 Dr. Jerrold Voss Scholarship. The comments of and useful discussions with my friend Erdogan Ozturk were always very helpful. Thank you Erdogan. My friends, Dr. Nuray Unlu and

Sally Eickholt,

were always with me in difficult times and their support was deeply appreciated. I also wish to thank Moriana Siri, whose warm friendship made Columbus home for me. Finally, I would like to thank the Turkish National Educational Ministry and the People of Turkey for financing my graduate education in the U.S.

5

VITA EDUCATION Sep.97- present

Ph. D. Student in City and Regional Planning. The Ohio State University, Knowlton School of Architecture, Graduate Program in City and Regional Planning, Columbus-Ohio. Master of City ( Transportation ) Planning. The University of Kansas, School of Architecture and Urban Design, Graduate Program in Urban Planning, Lawrence- Kansas. Non-Degree Advanced Level English and Orientation Course. Louisiana State University, English Language and Orientation Program, Baton Rouge- Louisiana. Master of City Planning. Mimar Sinan University, Faculty of Architecture, City and Regional Planning Department, Istanbul Turkey. (Discontinued to work with City of Mersin as a planner and advisor to the Mayor ) Bachelor of City Planning. Mimar Sinan University, Faculty of Architecture, City and Regional Planning Department, Istanbul Turkey.

Jan. 95 - Dec. 96 Oct. 94 - Dec. 96 Mar. 89 - Dec. 89

Sep. 83 - June 87

EMPLOYMENT Jan 00- present

Teaching and Research Assistant The Ohio State University, Columbus, OH. • • •

Aug.96- May. 97

Tasks: Teaching the course CRP 310 “Introduction to City and Regional Planning” (taught 3 quarters). Data preparation, manipulation, analysis, and reporting for a telecommunication research project. Assisting in the course CRP 780 “Workshop in City and Regional Planning Techniques” Research and Teaching Assistant The University of Kansas, Lawrence, KS.



Tasks: Data preparation, manipulation, and reporting for Central Place Theory Study for both Hokkaido Island, and Washington, Oregon, Idaho Region. 6



Jun. 96 - Aug. 96

Assisting Professor Dimitrios Dendrinos to teach the course “Advanced Seminar in Transportation” . Intern Transportation Planner. City of Excelsior Springs, Missouri.

• • May. 91 - Apr. 94

Tasks: Preparation of Excelsior Springs Sketch Transportation Plan. Graphic design of the maps used in Excelsior Springs Master Plan Study. City and Regional Planning Specialist. Prime Ministry of Republic of Turkey, Southeastern Anatolia Project, Regional Development Administration, Regional Directorate, Sanliurfa.



• •





Jan. 90 - May. 91

Tasks: Execution of the authority for urban development and infrastructure projects in the GAP (Southeastern Anatolian Project, a large-scale regional development enterprise including the construction of 22 hydro-power and irrigation dams on the Euphrates and Tigris Rivers) region covering 8 provinces and 140 county municipalities. Coordination of the projects run by various Government Agencies in the region. Supervision of a consultant consortium responsible for the preparation of a regional transportation study, urban infrastructure projects (including 25 urban master and implementation plans, water supply, sewer system, and waste water and solid waste treatment projects for 45 cities in the region), sub-regional development plans, and sectoral development studies. Executive Secretary of the committee responsible for the coordination and supervision of evacuation and resettlement activities of Halfeti and Hasankeyf Communities, which will be overwhelmed by dam constructions. Design of a Master Land-Use Plan, and preparation of zoning ordinances for a 200,000 people community, Karakopru Suburb, in the Sanliurfa Metropolitan Area. City Planner and Advisor to the Mayor. City of Mersin, Turkey.

• • •

Tasks: Participation in master plan amendment works, slum reclamation planning, landscape designs, and capital improvement program preparations. Monitoring planning and implementation activities as the representative of the mayor. Coordination of urban development and infrastructure project investments according to Cukurova Urban Growth Project financed by the World Bank.

7

Aug. 89 - Dec. 89

City Planner. Iller 71 City Planning Office, Istanbul. •

Tasks: Preparation and design of Istanbul Kucukbakkalkoy and Yeni Sahra Slum Reclamation Plans, and Istanbul Kumburgaz Urban Development Plan.

Apr. 88 - Jul. 89

Enlisted Artillery Officer. Turkish Army.

Nov. 87 - Mar. 88

City Planner and Urban Designer. Iller 71 City Planning Office, Istanbul. •

Tasks: Design and implementation of housing projects, preparation of a master plan and implementation plans, and landscape projects for an area covering 200 hectares.

FIELDS OF STUDY Major Field:

City and Regional Planning

Minor Fields: Quantitative Modeling Economic Planning Regional Planning Transportation Planning

8

TABLE OF CONTENTS Abstract………………. ……………………………………………………………………….

Page 2

Dedication…………… ………………………………………………………………………

4

Acknowledgments…. ………………………………………………………………………

5

Vita…………………….. ………………………………………………………………………

6

List of Tables……….. ………………………………………………………………………

11

Chapters: 1.

Introduction…………. ………………………………………………………………………

14

2.

Literature Review…..………………………………………………………………………

18

3.

4.

5.

2.1

Commodity Flow Models……. …………………………………………………… 18 2.1.1. Interregional Input-Output Models……………………………… 18 2.1.2. Spatial Interaction Models of Commodity Flows..………….. 20

2.2 2.3 2.4

Overview of Spatial Interaction Models………………………………………. International Trade Models… …………………………………………………… Summary………………………… ……………………………………………………

28 31 32

Modeling Methodology….…………………………………………………………………

34

3.1 3.2

Theoretical Background….. ……………………………………………………… The Empirical Commodity Flow Model..…………………………………….. 3.2.1 Variables……………………. …………………………………………. a Origin Variables…………… …………………………………………. b Destination Variables……. …………………………………………. c Geographical Variables…. …………………………………………. 3.2.2 Functional Form…………… ………………………………………….

34 38 39 39 40 40 42

Data Sources and Processing……………………………………………………………

44

4.1 4.2

44 50

Dependent Variable…………………… …………………………………………. Independent Variables……………….. ………………………………………….

Results…..……………………………………………………………………………………

5.1 Overview of U.S. Interregional Commodity Flows………………………………….. 9

54 54

5.2 Individual Commodity Equation Results……………… ……………………………… 5.2.1 Food and Kindred Products………… ……………………………… 5.2.2 Lumber and Wood Products……….. ……………………………… 5.2.3 Furniture and Fixture Products……. ……………………………… 5.2.4 Pulp, Paper and Allied Products….. ……………………………… 5.2.5 Chemical and Allied Products……… ……………………………… 5.2.6 Petroleum and Coal Products……… ……………………………… 5.2.7 Rubber & Misc. Plastic Products….. ……………………………… 5.2.8 Clay, Concrete, Glass & Stone Products…………………………… 5.2.9 Primary Metal Products……………… ……………………………… 5.2.10 Fabricated Metal Products………….. ……………………………… 5.2.11 Machinery Products (Non-Electrical)…. …………………………. 5.2.12 Electrical Machinery Products…….. ……………………………… 5.2.13 Transportation Equipment………….. ……………………………… 5.2.14 Precision Instruments……………….. ……………………………… 5.2.15 Miscellaneous Manufacturing Products…………………………… 5.2.16 Apparel Textile Leather Products…. ……………………………… 5.2.17 Synthesis………………………………… ……………………………… 5.3 Elasticity Analysis…………………………………………… ……………………………… 5.3.1 Competing Destination………………. ……………………………… 5.3.2 Intervening Opportunities…………… ……………………………… 5.3.3 Distance………………………………….. ……………………………… 5.3.4 Origin Personal Income per Capita…….. ……………………….. 5.3.5 Origin Population……………………… …………………………….. 5.3.6 Origin Employment……………………. ……………………………… 5.3.7 Origin Value-Added……………………. ……………………………… 5.3.8 Origin Wholesale Employment…….. ……………………………… 5.3.9 Origin Average Plant Size…………… ……………………………… 5.3.10 Destination Manufacturing Employment…… ………………….. 5.3.11 Destination Wholesale Employment…………. ………………….. 5.3.12 Destination Personal Income Per Capita…… ………………….. 5.3.13 Destination Population………………. ……………………………… 5.3.14 Summary………………………………… ………………………………

62 62 63 65 67 68 70 71 72 74 76 77 79 80 82 84 85 86 95 95 96 97 98 99 100 101 102 103 104 105 106 107 108

6.

Conclusions…………. ……………………………………………………………………….

109

Bibliography………. ………………………………………………………………………………

114

Appendixes : A B C D E

Standard Transportation Commodity Classification (STCC)….. Standard Classification Of Transported Goods (SCTG)………… Descriptive Statistics…………………………………………………….. List Of Us. Custom Districts…………………………………………… Descriptions Of Databases……………………………………………… E.1 1993 and 1997 Commodity Flow Surveys………………… E.1.1 1993 Commodity Flow Survey…………………… E.1.2 1997 Commodity Flow Survey…………………… E.2 County Business Patterns……………………………………… E.3 Census of Manufactures…………………………………………

10

……………….. ……………….. ……………….. ……………….. ……………….. ……………….. ……………….. ……………….. ……………….. ………………..

117 122 124 170 171 171 171 175 176 179

LIST OF TABLES Table

Page 2

2.1

Reed’s Supply Model Variables and Their R ……………………………………….

21

2.2

Huxley’s Forms of Gravity Model……………………………………………………….

25

4.1

Magnitudes of Missing Observations in the 1993 CFS……………………………

46

4.2

Magnitudes of Missing Observations in the 1997 CFS……………………………

47

4.3

Matching of the SCTG and STCC Groups……………………………………………

49

4.4

Commodity Groups Codes and Definitions…..……………………………………..

50

4.5

Descriptive Statistics for 1993-All Commodities Combined…………………..

52

4.6

Descriptive Statistics for 1997-All Commodities Combined……...……………

53

5.1

Total Value of Shipments and Their Shares Across Commodity Groups…….

55

5.2

1993 States’ In- and Out-shipments and Their Shares…………………………..

56

5.3

1997 States’ In- and Out-shipments and Their Shares…………………………..

58

5.4

1993 Custom States Imports and Export Shares…………………………………

60

5.5

1997 Custom States Imports and Export Shares………………………………….

61

5.6

Food and Kindred Products Regression Parameters……………………………..

63

5.7

Lumber and Wood Products Regression Parameters……………………………

64

5.8

Furniture and Fixture Products Regression Parameters…………………………

66

5.9

Pulp, Paper and Allied Products Regression Parameters……………………….

67

5.10

Chemical and Allied Products Regression Parameters…………………………..

69

5.11

Petroleum or Coal Products Regression Parameters……………………………..

70

5.12

Rubber and Misc. Rubber Products Regression Parameters…………………..

72

5.13

Clay, Concrete, and Stone Products Regression Parameters..…………………

73

5.14

Primary Metal Products Regression Parameters…………………………………..

75

5.15

Fabricated Metal Products Regression Parameters……………………………….

77

5.16

(Non-Electrical) Machinery Products Regression Parameters………………….

78

5.17

Electrical Machinery Products Regression Parameters………………………….

80

5.18

Transportation Equipment Regression Parameters……………………………….

81

11

5.19

Precision Instruments Regression Parameters…………………………………….

83

5.20

Miscellaneous Products Regression Parameters…………………………………..

84

5.21

Apparel, Textile, Leather Products Regression Parameters…………………….

86

5.22

The 1993 Model Variable Coefficients and Their Significance Levels Across Commodity Groups (20-32)……………………………………………………

5.23

The 1993 Model Variable Coefficients and Their Significance Levels Across Commodity Groups (33-75)……………………………………………………

5.24

92

The 1997 Model Variable Coefficients and Their Significance Levels Across Commodity Groups (20-32)……………………………………………………

5.25

91

93

The 1997 Model Variable Coefficients and Their Significance Levels Across Commodity Groups (33-75)……………………………………………………

94

5.26

Statistics for CD Elasticities……………………………………………………………..

96

5.27

Statistics for IO Elasticities………………………………………………………………

97

5.28

Statistics for DIST Elasticities…………………………………………………………..

98

5.29

Statistics for OPIPC Elasticities…………………………………………………………

99

5.30

Statistics for OPOP Elasticities…………………………………………………………

100

5.31

Statistics for OEMP Elasticities…………………………………………………………

101

5.32

Statistics for OVLAD Elasticities………………………………………………………..

102

5.33

Statistics for OWSEM Elasticities………………………………………………………

103

5.34

Statistics for OAPS Elasticities………………………………………………………….

104

5.35

Statistics for DMNEM Elasticities………………………………………………………

105

5.36

Statistics for DWSEM Elasticities………………………………………………………

106

5.37

Statistics for DPIPC Elasticities………………………………………………………..

107

5.38

Statistics for DPOP Elasticities…………………………………………………………

108

6.1

Performance of Custom District Variables in 1993……………………………….

111

C.1

Descriptive Statistics for 1993-Commodity 20…………………………………….

124

C.2

Descriptive Statistics for 1993-Commodity 24…………………………………….

125

C.3

Descriptive Statistics for 1993-Commodity 25…………………………………….

125

C.4

Descriptive Statistics for 1993-Commodity 26…………………………………….

126

C.5

Descriptive Statistics for 1993-Commodity 28…………………………………….

126

C.6

Descriptive Statistics for 1993-Commodity 29…………………………………….

127

C.7

Descriptive Statistics for 1993-Commodity 30…………………………………….

127

C.8

Descriptive Statistics for 1993-Commodity 32…………………………………….

128

C.9

Descriptive Statistics for 1993-Commodity 33…………………………………….

128

12

C.10

Descriptive Statistics for 1993-Commodity 34…………………………………….

129

C.11

Descriptive Statistics for 1993-Commodity 35…………………………………….

129

C.12

Descriptive Statistics for 1993-Commodity 36…………………………………….

130

C.13

Descriptive Statistics for 1993-Commodity 37…………………………………….

130

C.14

Descriptive Statistics for 1993-Commodity 38…………………………………….

131

C.15

Descriptive Statistics for 1993-Commodity 39…………………………………….

131

C.16

Descriptive Statistics for 1993-Commodity 75…………………………………….

132

C.17

Descriptive Statistics for 1997-Commodity 20…………………………………….

132

C.18

Descriptive Statistics for 1997-Commodity 24…………………………………….

133

C.19

Descriptive Statistics for 1997-Commodity 25……….……………………………

133

C.20

Descriptive Statistics for 1997-Commodity 26…………………………………….

134

C.21

Descriptive Statistics for 1997-Commodity 28…………………………………….

134

C.22

Descriptive Statistics for 1997-Commodity 29…………………………………….

135

C.23

Descriptive Statistics for 1997-Commodity 30……….……………………………

135

C.24

Descriptive Statistics for 1997-Commodity 32……….……………………………

136

C.25

Descriptive Statistics for 1997-Commodity 33……….……………………………

136

C.26

Descriptive Statistics for 1997-Commodity 34……….……………………………

137

C.27

Descriptive Statistics for 1997-Commodity 35……….……………………………

137

C.28

Descriptive Statistics for 1997-Commodity 36……….……………………………

138

C.29

Descriptive Statistics for 1997-Commodity 37……….……………………………

138

C.30

Descriptive Statistics for 1997-Commodity 38……….……………………………

139

C.31

Descriptive Statistics for 1997-Commodity 39……….……………………………

139

C.32

Descriptive Statistics for 1997-Commodity 75……….……………………………

140

C.33

1993 Flow Values and Percentages across States and Commodities……….

141

C.34

Three-Digit Breakdown of 1993 Commodity Flows………………..………………

165

E.1

Variables included in the CFS…………………………….…..…………..……………

174

E.2

Tables in the Commodity Flow Survey…………………..……………………………

175

E.3

Variables in the U.S. data files………………………………………………………….

177

E.4

Variables in States Data File……………………………….……………………………

177

E.5

Variables in County Data Files……………………………..……………………………

178

E.6

Geographic Area Series……………………………………..……………………………

180

E.7

Final General Summary Database Relevant Files …..……………………………

180

13

CHAPTER 1 INTRODUCTION While intercity and intracity passenger flows have been the subject of extensive research

in

the

field

of

urban

and

regional

planning,

geography,

economics,

and

engineering, commodity flows, or freight transportation, have not been analyzed to the same extent. One of the main reasons for this lack of research has been

the relative

unavailability of suitable data, even in developed countries. Although inter-industry

flow

data are available in input-output tables at the national level, this type of data configuration does not allow for interregional origin-destination analyses. Prior to 1993, the most recent commodity flows survey performed in the US was for the year 1977, with data difficult to access and not in electronic format. There has also been a dearth of such data in other countries, as demonstrated by the very limited number of related empirical studies described further on (e.g. India; Great Britain; Alberta, Canada). However, the Bureau of Transportation Statistics, a joint unit of the US department of Transportation and the US Bureau of the Census, has recently released the results of the 1993 and 1997 Commodity Flows Surveys (CFS), making them widely available in electronic form on CDROM. The structure of these data is very suitable for empirical origin-destination analyses of commodity flows, and makes it feasible to develop and test new empirical models aimed at explaining the variations of these flows. Why is it important to better understand the structure of interregional commodity flows?

Fist,

the

ability

to

forecast

such

flows

may be critical for transportation

infrastructure planning, whether highways, railroad tracks, or river/port facilities. Second, a better understanding of interregional and inter-industry dependencies may be important for designing regional development policies aimed at reducing regional disparities through judicious location of commodity-producing activities. Third, commodity flow analyses may be useful to assess the trading area of specific activities at specific locations, and thus may guide location decisions by entrepreneurs. 14

Spatial flows of commodities have been empirically investigated with three types of models: interregional input-output models (Leontief and Strout, 1963; Isard, 1951; Moses, 1955), linear programming (LP) models (Chisholm and O’Sullivan, 1973), and spatial interaction (gravity) models (Ashtakala and Murthy, 1988; Black, 1971 & 1972; Chisholm and O’Sullivan, 1973; Huxley, 1979; Reed, 1967). The main problem with input-output models is that they are data hungry and, since they assume that technology is fixed over time, they are not very dependable for long-term forecasting. LP models are only applicable to highly homogenous commodities, as they do not allow for cross-hauling. Spatial interaction models have had a very wide range of applications, since

they are

easier to implement, but generally lack theoretical foundations. An alternative to the above empirical approaches might be the development and numerical implementation of a spatial price equilibrium (SPE) model, as proposed initially by

Samuelson

(1952),

and

reformulated

as

a

quadratic

programming

problem

by

Takayama and Judge (1964). Samuelson shows that as long as there is a price differential between regions, commodity flows will take place from low-price regions to high-price regions until the price differentials are equal to the transportation costs between the regions. However, implementing such models requires enough observations to empirically estimate local demand and supply functions for any commodity considered. Such data requirements all but preclude the implementation of this approach. Further, the SPE model assumes perfectly homogenous commodities, as cross-hauling is impossible. This requirement is in conflict with the aggregate nature of available data, where cross-hauling patterns are prevalent. In light of the above considerations and of the availability of recent CFS data, the focus of this research is on the development and estimation of spatial interaction (SI) models of interregional commodity flows. While SI models have long lacked sound theoretical foundations, Brocker (1989) recently rehabilitated them by demonstrating that some forms of SI models can be viewed as reduced forms of Samuelson’s SPE model. Thus,

using

spatial

interaction

models

to

analyze

commodity

flows

has

become

theoretically justifiable, beyond mere empirical convenience. Using Brocker’s theoretical framework, this dissertation attempts to expand the empirical research on interregional commodity flows. It specifies a spatial interaction model that incorporates (1) variables similar to those used in studies,

(2)

variables

used

past commodity flow

in international trade models, and (3) a set of

completely

new variables. The selection of the variables is consistent with Brocker’s framework and with inter-industry transactions considerations. For instance, the origins and destinations 15

are characterized by proxy variables representing

final and intermediate demands as

mass variables, and adjacency and custom district dummy, distance, destination 1999)

competing

(Fotheringham, 1983a, 1983b), and intervening opportunities (Guldmann,

variables are also considered. From a specification viewpoint, instead of the

multiplicative

functional

form

used

in

the

past,

a

flexible

Box-Cox

transformation

specification is used in this research. The specified model is used to analyze and compare interregional commodity flows in 1993 and 1997. The geographical coverage is the 48 US continental states, and the industry coverage is 16 two-digit manufacturing sector product groups. Using the econometric estimates, an elasticity analysis of the commodity flows with respect to each independent variable is conducted. Some of the research questions to be answered by this research can be stated as follows: •

Does adjacency increase commodity flows between states?



Does having a custom district significantly affect a state in- or out-shipments ? Or, in other words, is

foreign trade an important determinant of interstate commodity

flows? •

What is the effect of distance on interregional trade flows, and how does this effect vary across commodities?



What kind of effects do the spatial configurations of the origins and destinations have on commodity flows? Are these effects “competitive” or “agglomerative”?



What are the effects of the final demand sectors at the origins and at the destinations on commodity flows?



Are

wholesale activities at the origins and destinations facilitators of interstate

commodity trade? •

What is the importance of the employment and value-added of the commodityproducing sector for commodity outshipments?



Are commodity outshipments impacted by scale or diversification effects at the establishment level?



What is the importance of intermediate demand at the destination in determining commodity flows ?



What are the sensitivities of commodity flows to all these variables ? How do these sensitivities change over the years ? 16

The remainder of the dissertation is organized as follows. Chapter 2 is devoted to a thorough literature review. Different types of commodity flow models are reviewed, including input-output,

optimization, and

spatial interaction models,

both empirical and

theoretical. Since the study uses spatial interaction modeling, the broader literature

and

some specific applications of these models are also reviewed. Chapter 3 discusses the theoretical

background

and

methodological

approach

of

the

study,

including

the

specification of the model and the selection rationale and expectation for the explanatory variables. Chapter 4 describes the data used in the study, including approaches to deal with missing observations. Chapter 5 presents and discusses the econometric estimates and elasticity analyses. Conclusions are summarized in Chapter 6.

17

CHAPTER 2 LITERATURE REVIEW Three main groups of articles related to (1) commodity flow models, (2) spatial interaction models, and (3) international trade models, are reviewed in this chapter. The articles related to commodity flows are centrally related to this research, and include input-output models, optimization models, and spatial interaction commodity models. The second group of articles is related to the general structure of spatial interaction models, particularly models accounting for the effect of the spatial structure. The third groups of article focus on international trade models, in particular

applications of spatial interaction

modeling to international trade, which can be viewed as a specific type of interregional commodity flows. 2. 1. Commodity Flows Models 2. 1. 1. Interregional Input-Output Models Input-output analysis, developed by Wassily Leontief in the late 1930’s, is a general framework to analyze interdependency among industries in an economy. Since then, many extensions of the input-output model have been developed to analyze interregional

and

inter-industrial

relationships

within

a

multi-region

and

multi-industry

economy (Miernick, 1963; Richardson 1972). Three main interregional input-output frameworks have been proposed. The first is presented in the article

“Interregional and Regional Input-Output Analysis: a Model of

Space Economy” by Isard (1951). This model is also known as the “pure interregional input-output model”, wherein each of the commodities in each region is treated as a distinct commodity. In other words, alcoholic beverages in Kansas, for example, are treated as different from alcoholic beverages in Louisiana. With this kind of specification, not only is the spatial interdependency of an economy analyzed, but so is also its industrial interdependency. In a sense, it is a simultaneous inter-spatial and interindustrial input-output model. The main disadvantage of this model is that it is 18

ambitious

in terms of

data

requirements. In order

to be able

to calculate

the

technical input

coefficients, commodity flow data from each region to all other regions for each specific industry are necessary. Another study, by Leontief and Strout (1963), entitled “Multiregional Input-Output Analysis”, does not intend “to provide a systematic theoretical description of the many factors and relationship that ultimately determine the pattern of a multiregional economic system; it is designed, rather, as a rough and ready working tool capable of making effective use of the limited amount of factual information” (Leontief & Strout, 1963). In contrast to the pure interregional input-output model, where each commodity in each different region is treated as a distinct commodity, in this framework “it is as if the producers of a specific commodity or service located in one particular region had merged their output in a single regional supply pool, and the users of that commodity or service located in a given region had ordered and received it thorough a regional demand pool. All inter-regional movement of a particular commodity or service within a multiregional economy can thus be visualized as shipment from regional supply to regional demand pools of that good” (Leontief & Strout, 1963).

A gravity-like formula represents the

distribution of the commodities from these pools:

X i . gh =

X i . go X i .oh X i .oo

Qi . gh ,

(2.1 )

where X i.gh = total shipment of good i from the supply pool in region g to its demand pool in region h, X i.oh = demand pool of good i in region h, X i.go = the supply pool of good i in region g, and Q i.gh = empirical constant. Four different estimating methods are suggested for determining the empirical constants. These are called the ‘Exact Solution’, the ‘Simple Solution’, the “Least Squares” and the ‘Point Estimate’ procedures. Parameters are estimated using 1954 US data for four commodity groups, and 13 regions. The model is used for forecasting flows in 1950, 1952, and 1958, and produces high error rates (Leontief & Strout, 1963), which make the model unreliable for long-term regional forecasts, since it is not possible to tell which method of estimation of the interregional coefficients is superior to the others in terms of commodity groups, types of movement, and forecast years. 19

Moses (1955) proposes The Column Coefficient Input-Output Model in his article “The Stability of Interregional Trading Patterns and Input-Output Analysis”.

Like the

Leontief & Strout model, this model is only interested in explaining the interregional flows of

commodities.

Inter-industrial

flows

among

regions,

except

for

intra-region

inter-

industrial flows, are not analyzed in this type of model. The mirror image of the previous model is the Row Coefficient Model. The difference between the models is that trade coefficients are estimated by using row sums, which

represent the total production of a

commodity, instead of using column sums, which represent

the total consumption of this

commodity. The main deficiency of the row coefficient input-output model, like the point estimate procedure of the multiregional input-output model (the Leontief & Strout model), is that it can produce inconsistent, negative flows, which are economically meaningless (Bon, 1984). Regardless of the framework, the main weakness of all input-output models is the assumption of fixed production and/or trade coefficients. Relative supply prices and the state of technology are not stable over time. Thus, any change in these factors alters the trade and/or production pattern of the economy. This fact brings limitations to the longterm predictive power of the input-output model. Besides these shortcomings, it is worth mentioning two of Hua’s critiques concerning the input-output model. According to Hua (1990), ”The matrix inverse, however, is the main attraction of an input-output model, because its elements measure detailed marginal effects of any change in final demand. In either case, the model user has little flexibility in using different kinds of information, in intervening and modifying the structure of the model, or in dividing the whole operation into smaller components. Second, in an input-output inter-regional trade model, only the final demand for goods in each region can be the driving force of inter-regional trade flows. This limitation is unsatisfactory from a theoretical point of view if the reason for using the model is to study market interactions or regional growth, when factors on the supply side need to be analyzed and formally linked. There is little room in the existing input-output inter-regional trade models for accommodating such a linkage, a serious shortcoming that seems to have been underemphasized in the input-output inter-regional trade literature” (Hua, 1990).

2. 1. 2. Spatial Interaction Models of Commodity Flows One of the first empirical studies in the commodity flow literature is

“Areal

Interaction in India: Commodity Flows of the Bengal Bihar Industrial Area” by Reed (1967).

The purpose of the study is to analyze the interaction of the Bengal Bihar area 20

with the rest of India. The study area is a sub-region of the Calcutta economic region and is located in the northeastern part of India. The data are collected at the 24 rail stations of the region during three months in 1962 (April, July, and October), and consist of origins, destinations, and tonnages of goods shipped-out or received by rail, for 47 distinct commodity groups. Two separate models, one for outflows from the Bengal-Bihar region (or demand), and one for inflows to this region (or supply), are used. As a first step, the outflows of all and each individual commodity groups to destination k, F1k, are regressed across the urban population, popk,

at

destination k, considered as the only demand

variable, and distance dk. The model is

F1 k = a ( pop k ) b1 (d k ) b 2 ,

(2.2 )

where a, b1, and b2 are parameters. The R2 is .53 for total outflow, and varies between .38 and .50 for individual commodity groups. The variables of the inflow model are the distance and urban population at the origin k for the total of all commodities, and the employment (or production) at k and distance for specific commodity groups. The inflow model is :

Fk 1 = a( employment / production k ) b1 ( d k ) b 2 ,

(2.3 )

The supply models, their mass variables, and their R2’s are given in Table 2.1. Table 2. 1. Reed’s Supply Model Variables and Their R 2

Dependent Variable All Commodities

Independent Mass Variable Urban Population

R2 .49

Iron and Steel

Steel Production

.40

Electrical Machinery

Electrical Machinery Employment

.52

Other Machinery

Other Machinery Employment

.48

Chemicals

Chemical Employment

.34

Cement

Cement Employment

.26

Limestone

Limestone Production

.37

Metallic Ores

Metallic Ore Production

.41

Non-Metallic Minerals

Non-Metallic Mineral Production

.34

Gram

Gram Production

.21

21

Reed suggests that governmental regulation of commodity pricing, transportation subsidies, competition among transportation industries, licensing practices, branch plant ownership, and other specific inter-firm linkages, could be sources of variation in commodity flows, and these factors are not easy to model. Further, he suggests that “the effects of intervening and otherwise competing supplies and demands may be possible sources of variation not effectively summarized by the distance measure used in this study…measures of relative location based upon the potential model with its distance variable recognize neither the uneven distributions of demand nor the directionality of movement found within an economic system” (Reed, 1967: 171-172). In this context two additional variables are defined to capture competition and intervening opportunities effects. The contribution of demand at a given point k to the total demand potential at i is formulated as

( Pk / d ik ) / ∑ ( Pj / d ij ) , and may be viewed as measuring the competitive j

position of k vis a vis all other points j for supplies available at i.

Alternatively, the

contribution of supplies at a given point k to the total supply potential at j is formulated as

( S k / d jk ) / ∑ ( S i / d ji ) , and may measure the competitive position of supplies at k vis a i

vis all other supply points i. Furthermore, two additional variables are defined to capture redistribution effects and concentration effects. It is stated that “along with the demand effect created by conditions within k itself, one might expect that an additional demand effect would be created by k’s ability to redistribute commodities received from elsewhere. Thus a better account of demand at k should result from the inclusion of both a measure of demand generated at k and a measure of k’s access to outside markets (redistribution effect) such as a market potential at k. Similarly, besides the supply produced locally at i, one might expect an additional supply effect based on i’s ability to concentrate within tiself supplies produced elsewhere. A better accounting of supply at i would then include a measure of supply produced at i and a measure of i’s access to other supplies (concentration effect) such as supply potential.” (Reed, 1967:181). The defined as

∑P

n

redistribution

d kn , and the concentration effect as

n

∑S

n

effect

is

mathematically

d kn .

n

Based on the above definitions, extended forms are proposed for both outflows and inflows. The extended full model is:

22

b1 i

b2 k b3 ik

S .P d

Fik = a .

Si Pk S n b6 Pn b7 d ik b4 d ik b5 . ( ) .( ) . (∑ ) . (∑ ) Sn Pn n d in n d kn ∑n d ∑n d kn in

(2.4 )

Since there is only one region, the outflow model becomes:

F1k = a .

b2 k b3 ik

P d

S1 d 1k Pn b7 . ( )b4 . ( ∑ ) Sn n d kn ∑n d kn

(2.5 )

and the inflow model becomes:

Pj Fk1 = a .

S d

b1 k b3 k1

.(

d k1 b5 S n b6 ) . (∑ ) Pn n d kn ∑n d kn

(2.6 )

Variables of supply and demand potentials, and variables of concentration and redistribution

effects

are

computed

using

various

supply

and

demand

measures.

Equations (2.5) and (2.6) are estimated using stepwise regression. Redistribution and concentration effects are not statistically significant.

Overall, Reed concludes that “the

expanded potential model did explain some additional variations in study area flows, strongly indicating that this model might prove useful in analyzing flows between all parts of an economic system.” (Reed, 1967:195) Another empirical commodity flow study is by Chisholm and O’Sullivan (1973). The title of the study is “Freight Flow and Spatial Aspects of the British Economy”.

The

authors, using U.K. Ministry of Transport 1962 and 1964 commodity flow data, try to analyze commodity flows in Britain over 78 zones and 13 commodity groups.

The main

difference between this study and Reed’s is that this one is divided into two steps. In the first step, the authors attempt to explain zonal freight generation at an aggregate level. In the second step,

two models are used, the Gravity Model (GM) and The Linear

Programming (LP) model. Because of its computational simplicity, the single-constrained GM is retained. The R2’s obtained by comparing actual and estimated flows using the gravity model for 13 commodity groups for road trips vary between .24 for steel and .62 for food. These results are not found satisfactory, and it is concluded that “the relatively low values of the R2 for the commodity classes indicate that the gravity model is not very suitable for examining flows disaggregated by type of goods” (Chisholm & O’Sullivan, 23

1973: 76). On the other hand, the R2’s for the LP solutions for road trips are higher, especially for homogenous commodity groups. The higher the

homogeneity, the higher

the R2. At the same time, the LP gives good R2 values for the flows between 24 major cities. The employed gravity model is

Tij = kOi D j

−α

d ij− β ,

(2.7)

where Tij represents the flow between origin i and destination j, Oi

and Dj the origin and

destination mass variables, and dij the distance between i and j. The mass variables used in the study are population and employment. The primal formulation of the LP is:

Min.C = ∑ ∑ Tij d ij i

s.t:

(2.8 )

j

∑T

ij

= Dj,

ij

= Oi ,

i

∑T j

Tij ≥ 0. The dual formulation is:

MaxV = ∑ V j D j − ∑ U i Oi j

(2.9 )

i

V j and Ui are the shadow prices associated to supplies at origins and requirements at destinations — reflecting locational advantages with respect to transport costs. The shadow prices obtained in solving the LP are thus indicative of geographical comparative cost advantages for production and consumption” (Chisholm & Sullivan, 1973:76). In paper entitled “Indirect Estimation of Interregional Trading Patterns for InputOutput Analysis: Empirical Results for the Gravity Model and Rail Freight Shipments”, Huxley (1979) tries to find the best estimator for the inter-regional flows

component of

the input-output table for 11 regions in Queensland (Australia). The data for this inputoutput study are gathered through a comprehensive survey. The goal is to find a more economical method for predicting future commodity flows among regions and for updating input-output tables. In a first step, the elements of all the main components of the inputoutput table, such as the interregional trade flows matrices, the intra-regional trade flows matrices, the final demand, the primary input, and the gross regional output vectors are regressed across regional employment. The results are, except for interregional trade 24

flows, satisfactory. The R2 is equal to .55 for interregional trade flows and above .99 for the other quantities. In a second step, Huxley attempts to specify a functional form for the gravity model to project interregional flows. The models are specified as linear and logarithmic. In these functional forms, the flows are directly regressed on the mass variables. Then, some interaction forms of these mass variables are defined, and included in the models, using a stepwise regression procedure. The specified models, and the interaction variables are presented in Table 2. 2. Table 2. 2. Huxley’s Forms of Gravity Model

Linear Forms Yij = a + b Xi

Logarithmic Forms Yij = aXib

Interaction Variables X 1 = Yij

Interaction Variables X 8 = Xi / Xj

Yij = a + bXj

Yij =aXjb

X 2 = Xi

X 9= Xj / Xi

Yij = a + bD

Yij =aD

b

X 3 = Xj

X 10 = Xi /D

Yij = a + b1X i + b2X j

X4 = D

X 11 = Xj /D

X 5 = Xi X j

X 12 = Xi X j /D

Yij = a + b1X j + b3D

Yij =aXib1X jb2 Yij =aXib1Db3 Yij =aXjb2Db3

X 6 = Xi D

X 13 = Xi X j D

Yij =a+b 1X i+b 2X j+b 3D

Yij =aXib1X jb2Db3

X 7 = Xj D

X 14 = (X i + Xj)/D

Yij = a + b1X i + b3D

X i and X j are variables that represent masses, and D is the highway distance.

The

following variables are used as measures of masses; Employment

:

Total regional employment at the origins and destinations

Population

:

Total population at the origins and destinations

GRP

:

Gross regional products at the origins and destinations

GRO

:

Gross regional output (includes inter-industry transactions)

PIIFDJ

:

Primary inputs of the region and final demands of the destinations

INTERMED

:

Intermediate sales of the origin and intermediate purchases of the destination

NTEMPL

:

Non-tertiary

regional

employment

at

the

origin

and

destination NTGRO

:

INTRAN

:

Non-tertiary gross regional outputs Intermediate sales of the origin minus intra-regional sales and intermediate purchases of the destinations minus intraregional purchases

EMPDENS

:

Total employees per square mile.

The fourth equation of the Linear Forms, and the seventh equation of the 25

Logarithmic Forms produce the best fit, with the employment as the measure of mass, with R2 of .56 and .57, respectively. However, the best fit with the employment interaction variables from the stepwise regression is reached with R2=0.78. The significant variables in the equation are X 2, X5, X6, X7, X9, X11, and X 13 (see Table 2.2).

The other mass

variables results are not reported. However, Huxley concludes that the gravity model is not a good predictive device for the indirect estimation of inter-industry trade flows among regions. As this study uses rail data, this conclusion confirms the findings of Chisholm and O’Sullivan. Another result of this study is that distance is not a significant variable to explain freight flows. This result is also reached by Reed (1967). study, a series of additional regressions are run.

In a further step of the

“In the first series, dollar flows were

regressed against tonnage flows coming from the opposite direction, based on the idea of the circular flow of income: goods flowing from region i to j should result in dollars flowing from j to i…a second set of regressions was also run, consisting of dollar flows against tonnage flows in the same direction” (Huxley, 1979:34). However, none of the results were significant. In two subsequent papers, Black (1971, 1972) analyzes the properties and determinants of the distance exponent in the gravity model, using the 1967 U.S. commodity flow survey for 24 major shipper groups. The first article, entitled “Utility of the Gravity Model and Estimates of its Parameters in Commodity Flow Studies”, defines the model as

T = k ij

S ik D kj Fijk

∑D

k j

Fijk

,

(2.10)

j

where, Tijk = Tons of commodity k produced in region i and shipped to region j; Sik = Total shipments of commodity k from region i; Djk = Total demand in region j for commodity k; and Fijk = A friction factor which is equal to 1/d ijb, where dij is the straight line distance between regions i and j, and b is an empirically-derived exponent which may vary across commodity groups. Three hypotheses are formulated concerning the exponent of the friction factor: (1)

The greater the regional specialization for a particular commodity, the lower the exponent of distance.

(2)

The greater the interregional flow of a particular commodity, the greater the 26

exponent of distance. (3)

The greater the per-unit value of a commodity, the lower the exponent of distance for that commodity. The correlation coefficient is 0.55 under hypothesis (1), 0.73 under hypothesis (2),

and 0.51 under hypothesis (3). In the second paper (Black, 1972),

entitled “Inter-Regional Commodity Flows:

Some Experiment with the Gravity Model”, the research questions are whether it is possible to determine the variables related to the variations in the distance exponent, whether it is possible to estimate the distance exponent using these variables, and whether the size of the study area has an effect on the distance exponent. Using the same data set and the same model as specified earlier, Black (1972) empirically obtains exponents values that are regressed across two variables: LMk = the local market for the kth commodity, or the share of the total flow n i the main diagonal of the kth commodity flow matrix; and CPk= the concentration of production for commodity k, or the share of total

flow shipped

by the largest shipping region. The R2 value for this regression is .93, and Black concludes that “(1) the greater the proportion of total shipments from the largest producer (or shipper), the lower the exponent, and (2) the greater the proportion of total flow which is local, the higher the exponent.” Ashtakala & Murthy (1988) use a production-constrained gravity model to forecast commodity flows in Alberta. They hypothesize that the flow between an origin and a destination is a function of production at the origin, consumption at the destination, and distance between them. The model is formulated as

Tij = Pi

C j d ij(λ )

∑C j

j

d ij( λ )

,

(2.11)

where Pi represents production at i, Cj the consumption at j, and dij is distance between i and j. Furthermore, dij (

d

(λ) ij

=

d ijλ − 1 λ

ë)

is the Box-Cox transformation, with

,

(2.12)

27

The only variable in the above model is ë, which is estimated through a grid iterative technique for each commodity group, minimizing the deviation between observed and computed flows. The maximum values of R2 are between 0.71 and 0.88 for six commodity groups. 2. 2. Overview of Spatial Interaction Modeling Spatial interaction models can be used for predicting behavior over space, and for design/planning purposes. According to Fotheringham and O’Kelly (1989), any movement or any communication over space as a result of a decision-making process establishes some sort of spatial interaction. The level of spatial interaction is a function of the level of economic development, and the level of economic development is a function of economic specialization. However, in the long run, these three phenomena are endogenous, and determinants of each other. There are four basic types of spatial interaction models: unconstrained, production constrained, attraction constrained, and doubly constrained.

Unconstrained models

provide information about origin and destination characteristics, production-constrained models address destination characteristics, and attraction-constrained models focus on origin characteristics. Doubly constrained (or origin and destination constrained) models are generally used for predictive purposes. All of the interaction models have four basic elements: a flow matrix, a friction or cost matrix, a matrix of origin propulsiveness measures, and a matrix of destination attractiveness variables

(Fotheringham and O’Kelly,

1989). The cost or friction function employed in spatial interaction models has either an exponential form or a power form. Fotheringham and O’Kelly make four suggestions concerning the use of the cost function. (1) The exponential function is scale sensitive, while the power function is relatively scale independent. Thus, a model with an exponential cost function calibrated for one location should not be used for another location. (2) If a multiplicative cost increase is expected for the cost matrix, an exponential cost function should be employed, and a power function should be used when an additive increase is expected. (3) The power function tends to overestimate low costs movements. (4) The exponential

function

produces

more

consistent

results

when

the

trip

makers

are

homogenous, while the inverse power function is better for a heterogeneous group of trip makers. In two subsequent articles, Fotheringham (1983a, 1983b) discusses a new dimension of spatial interaction modeling. According to Fotheringham, spatial interaction 28

models are often misspecified because the spatial structure is not properly represented in these models (1983a). The distance decay parameter is generally expected to represent spatial relationships, even though some studies indicate that distance is not a significant variable in some spatial interactions (Reed, 1967; Huxley, 1979). The basic issue is that two very different spatial configurations with identical origin/destination distances may produce identical flow patterns. To overcome the problem of representing the spatial configuration in the model, Fotheringham suggests the inclusion of an additional variable that represents the accessibility of the destination to all other destinations available to the origin, as perceived by the residents of the origin (Fotheringham, 1983a). Specifically, an origin-specific production-constrained competing destination model is given by

I ij = Z i Oi m j Aijδ i d ijβ i

(2.13)

where, Zi =

1 n

∑m

j

(2.14) δi ij

A d

βi ij

j

w

Aij =

∑m d k

αi ik

, (k i , k j)

(2.15)

k

Iij is the interaction volume between i and j; mj represents the attractiveness of destination j; dij is the distance between i

and j; Oi represents the known total outflow

from i; A ij is the competing destination variable; and Zi is a balancing factor that ensures that n

∑I j

n

^ ij =

∑I

(2.16)

ij

j

Although this was not noticed by Fotheringham, the competing destination variable Aij, was first mentioned by Reed in 1967 as being demand and supply potentials. While these variables are not exactly the same as those defined by Fotheringham, the idea of the inclusion of a variable that reflects the relative locations of

the destinations was

first mentioned by Reed in his study of commodity flows in India. In a second paper, Fotheringham (1983b) mentions two possible relationships among destinations: competition and agglomeration effects.

Under competition effects,

the interaction with a destination is lower if the destination is

part of a group of

competing destinations. The interaction is higher when the destination is part of a complementary destinations group, in which agglomeration effects take place. If either of these effects are present, the standard gravity model is misspecified. Interestingly, the 29

existence of different types of relationships over space is first mentioned by Edward Ullman (1967) in his “American Commodity Flow” study. Agglomeration effects are then termed “complementary”, and competition effects “intervening opportunities”. Ishikawa (1987), using Japan’s 1960 and 1980 migration, and 1980 university enrollment data, investigates the validity of the competing destination model. Two different

models,

(1)

the

production

constrained

model,

and

(2)

the

production

constrained competing destination model, are formulated. It is concluded that the production-constrained competing destination model provides a noticeable improvement in the R2. Fotheringham (1995) tests whether there is a directional variation in the distance decay parameter, using U.S.

migration data for 48 origins, and concludes that

there is no such directional variation. Finally, Guldmann (1999) tests Fotheringham’s competing

destination

model

with

telecommunication

data.

Following

Fotheringham

(1983a), he introduces into his model a competing destination (CD) factor that measures the accessibility of destination j to all (or a subset of) the other destinations. In addition, Guldmann also considers a similar intervening opportunities variable (IO), which basically represents the spatial configuration of origins. The model specified is:

ln MS ij = ln K 0 + α ln MS ji + β ln Dij + γ ln Pij + δ ln MSOi + ζ ln MSD j + ϑln Aij (2.17) ln MS ji = ln K 0 + α ln MS ij + β ln Dij + γ ln Pji + δ ln MSO j + ζ ln MSDi + ϑln A ji (2.18) where MSij is the flow variable, Dij the distance, Pij the price, MSO i and MSDj the total originating and terminating flows, respectively, and intervening opportunities

Aij the competing destinations or

variable. The system is estimated using the three-stage least-

squares procedure for simultaneous equations estimation, because of the reverse flow effect. The model is first estimated without any spatial structure variable, and later with four different specifications for such variables. The competing destination model provides the strongest improvement over the base model. 2. 3. International Trade Models Frankel and Wei (1998) employ an econometric model to gauge the effects that regional trade arrangements have on world trading patterns, since regional blocks may reduce world welfare if trade-diversion effects dominates trade-creation effects. The data set covers sixty-three countries (3,906 exporter-importer pairs). The specified model is:

30

log( Export ij ) = α + β1 log( GNPi ) + β2 log( GNPj ) + β3 log( GNPi / Pop i ) + β4 log( GNPj / Pop j ) + β5 log( Dist ij ) + β6 log( OverallDis t i ) + β7 log( OverallDis t j ) + β8 log( ADJACENCY ) + β9 log( LANGUAGE ) +

(2.19)

λ1 ( EC _ I ij ) + λ2 ( MERCOSUR _ I ij ) + λ3 ( ASEAN _ I ij ) Most of the variables in Equation (2.19) are self-explanatory. “Overall Distance” measures how far a country is from all other countries. “Adjacency” si a dummy variable for country pairs sharing a common border. “Language” is a dummy variable for countries have linguistic or colonial ties. “EC”, “MERCOSUR”, and “ASEAN” are also dummy variables for trade blocks, indicating the European Community, the Customs Union of the Southern Cone Countries in South America, and the Association of Southeast Asian Nations, respectively.

The results indicate that larger economies trade more but not

proportionately to their GNP. Bilateral distance also has a significant effect. Contiguity and having language commonality also facilitate trade. In terms of trade blocks, the results are mixed. Eichengreen and Irwin (1998) are also interested in whether regional trade blocks have

important effects on trade pattern, and test whether past trade has a significant

impact on current trade. The basic model is

Tij = β0 + β1 ln( Yi Y j ) + β2 ln( Pi Pj ) + β3 ln( DIST ij ) + β4 (CONTij ) (2.20) where Tij is the value of bilateral trade between countries i and j; YiY j is the product of the two countries’ national incomes (the so-called gravity variable); PiPj is the product of the two countries’ per capita incomes; DIST is the straight-line distance; and CONT is a dummy variable indicating whether the two countries are contiguous. After estimating the basic

model,

lagged

trade,

financial

block

and,

finally,

formal

colonial

relationship

variables are added to the basic model. It is concluded that these variables have an important effect In another study, Frankel & Romer (1999) empirically investigate the impact of international trade on standards of living. In the first part of the study, the geographic characteristics of countries are used to construct instruments for international trade, since geographic characteristics might have important effects on income and trade.

Then, these

constructed instruments are used to investigate the impact of trade on income. The estimated bilateral trade equation is

31

ln( Tij / GDPi ) = α0 + α1 ln Dij + α2 ln N i + α3 ln Ai + α4 ln N j + α5 ln A j + α6 ( Li + L j ) + α7 Bij + α8 Bij ln Dij + α9 Bij ln N i +

(2.21)

α10 Bij ln Aij + α11 βij ln N i + α12 Bij ln A j + α13 Bij ln( Li + L j ) where N is population, A is area, L a dummy variables for landlocked countries, and B a dummy variable for common border between two countries.

After

(2.21),

instrument

the

income

equation

is

estimated

using

the

estimating

Equation

constructed.

The

dependent variable, Y i, is the income per person. The model specification is

ln Y = a + bTi + c1 ln N i + c 2 ln Ai ,

(2.22)

where Ti is the trade share (Tij /GDP i), and Ni and A i are the population and the area of country i.

Frankel & Romer (1999) conclude that “trade raises income. The relationship

between the geographic component of trade and income suggests that a rise of one percentage point in the ratio of trade to GDP increases income per person by at least onehalf percent. Trade appears to raise income by spurring the accumulation of physical and human capital and by increasing output for given levels of capital…The point estimates suggest that increasing a country’s size and area by one percent raises income by onetenth of a percent or more.”

2. 4. Summary Although spatial interaction models have been extensively used in such areas as migration, commuting, shopping, and telecommunication, their application to empirical commodity flow modeling has remained very limited, most likely because of limited data availability. The few studies completed so far focus on best-fitting very simple models with little theoretical foundation, instead of searching for

better explanatory models with a

diversified number of variables. For instance, Black (1971 & 1972), Chisholm and O’Sullivan (1973), and Ashtakala and Murthy (1988),

employ a basic gravity model (with two mass and one friction

variables), and their focus is on estimating the exponent of the distance variable. Some try different proxies for the mass variables (Huxley, 1979) while others use total zonal inflows and outflows as mass variables. In addition to these simple models, Reed (1967) adds two more

variables:

supply/demand

potential,

and

redistribution/concentration

effects

variables. Except for Reed, no one has attempted to account for the effects of the spatial configuration of the nodes, despite the risk of model misspecification. Although the mathematical formulation of Reed’s spatial configuration variables is similar to the ones specified in Fotheringham (1983a-b), Guldmann (1999) , and Ishikawa (1987), these 32

variables were not intended to measure competing destinations effects but rather redistributive/concentration effects. The above discussion points to the need to consider expanded spatial interaction models of interregional commodity flows, both in terms of the set of explanatory variables and of the functional form used to relate flows to their determinants. The selection of the variables should be guided by a theoretical framework, and the choice of functional form should be guided by flexibility and best fit with the observed data. The research presented in the following chapter is consistent with these goals.

33

CHAPTER 3 MODELING METHODOLOGY The purpose of this study is to expand empirical research on interregional commodity flows. A theoretical framework is first developed, based upon the Spatial Price Equilibrium model developed by Samuelson (1952), and its reformulation by Brocker (1989). This framework provides guidance for the selection of the explanatory variables. Some of theses variables have been

used in previous commodity flow studies, including

population, sectoral employment, and distance. Other variables have been used in other applications of spatial interaction modeling, and are adapted, for the first time, to interregional

commodity

flow

modeling,

including

competing

destinations,

intervening

opportunities, per capita personal income, and adjacency dummy variables. A third group represents

variables that have never been used in

spatial interaction modeling of

commodity flows, including wholesale employment, sectoral value-added,

average plant

size, and custom district dummies. Finally instead of using the standard multiplicative functional form, a flexible Box-Cox transformation is used to select the best functional form.

3. 1. Theoretical Background The Spatial Price Equilibrium (SPE) Model developed by Samuelson (1952) provides

a

consistent

theoretical

framework

for

trade

in

a

multi-regional spatial

configuration, where commodity flows take place from high-price regions to low-price regions until equilibrium is reached, with price differentials between regions equal to transportation costs. This basic principle is valid for all commodities among all regions as long as the regions are economically connected. Samuelson (1952:290)

defines a net social pay-off (NSP) function. For a two-

region case, the defined NSP is E12

− E12

0

0

NSP = − ∫ s1 ( x ) dx −

∫s

2

( x) dx − t12 ( E12 ) 34

(3.1 )

where s1(x) and s2(x) are the excess supply functions of the two regions (differences between local

demand and supply functions).

E 12 represents the flow between the two

regions, and t12 is the transportation cost between the regions. A multi-regional NSP function is then formulated as (Samuelson, 1952:292): n

NSP = ∑ S i ( Ei ) − ∑∑ t ij ( Eij )

(3.2 )

i< j

1

where S i is the excess supply function of region i, E i is the total output of region i, and tij is the

the transportation cost between regions i and j. Given the excess supply functions and transportation

costs,

Samuelson

shows

that

solving

the

optimization

problem,

consisting in maximizing the NSP, produces (1) the equilibrium prices in each region, and (2) the equilibrium volumes and directions of trade between regions.

At the NSP

maximum, the following condition holds:

− Tij ≤ s i ( Ei ) − s j ( E j ) ≤ Tij ,

(3.3 )

(for all i, j= 1,…, n)

which states that flows among regions will take place as long as price differentials between regions are greater than transportation costs between regions. Despite its sound theoretical framework, this formulation presents three important practical problems: “first, the problem of specifying functional forms, which are a good compromise between the requirements of theoretical consistency, flexibility, parsimony, computational facility and factual conformity; second, the problem of estimating the model parameters with available data; and third, the problem of designing efficient algorithms for a numerical approximation of the equilibria” (Brocker, 1989:8). The first and second problems are related to the specification of the excess supply functions while the third problem is related to the solution of the maximization problem, for which Samuelson did not offer any specific solution algorithm or methodology. To overcome the latter, Takayama & Judge (1964) suggest a quadratic programming approach, assuming linear regional demand and supply functions. The estimation of regional functions with only one price variable would necessitate at least two observations for all regions and for the

commodity

at

stake.

The

goodness-of-fit

for

such

functions

would

be

highly

questionable. As the number of variables increases, the necessary data observations also increase. In a multi-regional and multi-commodity framework, it would be very difficult to obtain so many observations for an empirical study. However, beyond this empirical problem, there are other problems worth mentioning. (1) The basic assumption of the 35

model is that the commodity is perfectly homogenous and spatially uniform. However, commodities in real life are not

homogenous and uniform. The available commodity flow

data almost always do not have this fine level of disaggregation. (2) The model functions under the assumption of perfect spatial information, which fails in real life. (3) The previous factors cause cross-hauling of commodities in practice, which is theoretically impossible according to the SPE model. Brocker (1989) attempts to connect theory and empirical research in trade modeling, and shows that all forms of the gravity model (constrained, unconstrained, and elasticity constrained) are reduced forms of spatial price equilibria of interregional trade. To substantiate this claim, Brocker uses a modified version of the SPE model. In the standard SPE model, buyers satisfy demand at the cheapest supply points in terms of c.i.f prices. However, in Brocker’s model, buyers may choose certain suppliers depending on some characteristics other than c.i.f. prices. At each supply point i, there are firms supplying the commodities, and at each demand point j, there are firms and households demanding

certain quantities

(y1j, y2j, …, yij,… yIj)

from the supply points (i=1->I).

The supply firms are faced with f.o.b. prices, and the buying firms and households with c.i.f. prices. This model consists of four equations. A real-valued supply function is defined as follows:

S i = σi ( p i , s i ),

(3.4 )

where Si is the supply quantity at supply point i, pi is the f.o.b. price at i, and si is a vector of other variables, such as prices of other commodities. ái is monotone, non-decreasing in pi. For each demand point j, there is a demand correspondence, a point-to-set mapping which assigns the vector of O-D flows terminating at j, yj = (y1j ,…,yij ,…,yIj), to the c.i.f. price vector, qj =(q1j ,…,qij ,…,qIj), so that

y j ε δ j (q j , w, d j ),

(3.5 )

where w is a vector of parameters that measure the supply characteristics influencing purchase choices, dj

is a vector measuring demand characteristics, including income,

prices of other commodities, etc. The third equation defines c.i.f. prices:

q ij = p i + c ij ,

(3.6 )

36

where cij is the transportation cost between i and j. And the fourth equation states the equilibrium conditions:

∑y

ij

= S i ∀ i.

(3.7 )

j

A spatial price equilibrium is characterized by prices and quantities satisfying (3.4) –(3.7), which

represent the explicit (or structural) form of the trade model, containing both prices

and quantities as endogenous variables. If prices are eliminated, we obtain

the reduced

form of the model, where equilibrium flows are directly assigned to the vector of exogenous variables, (s, w, d, c) = (s1,…, sI, w 1,…, w I, d1,…dJ, c11,…, cIJ). The reduced form is denoted by æ =( æ matrix Y = *

(y*1,…,

y*J),

11

,…, æ

ij

,…, æ IJ), so that, for any equilibrium flow

we have

Y* = æ (s, w, d, c).

(3.8)

Of course, there is no closed mathematical formulation of æ. One way to think about this function is to solve the equilibrium problem for a wide range of combinations of values for the input parameters (s, w, d, c), for instance over a grid. The resulting flow values Yij* could then be regressed over the input parameters, providing an approximation of the function æ. However, this would remain a purely numerical exercise, unless we are able (1) to select the proper values of (s, w, d, c) characterizing a real interregional setting (e.g., the US and its states), and (2) to assess the goodness-of-fit of the approach. While actual flow data may be available, the proper selection of the input parameters is an extremely difficult task because of data unavailability. An alternative approach is to view (3.8) as a general guide for the selection of simpler, and empirically estimable functional forms. Brocker (1989) shows that the generalized gravity form æij (s, w, d, c) = aij (s, w, d, c) f(cij ) bi (s, w, d, c)

(3.9)

is consistent with (3.8). Equation (3.9) suggests that the origin and destination factors, ai and bj , may be functions of whole vectors (s, w, d, c), and not only of the components of

37

these vectors that are associated with i or j, exclusively. In the standard gravity model, we would have ai = ai (si, w i),

(3.10)

bj = bj (dj) ),

(3.11)

that is, the mass factor at the origin is only a function of the supply variables at the origin, and the mass factor at the destination is only a function of the demand variable at the destination. Equation (3.9) clearly suggests that supply and demand variables associated to other locations k ( i, j) may be included in ai and bj . The following section presents the adaptation of model (3.9) to an empirically estimable model. 3. 2. The Empirical Commodity Flow Model The goal of this research is to empirically identify the determinants of the flows of 16 commodity groups between origin and destination states across the continental US for two separate time periods, 1993 and 1997, and to compare the results. Commodity flows between any two points are expressed as a function of three vectors of variables: a vector that characterizes the supply, a vector that characterizes the demand, and a vector that characterizes the friction between the origin and destination points. The origins serve as supply points, and therefore the variables chosen to represents the origin should be proxies for supply conditions as well as demand conditions at the origin. Likewise, the destinations serve as demand points, and destination variables should mainly be proxies for commodity demands, both intermediate and final. Standard friction variables include distance and adjacency. However, demand and supply conditions at locations other than the origin and destination must also be accounted for. They depend upon the overall spatial configuration (structure) of all origin and destination points. Competing destinations and intervening opportunities

variables may then be used

to capture these effects. In the following sections, the specific variables that make up the above vectors are described, the rationale for their selection is presented, and the choice of functional form for Equation (3.9) and the estimation procedure are discussed.

3. 2. 1. Variables a) Origin Variables The

variables

characterizing

the

origins

(states)

are

employment

in

the

(commodity group) sector (oemp), value-added in the (commodity group) sector (ovlad), 38

wholesale employment (owsem), personal income per-capita (opipc), total population (opop), and average plant size (oaps). Sectoral employment (oemp) and sectoral value-added (ovlad) are used as proxy variables for sectoral production at the origin, and represent supply characteristics. Their expected signs are positive, indicating that as sectoral employment and/or sectoral valueadded increase(s), the outflow of the commodity increases for any given destination. Wholesale employment (owsem) is used to measure the effect of redistribution activities on commodity out-shipment at a given origin. In other words, out-shipment of this commodity may not only be a function of local production activities but also function of redistribution activities taking place at that origin. As

a

wholesale employment

increases, the out-shipment of this commodity is expected to increase. Wholesale activities may also facilitate consumption of the commodity by the final demand sector at the origin. Thus, the expected sign of the coefficient of owsem is positive. Total population (opop) and personal income per-capita (opipc) are two proxy variables for demand conditions at the origin. Although the origins are supposed to be associated with supply conditions for commodity out-shipment, local final demand at the origin may have significant effects. Their expected signs are negative, implying that, as local final demand increases, the out-shipment of the commodity decreases due to increased local consumption. Of course, the extend of this effect would depend on supply elasticity. The average plant size (oaps) is estimated by dividing total sectoral employment by the total number of establishments in that sector. It is intended to capture scale or diversification effects in the industry. Theoretically, as the plant scale of an industrial sector increases, total production and thus total out-shipment in that industry are supposed to increase due to increased production efficiency. However, the two-digit level aggregation may not reflect this effect properly. In other words, the total amount of out shipments by small firms may outrun the out-shipment of the larger firms, because many smaller firms may be characterized by more product diversity, more attractive to export markets than a few larger firms. For this reason, this variable may either (1) have a positive sign, indicating that scale effects control

out-shipments or

that the out-shipment

market is dominated by a few large firms , or (2) have a negative sign, implying that the diversification effect dominates the industry or the market is shared by many small-scale diversified companies. [For a theoretical discussion of these effects, see Krugman (1980)].

39

b) Destination Variables The destinations are approximated by four main variables characterizing the demand: manufacturing employment (dmnem), wholesale employment (dwsem), personal income per-capita (dpipc), and total population (dpop). Manufacturing

employment

(dmnem)

is

the

proxy

for

intermediate

demand;

personal income per-capita (dpipc) and total population (dpop) are to measure final demand conditions;

and wholesale employment (dwsem) is a proxy to measure

redistributions effects at the destination. All of their expected signs are positive, implying that any increase in either of these variables will increase commodity flows to the destinations. c) Geographical Variables A total of six variables characterizing the geographical structure of the interaction space are included in this study: distance (dist), competing destination variable (cd), intervening opportunities variable (io), adjacency dummy (adjncy), origin custom district dummy (ocddmy), and destination custom district dummy (dcddmy). Distance is the most conventional friction variable used in all spatial interaction models. It takes different forms, like highway distance, great circle distance, etc. In this study, the average distances of all hauled commodities are used. The expected sign for the distance variable is always negative, indicating that the interaction between the origin and the destinations decreases as the distance between them increases. Two

specific

variables

configuration of states vis a vis

are

employed

to

capture

the

effect

of

the

spatial

each other: competing destination (cd), and intervening

opportunities (io) variables. These variables may be viewed as integrating into the model the demand/supply effects at locations other than the origin (i) and destination (j). With reference to Eqs. (3.8) and (3.9), these variables represent at least a portion of the vectors (s, w, d).The competing destination variable measures the accessibility of a specific destination to all other destinations. It is estimated using a destination total employment and the distance between two destinations. Mathematically it is expressed as follows:

CDij = ∑ k TEk / d kj

(3.12)

k (i,j)

There is no presumption about the sign of this variable. A negative sign indicates that there is competition among destinations, and as other destinations k get closer to destination j, the amount of the commodity terminating at destination j decreases. The opposite case, a positive coefficient sign,

implies agglomeration effects: flows increase as

other destinations get closer to destination j, and thus make it more attractive to flows. 40

The intervening opportunities (io) variable is defined by a formula similar to (3.12). The distance used in Eq. (3.12) is taken as dki instead of dkj . According to the intervening opportunities concept, flows to a destination decrease when the opportunities between the origin and the destination increase. Just like clusters at destinations, the io variable may be used to describe the spatial configuration of the clusters around origins. According to this idea, a positive sign indicates that when other origins are getting closer, thus implying an economic concentration around the origin, the flow to destination increases. This would suggest possible agglomeration effects at the supply level. However, a negative sign would suggest that the destinations in the origin clusters may act as competing destinations, thus reducing the flow to the destination. An alternative interpretation of a negative sign could be linked to agglomeration diseconomies. The higher the cluster, the larger the negative effects (e.g., congestion), hence the lesser the demand and the flow to destination. In addition, three dummy variables are used. First, the adjacency dummy (adjncy) is intended to measure whether having a common physical border has an effect on commodity flows between states. The expectation for the sign of this parameter is positive, indicating that being adjacent increases trade flows between neighboring states, because of better business information, regional cultural commonalities, etc. Imports and exports are included in the 1993 and 1997 CFS, from and up to the custom districts where the commodity enters or leaves the US. For this reason, two dummy variables, the origin custom district dummy (ocddmy) and the destination custom district dummy (dcddmy), are intended to measure the effect of foreign trade at either origin or destination, on commodity flows. The magnitudes of these variables

depend on

the

significance of the foreign trade share in commodity flows. Those two variables may have either a negative or a positive sign. A positive ocddmy indicates that this sector may have a significant foreign import of the commodity, while a negative sign implies a significant foreign export of the commodity. A positive dcddmy, on the other hand, implies that the sector may have a significant foreign export, while a negative sign would point to imports. The list of the US Custom Districts and the states that include them are presented in Appendix D. States with custom districts are coastal (Ocean or Great Lakes) states and the states along the borders with Canada and Mexico (Montana, North Dakota, and Arizona).

41

3. 2. 2. Functional Form The commodity flow between two points can be written with

the variables

specified above, and may be expressed in the framework of Equation (3.9).

Fij = a i ( ovlad , oemp , opop , opipc, oaps , owsem ,io , ocddmy ) f ij ( dist , adjcny ) b j ( dwsem , dmnem , dpop , dpipc , dcddmy , cd ) where ai is the supply point factor, bj the demand point factor, and fij the

(3.13)

interaction

factor. Equation (3.13) could be, in line with past empirical research, represented by a multiplicative functional form, which would become linear when using the logarithms of the dependent and independent variables. However, other functional forms may be acceptable, and there are no strong theoretical reasons to prefer one functional form to another. In this situation, it is reasonable to allow for the endogenous selection of the functional form. The Box-Cox transformation, wherein the variable X is transformed into the variable X (

ë)

according to

X ( λ ) = ( X λ − 1) λ ,

(3.14)

is ideally suited to this purpose (Box and Cox, 1964). Two different transformation parameters are considered: one for all the independent variables (ë) and one for the dependent variable (è). Dummy variables, however, are not transformed. The Box-Cox model can be expressed as

X nλλ − 1 Y θθ − 1 X 2λλ − 1 = a 0 + a1 X 1 + a2 + .... + an +ε θ λ λ

(3.15)

where ª is assumed to be a normally distributed error term, with E(ª)=0 and E( ª ª )= á ’

I. The Box-Cox transformation (3.14) is continuous at ë =0, because X (

ë)

2

tends toward lnX

when ë ->0. Thus, the linear and log-linear functional forms are simply specific points (ë=1 and ë=0) on a continuum of forms allowing for different degrees of independence and interaction among the variables. The fundamental criterion for comparing the infinite number of a priori possible models is how well they are able to explain the data, with the best model maximizing the 42

likelihood of the original observations under the normality assumption for the error term ª. The optimal values of the parameters (a0, a1,…, an, ë, è) in Equation (3.15) maximize the log-likelihood function

LK = −

N N N [ln( 2π ) + 1 ] − ln[ σ 2 ( θ , λ )] + ( θ − 1 )∑ ln Yi , 2 2 i= 1

where N is the sample size, á (ë, è) is the estimated error variance, and Y i 2

(3.16)

the i-th

observation of the original dependent variable. The log likelihood function is a non-linear function, which is optimized using non-linear programming techniques (Green, 1997). Once the optimal functional form (ë , è ) has been determined, it is possible to test *

*

whether an alternate form (ë, è) is significantly different from the optimal one. If the two forms are equivalent, then the statistics 2[LK(ë , è*) - LK(ë, è)] is approximately distributed *

as a X2 with two degrees of freedom. The null hypothesis of functional form equivalence is accepted if LK(ë , è*) - LK(ë, è) < ½ X 2 (á) *

where á is the selected level of significance.

43

(3.17)

CHAPTER 4 DATA SOURCES AND PROCESSING Four main databases are used in this study: the 1993 and 1997 Commodity Flow Surveys (CFS); the 1993 and 1997 County Business Patterns; the 1992 and 1997 Censuses of Manufactures (Bureau of the Census); and the Annual State Personal Income (Bureau of Economic Analysis). The CFS provides the data for the dependent (commodity flow) and distance variables. The other databases provide the data for the independent variables. Except for the Annual State Personal Income database (selective downloading from the Internet for only the required data items),

detailed descriptions of the databases

are presented in Appendix E. 4. 1. Dependent Variable The dependent variables, flow93 and flow97, are drawn from the 1993 and the 1997 Commodity Flow Surveys. These variables measure the value of commodity outshipments in millions of US dollars. Table

9 in the 1993 CFS, and Table 14 in the 1997

CFS, include the values of these out-shipments from each origin state to every other state for each commodity group at the two-digit level. These two

tables are the main

commodity flow data source in this study. The main reason for choosing the two-digit level of commodity aggregation is that data suitable for O-D analysis in

both the 1993 and 1997 CFSs are not provided at any

more disaggregated levels. O-D flow data are, of course, available at the 1 digit level (i.e. total manufacturing). Using such data would involve less missing data (see below), but would preclude understanding the variations in the factors determining flows across sectors. From a geographical viewpoint, the 1993 CFS data are provided for all states and NTARs (National Transportation Analysis Region), while the 1997 CFS data are provided for all states and major metropolitan areas. The NTARs make up a complete coverage of the US, and each NTAR is made of a set of whole counties. Therefore, there is no areal 44

match between NTARs and metropolitan areas. Since the study is intended to provide comparison between 1993 and 1997, the smallest common spatial aggregation unit is the state. Both Table 9 and Table 14 include missing observations, because of data disclosure and data sampling problems. After analyzing the total magnitude of the missing observations problem in both databases, we conclude that the flow corresponding to these missing observations is not significantly high in either year, and therefore missing observations are eliminated from the database. In the 1993 CFS, there is a total of 83,232 flow observations, with 22,476 of them missing (27 %). In the 1997 CFS, 39,449 (34 %)

of

the 114,036

observations are

missing. However, these percentages do not properly assess the extent of the problem. Table 5, in both years, include the total out-shipments for each origin state and each commodity group, without missing data. Summing up the data in Table 5 across states provides the “unsuppressed” total out-shipments for each commodity group. A similar summation over Table 9 provides the corresponding “suppressed” sums, i.e., without accounting for missing data. The relative difference between these two sums provides a measure of the “flow loss” due to missing data. Table

4.1

presents

the

missing

data

assessment

for

1993.

The

two-digit

commodity codes in this table pertain to the Standard Transportation Commodity Classification (STCC), and code definitions are presented in Appendix A. This classification system is completely compatible with the two-digit Standard Industrial Classification (SIC) system. For all commodities combined, missing flows represent 12 % of the total flow. Across commodities, the highest percent is 36 %, for leather products (later included in group 75 together with groups 22 and 23), and the lowest is 3 %, for food and kindred products. The 1997 missing data analysis is presented in Table 4.2. The commodity codes in this table pertain to the Standard Classification of Transported Goods (SCTG), and commodity code definitions are presented in Appendix B. For all commodities combined, missing flows represents 15 % of the total flow. Across commodities, the highest percentage is 24 %, for pharmaceutical products, and for fertilizers and fertilizers materials, and the lowest is 10 %, for plastic and rubber products.

45

Table 4. 1. Magnitudes of Missing Observations in the 1993 CFS

Code

STCC Definitions

Table 9 Suppres sed Sums ($Million)

Table 5 Unsuppresse d Sums Difference ($Million) ($Million)

Share

20 Food and Kindred Products

856608

835131

21477

0.03

22 Textile Mill Products

100971

89679

11292

0.11

23 Finished Textile Products

291186

235206

55980

0.19

24 Lumber or Wood Products

125004

117579

7425

0.06

25 Furniture of Fixture 26 Pulp, Paper, Allied Products

69278

60182

9096

0.13

194992

186529

8463

0.04

28 Chemicals or Allied Products

529383

486948

42435

0.08

29 Petroleum or Coal Products

346044

311236

34808

0.10

30 Rubber of Plastics Products

168377

149832

18545

0.11

44984

28746

16238

0.36

91340

77613

13727

0.15

33 Primary Metal Products

228603

206442

22161

0.10

34 Fabricated Metal Products

237313

224969

12344

0.05

35 Machinery Products

440852

361511

79341

0.18

36 Electrical Machinery Products 37 Transportation Equipment

411573

350453

61120

0.15

627279

503392

123887

0.20

38 Precision Instruments

199400

145418

53982

0.27

39 Miscellaneous Shipment

142532

125180

17352

0.12

5105719

4496046

609673

0.12

31 Leather Products 32 Clay,Concrete,Glass Products

Total

46

Table 4. 2. Magnitudes of Missing Observations in the 1997 CFS

SCTG Descriptions Code 04 Feed, Cereal, Egg Products

Table 14

Suppressed sums ($Million)

Table 5

Unsuppressed Sums ($Million)

Difference ($Million)

Share

64219

54224

9995

0.16

05 Meat, Fish, Seafood Product

181582

151836

29746

0.16

06 Milled grain products

109497

88571

20926

0.19

07 Prepared foodstuffs

345876

318548

27328

0.08

08 Alcoholic beverages

87222

76337

10885

0.12

19 Petroleum & coal products

74666

60902

13764

0.18

20 Basic chemical

158302

125934

32368

0.20

21 Pharmaceutical products

210793

159458

51335

0.24

22 Fertilizer & materials 23 Chemical products n.e.c.

23017

17434

5583

0.24

209063

167688

41375

0.20

24 Plastics and rubber 25 Logs and other wood prdts

278530

257619

20911

0.08

13054

11327

1727

0.13

26 Wood products 27 Pulp & paper products

120907

114890

6017

0.05

106370

92411

13959

0.13

28 Paper & paperboard artcls 30 Textiles, leather products

97936

86422

11514

0.12

371281

304311

66970

0.18

31 Nonmetallic products

108949

98474

10475

0.10

32 Primary base metal

285254

252141

33113

0.12

33 Articles of base metal

226217

203392

22825

0.10

34 Machinery products

416305

363755

52550

0.13

35 Electrical equipment

854449

737083

117366

0.14

36 Vehicles

568928

485248

83680

0.15

37 Transportation equipment

127666

43098

84568

0.66

38 Precision instruments

143441

120625

22816

0.16

97232

81033

16199

0.17

420882

375150

45732

0.11

5701638

4847911

853727

0.15

39 Furniture products 40 Miscellaneous products Total

47

In order to provide for comparability between the two years, the 1997 commodity classification

system,

SCTG,

needs

to

be

converted

into

classification system, STCC. After analyzing and comparing both

the

1993

commodity

commodity classification

systems at the 3 and 4 digit levels, the conversion is done according to the matching presented in Table 4.3. Since SCTG is a more detailed system than STCC, generally one STCC group includes more than one SCTG group. The only exception is the SCTG group 30, which includes STCC groups 22, 23 and 31. This group is recoded as textile mill, apparel

and

group 75:

leather products. When regrouping SCTG-defined flows in 1997

(e.g. SCTG groups 20, 21, 22, and 23 are summed up, for any O-D pair, to obtain STCC group

28 for this O-D pair), if one SCTG flow is missing for an O-D pair, then this O-D pair

is discarded. This process, however inevitable, further reduces the size of the 1997 database. The analysis is finally done for the 16 commodity groups presented in Table 4. 4 . The geographical coverage of the study is the 48 continental states of the US. Shipments originating from businesses located in Puerto Rico and other territories, shipments traversing the U.S., and shipments from a foreign location to a U.S. location are not included in the CFS.

Imported products shipments are included after they leave the

importer’s domestic location for another location.

Export shipments are also included

until they reach the port of exit from the U.S. Shipments through a foreign country, with both the origin and destination in the U.S., are included. However, in the calculation of the mileages for these types of shipments, the foreign segment is not included. Zero flows have been recorded for several O-D pairs in both years. In the case of a linear model, such observations would not cause any computational problem. However, since this

study uses Box-Cox transformation estimations, zero-valued flows create

computational problems. For this reason, and because the data are

survey-derived rather

census-derived, flows with zero values are replaced by very low values (0.0001 million US dollars). Descriptive statistics for the dependent flow variables across all commodities are presented in Table 4. 5 and Table 4. 6 for 1993 and 1997, respectively. statistics by commodity group are presented in Appendix C.

48

Descriptive

Table 4. 3. Matching of the SCTG and STCC groups SCTG

DESCRIPTION

STCC

DESCRIPTION

1 LIVE ANIMALS AND FISH

1 FARM PRODUCTS

2 CEREAL GRAINS

1 FARM PRODUCTS

3 AGRICULTURAL PRODUCTS, EXCEPT LIVE ANIMALS, 4

PRODUCTS OF ANIMAL ORIGIN

1 FARM PRODUCTS 20 FOOD OR KINDRED PRODUCTS

5 MEAT, FISH, SEAFOOD, AND PREPARATIONS

20 FOOD OR KINDRED PRODUCTS

6 MILLED GRAIN AND BAKERY PRODUCTS

20 FOOD OR KINDRED PRODUCTS

7 PREPARED FOODSTUF, NEC AND FATS AND OILS

20 FOOD OR KINDRED PRODUCTS

8 ALCHOLIC BEVERAGES

20 FOOD OR KINDRED PRODUCTS

9 TOBACCO PRODUCTS

21 TOBACCO PRODUCTS

10 MONUMENTAL OR BUILDING STONE

32 CLAY, CONCRETE GLASS, OR STONE PRODUCTS

11 NATURAL SANDS

14 NONMETALLIC MINERALS

12 GRAVEL AND CRUSHED STONE

14 NONMETALLIC MINERALS

13 NONMETALLIC MINERALS NEC

14 NONMETALLIC MINERALS

14 METALLIC ORES

10 METALLIC ORES

15 COAL

11 COAL

17 GASOLINE AND AVIATION PRODUCTS

13 CRUDE PETROLEUM, NATURAL GAS, OR GASOLINE

18 FUEL OILS

13 CRUDE PETROLEUM, NATURAL GAS, OR GASOLINE

19 PRODUCTS OF PETROLEUM REFINIRING

29 PETROLEUM OR COAL PRODUCTS

20 BASIC CHEMICALS

28 CHEMICALS OR ALLIED PRODUCTS

21 PHARMACEUTICAL PRODUCTS

28 CHEMICALS OR ALLIED PRODUCTS

22 FERTILIZERS AND FERTILIZER MATERIALS

28 CHEMICALS OR ALLIED PRODUCTS

23 CHEMICAL PRODUCTS AND PREPARATIONS

28 CHEMICALS OR ALLIED PRODUCTS

24 PLASTICS AND RUBBER

30 RUBBER OF MISC. PLASTIC PRODUCTS

25 LOGS AND OTHER WOOD IN THE ROUGH

24 LUMBER OR WOOD PRODUCTS, EXCLUDING FURNITURE

26 WOOD PRODUCTS

24 LUMBER OR WOOD PRODUCTS, EXCLUDING FURNITURE

27 PULP, NEWSPRINT, PAPER, AND PAPERBOARD

26 PULP, PAPER, OR ALLIED PRODUCTS

28 PAPER AND PAPERBOARD ARTICLES

26 PULP, PAPER, OR ALLIED PRODUCTS

29 PRINTED PRODUCTS

27 PRINTED MATTER

30 TEXTILES, LEATHER, AND ARTICLES

75 TEXTILE MILL PRODUCTS, APPAREL , AND LEATHER PRODUCTS

31 NONMETALLIC MINERAL PRODUCTS

32 CLAY, CONCRETE GLASS, OR STONE PRODUCTS

32 BASE METAL IN PRIMARY FINISHED BASIC FORM

33 PRIMARY METAL PRODUCTS

33 ARTICLES OF BASE METAL

34 FABRICATED METAL PRODUCTS

34 MACHINERY

35 MACHINERY, EXCLUDING ELECTRICAL

35 ELECTRONIC AND OTHER ELECTRICAL EQUIPMENT

36 ELECTRICAL MACHINERY, EQUIPMENT, OR SUPPLY

36 VEHICLES

37 TRANSPORT EQUIPMENT

37 TRANSPORTATION EQUIPMENT, NEC

37 TRANSPORT EQUIPMENT

38 PRECISION INSTRUMENTS AND APPARATUS

38 INSTRUMENT, PHOTOGRAPHIC, OPTICAL GOODS

39 FURNITURE, MATTRESSES AND MATRESS SUPPORTS

25 FURNITURE OF FIXTURE

40 MISCELLANEOUS MANUFACTURED PFODUCTS

39 MISCELLANEOUS MANUFACTURED PFODUCTS

49

Table 4. 4. Commodity Groups Codes and Definitions

Codes

Definitions

20

Food and Kindred Products

24 25

Lumber or Wood Products Furniture of Fixture

26

Pulp, Paper, or Allied Products

28 29

Chemicals or Allied Products Petroleum or Coal Products

30

Rubber of Plastics Products

32 33

Clay, Concrete, Glass or Stone Products Primary Metal Products

34

Fabricated Metal Products

35 36

Machinery, excluding electrical, Products Electrical Machinery Products

37

Transportation Equipment

38 39

Precision Instruments Miscellaneous Freight Shipment

75

Textile, Apparel and Leather Products

4. 2. Independent Variables The adjacency dummy variable, adjncy,

is defined as equal to 1 if the origin and

destination states have a common border, and 0 otherwise. The custom district variables, ocddmy and dcddmy, are defined as equal to 1 if the state contains at least one custom district, and 0 otherwise. The list of the US custom districts is presented in Appendix D. All

the employment variables are drawn from the County Business Patterns (CBP)

database, and include (1) origin sectoral employment, oemp; (2) origin

wholesale

employment,

and

owsem;

(3)

destination

manufacturing

employment,

dmnem;

(4)

destination wholesale employment, dwsem. Some sectoral employment data are missing in the CBP because of data disclosure problems, and are replaced in the following manner. While

there are missing observations at the two-digit level, one-digit level state

employment data are all available. Summing up two-digit employment for each state up to the one-digit level, and

subtracting this sum from the exact one-digit employment, we

obtain the total value of all missing observations. For each one-digit sector, the total missing value is apportioned over the “missing” two-digit sectors using CBP data for earlier or later years. An illustrative example is presented in Appendix E. The origin average establishment size variable, oaps, is estimated by dividing the origin sectoral employment by the number of establishments in that sector. The numbers of establishments are drawn from the CBP (no missing observations). 50

The value-added variable, ovlad, is drawn from the 1992 and 1997 Censuses of Manufactures (Bureau of the Census). While it was initially planned to draw the valueadded variable for 1993 from the 1993 Annual Survey of Manufactures, this survey sometimes does not include

value-added and other data for some sectors,

although

theses sectors may have significant employment in the state. For this reason, the 1992 Census of Manufactures value-added values are used for the 1993 analysis. The errors introduced by this approximation are likely to be negligible. The methodology to replace missing value-added data involves computing per-employee value-added

for the US as

whole for each two-digit sector, and then multiply state sectoral employment by this US ratio in case of missing observations. The state personal income per-capita variables, opipc and dpipc, and the state population variables, opop and dpop, are drawn from the Annual State Personal Income database from

of

the

the Bureau of Economic Analysis (BEA). These data have been downloaded BEA

Website:

www.bea.doc.gov/bea/regional/spi

There

are

no

missing

observations for these data. The distance variable, dist, is directly derived from the 1993 CFS as average hauled distance, and drawn directly from the 1997 CFS. Table 9 in the 1993 CFS has both tonnage and ton-miles values for each commodity group. Dividing ton-miles

by ton values,

the average hauled distance for each commodity group between each O-D pair is estimated. Because individual commodity groups have missing observations, the average hauled distances for all commodities combined are uniformly used for each commodity group between each O-D pair. In the 1997 CFS, however, these average hauled distances are directly provided in Table 14. The competing destination variable, cd, and the intervening opportunities variable, io, are estimated using distance and total employment, as explained in Chapter 3. Total employment

is

drawn

from

the

CBP,

with

no

missing

observations

The

distance

computation has been explained previously. Table 4.5 and Table 4.6 present descriptive statistics for all the variables across all commodities. Statistics by commodity groups are presented in Appendix C.

51

Table 4. 5. Descriptive Statistics for 1993-All Commodities Combined

Variable

N

Mean

St. Dev.

Minimum

Maximum

Sum

flow93 ($M)

26023

92

280

0

7800

2385928

cd93

26023

201995

178751

16323

771064

5256512406

io93

26023

197308

179558

16323

771064

5134543484

dist93 (miles)

26016

1241

757

40

3519

32285248

opipc93 ($)

26023

20660

2841

15468

29602

537622216

opop93

26023 5473805 5785073

oemp93

26023

22100

32709

1

284042

575108111

owsem93

26023

143180

153831

7807

783658

3725983058

ovlad92 ($M)

26023

1730

2889

0

21698

45031541

oaps93

26023

71

94

2

1715

1754252

dmnem93

26023

417532

390355

11285

1898885

10865447978

dwsem93

26023

146218

153495

7807

783658

3805026717

dpipc93 ($)

26023

20736

2881

15468

29602

539601756

dpop93

26023 5585155 5766049

460000 30380000 142444834000

460000 30380000 145342500000

flow93: commodity flow, cd93: competing destination, io93: intervening opportunities, dist93: distance, opop93: origin state percapita income, opop93: origin state population, oemp93: origin state sectoral employment, owsem93: origin state wholesale employment, ovlad92: origin state sectoral value-added, oaps93: origin state sectoral average plant size, dmnem93: destination state manufacturing employment, dwsem93; destination state wholesale employment, dpipc93: destination state per-capita income, dpop93: destination state population

52

Table 4. 6. Descriptive Statistics for 1997-All Commodities Combined

Variable

N

Mean

St. Dev.

Minimum

Maximum

Sum

flow97 ($M)

17406

119

370

0

13664

2066083

cd97

17406

233802

201514

16357

867414

4069552686

io97

17406

227328

201688

16357

867414

3956863410

dist97 (miles)

17402

1194

735

47

3220

20786551

opipc97 ($)

17406

24782

3456

18885

35596

431362191

opop97

17406 5979589 6204436

oemp97

17406

24123

37131

1

293723

419890235

owsem97

17406

161246

168225

8624

856950

2806641762

ovlad97 ($M)

17406

2087

4262

0

65716

36330530

oaps97

17406

70

82

0

653

1194204

dmnem97

17406

451250

415059

13219

2019053

7854453186

dwsem97

17406

167169

170306

8624

856950

2909735446

dpipc97 ($)

17406

24832

3506

18885

35596

432226748

dpop97

17406 6187611 6279457

480000 32268000 104080723000

480000 32268000 107701561000

flow97: co mmodity flow, cd97: competing destination, io97: intervening opportunities, dist97: distance, opop97: origin state per-capita income, opop97: origin state population, oemp97: origin state sectoral employment, owsem93: origin state wholesale employment, ovlad97: origin state sectoral value-added, oaps97: origin state sectoral average plant size, dmnem97: destination state manufacturing employment, dwsem97; destination state wholesale employment, dpipc97: destination state per-capita income, dpop97: destination state population

53

CHAPTER 5 RESULTS After a brief overview of commodity flows and the general interregional trade structure over the continental US, the estimated sectoral equations, their significant variables, and possible explanations for the

findings are presented. These results are

synthesized to outline general patterns. Finally, a detailed elasticity analysis over the major model variables is conducted and its results analyzed.

5. 1. Overview of US Interregional Commodity Flows The total value of

the 16 commodity groups traded in the U.S. within and across

state-lines is $ 5,160 billions in 1993, and $ 5,780 billions in 1997. The largest share characterizes food and kindred products, with around 15 %. The second largest share pertains to transportation equipment, share is only 8 %

in 1993,

with around 12 %. While the electrical machinery

it increases to 15 % in 1997. The third largest

group is

chemical products, 10.5 %. As can be seen in Table 5.1, the other significant product groups are non-electrical machinery (8 %), and textile, leather and apparel (7.5 %). The highest per-ton value product group is precision instruments in both years, with $ 2,566 in 1993, and $ 5,812 in 1997. The cheapest or bulkiest product group is clay, concrete, glass and stone products, around $12. Other high-value product groups include electrical

and non-electrical machineries, transportation equipment, and textile,

leather, and apparel products. Lumber and wood products, coal and petroleum products, and primary metal products belong to the low value/bulky products groups. It is expected that low-value commodities are characterized by short hauls, and high-value ones by long hauls, reflecting the share of transportation costs in total production costs.

54

Table 5. 1 Total Value of Shipments and Their Shares Across Commodity Groups

Group Name

totsp93 ($ Million)

totsp97 Value (%) 93 $/ton ($ Million)

Value (%) 97 $/ton

20 Food & Kindred

852009

16.5

100

801416

13.9

90

24 Lumber & Wood

125982

2.4

19

141341

2.4

19

69289

1.3

427

97209

1.7

488

26 Paper & Pulp

194512

3.8

90

204899

3.5

96

28 Chemical

528455

10.2

103

624525

10.8

104

29 Coal & Petroleum

358617

6.9

19

74991

1.3

16

30 Rubber

172753

3.3

330

278393

4.8

214

32 Clay, Concrete

90862

1.8

11

112638

1.9

13

33 Primary Metals

228428

4.4

86

285577

4.9

88

34 Fabricated Metals 35 Machinery

237001

4.6

279

226694

3.9

226

441339

8.6

1295

416859

7.2

835

36 Electrical Mach.

411391

8.0

1363

857705

14.8

2162

37 Transportation

640169

12.4

748

705239

12.2

676

38 Precision Inst.

199355

3.9

2566

155537

2.7

5812

39 Miscellaneous

164645

3.2

757

420076

7.3

530

75 Textile & Leather

446096

8.6

1050

377028

6.5

827

Cmt.

25 Furniture

Total

5160903 100.0

5780127 100.0

Cmt: commodity group; totsp93: total outshipments in 1993; totsp97: total outshipments in 1997.

Tables 5.2. and 5.3. present total (all commodities) in- and out-shipments by state ranked by decreasing amount, in 1993 and 1997, respectively. Detailed data by commodity for 1993 are available in Appendix C, Table C.33. In- and out-shipments, and their percentages, display similar magnitudes, not only in the same year, but also across the two years. Thus, it appears that states have balanced and stable import and export patterns. With approximately 12 % in shipments share, California displays a spatial concentration

in

manufacturing

production

and

consumption,

possibly

at

both

intermediate and final levels. The second largest spatial concentration is in

the

Texas,

around 8 % of the U.S. interstate trade. The same 12 states, namely, California, Texas, Illinois, Ohio, New Jersey, Michigan, New York, Pennsylvania, North Carolina, Florida, and

Indiana,

Georgia,

export/import approximately 50 % to 60 % of all shipments in the

US, in both years. 55

Table 5. 2. 1993 States’ In- and Out -shipments and Their Shares

1993 Out -shipment State Code

State Name

6 California

Value ($ Million)

1993 In-shipment State Code

(%)

595577 11.3

State Name

6 California

Value ($ Million)

(%)

542548 11.7

48 Texas

410332

7.8

48 Texas

387569

8.4

17 Illinois

305139

5.8

39 Ohio

260159

5.6

39 Ohio

299750

5.7

36 New York

256718

5.5

34 N. Jersey

240792

4.6

17 Illinois

251640

5.4

26 Michigan

240462

4.6

26 Michigan

240836

5.2

36 New York

238580

4.5

12 Florida

215651

4.6

42 Penn

225637

4.3

42 Pennsylv.

202020

4.4

37 N. Carolina

202509

3.8

34 N. Jersey

162045

3.5

13 Georgia

197280

3.7

13 Georgia

159862

3.4

18 Indiana

160010

3.0

37 N. Carolina

148397

3.2

47 Tennessee

158612

3.0

18 Indiana

126337

2.7

12 Florida

153149

2.9

47 Tennessee

120394

2.6

55 Wisconsin

133409

2.5

51 Virginia

106781

2.3

29 Missouri

117522

2.2

55 Wisconsin

105800

2.3

53 Washington

114175

2.2

53 Washington

102327

2.2

25 Massach.

101357

1.9

29 Missouri

101349

2.2

51 Virginia

100539

1.9

25 Massach.

89781

1.9

21 Kentucky

99302

1.9

27 Minnesota

83261

1.8

27 Minnesota

94878

1.8

22 Louisiana

75984

1.6

24 Maryland

86728

1.6

24 Maryland

75928

1.6

22 Louisiana

85518

1.6

21 Kentucky

72423

1.6

1 Alabama

80772

1.5

45 S. Carolina

70252

1.5

45 S. Carolina

77908

1.5

1 Alabama

68271

1.5

56

Table 5. 2. 1993 States’ In- and Out -shipments and Their Shares (continued)

1993 Out -shipment State Code

State Name

19 Iowa

Value ($ Million)

1993 In-shipment State (%) Code

67069 1.3

State Name

Value ($ Million) (%)

4 Arizona

57324 1.2

9 Connect.

64928 1.2

19 Iowa

53257 1.1

4 Arizona

62014 1.2

41 Oregon

52152 1.1

41 Oregon

59505 1.1

8 Colorado

47756 1.0

20 Kansas

57604 1.1

9 Connect.

44283 1.0

5 Arkansas

53473 1.0

8 Colorado

52447 1.0

5 Arkansas

42164 0.9

28 Mississippi

48311 0.9

40 Oklahoma

41377 0.9

40 Oklahoma

44958 0.9

28 Mississippi

41169 0.9

31 Nebraska

36421 0.7

31 Nebraska

24446 0.5

49 Utah

31571 0.6

49 Utah

24339 0.5

54 W. Virginia

27363 0.5

32 Nevada

20532 0.4

23 Maine

18829 0.4

54 W. Virginia

19500 0.4

44 Rhode Isl.

17202 0.3

35 N. Mexico

15472 0.3

32 Nevada

15620 0.3

23 Maine

12336 0.3

33 New Hemp

15296 0.3

16 Idaho

11917 0.3

10 Delaware

15274 0.3

33 New Hamp

11173 0.2

16 Idaho

14539 0.3

10 Delaware

11095 0.2

35 N. Mexico

10060 0.2

30 Montana

8828 0.2

46 S. Dakota

7932 0.2

46 S. Dakota

7808 0.2

50 Vermont

7881 0.1

44 Rhode Is

7384 0.2

30 Montana

7692 0.1

38 N. Dakota

6383 0.1

38 N. Dakota

7333 0.1

50 Vermont

4747 0.1

56 Wyoming

5134 0.1

56 Wyoming

4155 0.1

57

20 Kansas

44149 1.0

Table 5. 3. 1997 States’ In- and Out -shipments and Their Shares

1997 Out -shipment State Code

State Name

6 California

Value ($ Million)

1997 In-shipment State (%) Code

675478 11.6

State Name

6 California

Value ($ Mil)

(%)

549805 13.7

48 Texas

481045

8.2

48 Texas

429945 10.7

39 Ohio

344921

5.9

26 Michigan

224397

5.6

17 Illinois

298171

5.1

39 Ohio

222371

5.5

26 Michigan

281782

4.8

36 New York

221116

5.5

36 New York

247692

4.2

17 Illinois

221048

5.5

37 N. Carolina

243340

4.2

12 Florida

191592

4.8

34 N. Jersey

242999

4.2

42 Pennsylv.

166358

4.1

42 Pennsylv.

239888

4.1

37 N. Carol

157843

3.9

13 Georgia

191284

3.3

13 Georgia

144867

3.6

18 Indiana

185496

3.2

18 Indiana

112917

2.8

12 Florida

169549

2.9

34 N. Jersey

112168

2.8

55 Wisconsin

157900

2.7

47 Tennessee

93316

2.3

47 Tennessee

141083

2.4

55 Wisconsin

88210

2.2

29 Missouri

128417

2.2

53 Washington

82947

2.1

53 Washington

127438

2.2

51 Virginia

77508

1.9

25 Massach.

127008

2.2

25 Massach.

76635

1.9

27 Minnesota

126244

2.2

27 Minnesota

75244

1.9

21 Kentucky

109555

1.9

29 Missouri

74919

1.9

51 Virginia

102937

1.8

45 S. Carolina

60600

1.5

45 S. Carolina

92466

1.6

21 Kentucky

54430

1.4

19 Iowa

89235

1.5

24 Maryland

53914

1.3

1 Alabama

88212

1.5

1 Alabama

53395

1.3

22 Louisiana

83677

1.4

22 Louisiana

50688

1.3

58

Table 5. 3. 1997 States’ In- and Out -shipments and Their Shares (continued)

1997 Out -shipment State Code

State Name

1997 In-shipment

Value State ($ Million) (%) Code

State Name

8 Colorado

Value ($ Million) (%)

41 Oregon

82450 1.4

4 Arizona

75476 1.3

41 Oregon

42398 1.1

9 Connect.

74770 1.3

4 Arizona

41547 1.0

8 Colorado

65838 1.1

24 Maryland

64063 1.1

9 Connect.

34726 0.9

5 Arkansas

60486 1.0

40 Oklahoma

29648 0.7

20 Kansas

60251 1.0

20 Kansas

28490 0.7

28 Mississippi

47941 0.8

5 Arkansas

24806 0.6

40 Oklahoma

47874 0.8

28 Mississippi

23813 0.6

31 Nebraska

46256 0.8

49 Utah

20688 0.5

49 Utah

35541 0.6

31 Nebraska

16428 0.4

33 New Hamp

28907 0.5

32 Nevada

12252 0.3

54 W. Virginia

28257 0.5

54 W. Virginia

10012 0.2

16 Idaho

21773 0.4

35 N. Mexico

8128 0.2

23 Maine

19314 0.3

16 Idaho

7834 0.2

32 Nevada

17971 0.3

33 New Hamp

7778 0.2

46 S. Dakota

14593 0.2

23 Maine

5966 0.1

44 Rhode Isl.

13331 0.2

38 N. Dakota

4943 0.1

10 Delaware

13249 0.2

44 Rhode Isl.

4500 0.1

35 N. Mexico

13076 0.2

30 Montana

4131 0.1

50 Vermont

11873 0.2

10 Delaware

3640 0.1

19 Iowa

42559 1.1

40446 1.0

38 N. Dakota

8028 0.1

50 Vermont

3507 0.1

30 Montana

7794 0.1

46 S. Dakota

3343 0.1

56 Wyoming

5131 0.1

56 Wyoming

1547 0.0

As mentioned earlier, foreign trade flows are

included in the CFSs in flows to and

from US custom states. Tables 5.4 and 5.5, present, for 1993 and 1997, respectively, shares

of

custom district foreign imports to custom district states’

(as foreign imports is included in these out-shipments), and

CFS out-shipments

the share of custom district

foreign exports out of all states out-shipments (potentially all states have some foreign

59

trade flows in their out-shipments). Furthermore, the last column of these tables presents the ratio of total out-shipments of custom district states to total out-shipments of all states. This ratio varies between 80 to 90 % , depending on the commodity group and in both years. All of the states that have custom districts are located either on the Table 5. 4. 1993 Custom States Imports and Export Shares

Cmd.

Group Name

All States Totsp93 Cdex93 (%)

Custom States Totsp93 Cdim93

20 Food & Kindred

852009 20789 2.4 692770

24 Lumber & Wood

125982

25 Furniture

69289

Cd. St./ All St. (%) (%)

21874

3.2

81

5282 4.2 107428

8656

8.1

85

3627 5.2

7940 13.1

88

60638

26 Paper & Pulp

194512

8886 4.6 169767

10388

6.1

87

28 Chemical

528455 35229 6.7 453635

25428

5.6

86

7248 2.0 322124

52423 16.3

90

172753 17137 9.9 144335

14651 10.2

84

78176

18893 24.2

86

228428 12962 5.7 196260

22994 11.7

86

29 Coal & Petroleum 358617 30 Rubber 32 Clay, Concrete 33 Primary Metals

90862 16116 17.7

34 Fabricated Metals 237001

2.6

87

441339 85813 19.4 382119

85303 22.3

87

411391 59883 14.6 359963

74481 20.7

87

37 Transportation

640169 74032 11.6 549739

88885 16.2

86

38 Precision Inst.

199355 21995 11.0 181591

20117 11.1

91

39 Miscellaneous

164645

12940

9.3

85

75 Textile & Leather

446096 15709 3.5 376529

56240 14.9

84

35 Machinery 36 Electrical Mach.

3700 1.6 205261

3800 2.3 139141

5300

Cmt: commodity group; totsp93: total outshipments in 1993; cdex93: custom district export in 1993, cdim93: custom district import in 1993

coasts or along the U.S.-Canada / U.S.-Mexico border (only Montana, North Dakota, and Arizona). These ratios reveal another important geographic fact: a major proportion of U.S. manufacturing is concentrated on coastal areas because of accessibility to water transportation. In terms of foreign export in 1993, non-electrical machinery products have the highest share (19.4 %). The other high-export sectors are clay, concrete, glass and stone products

(17.7

%), electrical machinery (14.6 %),

transportation equipment and

precision instruments (around 11 %). The product groups with high shares of foreign imports include clay, concrete, glass and stone products (24.2

%) ,

non-electrical

machinery (22.3 %), electrical machinery (20.7 %), transportation equipment (16.2 %), 60

coal and petroleum products (16 %), textile, leather and apparel (15 %), furniture and fixture products (13 %,), and precision equipments (11.1 %).

Table 5. 5. 1997 Custom States Imports and Export Shares

Cmd.

Group Name

All States Totsp97 Cdex97 (%)

Custom Cd.St./ States All St Totsp97 Cdim97 (%) (%)

20 Food & Kindred

801416

30000

3.7 632991

32155 5.10

79

24 Lumber & Wood

141341

5841

4.1 118752

13018 11.0

84

5.4

13868 16.5

86

25 Furniture

97209

5255

26 Paper & Pulp

204899

13743

6.7 178801

14256

8.0

87

28 Chemicals

624525

54526

8.7 534774

41369 7.70

.86

72188 110

88

29 Coal & Petroleum

74991

9038 12.1

84213

65988

30 Rubber

278393

28221 10.1 240577

22183

9.2

86

32 Clay, Concrete

112638

16883 15.0

95404

27137 28.4

85

33 Primary Metals

285577

21114 7.40 245685

37726 15.4

86

34 Fabricated Metals

226694

5773 2.50 196057

8548 4.40

86

35 Machinery

416859 136533 32.8 351348 140225 39.9

84

36 Electrical Mach.

857705 106310 12.4 785650 113969 14.5

92

37 Transportation

705239 100214 14.2 574515 120738 21.0

82

38 Precision Inst.

155537

34114 21.9 143729

28949 20.1

92

39 Miscellaneous

420076

5323 1.30 363945

19352 5.30

87

75 Textile & Leather

377028

24609 6.50 340521

77551 22.8

90

Cmt: commodity group; totsp97: total outshipments in 1997; cdex97: custom district export in 1997; cdim97: custom district import in 1997

In 1997, non-electrical products still have the highest export share (32.8 %). The other high-export sectors are precision instruments (21.9 %), transportation equipments (14.2), electrical machinery (12.4 %), and clay, concrete, glass and stone products (15 %). The highest import share in 1997 applies to petroleum products (110 %). This of course, reflects the fact that the US imports more that 50 % of its total petroleum consumption. Non-electrical machinery has the second highest import share (40 %), while textile and leather products (22.8 %), transportation equipments (21.0 %),

precision

instruments (20.1 %), electrical machinery (14.5 %), and clay, concrete, glass and stone products (15 %) are the other high-import product groups in 1997.

61

5. 2. Individual Commodity Equation Results 5. 2. 1. Food and Kindred Products The regression results for commodity group 20, food and kindred products, are presented in Table 5.6. Except for opipc, which is insignificant,

and dcddmy, which is

significant at the 10 % level only, all the other variables are significant at the 5 % level in the 1993 equation. The Box-Cox model performs significantly better than the log-linear model in 1993. In the 1997 equation, however, only occdmy and dist are significant. One possible explanation for this poor result, as discussed earlier, is that the conversion from the SCTG (Standard Classification of Transported Goods) system into the STCC (Standard Transportation

Commodity

Classification)

system

forced

the

elimination

of

many

observations, producing a final sample of only 320 observations (versus 1663 in 1993) In the 1993 equation, the sign of the adjncy coefficient is positive and significant. It indicates that sharing a common border increases food and kindred product flows between neighboring states.

Ocddmy is negative and dcddmy is positive, suggesting that

this commodity group may involve a significant amount of foreign export.

Although the US

foreign export share is only 3.7 % for this commodity, food products represent the largest US foreign commodity shipment, hence even a small export share may be significant. Cd and io are both negative and significant, indicating that destinations clusters at both ends compete in their demands for the commodity.

Dist is negative and highly significant as

expected: the flow decreases as distance (or transport cost) increases. A negative opop indicates that, as the origin population increases, the outflows of food products decrease because of increased local final consumption. Oaps is negative and significant, implying diversification effects. The two variables, oemp and ovlad, representing supply conditions, are positive as expected: the commodity outflow increases as the productive potential of the origin increases.

The redistribution variables, owsem and dwsem, are both positive

and significant: the wholesale sector appears to be an important facilitator in interregional food and kindred products trade. The destination final demand variables, dpipc and dpop, and the destination intermediate demand variable, dmnem, are positive and significant: commodity flows increase as

the manufacturing employment, personal income per capita,

and total population increase at the destination. The 3-digit breakdown of food and kindred products includes meat, poultry and fish, diary products, preserved fruit and vegetables, sugar, bakery products, beverages etc (see Table C.34 in Appendix C). These products are not only supplied for household consumption but also as other industries’ inputs. This explains that the final demand variables at the origin, opop, and at the destination, dpop and dpipc, and the intermediate 62

demand variable, dmnem, are significant and with the expected sign. Furthermore, since this is a product group that is supplied to both the final and intermediate demand sectors, wholesale companies are likely to be involved in this procurement. Table 5. 6. Food and Kindred Products Regression Parameters

20 constant adjncy ocddmy dcddmy

1993

T

-16.312 *

-2.89

-77.759 **

-1.70

1.624 *

6.92

0.106

0.07

-0.273 *

-2.18

0.213 **

1997

-1.992 *

T

-2.48

1.85

0.172

0.27

cd

-0.852 *

-4.58

-1.112

-0.92

io

-0.670 *

-4.32

-1.349

-1.06

dist

-2.199 *

-9.38

-3.379 **

-1.94

opipc

-0.086

-0.34

2.432

1.00

opop

-0.569 *

-3.39

-0.709

-0.76

oemp

0.872 *

4.64

0.838

0.88

owsem

0.896 *

4.11

1.840

1.11

ovlad

1.054 *

5.63

0.662

1.44

oaps

-0.876 *

-4.78

-0.430

-0.54

dmnem

0.194 *

2.77

0.488

0.94

dwsem

1.024 *

4.94

1.089

1.10

dpipc

1.045 *

3.64

2.271

1.00

dpop

0.474 *

2.78

0.440

0.50

lambda

0.078 *

5.08

0.088

1.29

theta

0.194 *

51.78

-0.133 *

MLHD

-6284.19

704.05

LLHD

-7411.55

663.50

X2 N

2254.72 *

27.59

1663

* Significant at 5 % level **

81.10 *

-8.13

27.59

320

Significant at 10 % level

5. 2. 2. Lumber and Wood Products Table 5.7 presents the regression results for commodity group 24, lumber and wood products. Adjncy is positive and significant in both years. Among the custom district variables, only

occdmy is significant at the

10 % level in 1993, pointing a weak export

effect. The US total foreign trade data show that the share and magnitude of import in this 63

product group are higher than for export. This may explain why foreign trade effect are not statistically noticeable. Both the co and io variables are negative and significant in

both

years, implying competition effects at both ends. Distance is negative and significant in Table 5. 7. Lumber and Wood Products Regression Parameters

24

1993

T

1997

T

-65.812 *

-5.87

-86.252 *

-4.25

1.817 *

7.82

2.118 *

5.49

-0.204 **

-1.70

0.132

0.80

0.65

-0.038

-0.23

constant adjncy ocddmy dcddmy

0.074

cd

-1.843 *

-3.62

-1.813 *

-2.40

io

-1.276 *

-3.25

-1.507 *

-2.53

dist

-2.218 *

-6.62

-2.355 *

-4.71

opipc

2.217 *

3.16

4.528 *

2.71

opop

0.087

0.29

0.156

0.29

oemp

-0.228

-0.85

1.643 *

3.68

owsem

0.423

1.49

ovlad

2.369 *

6.04

oaps

0.333

1.39

dmnem

1.016 *

3.67

0.813 *

2.21

dwsem

0.759 *

2.56

1.266 *

2.18

dpipc

3.398 *

4.08

2.993 *

2.55

dpop

1.123 *

2.26

1.002

1.35

lambda

0.016

0.75

0.007

0.25

theta

0.170 *

-0.144

0.121 *

-2220.28

-611.76

LLHD

-2682.28

-722.38

2

924.00 *

N * Significant at 5

as

1.188 *

MLHD X

both years,

33.37

-0.127

1599 % level

expected.

significant. The magnitude

27.59

In

**

221.25 *

-0.25 2.93 -0.37

15.94

27.59

894

Significant at 10 % level

contrast

of opipc,

to in

expectations,

opipc

is positive and

this case, may represent total productive

capacity at the origin rather than being a proxy for local final demand. For example, as can be seen in Appendix C, one of the main producers of this commodity, Oregon, outships nearly 10 % of all US shipments, and this product accounts for 20 % of Oregon total shipments. This share is also around 20 % for both Idaho and Montana. In other 64

words, this product group may be an income generator for some states. The sectoral employment variable oemp is positive and significant only in

1997. However, ovlad is

significant in each year as expected. Wholesale activity at the origin is not a significant redistributive factor in this case. The destination mass variables, namely dmnem, dwsem, and dpipc, are significant with the expected signs in each year. Dpop is not significant in 1997, but is so in 1993. Since lumber and wood products are generally consumed as intermediate products, and not directly by the final consumption sectors, this result may seem counter intuitive. However, one of the main demanders of this product in the US is naturally the construction sector, and this sector is close to the final demand sector. For this reason, the final demand sector may be an indirect determinant of the consumption of wood and lumber products via the construction sector. The Box-Cox model is statistically superior to the log-linear model for this product group.

5. 2. 3. Furniture and Fixture Products The regression parameters of commodity group 25,

furniture and fixture products,

are presented in Table 5.8. Adjncy is positive and significant in each year. Dcddmy is positive and significant at the 10 % level in 1993, pointing to a foreign export effect at destination. Cd and io

are negative and significant each year (io is significant at the 10

% level in 1997), indicating competition effects at both ends. Distance is negative and significant each year, as expected. Opop is negative and significant in 1997, but opipc is never significant, suggesting that there is little final demand effect at the origin. Oemp is significant each year, with the expected sign, implying that flow increases with origin sectoral employment. The other sectoral production variable, ovlad,

is only significant in

1997, with the expected positive sign. Owsem and dwsem are both positive and significant each year, suggesting that wholesale employment at both the origin and destination is a facilitator of flow. Oaps is positive and significant each year, implying that furniture outflows are encouraged by scale economies;

the destination final demand variables,

dpipc and dpop, are highly significant, with the expected signs, suggesting that the furniture industry mainly supplies

final demand. This is confirmed by the weak and sign-

inconsistent effect of the destination manufacturing variable dmnem.

65

Table 5. 8. Furniture and Fixture Products Regression Parameters

25 constant adjncy ocddmy dcddmy

1993

T

1997

-53.690 *

-5.70

-32.251 *

-4.41

0.502 *

2.24

0.840 *

4.08

-0.098 0.198 **

T

-0.88

-0.064

-0.56

1.80

0.154

1.35

cd

-1.133 *

-3.69

-0.494 *

-2.99

io

-0.578 *

-2.77

-0.256 **

-1.84

dist

-1.644 *

-6.99

-1.033 *

-6.46

opipc

0.095

0.27

-0.192

-0.74

opop

0.066

0.29

-0.486 *

-2.68

oemp

1.377 *

5.52

0.695 *

3.81

owsem

0.675 *

2.80

1.004 *

3.75

-0.50

0.335 *

3.25

2.38

0.518 *

2.35

-0.91

-0.167 *

-1.97

ovlad oaps dmnem

-0.060 0.451 * -0.088

dwsem

0.705 *

2.96

0.533 *

2.72

dpipc

2.424 *

4.03

1.236 *

3.24

dpop

1.250 *

2.81

0.688 *

2.50

lambda

0.040 *

1.99

0.066 *

3.16

theta

0.118 *

21.45

0.163 *

30.70

MLHD

-369.13

-1906.76

LLHD

-565.93

-2313.69

X2

393.60 *

N

1573

*

Significant at 5 % level

**

27.59

813.85 *

27.59

1148

Significant at 10 % level

The 3-digit breakdown of furniture products confirms that this group provides mainly for the final demand sector (see Table C.34 in Appendix C). This is supported by our findings in the case of the final demand variables at destination. One possible explanation for not having significant origin final demand variables is that the furniture industry in the US is spatially concentrated and displays economies of scale (as shown by the variable oaps). As can be seen in Appendix C, 15 states provide approximately 71 % of all US shipments. Since final demand dominates in this group, wholesalers may also facilitate this final demand.

66

The

Box-Cox models are statistically superior to the log-linear models in both

years.

5. 2. 4. Pulp, Paper and Allied Products Table 5.9 presents the regression results for commodity group 26: pulp, paper or allied products.

is positive and significant at the 5 % level in 1993, but not

Adjncy

significant in 1997. Among the custom district dummy variables, only ocddmy is positive and significant in 1993, implying an import trade effect in 1993. However, this effect is weak in 1997, as can be confirmed by the custom districts foreign trade data. Cd and io Table 5.9. Pulp, Paper and Allied Products Regression Parameters

1993

T

-15.176 *

-2.91

-7.481

-0.92

adjncy

0.511 *

2.20

0.515

1.61

ocddmy

0.271 *

1.97

0.285

1.28

26 constant

T

dcddmy

-0.014

-0.12

0.156

0.83

cd

-0.601 *

-3.55

-0.399 *

-2.00

io

-0.504 *

-3.35

-0.228 **

-1.80

dist

-1.617 *

-7.18

-1.329 *

-4.40

opipc

-0.373

-1.64

-0.381

-1.08

-0.086

-0.54

opop

0.223 **

1.77

oemp

0.518 *

3.05

0.699 *

2.70

owsem

0.149

1.16

0.205

0.99

ovlad

0.485 *

3.80

0.278 *

2.02

-1.19

-0.756 *

-3.11

oaps

-0.159

dmnem

0.367 *

3.67

0.231 *

2.02

dwsem

0.667 *

3.72

0.417 *

2.09

dpipc

0.838 *

3.12

0.584 **

1.71

dpop

0.167

1.47

0.180

1.12

lambda

0.094 *

4.67

0.115 *

3.41

theta

0.181 *

37.47

0.126 *

15.83

MLHD

-3387.18

-1005.18

LLHD

-3989.37

-1122.33

X

2

1204.37 *

N *

1997

Significant at 5 % level

27.59

1560 **

234.30 * 722

Significant at 10 % level

67

27.59

are both negative and significant each year. Competition effects dominate at destinations clustered around both ends. Distance is significant, with the expected sign in both years. 30 % of this product group shipments is paper, and the rest is

paper products used in

manufacturing industries, such as paperboards, boxes etc. For this reason, the final demand variables at both the origin and destination are not very significant. The sectoral production variables, oemp and ovlad, are positive and significant, as expected, year.

each

Oaps is only significant in 1997. However, the negative sign each year may indicate

strong diversification effects. The wholesale employment is significant, with the expected sign, only at the destination in each year. Since wholesale employment at the origin is never significant, we may be conclude that out-shipments are provided by producers, and not by wholesalers. The demand induced by wholesale trade at the destination is significant for intra-state redistribution. Also, the intermediate demand at the destinations is significant each year, as dmnem is positive and significant. It is also possible to say that wholesalers in this product group provide mainly the intermediate demand sectors. The Box-Cox model is statistically superior to the log-linear model for both years. 5. 2. 5. Chemical and Allied Products The regression results for commodity group 28, chemical and allied products, are presented in Table 5.10. Because of the sample reduction problem caused by the conversion of SCTG into STCC, as discussed earlier, the Box-Cox equation of 1997 is not significantly better than the log-linear model, and none of the variables in the equation is significant. For this reason, only the 1993 equation is discussed below. A positive and significant adjncy

suggests that a common physical border

increases the chemical products trade between states.

Foreign trade does not display a

significant effect on chemical product flows, since none of the custom district dummy variables is significant. Cd and io are both negative and significant, implying that competing effects dominate at both end destinations. Distance is significant and negative, as expected. The product composition of this sector includes industrial organic and inorganic chemicals, plastic materials and fibers, paints and enamels, and agricultural chemicals (see Table C.34 in Appendix C). All these product items are consumed by other industries. For this reason, the final demand variables,

opipc, opop, dpipc, and dpop, at both the

origins and destinations, are not significant. Oemp and ovlad are both positive and significant, as expected. Wholesale employment at the origin and destination is positive and significant, facilitating interregional chemical products flows. The negative sign of 68

oaps suggests that exports in the chemical industry are boosted by diversification effects. The intermediate demand effect, proxied by dmnem, are

also positive and significant, as

expected. Wholesale employment, at both the origin and the destination, is positive and significant, pointing a significant redistribution activity. Since the industry is characterized by mainly intermediate demand, wholesalers are probably providing to demand sectors. Table 5. 10. Chemical and Allied Products Regression Parameters

28

1993

constant

-3.078

T

1997

T

-0.71

-85.641

-0.87

adjncy

1.838 *

7.65

0.966

0.96

ocddmy

0.009

0.07

-0.156

-0.28

dcddmy

0.155

1.25

0.148

0.33

cd

-0.233 *

-2.94

-0.168

-0.29

io

-0.200 *

-2.82

-0.400

-0.42

dist

-1.288 *

-8.32

-2.077

-1.12

opipc

0.001

0.01

1.853

0.54

opop

-0.100

-1.22

2.028

0.41

oemp

0.447 *

2.91

0.169

0.25

owsem

0.226 *

1.96

-0.488

-0.37

ovlad

0.601 *

6.44

-0.335

-0.57

oaps

-0.707 *

-3.26

0.680

0.49

dmnem

0.102 *

2.16

-0.139

-0.29

dwsem

0.730 *

4.54

0.825

0.67

dpipc

-0.120

-0.70

0.774

0.41

dpop

0.050

0.65

1.147

0.45

lambda

0.111 *

6.43

0.050

0.38

theta

0.185 *

45.88

-0.012

-0.68

MLHD

-5769.75

89.48

LLHD

-6716.11

87.89

X2

1892.72 *

N

1568

* Significant at

5 % level

**

27.59

3.17 223

Significant at 10 % level

69

27.59

intermediate

5. 2. 6. Petroleum and Coal Products Table 5.11 presents the regression results for commodity group 29, petroleum and coal products. In each year, the flow of this commodity is mainly explained by

origin or supply conditions, since none of the destination

variables is significant.

Adjncy is positive and significant each year. Ocddmy is positive and significant in 1993, indicating a foreign import effect. Considering the US dependency on result seems

reasonable.

petroleum, this

The custom districts data also shows relatively high import

share for both years; however,

not such effect is uncovered

in 1997.

Distance is a

significant variable each year, with the expected negative sign. Cd is significant at the 10 Table 5. 11. Petroleum or Coal Products Regression Parameters

T

1997

T

64.297 *

3.18

278.711 *

3.53

adjncy

8.254 *

5.68

28.834 *

6.61

ocddmy

1.843 *

2.62

0.358

0.21

dcddmy

0.329

0.56

0.316

0.21

29

1993

constant

cd

-0.413 **

-1.80

-1.112

-1.57

io

-0.182

-1.20

0.329

0.87

dist

-2.998 *

-3.91

-6.641 *

-3.25

opipc

-2.188 *

-2.49

-11.057 *

-2.25

opop

-0.450

-1.63

-2.145

-1.52

oemp

1.949 *

3.25

4.535 *

2.78

owsem

0.780 *

1.99

3.132 **

1.79

0.470

0.70

ovlad

-0.122

-0.68

oaps

-2.848 *

-3.93

-3.33

dmnem

0.033

0.39

0.481

1.32

dwsem

0.278

1.26

0.859

1.07

dpipc

-0.104

-0.26

-1.793

-1.23

dpop

0.314

1.47

0.715

1.12

lambda

0.186 *

5.60

0.164 *

4.20

-0.240 *

-24.07

-0.391 *

-28.95

theta MLHD

4892.51

5328.21

LLHD

4502.05

4651.76

X

2

N *

-5.692 *

Significant at 5 % level

780.92 *

27.59

1557 **

1352.91 * 1321

Significant at 10 % level

70

27.59

% level in 1993, indicating a weak competition effect at destinations. Opipc is significant as expected: local consumption of petroleum products negatively affects out-shipments of this commodity. The sign of opop confirms this effect, but this variable is barely significant. A negative and significant oaps indicates that the sector is characterized by diversification effects. The variables that positively affect out-shipments in this commodity group are oemp and owsem. In both years, the Box-Cox models are statistically superior to the log-linear specifications. The results show that only the origin final demand sector variables are significant for this product group. When we consider that this group has a high distance coefficient, and thus is very sensitive to transportation, this result makes sense. One can expect that petroleum and coal products are bulkier and can relatively be more dangerous to haul over longer distances. For this reason, the supply at the origins may not be very sensitive to the demand at the destinations.

5. 2. 7. Rubber & Miscellaneous Plastic Products The regression results for rubber and other plastic products, commodity group 30, are

presented

in

Table

5.12. Adjncy

is significant, with the expected sign. The

insignificance of the custom district dummy variable suggests that there are no effects of foreign trade in this sector, although the share of foreign trade for this group is around 10 % for both imports and exports. The negative cd variable implies competition effects at the node cluster around destinations, but the io variable is not significant. highly significant, with the

expected sign

each year.

Opipc and

Distance is

opop are

both

negative and significant, as expected: outflows decrease with increasing local final consumption. Oemp is not significant in 1993, yet negatively significant in 1997.

On the

other hand, ovlad, the other sectoral production variable, is significant, with the expected sign, each year. Wholesale employments is significant at both the origin and destination, suggesting that redistribution effects are important for this commodity. The intermediate demand variable, dmnem is significant each year, and the final demand variables at the destination, dpipc and dpop, are significant and positive, as expected, in 1993, but not in 1997. The 3-digit breakdown shows that this product group includes rubber tires, plastic footwear, plastic hose, belting

and all other miscellaneous plastic products (see Table

C.34 in Appendix C). These products are used at both the final and intermediate demand 71

levels, which is supported by our statistical findings. Wholesalers of this commodity provide for both demand sectors. In both years, the Box-Cox models are statistically superior to the log-linear specifications. Table 5.12. Rubber and Misc. Rubber Products Regression Parameters

30 constant adjncy

1993

T

1997

T

-7.088 **

-1.78

9.425 *

2.60

1.142 *

5.64

0.882 *

4.47

ocddmy

-0.166

-1.63

0.036

0.32

dcddmy

-0.048

-0.48

0.111

1.00

cd

-0.422 *

-4.05

-0.155 *

-3.12

io

-0.093

-1.46

0.036

1.34

dist

-1.069 *

-8.37

-0.718 *

-6.81

opipc

-0.692 *

-3.53

-0.545 *

-3.59

opop

-0.326 *

-3.31

-0.150 *

-2.85

oemp

0.123

1.34

-0.152 *

-2.69

owsem

0.755 *

4.53

0.405 *

3.95

ovlad

0.679 *

5.55

0.594 *

5.16

oaps

-0.009

-0.09

-0.114

-0.90

dmnem

0.244 *

4.06

0.142 *

3.56

dwsem

0.422 *

3.84

0.276 *

3.71

dpipc

0.445 *

2.51

0.014

0.17

dpop

0.240 *

2.47

-0.006

-0.25

lambda

0.103 *

6.25

0.179 *

8.38

theta

0.193 *

44.43

0.196 *

38.76

MLHD

-3837.35

-5605.73

LLHD

-4684.41

-6526.13

X2 N

1694.11 *

27.59

1604

* Significant at 5 % level **

1840.82 *

27.59

1308

Significant at 10 % level

5. 2. 8. Clay, Concrete, Glass and Stone Products Table 5.13. presents the regression results for commodity group 32: clay, concrete, glass or stone products. The significant and positive adjncy indicates that sharing a physical border increases commodity trade. Ocddmy is negative and significant 72

in 1997, pointing to a significant export effect. The coefficient of this variable is also negative in 1993, but, insignificant. Foreign trade data show that the import in this sector is larger than its export, but this is not captured by these variables. The cd variable is negative and significant

each year, suggesting a competition effect at destinations.

However, the variable io is positive

each year, although not significant in 1997.

This

suggests that intervening opportunities around the origins do not decrease the shipments to farther destinations, unlike other commodity groups. One possible explanation for this Table 5. 13. Clay, Concrete, Glass and Stone Products Regression Parameters

32 constant adjncy

1993

T

T

-34.0399 *

-4.730

-23.159 *

-2.890

1.3662 *

5.757

1.391 *

4.218

-0.457 *

-2.630

-0.029

-0.179

-0.549 *

-2.178

ocddmy

-0.1157

-0.898

dcddmy

0.0778

0.662

cd

-0.7736 *

-3.049

io

0.2632 *

2.021

-1.4072 *

-5.867

dist

1997

0.170 -1.368 *

1.424 -4.275

opipc

0.1425

0.490

0.141

0.433

opop

-0.2930

-1.639

-0.116

-0.762

oemp

0.9571 *

2.943

owsem

0.0691

0.400

ovlad

0.3584 **

1.651

oaps

0.885 * -0.144 0.454 ** -0.858 *

2.485 -0.798 1.958

-0.6571 *

-2.250

dmnem

0.3606 *

2.903

0.188 **

1.809

dwsem

0.3458 *

2.062

0.547 *

2.087

dpipc

1.2599 *

3.060

0.710 **

1.873

dpop

0.6517 *

2.197

0.239

1.267

lambda

0.0684 *

2.795

0.096 *

2.872

theta

0.1306 *

23.301

0.088 *

10.973

MLHD

-756.08

-100.73

LLHD

-1001.00

-165.82

X

2

489.85 *

N * Significant at 5 % level **

27.59

1644

130.17 * 892

Significant at 10 % level

73

-2.251

27.59

situation is that, as for furniture products, this product group also displays a spatial concentration. Ten states produce approximately 70 % of total US shipments. Dist is negative and significant. The origin final demand variables, opipc and opop, are not significant, suggesting that this sector does not provide for the final demand sector at the origin. Oemp and ovlad, representing sectoral productive capacities, are positive and significant each year. This sector may also be characterized by diversification effects, since oaps is negative and significant each year. Dmnem and dwsem are both significant each year, with the expected signs, indicating that intermediate and redistributive demands at the destinations have significant effects on the trade of this commodity. The destination final demand variables, dpipc and dpop, are positive and significant. Two main items in this commodity group are glass and concrete based products. Although these items may not demanded directly by final demand sectors, they are mainly supplied for the construction industries, just like lumber and wood products, and they are demanded indirectly by the final demand sectors via the construction sector. Finally, the Box-Cox models in both years are statistically superior to the log-linear specifications. 5. 2. 9. Primary Metal Products The regression results for primary metal products, commodity group 33, are presented in Table 5.14. The variable adjncy is significant in both years. The foreign trade variables, ocddmy and dcddmy, are negative and significant, although at the 10 % level, indicating foreign trade effects in terms of both imports and exports. The foreign export and import shares are 6 % and 12 %

in 1993, and these effects are captured by the

custom district variables. Also, the nodes clustered around both origins and destinations display competition effects, since the variables cd and io are negative and significant. Dist is significant and negative, implying that flows decrease with distance. Only opipc is significant in 1997, with the expected signs at the 5 % level. Oemp, the sectoral production variable, is highly significant and positive in each year. The other sectoral production variable, ovlad, is negative in 1993, yet significant at the 10 % level. However, ovlad is significant with the expected sign in 1997. Owsem is positive and significant each year,

indicating that wholesale employment

at the origin has a significant

impact on trade flows. Oaps has a negative sign in both years, but is not significant in 1993.

Dmnem is positive and

significant

effects at destinations. Dwsem is also

each

year, pointing to intermediate

demand

positive and significant each year. Only one 74

Table 5.14. Primary Metal Products Regression Parameters

1993

T

-14.839 *

-2.32

0.848 *

ocddmy

33 constant

T

5.045

0.88

3.73

0.420 **

1.67

-0.222 **

-1.78

-0.265 **

-1.80

dcddmy

-0.358 *

-2.97

-0.234 **

-1.72

cd

-0.479 *

-3.18

-0.447 *

-3.14

io

-0.415 *

-3.06

-0.320 *

-3.17

dist

-1.574 *

-7.24

-1.482 *

-7.50

adjncy

opipc

0.039

0.14

-0.656 *

-2.12

opop

0.203

1.28

-0.267 **

-1.66

oemp

1.230 *

4.49

0.668 *

5.46

owsem

0.297 **

1.79

0.909 *

3.51

ovlad

-0.213 **

-1.69

0.233 *

2.70

oaps

-0.107

-0.82

-0.527 *

-3.68

dmnem

0.485 *

3.88

0.539 *

3.87

dwsem

0.290 **

1.88

0.302 **

1.86

dpipc

-0.450

-1.64

-0.339

-1.33

dpop

0.568 *

2.43

0.246

1.64

lambda

0.071 *

3.48

0.088 *

4.55

theta

0.176 *

38.97

0.143 *

28.37

MLHD

-3671.21

-3011.63

LLHD

-4319.39

-3366.60

X2

1296.36 *

N

1511

* Significant at 5 % level **

of the destination

1997

27.59

709.93 *

27.59

1178

Significant at 10 % level

final demand variables, dpop, is significant and positive in 1993. This

product group includes steel works, iron or steel castings, nonferrous metal products, metal based alloy castings, and other miscellaneous primary metal products. Obviously, all of these items are consumed at the intermediate demand level. Although one or two final demand variables display some degree of significance, the final demand proxies do

75

not show a systematic significance. On the other hand, the variable that measures the effect of intermediate demand is highly significant in both years. For this reason, wholesalers in this sector may be selling mainly to the intermediate demand sectors. In both years, the Box-Cox models are statistically superior to the log-linear models. 5. 1. 10. Fabricated Metal Products Table 5.15 presents the regression results for commodity group 34: fabricated metal products. As for all other commodity groups so far, adjncy is a positive and significant variable each year. Among foreign trade variables, only occdmy is negative and significant

in

1993,

implying

an export

effect. However, this product group is the

weakest foreign trade sector in the US. Its export share is about 2 %, and its import share is around 4 %.

Cd is negative and significant each year: competition effects dominate at

the destinations. Io is significant and negative in 1993, but is not significant in 1997. Dist is highly significant each year, with a negative sign. Opipc has a

negative sign

in

both

years, yet is only significant in 1997. On the other hand, opop is both negative and significant each year, as expected. Owsem is positive and significant each year, implying that wholesale employment at the origin facilitates the trade of fabricated metal products. The only significant sectoral production variables are ovlad in 1993 and oemp in 1997. Oaps is significant in 1997, with a negative sign,

implying that this product is

characterized by diversification effects. All destination variables, dmnem, dwsem, dpipc, and dpop are positive and significant each year, as expected. The main products in this group are metal cans, hand tools or general hardware, plumbing fixtures, structural metal products, bolts, nuts, screws, rivets, metal stampings, miscellaneous wire products, and other fabricated metals. (see Table C.34 in Appendix C) As the breakdown shows, this commodity final

demand

sectors

together.

group

supplies

both the intermediate

and

Our findings also support this intuition since both final

and intermediate demand proxy variables are significant. Since the wholesale trade sector is supplying

both demand categories, it is also significant in the flow of fabricated metal

products. The Box-Cox models are statistically superior to the log-linear specifications.

76

Table 5. 15. Fabricated Metal Products Regression Parameters

1993

34 constant

1997

T

-12.803 *

-2.83

1.218 *

6.88

ocddmy

-0.217 *

-2.26

-0.017

-0.14

dcddmy

0.110

1.25

0.120

1.10

cd

-0.537 *

-4.75

-0.709 *

-3.51

io

-0.215 *

-2.92

-0.032

-0.33

dist

-1.256 *

-9.68

-1.427 *

-7.09

opipc

-0.299

-1.49

-1.371 *

-3.49

opop

-0.313 *

-3.01

-0.675 *

-2.70

oemp

-0.040

-0.32

0.805 *

2.80

adjncy

9.103

1.53

0.996 *

5.09

owsem

0.896 *

5.46

1.000 *

3.86

ovlad

0.857 *

5.63

0.259

1.38

oaps

0.247

1.50

-0.980 *

-3.43

dmnem

0.236 *

4.38

0.246 *

2.93

dwsem

0.347 *

3.63

0.506 *

2.83

dpipc

0.368 *

2.30

0.087

0.38

dpop

0.430 *

3.45

0.575 *

2.48

lambda

0.096 *

6.50

0.076 *

3.68

theta

0.211 *

54.33

0.192 *

38.34

MLHD

-5429.08

-4430.00

LLHD

-6718.99

-5239.25

X2

2579.82 *

N *

T

Significant at 5 % level

27.59

1744 **

1618.50 *

27.59

1174

Significant at 10 % level

5. 2. 11. Machinery Products (Non-Electrical) The regression results for commodity group 35, non-electrical machinery, are presented in Table 5.16. Adjncy is positive and significant, implying increasing trade between contiguous states. Dcddmy is significant and positive in both years while ocddmy is negative and significant in 1997, which suggests that this product group is subject to significant foreign export effects, especially in 1997. Non-electrical machinery

has the

highest share of foreign trade in the whole US, both in terms of export and import. Cd and io are negative and significant each year, implying competition effects at both ends. 77

Dist

is again highly significant and negative, as expected. The origin variables, opipc and opop are

negative and

products

significant,

indicating

that

local

consumption

of

machinery

at the origin decreases out-shipments. Ovlad is significant and positive both

years, as expected. Oemp , on the other hand, is positive and significant only in 1993, and at the 10 % level.

The redistribution variables at the origin and destination, owsem and

dwsem, are both significant and positive

each year. The intermediate demand variable at

the destination, dmnem, has also a positive and significant effect on the flows of this commodity group. The only significant final demand variable at the destination is dpop. Table 5. 16. (Non-Electrical) Machinery Products Regression Parameters

35 constant

1993

T

-15.730 *

-3.53

adjncy

1.524 *

ocddmy dcddmy

1997

T

0.736

0.11

7.55

0.943 *

4.07

0.039

0.36

-0.250 *

-1.96

0.172 **

1.67

0.257 *

2.07

cd

-0.423 *

-3.95

-0.593 *

-3.24

io

-0.145 *

-2.51

-0.304 *

-2.15

dist

-0.860 *

-7.84

-1.329 *

-6.96

opipc

-0.261

-1.61

-1.175 *

-3.08

opop

-0.321 *

-3.29

-0.290

-1.60

oemp

0.165 **

1.90

0.213

1.50

owsem

0.877 *

4.99

0.877 *

3.70

ovlad

0.488 *

4.24

1.074 *

5.42

oaps

0.069

0.47

0.173

0.92

dmnem

0.091 *

2.41

0.185 *

2.14

dwsem

0.468 *

4.08

0.527 *

2.46

dpipc

0.262 **

1.73

dpop

0.329 *

2.83

0.695 *

2.47

lambda

0.114 *

6.42

0.066 *

3.19

theta

0.198 *

48.29

0.190 *

37.13

-0.268

MLHD

-6324.75

-5519.14

LLHD

-7520.52

-6356.18

X2

2391.54 *

N * Significant at 5 % level **

27.59

1563

1674.06 * 1271

Significant at 10 % level

78

-0.92

27.59

Sub-products

of

this

group

include

engines,

turbines,

farm

machinery,

construction and mining equipments, metal working machinery, office, computing or accounting machineries, service industry machines and other miscellaneous machineries (see Table C.34 in Appendix C). All of these products are used both demand sectors, and this facts supports our empirical findings that both of the demand proxy variables are significant as expected. Furthermore, the wholesale sector is likely supplying both demand sectors. The Box-Cox models

are statistically superior to the log-linear models in both

years.

5. 2. 12. Electrical Machinery Products Table 5.17 presents the regression results for electrical machinery products, product group 36. Adjncy is positive and significant, as in all other previous commodity groups. Among the custom district variables, only ocddmy is positive and significant in 1997, pointing to an import effect this year. However, this group had a significant amount of foreign imports and exports in 1993 as much as in 1997, but these 1993 flows are not accounted

for by the custom district variables. Cd is negative and significant each year,

implying competition effects at the clusters around destinations. However, io is only significant in 1997, with a negative sign. The final demand variables, opipc and opop, are negative and significant, as expected, except for opipc in 1997. Although oemp is positive and significant, the other sectoral production variable, ovlad, is negative and significant each year, an obviously counterintuitive result. Owsem and dwsem are significant and positive each year, pointing to the effects of redistributive activities in the electrical machinery sector. Oaps is negative and significant, implying diversification effects. The effects of the final demand variables at the destination, dpipc and dpop, each year, but the intermediate demand

are significant

variable at destinations, dmnem, is not, which is

surprising in light of the role of such equipment products in manufacturing. The sub-categories of this product group include household appliances, radio and television

sets,

communication

equipments,

electronic

components

or

accessories,

electrical industrial equipments and other electrical machinery and equipments (see Table C.34 in Appendix C). These products items are generally consumed by the final demand sectors, and, for this reason, only the final demand variables at both the origins and the destinations display significance, in addition to the wholesale employment variable. In this case, wholesalers are supplying the final demand sectors. 79

The Box-Cox models are statistically superior to the log-linear specifications in both years. Table 5. 17. Electrical Machinery Products Regression Parameters

36 constant

1993

T

1997

T

-30.241 *

-4.74

-25.810 *

-4.51

1.301 *

6.49

0.862 *

4.34

ocddmy

-0.178 **

-1.67

0.267 *

2.29

dcddmy

-0.075

-0.73

-0.138

-1.25

cd

-0.675 *

-4.28

-0.256 *

-3.45

io

-0.023

-0.28

-0.198 *

-2.60

dist

-0.957 *

-8.30

-0.649 *

-7.35

opipc

-1.159 *

-2.92

-0.045

-0.25

opop

-0.445 *

-2.65

-0.452 *

-3.40

oemp

1.537 *

6.06

0.929 *

5.46

owsem

0.805 *

4.29

0.831 *

4.51

ovlad

-0.360 *

-3.66

-0.235 *

-4.66

oaps

-1.027 *

-5.63

-0.520 *

-4.40

adjncy

dmnem

0.053

0.88

0.011

0.30

dwsem

0.562 *

3.62

0.391 *

3.48

dpipc

1.517 *

4.57

0.708 *

3.72

dpop

0.931 *

3.53

0.316 *

2.86

lambda

0.070 *

4.33

0.122 *

6.62

theta

0.185 *

45.51

0.156 *

29.48

MLHD

-5967.28

-6746.63

LLHD

-7053.58

-7418.27

X2

2172.61 *

N

1581

* Significant at 5 % level **

27.59

1343.29 *

27.59

1274

Significant at 10 % level

5. 2. 13. Transportation Equipment The regression results for transportation equipment, commodity group 37, are presented in Table 5.18. In contrast to all previous commodity groups, adjncy is not significant. Except for ocddmy in 1993, which implies a foreign export effect, none of the custom district variables are significant. However, this product group is one of the important foreign trade industries in the US, both in 1993 and 1993. This is not accounted for the foreign trade variables. One of the geographical structure variables, cd, is significant and negative, indicating competition effects at the destinations. Io is 80

Table 5. 18. Transportation Equipment Regression Parameters

37

1993

constant adjncy

T

1997

T

-5.709

-1.00

5.265

0.61

0.062

0.21

0.279

0.60

-3.71

-0.107

-0.40

0.45

-0.038

-0.15

-0.511 *

-2.25

ocddmy

-0.597 *

dcddmy

0.071

cd

-0.559 *

-3.90

io

0.294 *

3.63

dist

-1.241 *

-7.54

-1.220 *

-4.49

opipc

-0.791 *

-3.54

-0.194

-0.66

opop

-0.298 *

-3.00

-0.493 *

-2.11

oemp

0.151 *

2.02

-0.128

-0.94

owsem

0.411 *

3.19

0.758 *

3.02

ovlad

0.585 *

5.75

0.971 *

3.39

oaps

-0.804 *

-5.37

-1.148 *

-3.87

dmnem

0.155 *

2.81

0.073

0.97

dwsem

0.180 **

1.74

0.365 **

1.90

dpipc

0.760 *

3.21

0.047

0.17

dpop

0.424 *

2.95

0.324

1.59

lambda

0.131 *

6.91

0.135 *

4.42

theta

0.139 *

30.51

0.110 *

13.65

0.059

MLHD

-4061.13

-862.86

LLHD

-4488.01

-939.80

X2

853.76 *

N

1292

* Significant at 5 % level **

27.59

153.89 *

0.62

27.59

420

Significant at 10 % level

significant and positive only in 1993, indicating that intervening opportunities do not decrease the supply to farther away destinations. When we consider that adjncy is not significant for this product group, this result seems consistent with the fact that transportation equipment supply is not necessarily consumed by close-by destinations. Dist is negative and significant, as expected. Opipc and opop are negative and significant (except for opipc in 1997), pointing to local consumption effects at the origin, that reduce out-shipments. Oemp significant, as expected.

Oaps

in 1993, and ovlad in each year,

are positive and

is negative and significant, implying diversification effects.

Owsem and dwsem are positive and significant each year: redistributive activities appear 81

to be important in the trade of transportation equipment. The intermediate demand variable, dmnem, at the destination is significant only in 1993, while the

final demand

variables at destinations, dpipc and dpop, are both positive and significant each year. Transportation equipment includes motor vehicles and equipments, aircrafts or parts,

ships

or

boats,

railroad

equipments,

motorcycles,

bicycles,

and

other

transportation equipments (see Table C.34 in Appendix C). All these products are equally important for both the final and intermediate demand sectors, and the empirical findings support this intuition. Wholesalers of these products groups are selling to both demands sectors. The Box-Cox

models are statistically superior to the log-linear specifications in

both years. 5. 2. 14. Precision Instruments Table 5.19 presents the regression results for

commodity group 38, instruments:

photographic goods, optical goods, watches, or clocks. Adjncy is a significant variable in the trade of precision instruments: sharing a common border increases the trade between states. Foreign trade variables are not significant, except for dcddmy in 1993, which is positive and may point to an export effect. Although import in this sector is as important as export, import effects are not captured by the foreign trade variables. Cd is negative and significant

each year,

while io is significant and negative only in 1997: both end

destinations are characterized by competition effects. Dist is, as expected, negative and significant. The origin final demand on

the

trade

dpipc and dpop,

of

variables, opipc and

opop, have negative

effect

precision instruments, while the destination final demand variables,

have positive and significant effects in each year. Oemp is positive and

significant, as expected, each year. However, ovlad has a significant and negative coefficient in 1997.

Redistribution effects are significant and positive, as expected, at

both the origins and destinations each year. Oaps is negative and significant, implying diversification effects. The intermediate demand variable at destination, significant.

82

dmnem, is not

Table 5. 19. Precision Instruments Regression Parameters

38 constant

1993

T

1997

T

-18.672 *

-3.00

-38.356 *

-3.56

adjncy

1.515 *

6.63

0.898 *

3.94

ocddmy

0.180

1.51

0.012

0.09

dcddmy

0.228 *

2.01

0.105

0.87

cd

-0.258 *

-2.72

-0.392 *

-1.99

io

0.089

1.20

-0.484 *

-2.12

dist

-0.696 *

-7.44

-0.739 *

-5.71

opipc

-1.010 *

-3.16

-0.509

-0.86

opop

-0.981 *

-3.90

-1.830 *

-3.37

oemp

0.924 *

5.88

2.162 *

5.06

owsem

1.415 *

4.96

2.644 *

4.48

ovlad

0.074

0.83

-0.506 *

-4.53

oaps

-0.855 *

-5.06

-1.210 *

-4.70

dmnem

-0.105 **

-1.67

-0.159

-1.13

dwsem

0.492 *

3.45

0.555 **

1.84

dpipc

0.983 *

3.66

1.455 *

2.73

dpop

0.577 *

3.12

1.507 *

2.75

lambda

0.080 *

5.05

0.024

1.25

theta

0.165 *

37.56

0.153 *

MLHD

-2900.42

-2321.54

LLHD

-3471.28

-2724.45

X2

1141.71 *

N * Significant at 5

27.59

1421 % level

**

29.01

805.83 *

27.59

942

Significant at 10 % level

This product group is made of engineering, laboratory, or scientific instruments, measuring, controlling, or indicating instruments, optical instruments and lenses, surgical, medical, dental instruments and supply, opticians goods, photographic goods, and clocks and watches (see Table C.34 in Appendix C). These products are mainly consumed by either households or service sectors. In either case, they serve

final consumption. This

fact supports the model findings that only the final demand variables are significant in the out-shipment of this product, supported by a significant wholesale sector. The Box-Cox models are statistically superior to the log-linear specifications in both years. 83

5. 2. 15. Miscellaneous Manufacturing Products The

regression

results

for

miscellaneous

manufacturing

products,

commodity

group 39, are presented in Table 5.20. Adjncy is positive and significant, but the custom district variables are not. The US custom district data do not indicate an important foreign trade share for this group of products. Cd is negative and significant each year, indicating competition effects at destinations, while io is significant and negative only in 1993. Dist is again highly significant and negative. Although only significant in 1993, opipc has a positive sign in contrast to expectation. On the other hand, opop is negative but only Table 5. 20. Miscellaneous Products Regression Parameters

1993

T

1997

T

-47.395 *

-6.84

-44.502 *

-4.93

adjncy

1.469 *

7.57

1.229 *

7.27

ocddmy

0.008

0.08

0.098

1.00

dcddmy

0.134

1.40

0.078

0.87

cd

-0.581 *

-3.53

-1.160 *

-3.63

io

-0.162 **

-1.96

0.217

1.39

dist

-0.926 *

-7.03

-1.430 *

-7.28

0.695 *

2.54

0.537

1.23

-1.09

-1.000 *

-2.35

39 constant

opipc opop

-0.135

oemp

0.785 *

4.99

1.090 *

4.33

owsem

0.573 *

3.52

1.713 *

3.89

ovlad

0.261 *

3.28

0.110

1.13

oaps

0.037

0.23

-0.096

-0.42

dmnem

0.052

0.94

0.211 **

1.83

dwsem

0.316 *

2.42

0.987 *

3.23

dpipc

1.055 *

3.48

0.842 *

2.19

dpop

0.817 *

2.87

1.550 *

2.72

lambda

0.072 *

3.51

0.015

0.74

theta

0.191 *

44.87

MLHD

-4166.82

-7173.09

LLHD

-5089.58

-7998.06

X

2

N *

0.173 *

Significant at 5 % level

1845.53 *

27.59

1559 **

1649.93 * 1466

Significant at 10 % level

84

31.55

27.59

significant in 1997. Oemp is positive and significant each year, while ovlad is only significant and positive in 1993. Redistribution activities have a significant and positive effect on the trade of this commodity group, and dwsem variables

as implied by the positive signs of

owsem

each year. The other significant variables are the destination final

demand variables, opipc and opop, each year. The intermediate demand variable, dmnem, is positive and significant at the 10 % level only in 1997. Jewelry, silverware, musical instruments, toys, amusements, sporting or athletic goods, pens, pencils, artists and office supplies, are the main items in this product group, and obviously they are consumed at the final demand level (see Table C.34 in Appendix C). For this reason, intermediate demand proxy variables do not display a high degree of significance. The Box-Cox models are statistically superior to the log-linear specifications in both years. 5. 2. 16. Apparel, Textile, Leather Products The commodity group 75 is a combination of three similar product groups (apparel, textile and leather products) and Table 5.21 presents the corresponding regression results. Adjncy is positive and significant. Although this product group has relatively high imports in both 1993 and 1997, they are not captured by the model variables. Cd and io are both negative and significant each year, implying competition effects at both ends. Dist is again negative and highly significant each year. Although this commodity group basically provides for the final demand sectors, opop is negative, as expected, only in 1993. All the other final demand variables at the origin are positive and significant, in contrast to expectations. Oemp is positive and significant, as expected, but ovlad is insignificant or weakly significant. Redistribution activities have an impact on this trade (except for dwsem in 1993). Since this product is consumed at both demand levels, the intermediate and final demand variables at the destinations are positive and significant. Finally,

the

Box-

Cox

models

are

specifications in both years.

85

statistically

superior

to

the

log-linear

Table 5. 21. Apparel, Textile, Leather Products Regression Parameters

75

1993

constant adjncy

1997

T

-85.186 *

-4.03

-98.340 *

-6.34

1.045 *

2.72

0.550 *

2.54

ocddmy

-0.213

-1.07

-0.010

-0.08

dcddmy

-0.118

-0.62

0.142

1.24

cd

-2.612 *

-2.64

-1.066 *

-3.22

io

-1.122

-1.60

-1.438 *

-3.19

dist

-2.037 *

-5.07

-1.400 *

-6.82

1.04

4.330 *

4.09

opipc

1.591

opop

-3.104 **

-1.94

0.885 *

1.99

oemp

3.789 *

3.38

1.484 *

4.33

owsem

3.460 *

2.94

0.556 **

1.83

ovlad

-0.355 **

-1.65

-0.008

-0.05

oaps

-1.348 *

-3.50

0.153

0.93

dmnem

1.484 *

2.50

0.577 *

2.93

dwsem

0.643

0.97

0.550 **

1.90

dpipc

4.822 *

2.80

1.505 *

2.76

dpop

3.216 *

1.99

1.465 *

2.54

-1.01

0.021

1.01

20.87

0.172 *

lambda

-0.026

theta

*

T

0.125 *

MLHD

-1272.02

-5558.05

LLHD

-1457.25

-6343.57

X2

370.46 *

N

680

Significant at 5 % level

**

27.59

1571.05 *

34.06

27.59

1221

Significant at 10 % level

5. 2. 17. Synthesis In

addition

to

reviewing

each

commodity

instructive to review them in a comparative fashion.

group

equation

separately,

it

is

The coefficients of the variables and

their significance levels across commodity groups are presented in Tables 5.22 and 5.23 for 1993, and in Tables 5.24 and 5.25 for 1997. A common physical border significantly increases commodity exchange between contiguous states. For 15 out of the 16 commodity groups in 1993 the dummy variable Adjncy is significant at the 5 % level and positive. In 1997, for 11 out of the 16 commodity groups, the variable is positive and significant at the 5 % level, and for 3 86

groups it is so at the 10 % level. This result is consistent with the estimates of all empirical, gravity-type models of international trade that include a contiguity dummy variable. While an increasing distance guarantees a declining interaction, this decline is attenuated among contiguous states. The ability to obtain better business information about supplies and/or consumers, as well as possible cultural commonalities, are most likely factors explaining this phenomenon. It is also likely that a business trying to expand its market beyond state boundaries will first focus on neighboring states before expanding beyond, thus ensuring a differential advantage to these states. It is also possible that short-haul transportation between contiguous states may be different, and less expensive than in the case of greater distances. The foreign trade dummy variables, ocddmy and dcddmy, do not display the same level of consistency as adjncy. In 1993, only 5 commodity groups have a significant occdmy at the 5 % level, and 4 groups at the 10 % level. However, in 1997, ocddmy is significant in only 4 groups at the 5 % level, and 1 at the 10 % level. The other foreign trade variable, dcddmy, performs even more poorly: it is significant at the 5 % level for 3 commodity groups, and at the 10 % level for only 2

groups, in 1993. In 1997 only 1

sector is significant at the 5 % level, and 1 sector at the 10 % level. Focusing on ocddmy in 1993, we note that seven out of nine significant coefficients are negative (commodity groups 20, 24, 30, 33, 34, 36, 37), which suggests that foreign exports taking place at the origin node reduce the interregional commodity outflows from these nodes. These results are consistent with the significant foreign export volumes of sectors 36 (non-electrical machinery) and 37 (transportation equipment), and, to a lesser extent, of sector 20 (food), 30 (rubber and plastics), and 33 (primary metals) (see Table 5.4). In contrast, the coefficient is positive for commodity group 26 and 29, which suggests that when the origin node imports pulp and paper, and petroleum/coal products, these foreign imports stimulate interregional flows out of these origin nodes. In the case of petroleum, this result is very much consistent with the importance of foreign imports in the US economy. The competing destination variable, cd, is uniformly negative and significant in all groups in 1993: in 15 groups at the 5 % level, and 1 group at the 10 % level. In 1997, cd is significant at the 5 % level in 13 groups, and in 1 group at the 10 % level. These results suggest that competition effects at destinations are strong determinants of interregional commodity flows. As other destinations are physically closer (clustered) to a specific destination, the flow of commodities reaching this destination decreases. Every other factor remaining constant, this clustering absorbs part of the flow that would have ended at this destination under a less clustered configuration. This result is consistent 87

with similar effects empirically uncovered in the case of other spatial interactions (e.g., migrations, telecommunications). The

intervening

opportunities

consistent effects as cd. It

variable,

io, does not have the same highly

displays mostly negative signs.

In 1993, 10 groups have

negative signs at the 5 % significance level, 1 group has a negative sign at the 10 % level, and 2 groups have a positive and significant io coefficient. In 1997, 7 groups have a negative sign at the 5 % level, and 1 group at the 10 % level. Overall, competition effects at the supply level appear dominant. Destination nodes clustered around the origin serve as alternative destinations for the commodity, and absorb past of the flow that would have ended at the selected destination. The distance variable, dist, is always negative and highly significant for all commodity groups in both years. Distance can be viewed as a proxy for transportation cost, and increasing transportation costs are an obvious deterrent to trade. Also, from an information

viewpoint,

opportunities, and

the

farther

away

the

lesser

the

information

about

business

hence the lesser the interactions. The distance coefficients for sectors

35 through 39 are generally lower (in absolute terms) than those for the other sectors, which indicate that the shipping distances for these goods are greater. This is consistent with the value per weight of these commodities (see Table 5.1), which ranges from $748/ton to $ 5,566/ton in 1993. The latter characterizes precision instruments (group 38), which have the lowest distance coefficient (-0.70). The other groups, which have values per weight ranging from $11/ton to $427/ton, have distance coefficient varying from –1.07 to –2.20. The origin state personal income per capita, opipc, is presumed having a negative sign:

when

the

local

consumption

of

the

commodity

increases,

decreases. This is generally verified for the commodity groups that are

its

out-shipment

involved in final

consumption (e.g., petroleum, rubber, machinery, transportation equipment, precision instruments). In 1993, 6 groups have negative signs at the 5 % significance level, and 2 groups at the 10 % level. In 1997,

6 groups have negative signs at the 5 % level. The

origin state total population variable , opop, is similar to opipc: both opop and opipc are used as surrogate for local final consumption. In 1993, 8 commodity groups have negative and significant signs at the 5 % level, and 2 groups at the 10 % level. In 1997, 7 groups have a negative sign at the 5 % significance level, and 2 groups at the 10 % level. When significant, opop and opipc have generally the same sign, which supports their use as measures of origin final consumption. 88

The origin sectoral employment, oemp, is generally positive and significant, as hypothesized. In 1993, 13 commodity groups are positive at the 5 % significance level. In 1997, 11 groups are positive and significant at the 5 % level, and only 1 group has a negative sign, in contrast to expectation. The other sectoral production variable, ovlad, is also mostly positive. In 1993, 9 commodity groups have positive signs at the 5 % level, and 1 group at the 10 % level. In 1997, however, only 8 groups have positive signs at the 5 % level. In 1993, 3 groups, and in 1997, 2

groups, have negative

signs, a result

contrary to expectations The variable used to measure scale and diversification effects at the origin states, oaps, has generally a negative sign when significant: as plant size decreases (for a fixed total output), out-shipments increase. More plants suggest more diversified products being produced, making them more attractive to export markets. In 1993, 9 commodity groups have negative signs at the 5 % significance level, and

8 groups do so in 1997.

Only commodity group 25 (furniture) has a positive sign in 1993 and 1997. Another important result of the study is that wholesale employment, representing redistributive activities, is

important in facilitating interregional commodity flows. In

1993, 13 groups have a significant positive sign for owsem (12 of which are at the 5 % level), and 14 groups have a positive dwsem (13 of which are significant at the 5 % level). In 1997, 11 groups have a significant positive sign for owsem, (9 of which are at the 5 % level), and 13 groups have a positive dwsem (all of which are at the 5 % level). The destination manufacturing employment, dmnem, is a proxy for the effects of intermediate demand sectors at the destination, and it is presumed to have a positive sign. In 1993, this presumption is verified for 11 commodity groups at the 5 % significant level. In 1997, 9 groups have significant and positive signs (7 at the 5 % level). The variables representing final demand at destinations, dpipc and dpop, are expected to be positive. In 1993, this is verified for 13 commodity groups for both variables. In 1997, however, dpipc is significant and positive for 8 commodity groups only, and dpop is so for 9 groups only. When comparing the overall results for 1993 to 1997, it is clear that the 1993 models perform much better. The likely explanation is that the 1993 analysis uses the STCC system directly, while the 1997 analysis converts quantities defined in the SCTG system into the STCC system, and this conversion requires deleting many incomplete

89

observations, eventually leading to smaller, and possibly less representative, samples. Overall, and in particular in 1993, the selected variables are generally significant in explaining interregional commodity flows in line with expectations. It is also noteworthy that the optimized Box-Cox specification is always statistically superior to the log-linear specification, which has been a mainstay of past empirical work.

90

Table 5. 22. The 1993 Model Variable Coefficients and Their Significance Levels Across Commodity Groups (20-32)

1993 adjncy ocddmy dcddmy

20

24

25

26

28

29

30

32

1.62 *

1.82 *

0.50 *

0.51 *

1.84 *

8.25 *

1.14 *

1.37 *

-0.27 *

-0.20 **

0.27 *

0.01

1.84 *

-0.17 **

0.16

0.33

-0.05

0.21 **

0.07

-0.10 0.20 **

-0.01

-0.12 0.08

cd

-0.85 *

-1.84 *

-1.13 *

-0.60 *

-0.23 *

-0.41 **

-0.42 *

io

-0.67 *

-1.28 *

-0.58 *

-0.50 *

-0.20 *

-0.18

-0.09

dist

-2.20 *

-2.22 *

-1.64 *

-1.62 *

-1.29 *

-3.00 *

-1.07 *

opipc

-0.09

2.22 *

0.09

0.00

-2.19 *

-0.69 *

0.14

opop

-0.57 *

0.09

0.07

0.22 **

-0.10

-0.45 **

-0.33 *

-0.29

oemp

0.87 *

-0.23

1.38 *

0.52 *

0.45 *

1.95 *

0.12

0.96 *

owsem

0.90 *

0.42

0.68 *

0.15

0.23 *

0.78 *

0.75 *

0.07

ovlad

1.05 *

2.37 *

0.49 *

0.60 *

0.68 *

0.36 **

oaps

-0.88 *

0.33

dmnem

0.19 *

1.02 *

dwsem

1.02 *

0.76 *

dpipc

1.05 *

dpop Lambda Theta *

Significant at 5 % level

-0.06 0.45 *

-0.16

-0.71 *

-0.12 -2.85 *

-0.01

0.26 * -1.41 *

-0.66 *

0.37 *

0.10 *

0.03

0.24 *

0.36 *

0.71 *

0.67 *

0.73 *

0.28

0.42 *

0.35 *

3.40 *

2.42 *

0.84 *

-0.12

-0.10

0.44 *

1.26 *

0.47 *

1.12 *

1.25 *

0.17

0.05

0.31

0.24 *

0.65 *

0.08 *

0.02

0.04 *

0.09 *

0.11 *

0.19 *

0.10 *

0.07 *

0.19 *

0.17 *

0.12 *

0.18 *

0.18 *

-0.24 *

0.19 *

0.13 *

**

-0.09

-0.37

-0.77 *

Significant at 10 % level

91

Table 5. 23. The 1993 Model Variable Coefficients and Their Significance Levels Across Commodity Groups (33-75)

1993

33

34

35

36

0.85 *

1.22 *

1.52 *

1.30 *

ocddmy

-0.22 **

-0.22 *

dcddmy

-0.36 *

cd

-0.48 *

-0.54 *

-0.42 *

-0.67 *

io

-0.41 *

-0.21 *

-0.15 *

-0.02

dist

-1.57 *

-1.26 *

-0.86 *

-0.96 *

-1.24 *

adjncy

-0.18 **

0.17 **

-0.08

0.06 -0.60 * 0.07 -0.56 * 0.29 *

38

39

75

1.51 *

1.47 *

1.05 *

0.18

0.01

-0.21

0.23 *

0.13

-0.12

-0.26 *

-0.58 *

-2.61 *

-0.16 *

-1.12 **

-0.70 *

-0.93 *

-2.04 *

0.69 *

0.09

opipc

0.04

-0.30

-0.26 **

-1.16 *

-0.79 *

-1.01 *

opop

0.20

-0.31 *

-0.32 *

-0.45 *

-0.30 *

-0.98 *

oemp

1.23 *

-0.04

0.16 *

1.54 *

0.15 *

0.92 *

0.78 *

3.79 *

owsem

0.30 **

0.90 *

0.88 *

0.81 *

0.41 *

1.42 *

0.57 *

3.46 *

ovlad

-0.21 **

0.86 *

0.49 *

-0.36 *

0.59 *

0.07

0.26 *

-0.36 **

oaps

-0.11

0.25

0.07

-1.03 *

-0.80 *

-0.85 *

0.04

-1.35 *

0.05

1.48 *

-0.13

1.59 -3.10 *

dmnem

0.49 *

0.24 *

0.09 *

0.05

0.16 *

-0.11 **

dwsem

0.29 *

0.35 *

0.47 *

0.56 *

0.18 **

0.49 *

0.32 *

0.64

0.37 *

0.26 **

1.52 *

0.76 *

0.98 *

1.05 *

4.82 * 3.22 *

dpipc

*

0.11

0.04

37

-0.45

dpop

0.57 *

0.43 *

0.33 *

0.93 *

0.42 *

0.58 *

0.82 *

Lambda

0.07 *

0.10 *

0.11 *

0.07 *

0.13 *

0.08 *

0.07 *

Theta

0.18 *

0.21 *

0.20 *

0.18 *

0.14 *

0.16 *

0.19 *

Significant at 5 % level

**

Significant at 10 % level

92

-0.03 0.13 *

Table 5. 24. The 1997 Model Variable Coefficients and Their Significance Levels Across Commodity Groups (20-32)

1997 adjncy ocddmy dcddmy

20 0.11 -1.99 * 0.17

24

25

2.12 *

0.84 *

26 0.52 **

28

29

30

32

0.97

28.83 *

0.88 *

1.39 *

0.13

-0.06

0.29

-0.16

0.36

0.04

-0.46 *

-0.04

0.15

0.16

0.15

0.32

0.11

-0.03

cd

-1.11

-1.81 *

-0.49 *

-0.40 *

-0.17

io

-1.35

-1.51 *

-0.26 *

-0.23 **

-0.40

dist

-3.38 *

-2.35 *

-1.03 *

-1.33 *

-2.08

-6.64 *

-0.72 *

4.53 *

-0.19

-0.38

1.85

-11.06 *

-0.55 *

0.14

0.16

-0.49 *

-0.09

2.03

-2.15

-0.15 *

-0.12

opipc

2.43

opop

-0.71

oemp

0.84

owsem

1.84

ovlad

0.66

oaps

-0.43

1.64 * -0.13 1.19 * -0.14

0.69 *

0.70 *

1.00 *

-1.11 ** 0.33

-0.15 * 0.04

0.17

4.53 *

-0.15 *

0.21

-0.49

3.13 **

0.41 *

0.33 *

0.28 *

-0.34

0.47

0.59 *

0.52 *

-0.76 *

0.68

-5.69 *

-0.11

-0.55 * 0.17 -1.37 *

0.89 * -0.14 0.45 * -0.86 *

dmnem

0.49

0.81 *

-0.17 *

0.23 *

-0.14

0.48

0.14 *

0.19 **

dwsem

1.09

1.27 *

0.53 *

0.42 *

0.83

0.86

0.28 *

0.55 *

dpipc

2.27

2.99 *

1.24 *

0.58 **

0.77

-1.79

0.01

0.71 *

dpop

0.44

1.00

0.69 *

0.18

1.15

0.72

-0.01

Lambda

0.09

0.01

0.07 *

0.12 *

0.05

0.16 *

0.18 *

0.10 *

0.12 *

0.16 *

0.13 *

-0.01

-0.39 *

0.20 *

0.09 *

Theta *

Significant at 5 % level

-0.13 *

**

Significant at 10 % level

93

0.24

Table 5. 25. The 1997 Model Variable Coefficients and Their Significance Levels Across Commodity Groups (33-75)

1997 adjncy

0.42 **

34

35

36

37

38

39

75

1.00 *

0.94 *

0.86 *

0.28

0.90 *

1.23 *

0.55 *

0.27 *

-0.11

0.01

0.10

-0.01

-0.14

-0.04

0.11

0.08

0.14

-0.51 *

ocddmy

-0.27 **

-0.02

-0.25 *

dcddmy

-0.23 **

0.12

0.26 *

cd

-0.45 *

-0.71 *

-0.59 *

-0.26 *

io

-0.32 *

-0.03

-0.30 *

-0.20 *

dist

-1.48 *

-1.43 *

-1.33 *

-0.65 *

-1.22 *

-0.74 *

opipc

-0.66 *

-1.37 *

-1.18 *

-0.05

-0.19

-0.51

opop

-0.27 **

-0.67 *

-0.29 **

-0.45 *

-0.49 *

-1.83 *

-1.00 *

0.89 *

oemp

0.67 *

0.81 *

0.21

0.93 *

2.16 *

1.09 *

1.48 *

owsem

0.91 *

1.00 *

0.88 *

0.83 *

0.76 *

2.64 *

1.71 *

0.56 **

ovlad

0.23 *

0.26

1.07 *

-0.23 *

0.97 *

-0.51 *

0.11

-0.01

oaps

-0.53 *

-0.98 *

0.17

-0.52 *

-1.15 *

-1.21 *

-0.10

0.15

dmnem

0.54 *

0.25 *

0.19 *

0.01

0.07

dwsem

0.30 *

0.51 *

0.53 *

0.39 *

0.37 *

0.71 *

dpipc

*

33

-0.34

0.09

-0.27

0.06

-0.13

-0.39 * -0.48 *

-0.16

-1.16 * 0.22 -1.43 * 0.54

-1.07 * -1.44 * -1.40 * 4.33 *

0.21 **

0.58 *

0.56 *

0.99 *

0.55 *

0.05

1.46 *

0.84 *

1.50 *

dpop

0.25 **

0.57 *

0.69 *

0.32 *

0.32 **

1.51 *

1.55 *

1.47 *

Lambda

0.09 *

0.08 *

0.07 *

0.12 *

0.14 *

0.02

0.02

0.02

Theta

0.14 *

0.19 *

0.19 *

0.16 *

0.11 *

0.15 *

0.17 *

0.17 *

Significant at 5 % level

**

Significant at 10 % level

94

5. 3. Elasticity Analysis In a log-linear specification of an econometric equation, the estimated parameter of variable Xi represents the elasticity of the dependent variable Y with respect to

Xi.

However, in the case of a Box-Cox equation, the determination of the elasticity of Y with respect to X i requires some calculations. Mathematically, the elasticity is defined as

εXi =

∂Y X i ∂X i Y

(5.1)

As the Box-Cox equation is defined as

X λ −1 X λ −1 Yθ −1 = a 0 + a1 X 1 + a 2 2 + .... + a n n θ λ λ

,

(5.2)

the elasticity becomes

εX i = ai

X iλλ Y θθ

(5.3)

Note that when ë -> 0 and è -> 0, and Equation (5.2) becomes the log-linear specification, the elasticity becomes ai,

as expected. It is possible to estimate these

elasticities at the sample means. However, in this study, this would be complicated by the existence of three dummy variables, for which there are no sample means. In order to assess the range of variation of the elasticities, these were computed for all the sample observations, where the dependent variable (the commodity flow) values are strictly positive. This computation was carried out for both 1993 and 1997.

5. 3. 1. Competing Destinations The elasticities for the variable cd are presented

in Table 5.26. The mean

elasticities of both years are close for each commodity separately (e.g., -1.60 in 1993 and –1.59 in 1997 for commodity 24). In 1993, the mean elasticity varies across commodities from –0.45 (precision instruments) to –1.63 (transportation equipment). Food and kindred products have the highest negative elasticity in 1997 (-5.42), but the lowest elasticity in 1997 is for precision instruments ( -0.32). In general, it appears that the commodity groups with bulky products have higher elasticities: lumber and wood products (–1.60 and –1.59); clay, concrete, glass (–1.48 and –1.50); transportation equipment (–1.63 and – 1.65).

Table 5. 26. Statistics for CD Elasticities

Mean Elasticity CD

Min. Elasticity

Max. Elasticity

Group Name

93

97

93

97

93

97

20 Food & Kindred

-1.00

-5.42

-6.07

-33.80

-0.45

-0.76

24 Lumber & Wood

-1.60

-1.59

-10.28

-6.19

-0.58

-0.69

25 Furniture

-1.49

-0.73

-4.97

-2.40

-0.83

-0.41

-1.05

-1.02

-41.52

-6.07

-6.45

-0.66

-0.48

-0.33

-26.78

-0.37

-0.22

-0.26

29 Coal & Petroleum

-0.71

-0.35

-1.07

- 0.53

-0.32

-0.17

30 Rubber

-0.86

-0.66

-9.17

-2.10

-0.42

-0.34

32 Clay, Concrete

-1.48

-1.50

-7.84

-3.13

-0.74

-0.98

33 Primary Metals

-0.67

-0.78

-3.39

-2.71

-0.31

-0.46

34 Fabricated Metals

-0.91

-0.92

-6.75

-3.50

-0.42

-0.49

35 Machinery

-0.86

-0.61

-19.54

-4.22

-0.36

-0.34

36 Electrical Mach.

-0.82

-0.55

-6.65

-1.54

-0.39

-0.29

37 Transportation

-1.63

-1.65

-6.17

-4.99

-0.82

-1.00

-0.45

-0.32

-2.87

-1.16

-0.21

-0.18

-0.86

-0.72

-8.60

-5.29

-0.39

-0.38

-1.35

-0.74

-7.45

-2.15

-0.63

-0.40

26 Paper & Pulp 28 Chemical

38 Precision Inst. 39 Miscellaneous 75 Textile & Leather

5. 3. 2. Intervening Opportunities Table 5. 27 presents the elasticities for the variable io. The mean elasticity in 1993 varies between 0.86 (for transportation equipment) and –1.10 (for lumber and wood products), and, in 1997, between 0.46 (for clay, concrete, and glass products) and –6.48 (for

food products). Although most of the

elasticities are negative, there are some

positive elasticities, such as for transportation equipment (0.86), clay, concrete and stone products (0.50), and precision equipment products (0.10),

(0.16), in 1993, and coal and petroleum

rubber and plastic products (0.16); clay, concrete and glass products

(0.46), transportation equipment (0.19), and miscellaneous product (0.14), in 1997. These positive elasticities result from positive regression coefficients, and may point to agglomeration effects at the supply level.

96

Table 5. 27. Statistics for IO Elasticities

IO

Mean Elasticity Group Name

Min. Elasticity

Max. Elasticity

93

97

93

97

93

97

20 Food & Kindred

-0.78

-6.48

-4.38

-40.19

-0.34

-1.13

24 Lumber & Wood

-1.10

-1.32

-7.21

-5.19

-0.39

-0.57

25 Furniture

-0.77

-0.38

-2.52

-1.15

-0.42

-0.21

-0.89

-0.58

-28.43

-2.78

-0.42

-0.34

-0.41

-0.78

-21.70

-0.89

-0.17

-0.60

29 Coal & Petroleum

-0.29

0.10

-2.63

0.06

-1.28

0.13

30 Rubber

-0.19

0.16

-1.94

0.08

-0.09

0.53

32 Clay, Concrete

0.50

0.46

0.25

0.28

2.65

0.95

33 Primary Metals

-0.58

-0.56

-2.64

-2.01

-0.25

-0.33

34 Fabricated Metals

-0.37

-0.04

-2.63

-0.14

-0.16

-0.02

35 Machinery

-0.29

-0.31

-7.41

-2.08

-0.12

-0.17

36 Electrical Mach.

-0.03

-0.44

-0.23

-1.13

-0.01

-0.22

0.86

0.19

0.44

0.11

3.11

0.43

26 Paper & Pulp 28 Chemical

37 Transportation 38 Precision Inst. 39 Miscellaneous

0.16

-0.40

0.08

-1.46

0.96

-0.22

-0.24

0.14

-2.42

0.07

-0.11

1.00

75 Textile & Leather

-0.58

-1.00

-3.22

-2.87

-0.27

-0.53

5. 3. 3. Distance As can be seen in Table 5.28, the mean elasticities of the variable dist are always negative, varying in 1993 between –1.88 (clay, concrete and glass product) and –0.81 (precision instruments). In 1997, food and kindred products has the highest mean elasticity (–9.26) and

chemical products the second highest

(–3.04).

However, since

these product groups have lost many observations due to the commodity classification conversion, it is possible that these values represent overestimates. Except for these two sectors, the mean elasticity varies between –2.20 (clay, concrete and glass product) and – 0.53 (precision instruments) in 1997. As can be expected, bulky products have the highest elasticities: food products (–1.79 and –9.26); lumber and wood products

(–1.77 and –

1.98); furniture and fixture products (–1.75 and –1.07); pulp, paper, and allied products (– 1.79 and –1.70); transportation equipment (–1.85 and –1.83) in 1993 and 1997 respectively. These high values, of course reflect the impact of transportation costs.

97

Table 5. 28. Statistics for DIST Elasticities

Mean Elasticity DIST

Min. Elasticity

Max. Elasticity

Group Name

93

97

93

97

93

97

20 Food & Kindred

-1.73

-9.26

-11.92

-53.83

-0.63

-1.93

24 Lumber & Wood

-1.77

-1.98

-11.71

-7.87

-0.62

-0.85

25 Furniture

-1.75

-1.07

-5.96

-3.61

-0.85

-0.53

-1.79

-1.70

-61.99

-11.39

3.25

-0.82

-1.54

-3.04

-109.46

-3.40

-0.45

-2.66

29 Coal & Petroleum

-1.09

-0.86

-2.63

-1.09

-1.28

-0.60

30 Rubber

-1.31

-1.19

-15.44

-5.54

-0.44

-0.38

32 Clay, Concrete

-1.88

-2.20

-11.72

-5.67

-0.76

-1.17

33 Primary Metals

-1.52

-1.58

-8.23

-6.18

-0.54

-0.75

34 Fabricated Metals

-1.33

-1.24

-11.91

-5.11

-0.44

-0.53

35 Machinery

-1.00

-0.98

-26.88

-7.26

-0.34

-0.46

36 Electrical Mach.

-0.83

-0.74

-6.96

-2.66

-0.33

-0.31

37 Transportation

-1.85

-1.83

-8.17

-6.12

-0.65

-0.95

-0.81

-0.53

-5.40

-2.00

-0.31

-0.28

-0.97

-0.82

-11.59

-6.33

-0.35

-0.40

-1.20

-0.87

-6.29

-2.58

-0.62

-0.43

26 Paper & Pulp 28 Chemical

38 Precision Inst. 39 Miscellaneous 75 Textile & Leather

5. 3. 4. Origin Personal Income per Capita The variable opipc is expected to have a negative sign, yet there are a few cases of positive parameter estimates. As we have no clear explanation for this result, the elasticities have not been computed in these cases. As can be seen in Table 5. 29, in 1993 the mean elasticities vary between –0.09 (food products) and –2.44 (coal and petroleum products), and, in 1997, between –2.50 (coal and petroleum products) and –0.08 (electrical

machinery

products).

Other

high

mean

elasticities

in

1993

are

–1.78

(transportation equipments), –1.50 (precision instruments), -1.24 (electrical machinery products), and –1.16 (rubber and plastic products). In 1997, the high

mean elasticity

commodity groups are rubber and plastic products (–1.63), fabricated metal products (– 1.54), and machinery products (-1.08).

98

Table 5. 29. Statistics for OPIPC Elasticities

Mean Elasticity OPIPC

Min. Elasticity

Max. Elasticity

Group Name

93

97

93

97

93

97

20 Food & Kindred

-0.09

***

-0.56

***

-0.56

***

24 Lumber & Wood

***

***

***

***

***

***

25 Furniture

***

-0.25

***

-0.80

***

-0.14

-0.56

-0.74

-20.82

-4.29

1.09

-0.45

***

***

***

***

***

***

29 Coal & Petroleum

-2.44

-2.50

-2.99

-3.12

-1.56

-1.75

30 Rubber

-1.16

-1.63

-14.23

-6.19

-0.51

-0.85

32 Clay, Concrete

***

***

***

***

***

***

33 Primary Metals

26 Paper & Pulp 28 Chemical

***

-0.95

***

-3.54

***

-0.53

34 Fabricated Metals

-0.42

-1.54

-3.62

-5.92

-0.18

-0.79

35 Machinery

-0.42

-1.08

-12.11

-7.86

-0.19

-0.57

36 Electrical Mach.

-1.24

-0.08

-11.50

-0.25

-0.58

-0.04

37 Transportation

-1.78

-0.47

-7.91

-1.33

-0.86

-0.29

-1.50

-0.40

-10.75

-1.48

-0.69

-0.22

***

***

***

***

***

***

***

***

***

***

***

***

38 Precision Inst. 39 Miscellaneous 75 Textile & Leather

5. 3. 5. Origin Population Table 5.30 presents elasticities for the variable opop. As for the variable opipc, we have not computed elasticities when the estimated parameter is positive. The mean elasticities, in 1993, vary between –2.24 (precision instruments) and –0.25 (miscellaneous products). In 1997, the highest mean elasticity is for food products (-4.51), and the lowest for machinery products (–0.27). Some other high mean elasticities in 1993 are: –1.47 (textile

and

leather

products),

–1.35

(transportation

equipment),

-1.36

(coal

and

petroleum products), -0.94 (rubber and plastic products, and machinery products), and –0.86 (food products). In 1997, transportation equipment had the second highest mean elasticity (–2.46), followed by precision instruments (-1.62), coal and petroleum products (-1.49), electrical machinery products (–1.47), fabricated metal products (–1.13), and rubber and plastic products (-1.14)..

99

Table 5. 30. Statistics for OPOP Elasticities

Mean Elasticity OPOP

Min. Elasticity

Max. Elasticity

Group Name

93

97

93

97

93

97

20 Food & Kindred

-0.86

-4.51

-4.88

-27.52

-0.39

-0.80

24 Lumber & Wood

***

***

***

***

***

***

25 Furniture

***

-0.90

***

-2.73

***

-0.50

26 Paper & Pulp 28 Chemical

***

-0.31

***

-1.61

***

-0.19

-0.30

***

-15.03

***

-0.12

***

29 Coal & Petroleum

-1.36

-1.49

-2.30

-1.65

-0.78

-0.75

30 Rubber

-0.94

-1.14

-9.89

-4.38

-0.43

-0.60

32 Clay, Concrete

-0.70

-0.43

-3.80

-0.89

-0.35

-0.27

33 Primary Metals

***

-0.62

***

-2.25

***

-0.36

34 Fabricated Metals

-0.73

-1.13

-5.32

-3.72

-0.32

-0.61

35 Machinery

-0.95

-0.37

-25.00

-2.53

-0.44

-0.20

36 Electrical Mach.

-0.69

-1.47

-5.76

-4.06

-0.33

-0.79

37 Transportation

-1.35

-2.46

-4.90

-5.86

-0.68

-1.57

-2.24

-1.62

-13.91

-5.95

-1.06

-0.91

-0.25

-0.65

-2.49

-4.87

-0.11

-0.34

-1.47

***

-8.15

***

-0.61

***

38 Precision Inst. 39 Miscellaneous 75 Textile & Leather

5. 3. 6. Origin Employment Table 5. 31 presents the elasticities of the variable oemp. This variable is expected to have a positive parameter estimate and thus positive elasticities. The elasticities have not been computed for negative parameters. Textile, apparel and leather products have the highest mean elasticity in 1993 (2.08), which is not surprising since this product is labor intensive. In 1997, the highest elasticity applies to food products (3.47). The lowest mean elasticities

in

both years are for commodity group 35, machinery products (0.27

in 1993 and 0.20 in 1997), which

can be

explained by the capital intensiveness of the

sector. Commodity group 36, electrical machinery, is another high-labor elasticity group (1.65 in 1993 and 1.62 in 1997), followed, in 1993, by furniture and fixture products (group 25) with 1.62,

clay concrete and glass products (group 32) with 1.51,

primary

metal products (group 33) with 1.40, and precision instruments, (group 38) with 1.31. Other high-elasticity sectors in 1997 are group 32 (1.84), (1.62), group 24 (1.41), and group 26 (1.30). 100

group 38 (1.67),

group 36

Table 5. 31. Statistics for OEMP Elasticities

Mean Elasticity OEMP

Min. Elasticity

Max. Elasticity

Group Name

93

97

93

97

93

97

20 Food & Kindred

0.86

3.47

0.40

0.55

5.10

20.71

24 Lumber & Wood

***

1.41

***

0.62

***

5.53

1.62

0.85

0.91

0.49

5.05

2.46

0.72

1.30

0.36

0.75

18.35

5.07

0.70

0.30

0.30

0.21

31.80

0.33

29 Coal & Petroleum

1.36

0.68

0.64

0.36

2.62

1.62

30 Rubber

0.19

***

0.10

***

1.61

****

32 Clay, Concrete

1.51

1.84

0.77

1.17

8.16

3.75

33 Primary Metals

1.40

0.92

.069

0.53

6.19

3.19

25 Furniture 26 Paper & Pulp 28 Chemical

34 Fabricated Metals

***

0.91

***

0.47

***

2.97

35 Machinery

0.27

0.20

0.12

0.11

6.18

1.23

36 Electrical Mach.

1.65

1.62

0.80

0.86

11.36

3.81

37 Transportation

0.34

***

0.16

***

1.01

***

1.31

1.67

0.65

0.94

6.37

6.04

0.93

0.65

0.42

0.33

9.45

4.81

2.08

0.99

0.96

0.52

11.82

2.75

38 Precision Inst. 39 Miscellaneous 75 Textile & Leather

5. 3. 7. Origin Value-Added The elasticities of the other sectoral production variable, ovlad, are presented in Table 5. 32. Like oemp, this variable is expected to have a positive sign, and thus positive elasticities. They

have not been computed in the cases of negative estimates. Commodity

group 24 (lumber and wood products) has the highest mean elasticity (1.88) in 1993, and commodity group 38 (precision instruments) has the lowest (0.09). The other high mean elasticities correspond to

transportation equipment (0.95), fabricated metal products

(0.91), food products (0.88), and rubber and plastic products (0.81). In 1997, food products have the highest mean elasticity (2.23), and the lowest one is very close to zero for textile and leather products. Transportation equipment (1.80),

rubber and plastic

products (1.06), lumber and wood products (1.00), machinery products (0.83), and clay, concrete and glass products (0.74) are the other high-elasticity product groups.

101

Table 5. 32. Statistics for OVLAD Elast icities

Mean Elasticity OVLAD

Min. Elasticity

Max. Elasticity

Group Name

93

97

93

97

93

97

20 Food & Kindred

0.88

2.23

0.40

0.33

5.01

13.70

24 Lumber & Wood

1.88

1.00

0.70

0.44

12.16

3.90

***

0.34

***

0.20

***

0.97

0.53

0.40

0.00

0.24

3.11

1.50

0.75

***

0.33

***

31.57

***

25 Furniture 26 Paper & Pulp 28 Chemical 29 Coal & Petroleum

***

0.06

***

0.03

***

0.10

30 Rubber

0.81

1.06

0.41

0.50

6.47

3.46

32 Clay, Concrete

0.47

0.74

0.24

0.48

2.43

1.53

33 Primary Metals

***

0.26

***

0.15

***

0.89

34 Fabricated Metals

0.91

0.24

0.42

0.13

5.84

0.80

35 Machinery

0.59

0.83

0.26

0.47

11.69

4.92

***

***

***

***

***

***

0.95

1.80

0.47

1.12

2.64

3.12

0.09

***

0.04

***

0.45

***

0.25

0.06

0.11

0.03

2.52

0.46

***

0.00

***

-0.01

***

0.00

36 Electrical Mach. 37 Transportation 38 Precision Inst. 39 Miscellaneous 75 Textile & Leather

5. 3. 8. Origin Wholesale Employment The variable owsem represents redistributive activities and is expected to have positive parameter and elasticity estimates. Elasticities have not been computed for the few cases (sector 24, 28, and 32 in 1997) where the estimate is negative. The elasticities are presented in Table 5.33. The precision products sector has the highest mean elasticity in 1993 (2.41), and pulp, paper and allied products the lowest one (0.25). The other highelasticity

commodity

groups

in

1993

include

textile

and

leather

products

(1.81),

machinery products (1.71), fabricated metal products (1.47), rubber and plastic products (1.49),

transportation equipments (1.15), and food products (1.00). In 1997, food

products have the highest mean elasticity (8.54); however, this value is possibly an overestimate. Beside this value, the transportation equipment sector has a high mean elasticity (2.33), and textile and leather products the lowest one

(0.38). The other high-

elasticity sectors in 1997 are precision instruments (2.15), food products (2.09), electrical machinery products (1.74), rubber and plastic products (1.62), primary metal products (1.54),furniture, fixture products (1.45), and fabricated metal products (1.27). 102

Table 5. 33. Statistics for OWSEM Elasticities

Mean Elasticity OWSEM

Min. Elasticity

Max. Elasticity

Group Name

93

97

93

97

93

97

20 Food & Kindred

1.02

8.54

0.46

1.46

5.87

51.75

24 Lumber & Wood

0.36

***

0.13

***

2.38

***

25 Furniture

0.88

1.45

0.48

0.81

2.84

4.38

0.25

0.49

0.12

0.29

8.46

2.44

0.44

****

0.18

***

22.38

***

29 Coal & Petroleum

1.19

0.91

0.67

0.62

2.02

1.33

30 Rubber

1.49

1.62

0.67

0.82

15.58

6.15

32 Clay, Concrete

0.13

***

0.06

***

0.68

***

33 Primary Metals

0.40

1.54

0.17

0.88

1.86

5.48

34 Fabricated Metals

1.47

1.27

0.64

0.70

10.99

4.15

35 Machinery

1.71

0.89

0.81

0.48

43.44

5.98

36 Electrical Mach.

0.96

1.74

0.46

0.94

8.01

4.70

37 Transportation

1.15

2.33

0.58

1.47

4.13

5.40

2.41

2.15

1.16

1.22

15.24

7.85

0.82

1.06

0.39

0.55

8.09

7.89

1.81

0.38

0.85

0.20

9.95

1.11

26 Paper & Pulp 28 Chemical

38 Precision Inst. 39 Miscellaneous 75 Textile & Leather

5. 3. 9. Origin Average Plant Size The estimated elasticities for the variable oaps are presented in Table 5. 34. While there was no initial expectations concerning the sign of this variable, they turn out to be, in general, negative, and thus so are the elasticities In 1993, negative elasticities vary between –1.04 (coal and petroleum products) and –0.01 (rubber and plastic products), and

positive elasticities vary between 0.03 (miscellaneous products) and 0.43 (furniture

and fixtures). Other high negative elasticities

in 1993 are –0.92 (transportation

equipment), -0.86 (textile and leather products), –0.80 (precision instruments), and -0.76 (electrical machinery products). In 1997,

negative elasticities vary between –1.34 (again

transportation equipment) and –0.05 (miscellaneous products), and positive elasticities vary between 0.09 (textile and leather products), and 0.89 (chemical products). Other high negative elasticities in 1997 are –1.03 (food products), -1.02 (clay, concrete and glass

103

Table 5. 34. Statist ics for OAPS Elasticities

Mean Elasticity OAPS

Min. Elasticity

Max. Elasticity

Group Name

93

97

93

97

93

97

20 Food & Kindred

-0.57

-1.03

-3.59

-6.22

-0.24

-0.16

24 Lumber & Wood

0.25

-0.12

0.09

-0.47

1.67

-0.05

25 Furniture

0.43

0.42

0.24

0.24

1.43

1.31

-0.14

-0.80

-1.03

-4.12

-0.06

-0.48

-0.62

0.89

-37.48

0.71

-0.25

1.02

29 Coal & Petroleum

-1.04

-0.47

-1.56

-0.70

-0.60

-0.28

30 Rubber

-0.01

-0.11

-0.09

-0.41

0.00

-0.06

32 Clay, Concrete

-0.70

-1.02

-4.02

-2.57

-0.34

-0.64

33 Primary Metals

26 Paper & Pulp 28 Chemical

-0.09

-0.48

-0.43

-1.74

-0.03

-0.27

34 Fabricated Metals

0.19

-0.65

0.08

-2.52

1.60

-0.33

35 Machinery

0.05

0.10

-0.02

-0.06

1.44

0.75

36 Electrical Mach.

-0.76

-0.51

-6.08

-1.40

-0.34

-0.27

37 Transportation

-0.92

-1.34

-3.48

-2.89

-0.43

-0.87

-0.80

-0.81

-5.04

-3.06

-0.36

-0.45

0.03

-0.05

0.01

0.40

0.33

-0.03

-0.86

0.09

-4.59

0.05

-0.43

0.26

38 Precision Inst. 39 Miscellaneous 75 Textile & Leather

products), – 0.81 (precision instruments),

-0.80 (paper and allied product), -0.65

(fabricated metal products), and –0.48 (primary metal products). 5. 3. 10. Destination Manufacturing Employment The variable dmnem is a proxy for the effects of intermediate demand at the destination, and the expectation about its sign is positive. The elasticities are presented in Table 5. 35. Elasticities have not been computed in

a few cases (sectors 25 and 38 in

1993, 25, 28 and 38 in 1997) when the estimates are negative. The highest mean elasticity in 1993 is 0.89 (lumber and wood products), and the lowest 0.07 for electrical machinery. Other high-elasticity sectors in 1993 include textile and leather products (0.76), clay, concrete and glass products (0.72), paper and allied product (0.68), and primary metal products (0.71). The highest mean elasticity in 1997 is 0.99 (primary metal products), and the lowest, as in 1993, is 0.02 for electrical machinery. Lumber and wood products (0.71), rubber and plastic products (0.67), paper and allied products (0.63), and clay, concrete, and glass products (0.55) are the other high-elasticity sectors in 1997. 104

Table 5. 35. Statistics for DMNEM Elasticities

Mean Elasticity DMNEM

Min. Elasticity

Max. Elasticity

Group Name

93

97

93

97

93

97

20 Food & Kindred

0.24

2.54

0.10

0.34

1.44

16.22

24 Lumber & Wood

0.89

0.71

0.32

0.31

5.63

2.78

***

***

***

***

***

***

0.68

0.63

0.35

0.40

4.05

3.36

0.23

***

0.10

***

13.81

***

29 Coal & Petroleum

0.07

0.17

0.03

0.07

7.56

6.89

30 Rubber

0.53

0.67

0.25

0.31

5.46

2.27

32 Clay, Concrete

0.72

0.55

0.35

0.36

3.87

1.09

33 Primary Metals

0.71

0.99

0.33

0.58

3.57

3.38

34 Fabricated Metals

0.42

0.34

0.19

0.18

3.27

1.26

35 Machinery

0.20

0.15

0.08

0.11

4.02

1.36

36 Electrical Mach.

0.07

0.02

0.03

0.01

0.50

0.07

37 Transportation

0.50

0.26

0.23

0.14

1.82

0.73

25 Furniture 26 Paper & Pulp 28 Chemical

38 Precision Inst. 39 Miscellaneous

***

***

***

***

***

***

0.08

0.13

0.04

0.07

0.84

0.96

75 Textile & Leather

0.76

0.41

0.36

0.22

4.12

1.19

5. 3. 11. Destination Wholesale Employment Table 5.36 presents the estimated elasticities for the variable dwsem. This variable is a proxy for the effects of redistributive activities at the destination, and it is presumed to have a positive parameter estimate. This is the case of all sectors in both years. Chemical products have the highest mean elasticities in both years: 1.46 in 1993, 1.59 in 1997 (except food products in 1997 with 5.14, possibly another overestimate). Coal and petroleum products have the lowest mean elasticity in 1993 (0.26), and textile and leather products the lowest in 1997 (0.38). The high-elasticity sectors in 1993 include paper and allied products (1.13), food products (1.16), furniture and fixture products (0.92), and machinery products (0.91). In 1997, the high elasticity sectors include clay, concrete and glass products (1.44), transportation equipment (1.12), lumber and wood products (1.10), rubber and plastic products (1.10), and paper and allied products (1.02).

105

Table 5. 36. Statistics for DWSEM Elasticities

Mean Elasticity DWSEM

Min. Elasticity

Max. Elasticity

Group Name

93

97

93

97

93

97

20 Food & Kindred

1.16

5.14

0.47

0.74

6.85

31.93

24 Lumber & Wood

0.65

1.10

0.24

0.48

4.18

4.32

25 Furniture

0.92

0.77

0.50

0.42

3.02

2.45

1.13

1.02

0.59

0.65

7.10

5.76

1.46

1.59

0.62

1.29

88.96

1.78

29 Coal &Petroleum

0.44

0.26

0.19

0.13

7.30

3.71

30 Rubber

0.83

1.10

0.40

0.59

8.77

3.31

32 Clay, Concrete

0.65

1.44

0.32

0.97

3.44

3.16

33 Primary Metals

0.39

0.51

0.18

0.30

1.95

1.71

34 Fabricated Metals

0.57

0.64

0.26

0.35

4.38

2.46

35 Machinery

0.91

0.53

0.40

0.30

19.88

3.69

36 Electrical Mach.

0.67

0.81

0.33

0.46

5.25

2.16

37 Transportation

0.50

1.12

0.26

0.70

1.93

3.22

0.83

0.45

0.41

0.26

5.29

1.64

0.45

0.61

0.21

0.32

4.69

4.46

0.34

0.38

0.16

0.20

1.84

1.10

26 Paper & Pulp 28 Chemical

38 Precision Inst. 39 Miscellaneous 75 Textile & Leather

5.3. 12. Destination Personal Income Per Capita The variable dpipc measures the effect of final demand at the destination, and presumed

is

to have positive parameter estimates. Elasticities have not been calculated in

the few cases where the estimates are negative (sectors 28, 29, 33 in 1993, and 29, 33, 35 in 1997). The elasticities are presented in Table 5.37. The highest mean elasticity in 1993 is 2.94, for furniture and fixture products, while the lowest one is 0.43, for machinery products. The other high mean elasticities of 1993 are 2.85 (lumber and food products), 2.63 (textile and leather products), 2.09 (clay, concrete and glass products), 1.71 (transportation equipment), and 1.63 (electrical machinery products). In 1997, however, lumber and wood products have the highest mean elasticity (2.58) (except food products, 8.97, which is possibly another overestimate), while af bricated metal products have the lowest elasticity (0.10). The other high mean elasticities in 1997 correspond to furniture and fixture products (1.60), clay, concrete and glass products (1.59), paper and allied products (1.14),

and textile and leather products (1.01). Comparing the 1993 and

1997 elasticities suggests that the effect of this variable has declined from 1993 to 1997. 106

Table 5. 37. Statistics for DPIPC Elasticities

Mean Elasticity DPIPC

Min. Elasticity

Max. Elasticity

Group Name

93

97

93

97

93

97

20 Food & Kindred

1.05

8.97

0.46

-1.56

6.90

54.68

24 Lumber & Wood

2.85

2.58

1.04

1.13

19.01

10.22

25 Furniture

2.94

1.60

1.59

0.88

9.95

5.18

1.21

1.14

0.53

0.70

47.80

6.85

***

1.34

***

1.18

***

1.46

26 Paper & Pulp 28 Chemical 29 Coal & Petroleum

***

***

***

***

***

***

30 Rubber

0.74

0.04

0.33

0.02

9.16

0.16

32 Clay, Concrete

2.09

1.59

1.00

0.99

12.07

4.06

33 Primary Metals

***

***

***

***

***

***

34 Fabricated Metals

0.52

0.10

0.22

0.05

4.46

0.38

35 Machinery

0.43

***

-9.47

***

12.40

***

36 Electrical Mach.

1.63

1.21

0.76

0.65

15.17

3.85

37 Transportation

1.71

0.11

0.85

0.07

7.58

0.33

1.46

1.13

0.67

0.63

10.50

4.28

1.37

0.51

0.59

0.26

15.20

3.86

2.63

1.01

1.28

0.53

13.91

2.95

38 Precision Inst. 39 Miscellaneous 75 Textile & Leather

5. 3. 13. Destination Population The other destination final demand variable, dpop, is also presumed to have positive parameter estimates and positive elasticities. This is the case of all sectors in both years. Elasticity statistics are presented in table 5. 38. Overall it appears mean elasticity has

declined from 1993 to 1997, suggesting that the effect of the final

demand at destinations on commodity flows has made in

that the

declined, supporting the observation

the case of the variable dpipc. The highest mean elasticity in 1993 is 1.92

(transportation equipment), and the lowest is 0.15 (chemical products). The other high mean elasticities are 1.89 (furniture and fixture products), 1.56 (clay, concrete and glass product),

1.53

(miscellaneous,

and

textile

and

leather

products),

1.44

(electrical

machinery products), and 1.32 (precision instruments). In 1997, food products have the highest mean elasticity (2.85) and primary metal products the lowest (0.57). Chemical products (2.64), transportation equipment (1.62), precision instruments (1.33), furniture and fixture products (1.27), textile and leather products (1.09), and miscellaneous products (1.01) are the other high-mean elasticities in 1997. 107

Table 5. 38. Statist ics for DPOP Elasticities

Mean Elasticity DPOP

Min. Elasticity

Max. Elasticity

Group Name

93

97

93

97

93

97

20 Food & Kindred

0.72

2.85

0.32

0.42

4.07

17.85

24 Lumber & Wood

1.02

0.90

0.37

0.39

6.73

3.52

25 Furniture

1.89

1.27

1.02

0.70

6.19

4.06

0.40

0.67

0.19

0.43

13.09

3.83

0.15

2.64

0.06

2.18

7.55

2.96

29 Coal & Petroleum

0.94

0.39

0.54

0.20

6.10

5.60

30 Rubber

0.69

***

0.31

***

7.28

***

32 Clay, Concrete

1.56

0.89

0.77

0.60

8.44

1.90

33 Primary Metals

1.00

0.57

0.39

0.33

4.69

1.92

34 Fabricated Metals

1.00

0.96

0.44

0.52

7.33

3.66

35 Machinery

0.98

0.89

0.46

0.49

25.63

6.07

36 Electrical Mach.

1.44

1.02

0.70

0.59

12.05

2.72

37 Transportation

1.92

1.62

0.96

1.04

6.96

4.59

1.32

1.33

0.63

0.76

8.19

4.88

1.53

1.01

0.69

0.53

15.06

7.44

1.53

1.09

0.72

0.58

8.45

3.17

26 Paper & Pulp 28 Chemical

38 Precision Inst. 39 Miscellaneous 75 Textile & Leather

5. 3. 14. Summary The estimated mean elasticities for both years across the commodity groups have been presented in Tables 5.26 through 5.38. Although the mean elasticities of some variables display systematic increases or decreases between 1993 and 1997, for instance owsem,

dpipc, and dpop, the others do not display such a pattern. As a general

observation, it is also possible to say that the variables with insignificant t-statistics values have relatively small elasticities, as can be expected. Although in many cases the magnitudes of the elasticities are similar,

sometimes they show significant variations

between the two years. These variations can be explained either because different samples have been used (in terms of time and size), or because the industry group that displays such variations has had structural changes over the 1993-1997 period.

108

CHAPTER 6 CONCLUSIONS

The overall results of this research can be summarized as follows. (1) Although past spatial interaction models of commodity flows have generally used

gravity-type log-

linear specifications, the optimized Box-Cox specification proves to be superior to the loglinear one for all commodity groups in both years (except in one case in 1997 because of a much smaller sample). (2) While past empirical research has tried to best-fit simple functional forms with very few variables, that provide the highest fit (Reed, 1967; Huxley, 1979; Chisholm & O’Sullivan, 1973), this study, on the other hand, selects a large number of explanatory variables in a more directed process. First, it chooses the variables representing the origin’s productive capacities, the origin’s demand conditions, and destination’s demands characteristics (at both the intermediate and

the

final levels),

consistent with the input-output framework. Next, besides the variables that represent the origin and destination characteristics, distance, contiguity and

variables that represent

nodal spatial configurations are considered. Although Reed (1967) had also tried to include

similar spatial configuration variables, his study focused on only one region. Here

we apply to commodity flows the concept of

“competing destinations” (Fotheringham,

1983a & 1983b) and “intervening opportunities” (Ullman, 1967; Guldmann, 1999) used in other applications of spatial interaction modeling. The empirical results show that, as in the case of international trade flow models (Frankel & Wei, 1998), state adjacency significantly affects interstate trade flows. In other words, neighboring states trade more with one another

than non-neighboring ones, even

after accounting for distance. The dummy variable measuring the effect of adjacency is significant for 15 commodity groups out of 16 in 1993, and for 13 groups in 1997. Beyond possibly lower transportation cost, cultural similarities, and ease in business information gathering between neighboring states are possible reasons for this result. The

findings

about

distance,

as

proxy

for

transportation

cost,

confirm

the

importance of this factor: the distance coefficient is negative and highly significant for all 109

commodity

groups

in

both

1993,

and

1997,

with

bulkier products having larger

coefficients, that is, being hauled over shorter distances, while higher value-per-weight commodities are shipped over longer distances. In contrast, Reed (1967) and Huxley (1979) did not find distance a significant variable in their studies. The competing destination variable has a negative and significant effect for all commodity groups in 1993, and for 14 groups in 1997. This result implies a “competition effect” at the state destinations, that is, the spatial proximity of state destinations around the terminating state causes flows to decrease to this terminating state. The same argument is valid for the spatial

configurations of destinations around the origins as

measured by the intervening opportunities variable. In 1993, 11 out of 16 commodity groups have negative and significant intervening opportunities variables, and 8 do so in 1997, pointing to a

“competition effect” in the clusters around supply nodes. Looking at

these results, it is possible to conclude that manufacturing activities are characterized by competitive destination effects. In geographical term, when other destinations get closer to either the supply or destination state, the amount of manufactured products (almost in all two-digit sub-categories) shipped to this destination state decreases. The custom district variables, ocddmy and dcddmy, are to measure foreign trade effects on commodity flows within the continental US, but are not consistent in capturing these effects. The foreign trade flows, their ratios to US total and custom district states total out-shipments, and the significant coefficients of the dummy variables, are presented in Table 6.1 for 1993. Some of the effects of exports and imports are captured by these variables. For example, the effects of export are captured in the case of food products (group 20), rubber (group 30), primary metal (group 33), electrical machinery (group 36) and transportation equipment (group 37), which are all large exporting sectors. In 1997, these variables are less consistent in terms of capturing the effects of foreign trade, possibly due to sampling problems.

110

Table 6. 1. Performance of Custom District Variables in 1993

Cmd.

Group Name

Custom States Totsp93 Cdim93

Cdex93 (%)

(%) ocddmy

20 Food & Kindred

20789 2.4 692770

21874

24 Lumber & Wood

5282 4.2 107428

8656

25 Furniture

3627 5.2

7940 13.1

26 Paper & Pulp

8886 4.6 169767

10388

6.1 (0.27)*

35229 6.7 453635

25428

5.6

28 Chemicals 29 Coal & Petroleum

60638

7248 2.0 322124

3.2 (-0.27)** (0.21)* 8.1 (-0.20)** -

-

14651 10.2 (-0.17)**

32 Clay, Concrete

16116 17.7

18893 24.2

33 Primary Metals

12962 5.7 196260 3700 1.6 205261

(0.20)**

52423 16.3 (1.84)*

17137 9.9 144335

34 Fabricated Metals

-

-

30 Rubber

78176

dcddmy

-

-

22994 11.7 (-0.22)** (-0.36)* 5300

2.6 (-0.22)*

35 Machinery

85813 19.4 382119

85303 22.3

36 Electrical Mach.

59883 14.6 359963

74481 20.7 (-0.18)**

-

37 Transportation

74032 11.6 549739

88885 16.2 (-0.60)*

-

38 Precision Inst.

21995 11.0 181591

20117 11.1

-

(0.23)

39 Miscellaneous

3800 2.3 139141

12940

9.3

-

-

56240 14.9

-

-

75 Textile & Leather

15709 3.5 376529

-

(0.17)**

* Significant at 5 % level ** Significant at 10 % level

Totsp93: Total Shipments in 1993; Cdex93: Custom District States Exports; Cdim93: Custom District States Imports

The variables measuring the final demand effects at the origins, the origin state per capita income and the origin state population, have negative effects in most cases, as expected. This result confirms the initial expectation that, as consumption opportunities at the origin increase,

less

export shipments

take place. On the other hand, increasing

consumption opportunities at the destinations were expected to positively affect the outflows, and this is also confirmed by many positive and significant parameter estimates for destination state per capita income and destination state population. One important finding is that wholesale activities at both the origins and destinations are important facilitators of commodity flows, by buying commodities from the production sector and reselling them to the intermediate and final demand sectors. Origin state wholesale employment is significant and positive in 13 groups in 1993 and 11 in 1997, while destination state wholesale employment is significant in 14 groups in 1993 and 13 in 1997.

111

Based on their likely demand structure, it is possible to group commodities into three main groups. (1) Those that are mainly inputs to final demand sectors, (2) those that are mainly inputs to intermediate demand sectors, and (3) those

that are inputs to

both demand sectors. (1) Product group 25, furniture and fixture; product group 29, coal and petroleum; product group 36, electrical machinery; product group 38, precision instruments; and products group 39, miscellaneous manufactured products, are primarily inputs to final demand sectors. According to the empirical findings, in none of these groups is the intermediate demand proxy variable, dmnem, very significant, confirming the hypothesis. The origin local demand for product group 25 is not significant, probably because the production of this commodity is spatially concentrated, with economies scale (oaps is positive and significant in only this product group). Product group 29 however, is very sensitive to origin local demand conditions but not to destination local demand, probably because this product is highly sensitive to shipping distance. The other groups, however, are sensitive to both origin and destination final demands. (2) Product group 24, lumber and wood; product group 28, chemical products; product group 32, clay, concrete, glass and stone products, and products group 33, primary metal products are the sectors that are producing primarily for intermediate demand sectors, and the variable dmnem is very significant for

all four groups, whereas

the final demand variables are not significant. Although lumber and wood; clay, concrete, glass and stone; and primary metal products may also be assumed to be important for final demand sectors via the construction sector, the performances of the final demand variables do not support this assumption. (3) Product group 20, food and kindred products; product group 26, pulp and paper products; product group 30, rubber and plastic products; product group 34 fabricated metal products; product group 35, non-electrical machinery; product group 37, transportation equipment; and product group 75, textile, apparel and leather products, on the other hand, are demanded by and supplied to both intermediate and final demand sectors. Variables representing both final and intermediate demand sectors are highly significant in these products groups. The supply variables at the origin, state sectoral employment and state sectoral value added, are expected to positively impact out-shipments. Although there are a few cases where these variables have negative signs, they are positive and significant for most commodities, confirming initial expectations. 112

Finally, scale or diversification effects for origin state sectoral establishments have been tested using the average establishment size variable, oaps. In both years, 8 of the 16 group have significant negative signs, implying that these sectors are characterized by “diversification effects”. Only 1 group, furniture and fixture products, displays a positive and significant coefficient in both years, pointing to economies of scale effect. Overall the results confirm the validity of the approach used to select the explanatory variables and the functional form of the model. The results also point to further research into better explaining the structure of interregional commodity flows. Depending upon

data availability, new explanatory variables could be introduced into the

model, that would better represent (1) final and intermediate demand, (2) wholesale sector effects, (3) supply potential, (4) spatial structure effects, and (5) actual foreign imports and exports. The cd and io variables have been computed with total state employment

data,

but

other

formulations

are

possible,

using,

for

instance,

total

shipments, or total value added, or specific sectoral employments. While this research shows that the Box-Cox specification is superior to the historically-preferred log-linear specifications, further research could focus on alternate expanded specification, such as the quadratic or translog ones. Finally research could be conducted towards a better understanding of the relationship between commodity flows and modes of transportation. The modal data available in the CFS should be useful for such research.

113

BIBLIOGRAPHY Ashtakala, B. and Murthy, A. S. N., 1988, “Optimized Gravity Models for Commodity Transportation”, Journal of Transportation Engineering, Vol. 114, 4: 393-408. Black, W. R., 1971, “The Utility of the Gravity Model and Estimates of its Parameters in Commodity Flow Studies, Proceedings of the Association of American Geographers, V.3, 132-143. Black, W. R., 1972, “lnterregional Commodity Flows: Some Experiments with the Gravity Model”, Journal of Regional Science, Vol. 12, 1: 5-21. Bon, R., 1984, “Comparative Stability Analysis of Multiregional Input-output Models. Column, Row, and Leontief-Strout Gravity Coefficient Models”, The Quarterly Journal of Economics, Vol. 99, 4: 212-235.

Box, G. E. P. and Cox, D. R., 1964, “An Analysis of Transformations”, Journal of the Royal Statistics Society, 26, 3:211-243. Brocker, J., 1989, “Partial Equilibrium Theory of lnterregional Trade and The Gravity Model”, U Papers of the Regional Science Association, V. 66, 7-18. Chisholm, M., and O’Sullivan, P., 1973, Freight Flows and Spatial Aspects of the British Economy , Cambridge University Press, New York and London.

Eichengreen, B., and Irwin, D. A., 1998, “The role of History in Bilateral Trade Flows” ”, in the Regionalization of the World Economy, ed. Frankel, J. A., The University of Chicago Press, Chicago. Frankel, J. A., and Wei, S. J., 1998, “Regionalization of World Trade and Currencies: Economics and Politics”, in The Regionalization of the World Economy, ed. Frankel, J. A., The University of Chicago Press, Chicago. Frankel, J. A., and Romer, D., 1999, “Does Trade Cause Growth?”, The American Economic Review, V. 89, 3: 155-169. 114

Fotheringham, A. S., and O’Kelly M. E., 1989, Spatial Interaction Models: Formulation and Applications, Kluwer Academic Publishers, Dordrecht/Boston/London. Fotheringham, A. S., 1983a, “A New Set of Spatial-Interaction Models: the Theory of Competing Destinations”, Environment and Planning A, V. 15, 15-36. Fotheringham, A. S., 1983b, “Some Theoretical Aspects of Destination Choice and Their Relevance to Production-Constrained Gravity Models”, Environment and Planning A, V. 15, 1121 -1132. Fotheringham, A. S., and Pitts, T. C., 1995, “Directional Variation in Distance Decay”, Environment and Planning A, V. 27, 715-729. Green, W. H., 1997, Econometric Analysis, Third Edition, Prentice Hall, N. Jersey. Guldmann, J M., 1999, “Competing Destinations and Intervening Opportunities Interaction Models of Inter-City Telecommunication Flows”, Paper in Regional Science, Vol. 78, 179194. Hua, C., 1990, “A Flexible and Consistent System for Modeling lnterregional Trade Flows”, Environment and Planning A, Vol. 22, 98-121. Huxley, S. J.,1979, “Indirect Estimation of lnterregional Trading Patterns for Input-Output Analysis: Empirical Results for the Gravity Model and Rail Freight Shipments”, Presented to the 26th North American Meeting of the Regional Science Association, Los Angles. Isard, W., 1951, “lnterregional arid Regional Input-Output Analysis: A Model of a SpaceEconomy, The Review of Economics and Statistics, Vol. 33, 4: 157-169. Ishikawa, Y., 1987, “An Empirical Study of the Competing Destinations Model Using Japanese Interaction Data”, Environment and Planning A, V. 19, 1359-1373. Krugman, P., 1980, “Scale Economies, Product Differentiation, and the Pattern of Trade”, American Economic Review, V.70, 5:950-959. Leontief, W. and Strout, A., 1963, “Multiregional Input-Output Analysis”, Structural Dependence and Economic Development, ed. by Barna Tibor, St Martin Press, New York. Miernick, W. H., 1965, The Elements of Input-Output Analysis, Random House.

115

Moses, L. N., 1955, “The Stability of Interregional Trading Patterns and lnput-Output Analysis”, The American Economic Review, December 1955. Reed, W. E., 1967, Areal Interaction in India, Duke University Research Papers, Chicago, Illinois. Richardson, H. W., 1972, Input Output and Regional Economics, Wiley, New York. Samuelson, P. A., 1952, “Spatial Price Equilibrium arid Linear Programming”, American Economic Review, V.42. 67-93. Takayama, T. and Judge, G. G., 1964, “Spatial Equilibrium and Quadratic Programming”, Journal of Farm Economics, V.46:1, p. 67-93. Ullman, E. L., 1967, American Commodity Flow, University of Washington Press, Seattle and London.

116

APPENDIX A STANDARD TRANSPORTATION COMMODITY CLASSIFICATION (STCC)∗∗ Code 00 01 011 012 013 014 015 019 08 084 086 09 091 098 10 101 102 103 104 105 106 107 108 109 11 111 112 13 131 132 14 141 142 144 145 147 ∗

Description ALL COMMODITIES Farm products Field crops Fresh fruits or tree nuts Fresh vegetables Livestock or livestock products Poultry or poultry products Miscellaneous farm products Forest products Barks or gums, crude Miscellaneous forest products Fresh fish or other marine products Fresh fish or other marine products Fish hatcheries, farms or preserves Metallic ores Iron ores Copper ores Lead or zinc ores Gold or silver ores Bauxite ores or other aluminum ores Manganese ores Tungsten ores Chromium ores Miscellaneous metal ores Coal Anthracite Bituminous coal or lignite Crude petroleum, natural gas, or gasoline Crude petroleum or natural gas Natural gasoline Nonmetallic minerals Dimension stone, quarry Crushed or broken stone or riprap Gravel or sand Clay, ceramic, or refractory minerals Chemical and fertilizer minerals

US Department of Commerce, Bureau of the Census, 1997, CFS CD-ROM: CD-THE CFS-93-2, Washington D.C.

117

149 19 191 192 193 194 195 196 199 20 201 202 203 204 205 206 207 208 209 21 211 212 213 214 22 221 222 223 224 225 227 228 229 23 231 233 235 237 238 239 24 241 242 243 244 249 25 251 253 254 259

Miscellaneous nonmetallic minerals Ordnance or accessories Guns, howitzers, mortars, related equipment, or parts, 30 mm Ammunition, over 30 mm Full tracked combat vehicles or parts Military sighting or fire control equipment Small arms, 30 mm or under, or parts Small arms ammunition, 30 mm or under Miscellaneous ordnance, accessories, or parts Food or kindred products Meat, poultry or small game, fresh, chilled, or frozen Dairy products Canned or preserved fruits, vegetables, or seafood Grain mill products Bakery products Sugar, beet, or cane Confectionery or related products Beverages or flavoring extracts Miscellaneous food preparations or kindred products Tobacco products, excluding insecticides Cigarettes Cigars Chewing or smoking tobacco, or snuff Stemmed or re-dried tobacco Textile mill products Cotton broad-woven fabrics Man-made fiber or silk broad-woven fabrics Wool broad-woven fabrics Narrow fabrics, cotton, silk or wool, or man-made fabrics Knit fabrics Floor coverings Thread or yarn Miscellaneous textile goods Apparel or other finished textile products Men’s', youths', or boys' clothing or uniforms Women’s', misses', children’s, or infants clothing Caps, hats or millinery, or hat bodies Fur goods Miscellaneous apparel or accessories Miscellaneous fabricated textile products Lumber or wood products, excluding furniture Primary forest or wood raw materials Sawmill or planning mill products Millwork or prefabricated wood products or plywood or veneer Wooden containers Miscellaneous wood products Furniture or fixtures Household or office furniture Public building or related furniture Lockers, partitions, or shelving Miscellaneous furniture or fixtures 118

26 261 262 263 264 265 266 27 271 272 273 274 276 277 278 279 28 281 282 283 284 285 286 287 289 29 291 295 299 30 301 302 303 304 306 307 31 311 312 313 314 315 316 319 32 321 322 324 325 326 327 328

Pulp, paper, or allied products Pulp or pulp mill products Paper Fiberboard, paperboard, or pulp board Converted paper or paperboard products Containers or boxes, paperboard, fiberboard, or pulp board Building paper or building board Printed matter Newspapers Periodicals Books Miscellaneous printed matter Manifold business forms Greeting cards, seals, labels, or tags Blank books, loose-leaf binders, or devices Service industries for printing trades Chemicals or allied products Industrial inorganic or organic chemicals Plastic materials or synthetic fibers, resins, or rubber Drugs Soap or other detergents, cleaning preparations, cosmetics, perfumes Paints, enamels, lacquers, shellacs, or varnishes Gum or wood chemicals Agricultural chemicals Miscellaneous chemical products Petroleum or coal products Products of petroleum refining Paving or roofing materials Miscellaneous coal or petroleum products Rubber or miscellaneous plastics products Rubber tires or inner tubes Rubber or plastic footwear Reclaimed rubber Rubber or plastic hose or belting Miscellaneous fabricated rubber products Miscellaneous plastic products Leather or leather products Leather Industrial leather belting Boot or shoe cut stock or findings Footwear, leather, or other materials Leather gloves or mittens Luggage or handbags, leather, or other materials Leather goods, etc. Clay, concrete, glass, or stone products Flat glass Glass and glassware, pressed or blown Hydraulic cement Structural clay products Pottery or related products Concrete, gypsum, or plaster products Cut stone or stone products 119

329

Abrasive, asbestos products, or miscellaneous nonmetallic

33 331 332 333 335 336 339 34 341 342 343 344 345 346 348 349 35 351 352 353 354 355 356 357 358 359 36 361 362 363 364 365 366 367 369 37 371 372 373 374 375 376 379 38 381 382 383 384 385 386

Primary metal products Steel works, rolling mill, or other reduction plant products Iron or steel castings Nonferrous metal primary smelted products Nonferrous metal basic shapes Nonferrous metal or nonferrous metal base alloy castings Miscellaneous primary metal products Fabricated metal products Metal cans Cutlery, hand tools, or general hardware Plumbing fixtures or heating apparatus Fabricated structural metal products Bolts, nuts, screws, rivets, washers, or other industrial fasteners Metal stampings Miscellaneous fabricated wire products Miscellaneous fabricated metal products Machinery, excluding electrical Engines or turbines Farm machinery or equipment Construction, mining, or materials handling machinery, or equipment Metal working machinery or equipment Special industry machinery General industrial machinery or equipment Office, computing, or accounting machines Service industry machines Miscellaneous machinery or parts Electrical machinery, equipment, or supplies Electrical transmission or distribution equipment Electrical industrial apparatus Household appliances Electric lighting or wire equipment Radio or television receiving sets Communication equipment Electronic components or accessories Miscellaneous electrical machinery, equipment, or supplies Transportation equipment Motor vehicles or equipment Aircraft or parts Ships or boats Railroad equipment Motorcycles, bicycles, or parts Guided missile or space vehicle parts Miscellaneous transportation equipment Instruments, photographic goods, optical goods, watches, or clocks Engineering, laboratory, or scientific instruments Measuring, controlling, or indicating instruments Optical instruments or lenses Surgical, medical, or dental instruments, or supplies Ophthalmic or opticians goods Photographic equipment or supplies 120

387

Watches, clocks, clockwork operated devices, or parts

39 391 393 394 395 396 399 40 401 402 41 411 412 42 429 48 489 99

Miscellaneous products or manufacturing Jewelry, silverware, or plated ware Musical instruments or parts Toys, amusements, sporting, or athletic goods Pens, pencils, or other office materials, or artists' materials Costume jewelry, buttons, novelties, or notions Miscellaneous manufactured products Waste or scrap materials Ashes Waste or scrap Miscellaneous freight shipment Miscellaneous freight shipments Special commodities not taken in regular freight service Containers, carriers or devices, shipping, returned empty Containers, carriers or devices, shipping, returned empty Waste hazardous materials or waste hazardous substances Hazardous materials Commodity unknown

121

APPENDIX B STANDARD CLASSIFICATION OF TRANSPORTED GOODS (SCTG)∗∗

01-05 Agricultural products and fish 01

Live animals and live fish

02

Cereal grains

03

Agricultural products, except live animals, cereal grains and forage products

04

Animal feed and feed ingredients, cereal, straw, and eggs and other products of animal origin, n.e.c.

05

Meat, fish, seafood, and preparations

06-09 Grains, alcohol, and tobacco products 06

Milled grain products and preparations and bakery products

07

Prepared foodstuffs, n.e.c. and fats and oils

08

Alcoholic beverages

09

Tobacco products

10-14

Stone, nonmetallic minerals, and metallic ores

10

Monumental or building stone

11

Natural sands

12

Gravel and crushed stone

13

Nonmetallic minerals, n.e.c.

14

Metallic ores

15-20 Coal and petroleum products 15

Coal

17

Gasoline and aviation turbine fuel

18

Fuel oils

US Department of Transportation, Bureau of The Transportation Statistics & US Department of Commerce, the Bureau of the Census, 2000, Commodity Flow Survey CD-ROM, CD-EC97-CFS, Washington D.C., ∗

122

19

Products of petroleum refining, n.e.c. and coal products

20

Basic chemical

21-24 Pharmaceutical and chemical products 21

Pharmaceutical products

22

Fertilizer and fertilizer materials

23

Chemical products and preparations, n.e.c.

24

Plastics and rubber

25-30 Wood products and textiles and leather 25

Logs and other wood in the rough

26

W ood products

27

Pulp, newsprint, paper, and paperboard

28

Paper or paperboard articles

29

Printed products

30

Textiles, leather, and articles

31-34 Base metal and machinery 31

Nonmetallic mineral products

32

Base metal in primary or semi-finished forms and in finished basic shapes

33

Articles of base metal

34

Machinery

35-38 Electronics, motorized vehicles, and precision instruments 35

Electronic and other electrical equipment and components, and office equipment

36

Vehicles

37

Transportation equipment, n.e.c.

38

Precision instruments and apparatus

39-43 Furniture and miscellaneous manufactured products 39

Furniture, mattresses and mattress supports, lamps, lighting fittings, and illuminated signs

40

Miscellaneous manufactured products

41

Waste and scrap

43

Mixed freight 123

APPENDIX C DESCRIPTIVE STATISTICS

Table C. 1. Descriptive Statistics for 1993-Commodity 20

Variable

N

Mean

St. Dev.

Minimum

Maximum

Sum

flow93 ($M)

1663

225

462

0

7010

373382

cd93

1663

206288

178933

16323

771064

343056362

io93

1663

201798

179309

16323

771064

335590892

dist93 (miles)

1662

1187

752

40

3500

1972551

opipc93 ($)

1663

20680

2827

15468

29602

34391391

opop93

1663

5589936

5818829

460000

30380000

9296063000

oemp93

1663

38028

34636

1067

184324

63240646

owsem93

1663

146655

154839

7807

783658

243887508

ovlad92 ($M)

1663

3588

3797

87

19585

5966710

oaps93

1663

84

38

23

205

140231

dmnem93

1663

426175

395305

11285

1898885

708728698

dwsem93

1663

150445

156146

7807

783658

250189394

dpipc93 ($)

1663

20748

2846

15468

29602

34503153

dpop93

1663

5736646

5868199

460000

30380000

9540042000

124

Table C. 2. Descriptive Statistics for 1993-Commodity 24

Variable

N

Mean

St. Dev.

Minimum

Maximum

Sum

flow93 ($M)

1599

32

88

0

1908

50573

cd93

1599

201990

179893

16323

771064

322982316

io93

1599

186602

175444

16323

771064

298375944

dist93 (miles)

1599

1257

776

40

3519

2010599

opipc93 ($)

1599

20469

2770

15468

29602

32729391

opop93

1599

5291226

5853632

460000

30380000

8460670000

oemp93

1599

18735

15524

581

59138

29957457

owsem93

1599

137035

154807

7807

783658

219118478

ovlad92 ($M)

1599

740

733

27

3267

1182676

oaps93

1599

22

6

13

45

35589

dmnem93

1599

415107

389168

11285

1898885

663756759

dwsem93

1599

145861

153633

7807

783658

233231221

dpipc93 ($)

1599

20764

2934

15468

29602

33201301

dpop93

1599

5575505

5779705

460000

30380000

8915233000

Table C. 3. Descriptive Statistics for 1993-Commodity 25

Variable

N

Mean

St. Dev.

Minimum

Maximum

Sum

flow93 ($M)

1573

22

51

0

429

35241

cd93

1573

198293

178190

16323

771064

311915433

io93

1573

188969

174030

16323

771064

297248408

dist93 (miles)

1573

1257

752

40

3500

1977876

opipc93 ($)

1573

20503

2849

15468

29602

32251602

opop93

1573

5289980

5685506

460000

30380000

8321139000

oemp93

1573

12100

16474

147

81743

19033484

owsem93

1573

136959

150905

7807

783658

215437079

ovlad92 ($M)

1573

521

673

0

2765

819306

oaps93

1573

44

27

11

118

69382

dmnem93

1573

407945

382204

11285

1898885

641697263

dwsem93

1573

142627

149683

7807

783658

224351561

dpipc93 ($)

1573

20723

2901

15468

29602

32597573

dpop93

1573

5454962

5622184

460000

30380000

8580656000

125

Table C. 4. Descriptive Statistics for 1993-Commodity 26

Variable

N

Mean

St. Dev.

Minimum

Maximum 1563

Sum

flow93 ($M)

1560

68

134

0

105745

cd93

1560

204749

179034

16323

771064 319408369

io93

1560

200209

178931

16323

771064 312326373

dist93 (miles)

1559

1181

735

40

3500

1841585

opipc93 ($)

1560

20604

2855

15468

29602

32142726

opop93

1560

5476608

5586756

460000

oemp93

1560

15546

13802

7

owsem93

1560

143533

149263

7807

ovlad92 ($M)

1560

1343

1227

0

5182

2095757

oaps93

1560

122

75

18

462

185287

dmnem93

1560

424016

393666

11285

1898885 661464705

dwsem93

1560

147988

155483

7807

783658 230861582

dpipc93 ($)

1560

20756

2855

15468

dpop93

1560

5662519

5838848

460000

30380000 8543508000 51953

24252414

783658 223911790

29602

32378924

30380000 8833530000

Table C. 5. Descriptive Statistics for 1993-Commodity 28

Variable

N

Mean

St. Dev.

Minimum

Maximum

flow 93 ($M)

1568

180

387

0

cd93

1568

210865

180195

16323

771064 330636055

io93

1568

207353

179427

16323

771064 325129616

dist93 (miles)

1568

1186

749

45

3519

1859002

opipc93 ($)

1568

20770

2944

15468

29602

32567483

opop93

1568

5768788

5938802

oemp93

1568

25364

27306

199

owsem93

1568

151342

158213

7807

ovlad92 ($M)

1568

3861

4346

13

18381

6054550

oaps93

1568

87

89

10

667

136140

dmnem93

1568

437871

396807

11285

1898885 686581817

dwsem93

1568

152890

157067

7807

783658 239731417

dpipc93 ($)

1568

20851

2923

15468

dpop93

1568

5824610

5898777

126

4208

Sum 281612

460000 30380000 9045459000 108096

39770458

783658 237304911

29602

32693696

460000 30380000 9132989000

Table C. 6. Descriptive Statistics for 1993-Commodity 29

Variable

N

Mean

St. Dev.

Minimum

Maximum 2194

Sum

flow93 ($M)

1557

25

127

0

39110

cd93

1557

178204

168813

16323

771064 277463495

io93

1557

168414

167279

16323

771064 262220643

dist93 (miles)

1557

1378

759

45

3519

2145590

opipc93 ($)

1557

20598

2871

15468

29602

32071490

opop93

1557

4638593

5151814

oemp93

1557

2549

5109

4

owsem93

1557

120886

137338

7807

ovlad92 ($M)

1557

410

828

0

4566

638705

oaps93

1557

51

45

3

233

77467

dmnem93

1557

366182

363566

11285

1898885 570145829

dwsem93

1557

126997

139343

7807

783658 197733742

dpipc93 ($)

1557

20549

2839

15468

dpop93

1557

4870905

5206945

460000 30380000 7222290000 29893

3969019

783658 188220076

29602

31994131

460000 30380000 7583999000

Table C. 7. Descriptive Statistics for 1993-Commodity 30

Variable

N

Mean

St. Dev.

Minimum

Maximum

flow93 ($M)

1604

59

108

0

cd93

1604

207385

180764

16323

771064 332645016

io93

1604

202408

181540

16323

771064 324663035

dist93 (miles)

1603

1251

768

40

3519

2005020

opipc93 ($)

1604

20606

2822

15468

29602

33052611

opop93

1604

5627241

5856473

oemp93

1604

20840

21281

227

owsem93

1604

146925

155960

7807

ovlad92 ($M)

1604

1339

1369

10

5314

2148221

oaps93

1604

60

27

10

120

95472

dmnem93

1604

430793

399722

11285

1898885 690991984

dwsem93

1604

151863

158215

7807

783658 243587673

dpipc93 ($)

1604

20782

2867

15468

dpop93

1604

5782890

5959983

127

969

Sum 94640

460000 30380000 9026095000 91648

33426685

783658 235667872

29602

33333952

460000 30380000 9275756000

Table C. 8. Descriptive Statistics for 1993-Commodity 32

Variable

N

Mean

St. Dev.

Minimum

Maximum 569

Sum

flow93 ($M)

1644

22

49

0

35659

cd93

1644

198902

179700

16323

771064 326994791

io93

1644

196283

181873

16323

771064 322689406

dist93 (miles)

1644

1252

768

40

3519

2058272

opipc93 ($)

1644

20584

2852

15468

29602

33839343

opop93

1644 5457164

5855273

oemp93

1644

12183

12584

691

owsem93

1644

142185

155859

7807

ovlad92 ($M)

1644

757

797

31

3115

1244567

oaps93

1644

31

10

10

53

51006

dmnem93

1644

409853

389239

11285

1898885 673798230

dwsem93

1644

143649

153377

7807

783658 236159080

dpipc93 ($)

1644

20736

2884

15468

dpop93

1644 5499333

5768369

460000 30380000 8971578000 51147

20029656

783658 233751896

29602

34089484

460000 30380000 9040904000

Table C. 9. Descriptive Statistics for 1993-Commodity 33

Variable

N

Mean

St. Dev.

Minimum

Maximum

flow93 ($M)

1511

84

212

0

cd93

1511

206818

178779

16323

771064 312501511

io93

1511

206977

186843

16323

771064 312742721

dist93 (miles)

1511

1187

751

40

3500

1793938

opipc93 ($)

1511

20727

2842

15468

29602

31318289

opop93

1511

5648388

5807533

oemp93

1511

15986

21028

9

owsem93

1511

147441

154157

7807

ovlad92 ($M)

1511

1211

1683

0

6975

1830549

oaps93

1511

95

54

9

321

144069

dmnem93

1511

430735

392098

11285

1898885 650840534

dwsem93

1511

149660

154148

7807

783658 226135652

dpipc93 ($)

1511

20664

2855

15468

dpop93

1511

5721323

5783419

128

3060

Sum 126686

460000 30380000 8534714000 90671

24154712

783658 222782707

29602

31223261

460000 30380000 8644919000

Table C. 10. Descriptive Statistics for 1993-Commodity 34

Variable

N

Mean

St. Dev.

Minimum

Maximum

Sum

flow93 ($M)

1744

75

150

0

1766

131075

cd93

1744

205250

179546

16323

771064

357955830

io93

1744

208015

181494

16323

771064

362778533

dist93 (miles)

1743

1216

760

40

3519

2119993

opipc93 ($)

1744

20777

2907

15468

29602

36235828

opop93

1744

5748641

5918395

oemp93

1744

32372

34546

451

123315

56456256

owsem93

1744

150846

157328

7807

783658

263076089

ovlad92 ($M)

1744

1974

2163

23

8040

3443345

oaps93

1744

37

13

13

66

64815

dmnem93

1744

424354

393493

11285

1898885

740073640

dwsem93

1744

148475

154435

7807

783658

258940155

dpipc93 ($)

1744

20752

2896

15468

29602

36192113

dpop93

1744

5677553

5807918

460000 30380000

9901652000

460000 30380000 10025630000

Table C. 11. Descriptive Statistics for 1993-Commodity 35

Variable

N

Mean

St. Dev.

Minimum

Maximum

Sum

flow93 ($M)

1563

147

298

0

3721

229194

cd93

1563

206782

177771

16323

771064

323200402

io93

1563

208491

179976

16323

771064

325871221

dist93 (miles)

1563

1194

750

40

3500

1866223

opipc93 ($)

1563

20777

2776

15468

29602

32474343

opop93

1563

5732906

5813970

460000 30380000

8960532000

oemp93

1563

47182

47880

934

198366

73745358

owsem93

1563

150969

154357

7807

783658

235964459

ovlad92 ($M)

1563

3162

3811

35

19854

4942780

oaps93

1563

36

12

10

63

56608

dmnem93

1563

430299

390085

11285

1898885

672557204

dwsem93

1563

149630

153309

7807

783658

233871227

dpipc93 ($)

1563

20750

2879

15468

29602

32431681

dpop93

1563

5709766

5764197

460000 30380000

8924365000

129

Table C. 12. Descriptive Statistics for 1993-Commodity 36

Variable

N

Mean

St. Dev.

Minimum

Maximum

Sum

flow93 ($M)

1581

149

341

0

5606

236180

cd93

1581 206543

178203

16323

771064

326543695

io93

1581 210263

183965

16323

771064

332426366

dist93 (miles)

1580

1221

755

40

3500

1928688

opipc93 ($)

1581

20850

2867

15468

29602

32964388

opop93

1581 5828548

5938516

460000

30380000

9214934000

oemp93

1581

36454

41199

115

219375

57634344

owsem93

1581 153134

157542

7807

783658

242104256

ovlad92 ($M)

1581

2843

3770

0

21356

4494315

oaps93

1581

102

54

11

248

161790

dmnem93

1581 426134

386032

11285

1898885

673717702

dwsem93

1581 148993

152145

7807

783658

235557429

dpipc93 ($)

1581

20826

2960

15468

29602

32925678

dpop93

1581 5686082

5707851

460000

30380000

8989696000

Table C. 13. Descriptive Statistics for 1993-Commodity 37

Variable

N

Mean

St. Dev.

Minimum

Maximum

Sum

flow93 ($M)

1292

252

646

0

7800

325979

cd93

1292 205882

179429

16323

771064

265999044

io93

1292 201563

180279

16323

771064

260419375

dist93 (miles)

1292

1181

748

40

3519

1524304

opipc93 ($)

1292

20599

2748

15468

29602

26613480

opop93

1292 5653083

5948699

460000 30380000

7303783000

oemp93

1292

40071

55798

236

278083

51772342

owsem93

1292 147735

157826

7807

783658

190873599

ovlad92 ($M)

1292

3525

4952

12

21698

4554884

oaps93

1292

134

90

13

537

172903

dmnem93

1292 430360

395014

11285

1898885

556024616

dwsem93

1292 150099

155614

7807

783658

193927709

dpipc93 ($)

1292

20706

2793

15468

29602

26751533

dpop93

1292 5705684

5832969

460000 30380000

7371744000

130

Table C. 14. Descriptive Statistics for 1993-Commodity 38

Variable

N

Mean

St. Dev.

Minimum

Maximum

Sum

flow93 ($M)

1421

61

153

0

1999

87165

cd93

1421 207569

183648

16323

771064

294955471

io93

1421 207493

184816

16323

771064

294847176

dist93 (miles)

1421

1290

766

40

3519

1833786

opipc93 ($)

1421

20856

2816

15468

29602

29636386

opop93

1421 5893861

6302444

460000 30380000

8375177000

oemp93

1421

22302

36003

79

184992

31690826

owsem93

1421 154789

166963

7807

783658

219955276

ovlad92 ($M)

1421

2215

3607

0

16790

3147951

oaps93

1421

67

31

10

164

95365

dmnem93

1421 427322

398113

11285

1898885

607223898

dwsem93

1421 151028

157855

7807

783658

214611106

dpipc93 ($)

1421

20798

2902

15468

29602

29554395

dpop93

1421 5770931

5940739

460000 30380000

8200493000

Table C. 15. Descriptive Statistics for 1993-Commodity 39

Variable

N

Mean

St. Dev.

Minimum

Maximum

Sum

flow93 ($M)

1559

53

125

0

2062

82468

cd93

1559 205015

179546

16323

771064

319618503

io93

1559 202109

181203

16323

771064

315088202

dist93 (miles)

1559

1252

763

40

3519

1951995

opipc93 ($)

1559

20745

2857

15468

29602

32342206

opop93

1559 5594112

5893783

460000 30380000

8721220000

oemp93

1559

10105

10540

238

47003

15753520

owsem93

1559 146708

156686

7807

783658

228717318

ovlad92 ($M)

1559

536

572

14

2272

836036

oaps93

1559

27

10

9

51

42240

dmnem93

1559 421390

391436

11285

1898885

656946247

dwsem93

1559 149912

155571

7807

783658

233712898

dpipc93 ($)

1559

20772

2860

15468

29602

32383591

dpop93

1559 5717928

5840324

460000 30380000

8914250000

131

Table C. 16. Descriptive Statistics for 1993-Commodity 75

Variable

N

Mean

St. Dev.

Minimum

Maximum

Sum

flow93 ($M)

680

209

517

0

5670

141929

cd93

680 209695

189956

16323

771064

142592566

io93

680 184085

185481

16323

771064

125177916

dist93 (miles)

680

1242

746

47

3500

844415

opipc93 ($)

680

20358

2722

15468

29602

13843471

opop93

680 5107374

5829019

460000 30380000

3473014000

oemp93

680

42844

62839

401

284042

29133950

owsem93

680 134274

156050

7807

783658

91306072

ovlad92 ($M)

680

1718

2522

3

10913

1168183

oaps93

680

59

44

11

167

40300

dmnem93

680 428490

410954

11285

1898885

291373351

dwsem93

680 148676

159023

7807

783658

101099960

dpipc93 ($)

680

20728

2866

15468

29602

14095252

dpop93

680 5689369

5998679

460000 30380000

3868771000

Table C. 17. Descriptive Statistics for 1997-Commodity 20

Variable

N

Mean

St. Dev.

Minimum

Maximum

Sum

flow97 ($M)

319

197

491

0

3559

62973

cd97

319 175732

185426

16357

867414

56058374

io97

319 139307

167658

16357

867414

44438823

dist97 (miles)

319

1489

772

75

3075

475100

opipc97 ($)

319

24176

3400

18885

35596

7712048

opop97

319 3626408

4427888

480000 32268000

1156824000

oemp97

319

22125

27126

1057

182745

7057906

owsem97

319

95434

124643

8624

856950

30443434

ovlad97 ($M)

319

2077

2902

18

16831

662420

oaps97

319

63

42

19

205

20186

dmnem97

319 346108

390158

13219

2019053

110408381

dwsem97

319 121858

150736

8624

856950

38872826

dpipc97 ($)

319

24029

3120

18885

35596

7665103

dpop97

319 4553279

5500129

480000 32268000

1452496000

132

Table C. 18. Descriptive Statistics for 1997-Commodity 24

Variable

N

Mean

St. Dev.

Minimum

Maximum

Sum

flow97 ($M)

894

30

112

0

1954

26930

cd97

894 213777

193456

16357

867414

191116839

io97

894 180141

175210

16357

867414

161045900

dist97 (miles)

894

1317

744

47

3185

1176435

opipc97 ($)

894

24653

3355

18885

35596

22040223

opop97

894 4899326

5541071

480000 32268000

4379997000

oemp97

894

17765

16861

523

64312

15882009

owsem97

894 132368

151360

8624

856950

118336750

ovlad97 ($M)

894

653

658

29

2667

584129

oaps97

894

51

19

20

97

45765

dmnem97

894 403395

384173

13219

2019053

360634825

dwsem97

894 148936

156323

8624

856950

133148915

dpipc97 ($)

894

24764

3592

18885

35596

22138898

dpop97

894 5513568

5745358

480000 32268000

4929130000

Table C. 19. Descriptive Statistics for 1997-Commodity 25

Variable

N

Mean

St. Dev.

Minimum

Maximum

Sum

flow97 ($M)

1148

38

72

0

605

44001

cd97

1148 241803

204762

16357

867414

277590214

io97

1148 227350

193200

16357

867414

260997648

dist97 (miles)

1147

1152

738

57

3206

1321562

opipc97 ($)

1148

24530

3353

18885

35596

28160909

opop97

1148 5948402

5994118

480000 32268000

6828766000

oemp97

1148

14210

17809

230

77898

16312629

owsem97

1148 158907

162813

8624

856950

182425264

ovlad97 ($M)

1148

880

1023

13

3890

1010503

oaps97

1148

27

18

8

74

30752

dmnem97

1148 470930

431254

13219

2019053

540627894

dwsem97

1148 176321

177894

8624

856950

202416778

dpipc97 ($)

1148

24992

3525

18885

35596

28690600

dpop97

1148 6508839

6579843

480000 32268000

7472147000

133

Table C. 20. Descriptive Statistics for 1997-Commodity 26

Variable

N

Mean

St. Dev.

Minimum

Maximum

Sum

flow97 ($M)

722

110

191

0

1429

79061

cd97

722 243448

212553

16357

867414

175769516

io97

722 207828

207046

16357

867414

150051803

dist97 (miles)

722

1160

768

70

3185

836010

opipc97 ($)

722

24242

3256

18885

35596

17502775

opop97

722 5453819

5902962

480000 32268000

3937657000

oemp97

722

14823

14445

11

53251

10702220

owsem97

722 144779

159897

8624

856950

104530789

ovlad97 ($M)

722

1532

1480

0

5813

1106231

oaps97

662

132

67

28

436

87528

dmnem97

722 467242

449128

13219

2019053

337348825

dwsem97

722 173284

184837

8624

856950

125111223

dpipc97 ($)

722

24962

3573

18885

35596

18022879

dpop97

722 6425677

6832812

480000 32268000

4639339000

Table C. 21. Descriptive Statistics for 1997-Commodity 28

Variable

N

Mean

St. Dev.

Minimum

Maximum

Sum

flow97 ($M)

223

227

456

0

3716

50559

cd97

223 177063

180557

16357

867414

39485081

io97

223 197871

228843

16357

867414

44125174

dist97 (miles)

223

1425

726

80

3026

317721

opipc97 ($)

223

24408

3827

18885

35596

5442912

opop97

223 5437453

7078573

480000 32268000

1212552000

oemp97

223

22400

27939

198

97010

4995141

owsem97

223 147921

193518

8624

856950

32986482

ovlad97 ($M)

223

4891

6765

40

27993

1090626

oaps97

223

84

90

10

444

18681

dmnem97

223 362399

416835

13219

2019053

80815042

dwsem97

223 135882

172745

8624

856950

30301608

dpipc97 ($)

223

24027

3311

18885

35596

5358029

dpop97

223 5031242

6282182

480000 32268000

1121967000

134

Table C. 22. Descriptive Statistics for 1997-Commodity 29

Variable

N

Mean

St. Dev.

Minimum

Maximum

Sum

flow97 ($M)

1321

10

63

0

1459

13232

cd97

1321 191524

187584

16357

867414

253003453

io97

1321 170545

172937

16357

867414

225289751

dist97 (miles)

1321

1400

730

70

3220

1848077

opipc97 ($)

1321

24558

3501

18885

35596

32441551

opop97

1321 4517860

5241297

480000 32268000

5968093000

oemp97

1321

2138

4119

1

26119

2824705

owsem97

1321 120798

141984

8624

856950

159573665

ovlad97 ($M)

1321

554

1326

0

8878

731756

oaps97

1321

55

43

6

200

59202

dmnem97

1321 360703

369019

13219

2019053

476488240

dwsem97

1321 133184

147871

8624

856950

175935762

dpipc97 ($)

1321

24527

3540

18885

35596

32400141

dpop97

1321 4949058

5409779

480000 32268000

6537706000

Table C. 23. Descriptive Statistics for 1997-Commodity 30

Variable

N

Mean

St. Dev.

Minimum

Maximum

Sum

flow97 ($M)

1308

124

201

0

1844

161892

cd97

1308 253076

202850

16357

867414

331023024

io97

1308 260107

205136

16357

867414

340219636

dist97 (miles)

1308

1079

692

57

3220

1411669

opipc97 ($)

1308

24944

3436

18885

35596

32626360

opop97

1308 6887646

6629288

480000 32268000

9009041000

oemp97

1308

26624

23697

358

93222

34824722

owsem97

1308 186398

179142

8624

856950

243808493

ovlad97 ($M)

1308

2190

1922

11

6944

2864804

oaps97

1308

61

21

12

107

79391

dmnem97

1308 492830

425211

13219

2019053

644621881

dwsem97

1308 182899

175470

8624

856950

239231939

dpipc97 ($)

1308

25024

3459

18885

35596

32730779

dpop97

1308 6748151

6487591

480000 32268000

8826582000

135

Table C. 24. Descriptive Statistics for 1997-Commodity 32

Variable

N

Mean

St. Dev.

Minimum

Maximum

Sum

flow97 ($M)

892

20

42

0

395

18064

cd97

892 216163

194302

16357

867414

192817623

io97

892 192083

180344

16357

867414

171337825

dist97 (miles)

892

1247

739

75

3220

1112056

opipc97 ($)

892

24330

3628

18885

35596

21702166

opop97

892 5058957

5728891

480000 32268000

4512590000

oemp97

892

11049

11737

720

50758

9856029

owsem97

892 136391

157603

8624

856950

121660666

ovlad97 ($M)

892

939

1018

32

4124

837848

oaps97

892

32

10

13

53

28149

dmnem97

892 417859

403036

13219

2019053

372730476

dwsem97

892 153307

163464

8624

856950

136749676

dpipc97 ($)

892

24605

3469

18885

35596

21947709

dpop97

892 5714849

6014271

480000 32268000

5097645000

Table C. 25. Descriptive Statistics for 1997-Commodity 33

Variable

N

Mean

St. Dev.

Minimum

Maximum

Sum

flow97 ($M)

1178

127

305

0

5013

149157

cd97

1178 239813

205359

16357

867414

282499337

io97

1178 240029

208532

16357

867414

282754666

dist97 (miles)

1178

1129

732

47

3168

1329405

opipc97 ($)

1178

24665

3382

18885

35596

29055912

opop97

1178 6291378

6379829

480000 32268000

7411243000

oemp97

1178

18761

22623

11

92482

22100561

owsem97

1178 168934

173358

8624

856950

199003688

ovlad97 ($M)

1178

1850

2469

3

10800

2178942

oaps97

1178

141

72

6

329

166464

dmnem97

1178 458715

409735

13219

2019053

540366539

dwsem97

1178 168325

168059

8624

856950

198286768

dpipc97 ($)

1178

24855

3581

18885

35596

29278699

dpop97

1178 6243575

6203342

480000 32268000

7354931000

136

Table C. 26. Descriptive Statistics for 1997-Commodity 34

Variable

N

Mean

St. Dev.

Minimum

Maximum

Sum

flow97 ($M)

1174

95

177

0

1908

111193

cd97

1174 249765

203565

16357

867414

293224511

io97

1174 262020

205214

16357

867414

307611525

dist97 (miles)

1174

1067

702

47

3206

1252460

opipc97 ($)

1174

25087

3578

18885

35596

29452156

opop97

1174 6839641

6506866

480000 32268000

8029739000

oemp97

1174

41257

38965

687

131948

48435694

owsem97

1174 184652

175854

8624

856950

216781358

ovlad97 ($M)

1174

3612

3471

52

13940

4240695

oaps97

1174

25

7

8

39

28940

dmnem97

1174 482187

415941

13219

2019053

566087697

dwsem97

1174 178189

171326

8624

856950

209193798

dpipc97 ($)

1174

24851

3497

18885

35596

29175255

dpop97

1174 6589895

6304310

480000 32268000

7736537000

Table C. 27. Descriptive Statistics for 1997-Commodity 35

Variable

N

Mean

St. Dev.

Minimum

Maximum

Sum

flow97 ($M)

1271

173

310

0

4667

220077

cd97

1271 248732

197583

16357

867414

316137739

io97

1271 247576

203104

16357

867414

314669124

dist97 (miles)

1271

1110

714

57

3220

1411043

opipc97 ($)

1271

24850

3330

18885

35596

31584318

opop97

1271 6408937

6184452

480000 32268000

8145759000

oemp97

1271

56775

55146

1224

237065

72161135

owsem97

1271 173852

167867

8624

856950

220965471

ovlad97 ($M)

1271

3589

3545

40

13216

4561879

oaps97

1271

75

23

27

157

94702

dmnem97

1271 483340

409209

13219

2019053

614325331

dwsem97

1271 177730

167604

8624

856950

225895349

dpipc97 ($)

1271

24984

3454

18885

35596

31754780

dpop97

1271 6541420

6181908

480000 32268000

8314145000

137

Table C. 28. Descriptive Statistics for 1997-Commodity 36

Variable

N

Mean

St. Dev.

Minimum

Maximum

Sum

flow97 ($M)

1274

356

863

0

13664

452990

cd97

1274 255316

204382

16357

867414

325272805

io97

1274 260495

205727

16357

867414

331871255

dist97 (miles)

1274

1144

735

47

3206

1457713

opipc97 ($)

1274

25227

3473

18885

35596

32139473

opop97

1274 6788222

6427557

480000 32268000

8648195000

oemp97

1274

45113

49253

222

263474

57473923

owsem97

1274 184358

173658

8624

856950

234871505

ovlad97 ($M)

1274

1522

1347

0

5306

1938890

oaps97

1234

266

114

43

653

328665

dmnem97

1274 496430

422865

13219

2019053

632451859

dwsem97

1274 184294

175028

8624

856950

234790847

dpipc97 ($)

1274

25028

3470

18885

35596

31886008

dpop97

1274 6797768

6479000

480000 32268000

8660357000

Table C. 29. Descriptive Statistics for 1997-Commodity 37

Variable

N

Mean

St. Dev.

Minimum

Maximum

Sum

flow97 ($M)

420

407

1088

0

12136

171138

cd97

420 238232

220128

16357

867414

100057553

io97

420 217726

198255

16357

867414

91444831

dist97 (miles)

420

1179

743

57

3186

495172

opipc97 ($)

420

24738

2882

18885

35596

10390033

opop97

420 5943695

6695202

480000 32268000

2496352000

oemp97

420

45670

65457

318

293723

19181441

owsem97

420 159881

180513

8624

856950

67150037

ovlad97 ($M)

420

5609

8819

15

39045

2355952

oaps97

420

134

82

15

324

56077

dmnem97

420 460149

449771

13219

2019053

193262712

dwsem97

420 177012

192995

8624

856950

74344983

dpipc97 ($)

420

24888

3441

18885

35596

10452954

dpop97

420 6536612

7123164

480000 32268000

2745377000

138

Table C. 30. Descriptive Statistics for 1997-Commodity 38

Variable

N

Mean

St. Dev.

Minimum

Maximum

Sum

flow97 ($M)

942

82

161

0

1649

76825

cd97

942 258334

205935

16357

867414

243350587

io97

942 256306

216966

16357

867414

241440295

dist97 (miles)

942

1205

739

70

3206

1134806

opipc97 ($)

942

25232

3562

18885

35596

23768154

opop97

942 6768212

6987278

480000 32268000

6375656000

oemp97

942

24988

37483

73

179674

23538773

owsem97

942 183774

187193

8624

856950

173115106

ovlad97 ($M)

942

6671

13010

9

65716

6284218

oaps97

942

51

25

5

130

47575

dmnem97

942 499904

431407

13219

2019053

470910013

dwsem97

942 186627

179844

8624

856950

175802240

dpipc97 ($)

942

25043

3508

18885

35596

23590322

dpop97

942 6886877

6644480

480000 32268000

6487438000

Table C. 31. Descriptive Statistics for 1997-Commodity 39

Variable

N

Mean

St. Dev.

Minimum

Maximum

Sum

flow97 ($M)

1466

158

271

0

2373

231334

cd97

1466 243681

201364

16357

867414

357235771

io97

1466 251070

202768

16357

867414

368069284

dist97 (miles)

1466

1135

723

47

3206

1664098

opipc97 ($)

1466

25019

3538

18885

35596

36677892

opop97

1466 6543836

6250388

480000 32268000

9593264000

oemp97

1466

12136

12255

350

62926

17791326

owsem97

1466 177090

168815

8624

856950

259613674

ovlad97 ($M)

1466

1569

1720

18

9446

2300450

oaps97

1466

16

6

7

32

24156.33

dmnem97

1466 472926

421644

13219

2019053

693309914

dwsem97

1466 176408

173967

8624

856950

258614642

dpipc97 ($)

1466

24934

3500

18885

35596

36552749

dpop97

1466 6536293

6436029

480000 32268000

9582205000

139

Table C. 32. Descriptive Statistics for 1997-Commodity 75

Variable

N

Mean

St. Dev.

Minimum

Maximum

Sum

flow97 ($M)

1221

158

324

0

3424

192385

cd97

1221 246561

203193

16357

867414

301051044

io97

1221 268152

205883

16357

867414

327413098

dist97 (miles)

1221

1138

726

47

3220

1389980

opipc97 ($)

1221

25189

3534

18885

35596

30756129

opop97

1221 7080614

6686546

480000 32268000

8645430000

oemp97

1221

46256

57417

428

232810

56478128

owsem97

1221 191510

180779

8624

856950

233833608

ovlad97 ($M)

1221

2245

2816

11

11088

2740702

oaps97

1221

60

38

15

146

73318

dmnem97

1221 478865

421829

13219

2019053

584694204

dwsem97

1221 176578

174340

8624

856950

215601740

dpipc97 ($)

1221

24869

3453

18885

35596

30365450

dpop97

1221 6539216

6442263

480000 32268000

7984383000

140

Table C. 33. 1993 Flow Values and Percentages Across States and Commodities State Obs Code State 1

6 California

totsp93 Cmdt ($ Mil)

stprc93 (%)

cmprc93 (%)

20

89385

10.5

15

2

17 Illinois

20

56826

6.7

18.6

3

48 Texas

20

53908

6.3

13.1

4

36 New York

20

42011

4.9

17.6

5

42 Penn

20

42103

4.9

18.7

6

12 Florida

20

37300

4.4

24.4

7

26 Michigan

20

34748

4.1

14.5

8

39 Ohio

20

33419

3.9

11.1

9

34 N. Jersey

20

32343

3.8

13.4

10

55 Wisconsin

20

31961

3.8

24

11

19 Iowa

20

25355

3

37.8

12

13 Georgia

20

23971

2.8

12.2

13

27 Minnesota

20

24038

2.8

25.3

14

29 Missouri

20

22837

2.7

19.4

15

24 Maryland

20

22205

2.6

25.6

16

37 N. Carolina

20

21710

2.5

10.7

17

47 Tennessee

20

19656

2.3

12.4

18

53 Washington

20

19256

2.3

16.9

19

31 Nebraska

20

18771

2.2

51.5

20

51 Virginia

20

17699

2.1

17.6

21

18 Indiana

20

16958

2

10.6

22

20 Kansas

20

17202

2

29.9

23

5 Arkansas

20

14719

1.7

27.5

24

8 Colorado

20

12524

1.5

23.9

25

1 Alabama

20

12152

1.4

15

26

25 Mass.

20

12259

1.4

12.1

27

21 Kentucky

20

10955

1.3

11

28

22 Louisiana

20

11401

1.3

13.3

29

41 Oregon

20

9459

1.1

15.9

30

4 Arizona

20

8187

1

13.2

31

28 Mississippi

20

7731

0.9

16

32

40 Oklahoma

20

6744

0.8

15

Cmdt: Commodity; totsp93: total shipment in 1993; cmprc93: Commodity percentage in 1993 for the state ; stprc93: State percentage in 1993 for the commodity.

141

Table C.33. 1993 Flow Values and Percentages Across States and Commodities (Continued) 33 45 S. Carolina

20

6785

0.8

8.7

34

20

5705

0.7

8.8

35 49 Utah

20

4760

0.6

15.1

36 16 Idaho

20

4468

0.5

30.7

37 23 Maine

20

2425

0.3

12.9

38 38 N. Dakota

20

2666

0.3

36.4

39 54 W. Virginia

20

2318

0.3

8.5

40 10 Delaware

20

1928

0.2

12.6

41 30 Montana

20

1713

0.2

22.3

42 32 Nevada

20

1824

0.2

11.7

43 33 New Hemp

20

1308

0.2

8.6

44 35 New Mexico 20

1502

0.2

14.9

45 44 Rhode Is

20

1282

0.2

7.5

46 46 S. Dakota

20

2040

0.2

25.7

47 50 Vermont

20

1184

0.1

15

48 56 Wyoming

20

308

0

6

49 48 Texas

21 29737

27.7

7.2

50 37 N. Carolina

21 18330

17.1

9.1

51

21 17858

16.6

3

9 Connect.

6 California

52 51 Virginia

21

8894

8.3

8.8

53 21 Kentucky

21

7167

6.7

7.2

54 34 N. Jersey

21

2208

2.1

0.9

55 27 Minnesota

21

2184

2

2.3

56 17 Illinois

21

1862

1.7

0.6

57 12 Florida

21

1361

1.3

0.9

58 36 New York

21

1350

1.3

0.6

59 26 Michigan

21

1313

1.2

0.5

60 20 Kansas

21

1186

1.1

2.1

61 29 Missouri

21

1191

1.1

1

62 47 Tennessee

21

1217

1.1

0.8

63 18 Indiana

21

1094

1

0.7

64

21

783

0.7

1

65 25 Mass.

21

722

0.7

0.7

66 39 Ohio

21

793

0.7

0.3

1 Alabama

142

Table C.33. 1993 Flow Values and Percentages Across States and Commodities (Continued) 67 42 Penn

21

765

0.7

0.3

68

21

659

0.6

1

69 22 Louisiana

21

615

0.6

0.7

70

21

491

0.5

0.8

71 49 Utah

21

556

0.5

1.8

72 55 Wisconsin

21

516

0.5

0.4

73

21

455

0.4

0.9

74 28 Mississippi 21

454

0.4

0.9

75 45 S. Carolina

21

455

0.4

0.6

76 19 Iowa

21

275

0.3

0.4

77 24 Maryland

21

321

0.3

0.4

78 54 W. Virginia

21

360

0.3

1.3

79 13 Georgia

21

248

0.2

0.1

80 32 Nevada

21

228

0.2

1.5

81 40 Oklahoma

21

267

0.2

0.6

82 41 Oregon

21

185

0.2

0.3

83 44 Rhode Is

21

185

0.2

1.1

84

8 Colorado

21

75

0.1

0.1

85 10 Delaware

21

95

0.1

0.6

86 16 Idaho

21

72

0.1

0.5

87 23 Maine

21

120

0.1

0.6

88 30 Montana

21

54

0.1

0.7

89 31 Nebraska

21

110

0.1

0.3

90 33 New Hemp

21

82

0.1

0.5

91 35 New Mexico 21

137

0.1

1.4

9 Connect.

4 Arizona

5 Arkansas

92 46 S. Dakota

21

90

0.1

1.1

93 50 Vermont

21

88

0.1

1.1

94 53 Washington 21

106

0.1

0.1

95 56 Wyoming

21

123

0.1

2.4

96 38 N. Dakota

21

53

0

0.7

97 41 Oregon

24 11832

9.4

19.9

98

24 11294

9

1.9

9402

7.5

8.2

5346

4.2

2.6

6 California

99 53 Washington 24 100 37 N. Carolina

24

143

Table C.33. 1993 Flow Values and Percentages Across States and Commodities (Continued) 101 13 Georgia

24 4680

3.7

2.4

102 48 Texas

24 4596

3.6

1.1

103 42 Penn

24 4405

3.5

2

104 55 Wisconsin

24 4272

3.4

3.2

105

24 3977

3.2

4.9

106 27 Minnesota 24 3759

3

4

1 Alabama

107 39 Ohio

24 3402

2.7

1.1

108 51 Virginia

24 3361

2.7

3.3

109 18 Indiana

24 3235

2.6

2

110 12 Florida

24 3138

2.5

2

111 28 Mississipi

24 3124

2.5

6.5

112 26 Michigan

24 2974

2.4

1.2

113

24 2895

2.3

5.4

114 16 Idaho

24 2926

2.3

20.1

115 22 Louisiana

24 2817

2.2

3.3

116 17 Illinois

24 2662

2.1

0.9

117 34 N. Jersey

24 2639

2.1

1.1

118 47 Tennessee 24 2683

2.1

1.7

119 45 S. Carolina 24 2446

1.9

3.1

120 25 Mass.

24 2048

1.6

2

121 29 Missouri

24 2020

1.6

1.7

122 36 New York

24 2040

1.6

0.9

123 21 Kentucky

24 1725

1.4

1.7

124 24 Maryland

24 1727

1.4

2

125 23 Maine

24 1586

1.3

8.4

126 30 Montana

24 1542

1.2

20

127

24 1171

0.9

2.2

128 19 Iowa

24 1171

0.9

1.7

129

24 1034

0.8

1.7

130 44 Rhode Is

24

985

0.8

5.7

131

24

774

0.6

1.2

132 33 New Hemp 24

711

0.6

4.6

133 54 W. Virginia 24

814

0.6

3

134 56 Wyoming

654

0.5

12.7

5 Arkansas

8 Colorado

4 Arizona

9 Connect.

24

144

Table C.33. 1993 Flow Values and Percentages Across States and Commodities (Continued) 135 20 Kansas

24

564

0.4

1

136 35 New Mexico 24

458

0.4

4.6

137 40 Oklahoma

24

542

0.4

1.2

138 49 Utah

24

537

0.4

1.7

139 50 Vermont

24

452

0.4

5.7

140 10 Delaware

24

432

0.3

2.8

141 31 Nebraska

24

392

0.3

1.1

142 46 S. Dakota

24

361

0.3

4.6

143 38 N. Dakota

24

194

0.2

2.6

144 32 Nevada

24

183

0.1

1.2

145

25 7592

11

1.3

146 37 N. Carolina

25 6292

9.1

3.1

147 26 Michigan

25 5238

7.6

2.2

148 48 Texas

25 4305

6.2

1

149 42 Penn

25 3386

4.9

1.5

150 18 Indiana

25 3120

4.5

2

151 51 Virginia

25 2985

4.3

3

152 39 Ohio

25 2936

4.2

1

153 17 Illinois

25 2556

3.7

0.8

154 34 N. Jersey

25 2563

3.7

1.1

155 36 New York

25 2514

3.6

1.1

156 28 Mississipi

25 2404

3.5

5

157 47 Tennessee

25 2406

3.5

1.5

158 12 Florida

25 1877

2.7

1.2

159 13 Georgia

25 1856

2.7

0.9

160 29 Missouri

25 1541

2.2

1.3

161 55 Wisconsin

25 1521

2.2

1.1

162

1 Alabama

25 1330

1.9

1.6

163

4 Arizona

25 1311

1.9

2.1

164 25 Mass.

25 1233

1.8

1.2

165 27 Minnesota

25 1151

1.7

1.2

166

25

948

1.4

1.8

167 53 Washington 25

971

1.4

0.9

168 24 Maryland

909

1.3

1

6 California

5 Arkansas

25

145

Table C.33. 1993 Flow Values and Percentages Across States and Commodities (Continued) 169 45 S. Carolina

25

749

1.1

1

170 32 Nevada

25

720

1

4.6

171

25

620

0.9

1.2

172 19 Iowa

25

601

0.9

0.9

173 31 Nebraska

25

493

0.7

1.4

174 41 Oregon

25

441

0.6

0.7

175 49 Utah

25

388

0.6

1.2

176

25

332

0.5

0.5

177 21 Kentucky

25

367

0.5

0.4

178 40 Oklahoma

25

265

0.4

0.6

179 20 Kansas

25

226

0.3

0.4

180 10 Delaware

25

126

0.2

0.8

181 22 Louisiana

25

127

0.2

0.1

182 38 N. Dakota

25

112

0.2

1.5

183 44 Rhode Is

25

133

0.2

0.8

184 54 W. Virginia

25

142

0.2

0.5

185 16 Idaho

25

53

0.1

0.4

186 23 Maine

25

93

0.1

0.5

187 30 Montana

25

37

0.1

0.5

188 33 New Hemp

25

81

0.1

0.5

189 35 New Mexico 25

65

0.1

0.6

190 50 Vermont

25

92

0.1

1.2

191 56 Wyoming

25

51

0.1

1

192 46 S. Dakota

25

30

0

0.4

193

6 California

26 16289

8.4

2.7

194 55 Wisconsin

26 11840

6.1

8.9

195 42 Penn

26 11325

5.8

5

196 17 Illinois

26 11083

5.7

3.6

197 39 Ohio

26

9424

4.8

3.1

198 13 Georgia

26

9119

4.7

4.6

199 36 New York

26

8695

4.5

3.6

200 48 Texas

26

7287

3.7

1.8

201 34 N. Jersey

26

7069

3.6

2.9

202 37 N. Carolina

26

7097

3.6

3.5

8 Colorado

9 Connect.

146

Table C.33. 1993 Flow Values and Percentages Across States and Commodities (Continued) 203 26 Michigan

26 6552

3.4

2.7

204

26 6271

3.2

7.8

205 53 Washington 26 5399

2.8

4.7

206 12 Florida

26 5212

2.7

3.4

207 29 Missouri

26 5282

2.7

4.5

208 25 Mass.

26 5131

2.6

5.1

209 47 Tennessee

26 4977

2.6

3.1

210 51 Virginia

26 4868

2.5

4.8

211 45 S. Carolina

26 4393

2.3

5.6

212 22 Louisiana

26 4301

2.2

5

213 27 Minnesota

26 4279

2.2

4.5

214

26 3869

2

7.2

26 3727

1.9

19.8

216 28 Mississippi 26 3466

1.8

7.2

217 41 Oregon

26 3357

1.7

5.6

218 18 Indiana

26 3194

1.6

2

219 24 Maryland

26 3132

1.6

3.6

220

26 2516

1.3

3.9

221 21 Kentucky

26 2422

1.2

2.4

222 19 Iowa

26 1852

1

2.8

223

26 1539

0.8

2.9

224 20 Kansas

26 1516

0.8

2.6

225 40 Oklahoma

26 1425

0.7

3.2

226 49 Utah

26 1004

0.5

3.2

227

26

859

0.4

1.4

228 16 Idaho

26

739

0.4

5.1

229 33 New Hemp

26

806

0.4

5.3

230 31 Nebraska

26

631

0.3

1.7

231 50 Vermont

26

536

0.3

6.8

232 10 Delaware

26

375

0.2

2.5

233 30 Montana

26

362

0.2

4.7

234 32 Nevada

26

325

0.2

2.1

235 35 New Mexico 26

156

0.1

1.6

236 38 N. Dakota

128

0.1

1.7

1 Alabama

5 Arkansas

215 23 Maine

9 Connect.

8 Colorado

4 Arizona

26

147

Table C.33. 1993 Flow Values and Percentages Across States and Commodities (Continued) 237 44 Rhode Is

26

250

0.1

1.5

238 46 S. Dakota

26

143

0.1

1.8

239 54 W. Virginia

26

270

0.1

1

240 56 Wyoming

26

20

0

0.4

241 48 Texas

28 59862

11.3

14.6

242 17 Illinois

28 46167

8.7

15.1

243

6 California

28 43637

8.3

7.3

244 34 N. Jersey

28 35773

6.8

14.9

245 39 Ohio

28 29223

5.5

9.7

246 42 Penn

28 26369

5

11.7

247 22 Louisiana

28 22762

4.3

26.6

248 37 N. Carolina 28 22030

4.2

10.9

249 36 New York

28 18936

3.6

7.9

250 13 Georgia

28 18185

3.4

9.2

251 47 Tennessee

28 17486

3.3

11

252 12 Florida

28 14630

2.8

9.6

253 26 Michigan

28 14867

2.8

6.2

254 29 Missouri

28 13660

2.6

11.6

255 18 Indiana

28 11474

2.2

7.2

256 21 Kentucky

28 10289

1.9

10.4

257 51 Virginia

28 10196

1.9

10.1

258 25 Mass.

28

9362

1.8

9.2

259 45 S. Carolina

28

9201

1.7

11.8

260 55 Wisconsin

28

8505

1.6

6.4

261

28

8079

1.5

10

262 19 Iowa

28

6671

1.3

9.9

263 24 Maryland

28

6956

1.3

8

264

28

6322

1.2

9.7

265 27 Minnesota

28

6218

1.2

6.6

266 54 W. Virginia

28

6080

1.2

22.2

267 28 Mississippi 28

4482

0.8

9.3

268 53 Washington 28

4233

0.8

3.7

269 20 Kansas

28

3888

0.7

6.8

270 49 Utah

28

3854

0.7

12.2

1 Alabama

9 Connect.

148

Table C.33. 1993 Flow Values and Percentages Across States and Commodities (Continued) 271 10 Delaware

28

3409

0.6

22.3

272 40 Oklahoma

28

3319

0.6

7.4

273 41 Oregon

28

3232

0.6

5.4

274

4 Arizona

28

2732

0.5

4.4

275

5 Arkansas

28

2724

0.5

5.1

276 31 Nebraska

28

2412

0.5

6.6

277

28

2240

0.4

4.3

278 16 Idaho

28

1477

0.3

10.2

279 32 Nevada

28

1645

0.3

10.5

280 44 Rhode Is

28

864

0.2

5

281 46 S. Dakota

28

1262

0.2

15.9

282 23 Maine

28

732

0.1

3.9

283 30 Montana

28

411

0.1

5.3

284 33 New Hemp

28

581

0.1

3.8

285 35 New Mexico 28

613

0.1

6.1

286 38 N. Dakota

28

520

0.1

7.1

287 50 Vermont

28

366

0.1

4.6

288 56 Wyoming

28

519

0.1

10.1

289 48 Texas

29 72090

20.1

17.6

290

6 California

29 44750

12.5

7.5

291 22 Louisiana

29 26179

7.3

30.6

292 17 Illinois

29 15566

4.3

5.1

293 39 Ohio

29 15156

4.2

5.1

294 13 Georgia

29 12379

3.5

6.3

295 34 N. Jersey

29 12211

3.4

5.1

296 26 Michigan

29 11686

3.3

4.9

297 42 Penn

29 11674

3.3

5.2

298 12 Florida

29

9515

2.7

6.2

299 18 Indiana

29

9008

2.5

5.6

300 36 New York

29

8210

2.3

3.4

301 37 N. Carolina

29

8069

2.3

4

302 53 Washington 29

7897

2.2

6.9

303 40 Oklahoma

29

6083

1.7

13.5

304 51 Virginia

29

6020

1.7

6

8 Colorado

149

Table C.33. 1993 Flow Values and Percentages Across States and Commodities (Continued) 305 21 Kentucky

29

5595

1.6

5.6

306

29

5181

1.4

6.4

29

4845

1.4

8.4

308 28 Mississippi 29

4977

1.4

10.3

309 41 Oregon

29

5009

1.4

8.4

310 47 Tennessee

29

4970

1.4

3.1

311 55 Wisconsin

29

4996

1.4

3.7

312 29 Missouri

29

3693

1

3.1

313 24 Maryland

29

3144

0.9

3.6

314 25 Mass.

29

3238

0.9

3.2

315 45 S. Carolina

29

3233

0.9

4.2

316 54 W. Virginia

29

3159

0.9

11.5

317 27 Minnesota

29

3028

0.8

3.2

318 44 Rhode Is

29

2727

0.8

15.9

319 19 Iowa

29

2476

0.7

3.7

320 35 New Mexico 29

2073

0.6

20.6

321 49 Utah

29

2313

0.6

7.3

322

4 Arizona

29

1895

0.5

3.1

323

5 Arkansas

29

1889

0.5

3.5

324

9 Connect.

29

1661

0.5

2.6

325 30 Montana

29

1850

0.5

24.1

326

8 Colorado

29

1371

0.4

2.6

327 10 Delaware

29

1408

0.4

9.2

328 56 Wyoming

29

1335

0.4

26

329 23 Maine

29

968

0.3

5.1

330 32 Nevada

29

953

0.3

6.1

331 33 New Hamp. 29

1209

0.3

7.9

1 Alabama

307 20 Kansas

332 38 N. Dakota

29

977

0.3

13.3

333 31 Nebraska

29

884

0.2

2.4

334 16 Idaho

29

469

0.1

3.2

335 46 S. Dakota

29

468

0.1

5.9

336 50 Vermont

29

130

0

1.7

337

30 14892

8.6

2.5

30 13596

7.9

4.5

6 California

338 39 Ohio

150

Table C.33. 1993 Flow Values and Percentages Across States and Commodities (Continued) 339 34 N. Jersey

30 10988

6.4

4.6

340 17 Illinois

30 10829

6.3

3.5

341 42 Penn

30

8853

5.1

3.9

342 26 Michigan

30

8483

4.9

3.5

343 48 Texas

30

8168

4.7

2

344 37 N. Carolina 30

7244

4.2

3.6

345 13 Georgia

30

6047

3.5

3.1

346 21 Kentucky

30

5624

3.3

5.7

347 36 New York

30

5509

3.2

2.3

348 47 Tennessee

30

5538

3.2

3.5

349 18 Indiana

30

5109

3

3.2

350 55 Wisconsin

30

5233

3

3.9

351 20 Kansas

30

4707

2.7

8.2

352 27 Minnesota

30

4380

2.5

4.6

353 25 Mass.

30

4163

2.4

4.1

354 12 Florida

30

3560

2.1

2.3

355 51 Virginia

30

3466

2

3.4

356 29 Missouri

30

3285

1.9

2.8

357

1 Alabama

30

3078

1.8

3.8

358

5 Arkansas

30

2966

1.7

5.5

359 45 S. Carolina

30

2772

1.6

3.6

360 24 Maryland

30

2564

1.5

3

361 40 Oklahoma

30

2533

1.5

5.6

362 28 Mississippi 30

2422

1.4

5

363 41 Oregon

30

2043

1.2

3.4

364 53 Washington 30

2010

1.2

1.8

365

30

1829

1.1

2.8

366 19 Iowa

30

1957

1.1

2.9

367 22 Louisiana

30

1200

0.7

1.4

368

8 Colorado

30

1067

0.6

2

369

4 Arizona

30

813

0.5

1.3

370 31 Nebraska

30

862

0.5

2.4

371 49 Utah

30

860

0.5

2.7

372 10 Delaware

30

551

0.3

3.6

9 Connect.

151

Table C.33. 1993 Flow Values and Percentages Across States and Commodities (Continued) 373 23 Maine

30

507

0.3

2.7

374 33 New Hamp. 30

581

0.3

3.8

375 44 Rhode Is

30

538

0.3

3.1

376 16 Idaho

30

279

0.2

1.9

377 32 Nevada

30

413

0.2

2.6

378 54 W. Virginia

30

407

0.2

1.5

379 30 Montana

30

105

0.1

1.4

380 35 New Mexico 30

188

0.1

1.9

381 38 N. Dakota

30

129

0.1

1.8

382 46 S. Dakota

30

246

0.1

3.1

383 50 Vermont

30

140

0.1

1.8

384 56 Wyoming

30

19

0

0.4

385 48 Texas

32 10521

11.6

2.6

386

32

8831

9.7

1.5

387 39 Ohio

32

6620

7.3

2.2

388 42 Penn

32

5746

6.3

2.5

389 13 Georgia

32

4322

4.8

2.2

390 17 Illinois

32

4016

4.4

1.3

391 36 New York

32

3897

4.3

1.6

392 37 N. Carolina

32

3882

4.3

1.9

393 12 Florida

32

3595

4

2.3

394 34 N. Jersey

32

3195

3.5

1.3

395 26 Michigan

32

3017

3.3

1.3

396 18 Indiana

32

2748

3

1.7

397 47 Tennessee

32

2162

2.4

1.4

398 51 Virginia

32

2097

2.3

2.1

399 29 Missouri

32

2008

2.2

1.7

400 21 Kentucky

32

1631

1.8

1.6

401 24 Maryland

32

1552

1.7

1.8

402 55 Wisconsin

32

1422

1.6

1.1

403 27 Minnesota

32

1406

1.5

1.5

404

32

1284

1.4

1.6

32

1293

1.4

2.9

406 53 Washington 32

1310

1.4

1.1

6 California

1 Alabama

405 40 Oklahoma

152

Table C.33. 1993 Flow Values and Percentages Across States and Commodities (Continued) 407 19 Iowa

32

1083

1.2

1.6

408 22 Louisiana

32

1091

1.2

1.3

409 25 Mass.

32

1091

1.2

1.1

410

32

975

1.1

1.9

411 20 Kansas

32

986

1.1

1.7

412 45 S. Carolina

32

958

1.1

1.2

413 54 W. Virginia

32

918

1

3.4

414

4 Arizona

32

845

0.9

1.4

415

9 Connect.

32

781

0.9

1.2

32

644

0.7

1.1

417 28 Mississippi 32

576

0.6

1.2

418 49 Utah

32

509

0.6

1.6

419

5 Arkansas

32

457

0.5

0.9

420 31 Nebraska

32

474

0.5

1.3

421 32 Nevada

32

498

0.5

3.2

422 23 Maine

32

255

0.3

1.4

423 33 New Hamp. 32

277

0.3

1.8

424 35 New Mexico 32

305

0.3

3

425 46 S. Dakota

32

273

0.3

3.4

426 50 Vermont

32

276

0.3

3.5

427 56 Wyoming

32

254

0.3

4.9

428 10 Delaware

32

148

0.2

1

429 16 Idaho

32

153

0.2

1.1

430 30 Montana

32

172

0.2

2.2

431 44 Rhode Is

32

203

0.2

1.2

432 38 N. Dakota

32

105

0.1

1.4

433 39 Ohio

33 26226

11.5

8.7

434 18 Indiana

33 17485

7.7

10.9

435 42 Penn

33 17639

7.7

7.8

436 17 Illinois

33 17053

7.5

5.6

437

33 14998

6.6

2.5

438 48 Texas

33 14453

6.3

3.5

439 26 Michigan

33 14241

6.2

5.9

440

33

3.6

10.3

8 Colorado

416 41 Oregon

6 California

1 Alabama

8307

153

Table C.33. 1993 Flow Values and Percentages Across States and Commodities (Continued) 441 36 New York

33 7186

3.1

3

442 21 Kentucky

33 5861

2.6

5.9

443

33 5182

2.3

8.4

444 47 Tennessee

33 5300

2.3

3.3

445 55 Wisconsin

33 4996

2.2

3.7

446 13 Georgia

33 4472

2

2.3

447 24 Maryland

33 4454

1.9

5.1

448 34 N. Jersey

33 4443

1.9

1.8

449 53 Washington 33 4431

1.9

3.9

450 54 W. Virginia

33 4360

1.9

15.9

451 29 Missouri

33 4129

1.8

3.5

452 12 Florida

33 3519

1.5

2.3

453 37 N. Carolina

33 3504

1.5

1.7

454

5 Arkansas

33 3108

1.4

5.8

455

9 Connect.

33 3062

1.3

4.7

456 19 Iowa

33 2497

1.1

3.7

457 27 Minnesota

33 2576

1.1

2.7

458 41 Oregon

33 2491

1.1

4.2

459 22 Louisiana

33 2293

1

2.7

460 49 Utah

33 2317

1

7.3

461 51 Virginia

33 2204

1

2.2

462 25 Mass.

33 2104

0.9

2.1

463 40 Oklahoma

33 2132

0.9

4.7

464 45 S. Carolina

33 2096

0.9

2.7

465 28 Mississippi 33 1851

0.8

3.8

466

33 1367

0.6

2.6

33

825

0.4

1.4

468 35 New Mexico 33

936

0.4

9.3

469 44 Rhode Is

33

942

0.4

5.5

470 31 Nebraska

33

702

0.3

1.9

471 33 New Hamp. 33

629

0.3

4.1

472 10 Delaware

33

505

0.2

3.3

473 30 Montana

33

388

0.2

5

474 32 Nevada

33

369

0.2

2.4

4 Arizona

8 Colorado

467 20 Kansas

154

Table C.33. 1993 Flow Values and Percentages Across States and Commodities (Continued) 475 16 Idaho

33

176

0.1

1.2

476 23 Maine

33

254

0.1

1.3

477 46 S. Dakota

33

150

0.1

1.9

478 38 N. Dakota

33

65

0

0.9

479 50 Vermont

33

79

0

1

480 56 Wyoming

33

71

0

1.4

481

6 California

34 21397

9

3.6

482 26 Michigan

34 20425

8.6

8.5

483 17 Illinois

34 19407

8.2

6.4

484 39 Ohio

34 18539

7.8

6.2

485 48 Texas

34 14651

6.2

3.6

486 42 Penn

34 14433

6.1

6.4

487 18 Indiana

34 10363

4.4

6.5

488 36 New York

34

9380

4

3.9

489 37 N. Carolina 34

8007

3.4

4

490 55 Wisconsin

34

7103

3

5.3

491 34 N. Jersey

34

6880

2.9

2.9

492 13 Georgia

34

6476

2.7

3.3

493 29 Missouri

34

6245

2.6

5.3

494 12 Florida

34

5990

2.5

3.9

495 47 Tennessee

34

5957

2.5

3.8

496 27 Minnesota

34

4852

2

5.1

497

34

3929

1.7

6.1

34

4105

1.7

4.1

499 53 Washington 34

3791

1.6

3.3

500 21 Kentucky

34

3223

1.4

3.2

501 45 S. Carolina

34

3213

1.4

4.1

502

34

3053

1.3

3.8

503 19 Iowa

34

3181

1.3

4.7

504 24 Maryland

34

3031

1.3

3.5

505 51 Virginia

34

2989

1.3

3

506 40 Oklahoma

34

2689

1.1

6

507

5 Arkansas

34

2405

1

4.5

508

8 Colorado

34

2160

0.9

4.1

9 Connect.

498 25 Mass.

1 Alabama

155

Table C.33. 1993 Flow Values and Percentages Across States and Commodities (Continued) 509 20 Kansas

34

2082

0.9

3.6

510 41 Oregon

34

2020

0.9

3.4

511 22 Louisiana

34

1805

0.8

2.1

512

34

1633

0.7

2.6

513 28 Mississippi 34

1742

0.7

3.6

514 31 Nebraska

34

1737

0.7

4.8

515 49 Utah

34

1358

0.6

4.3

516 32 Nevada

34

1074

0.5

6.9

517 33 New Hamp. 34

1071

0.5

7

518 54 W. Virginia

34

1244

0.5

4.5

519 23 Maine

34

678

0.3

3.6

520 44 Rhode Is

34

618

0.3

3.6

521 46 S. Dakota

34

410

0.2

5.2

522 50 Vermont

34

430

0.2

5.5

523 10 Delaware

34

174

0.1

1.1

524 16 Idaho

34

273

0.1

1.9

525 30 Montana

34

173

0.1

2.2

526 35 New Mexico 34

265

0.1

2.6

527 38 N. Dakota

34

214

0.1

2.9

528 56 Wyoming

34

126

0.1

2.5

529

35 61794

14

10.4

530 48 Texas

35 46901

10.6

11.4

531 17 Illinois

35 33302

7.5

10.9

532 39 Ohio

35 25775

5.8

8.6

533 37 N. Carolina

35 18633

4.2

9.2

534 36 New York

35 17904

4.1

7.5

535 42 Penn

35 17163

3.9

7.6

536 26 Michigan

35 16627

3.8

6.9

537 55 Wisconsin

35 15576

3.5

11.7

538

35 12233

2.8

18.8

539 27 Minnesota

35 11689

2.6

12.3

540 47 Tennessee

35 11232

2.5

7.1

541 12 Florida

35 10729

2.4

7

542 13 Georgia

35 10448

2.4

5.3

4 Arizona

6 California

9 Connect.

156

Table C.33. 1993 Flow Values and Percentages Across States and Commodities (Continued) 543 18 Indiana

35 9504

2.2

5.9

544 34 N. Jersey

35 9515

2.2

4

545 25 Mass.

35 9087

2.1

9

546

8 Colorado

35 7931

1.8

15.1

547 29 Missouri

35 7511

1.7

6.4

548 19 Iowa

35 7199

1.6

10.7

549 21 Kentucky

35 7249

1.6

7.3

550 45 S. Carolina

35 7010

1.6

9

551 51 Virginia

35 6619

1.5

6.6

552 53 Washington 35 6167

1.4

5.4

553

35 5147

1.2

8.3

554 24 Maryland

35 5347

1.2

6.2

555 40 Oklahoma

35 4787

1.1

10.6

556

35 4094

0.9

5.1

557 41 Oregon

35 4084

0.9

6.9

558 20 Kansas

35 3491

0.8

6.1

559 22 Louisiana

35 2995

0.7

3.5

560 28 Mississippi 35 3067

0.7

6.3

561

4 Arizona

1 Alabama

5 Arkansas

35 2728

0.6

5.1

562 31 Nebraska

35 2499

0.6

6.9

563 16 Idaho

35 2152

0.5

14.8

564 49 Utah

35 2398

0.5

7.6

565 33 New Hamp. 35 1949

0.4

12.7

566 35 New Mexico 35 1125

0.3

11.2

567 38 N. Dakota

35 1119

0.3

15.3

568 54 W. Virginia

35 1425

0.3

5.2

569 23 Maine

35

843

0.2

4.5

570 32 Nevada

35

868

0.2

5.6

571 44 Rhode Is

35 1068

0.2

6.2

572 46 S. Dakota

35 1026

0.2

12.9

573 10 Delaware

35

256

0.1

1.7

574 30 Montana

35

224

0.1

2.9

575 50 Vermont

35

463

0.1

5.9

576 56 Wyoming

35

386

0.1

7.5

157

Table C.33. 1993 Flow Values and Percentages Across States and Commodities (Continued) 577

6 California

36 81196

19.7

13.6

578 34 N. Jersey

36 27309

6.6

11.3

579 17 Illinois

36 23523

5.7

7.7

580 48 Texas

36 23017

5.6

5.6

581 39 Ohio

36 21863

5.3

7.3

582 36 New York

36 17964

4.4

7.5

583 18 Indiana

36 15914

3.9

9.9

584 25 Mass.

36 13977

3.4

13.8

585 42 Penn

36 14035

3.4

6.2

586 12 Florida

36 13755

3.3

9

587 47 Tennessee

36 12271

3

7.7

588 37 N. Carolina 36 12046

2.9

5.9

589 13 Georgia

36 11533

2.8

5.8

590 55 Wisconsin

36

9052

2.2

6.8

591 29 Missouri

36

8412

2

7.2

592

4 Arizona

36

7507

1.8

12.1

593 51 Virginia

36

7434

1.8

7.4

594 27 Minnesota

36

7192

1.7

7.6

595 24 Maryland

36

6668

1.6

7.7

596 26 Michigan

36

6490

1.6

2.7

597 21 Kentucky

36

6228

1.5

6.3

598

1 Alabama

36

5229

1.3

6.5

599

8 Colorado

36

5190

1.3

9.9

36

5311

1.3

6.8

601 53 Washington 36

4675

1.1

4.1

602

5 Arkansas

36

4003

1

7.5

603

9 Connect.

36

4249

1

6.5

36

4215

1

6.3

605 28 Mississippi 36

3664

0.9

7.6

606 40 Oklahoma

36

3510

0.9

7.8

607 41 Oregon

36

3323

0.8

5.6

608 33 New Hamp. 36

2970

0.7

19.4

609 20 Kansas

36

2430

0.6

4.2

610 50 Vermont

36

2293

0.6

29.1

600 45 S. Carolina

604 19 Iowa

158

Table C.33. 1993 Flow Values and Percentages Across States and Commodities (Continued) 611 54 W. Virginia

36

2271

0.6

8.3

612 49 Utah

36

2150

0.5

6.8

613 22 Louisiana

36

1615

0.4

1.9

614 31 Nebraska

36

1641

0.4

4.5

615 44 Rhode Is

36

1328

0.3

7.7

616 23 Maine

36

817

0.2

4.3

617 32 Nevada

36

741

0.2

4.7

618 10 Delaware

36

213

0.1

1.4

619 16 Idaho

36

550

0.1

3.8

620 35 New Mexico 36

575

0.1

5.7

621 46 S. Dakota

36

427

0.1

5.4

622 56 Wyoming

36

324

0.1

6.3

623 30 Montana

36

119

0

1.5

624 38 N. Dakota

36

172

0

2.3

625 26 Michigan

37 84208

13.2

35

626

37 67297

10.5

11.3

627 39 Ohio

37 61951

9.7

20.7

628 13 Georgia

37 35161

5.5

17.8

629 18 Indiana

37 34401

5.4

21.5

630 53 Washington 37 34510

5.4

30.2

631 17 Illinois

37 30443

4.8

10

632 29 Missouri

37 23930

3.7

20.4

633 48 Texas

37 22786

3.6

5.6

634 21 Kentucky

37 21764

3.4

21.9

635 34 N. Jersey

37 21302

3.3

8.8

636 36 New York

37 19762

3.1

8.3

637 12 Florida

37 17400

2.7

11.4

638 47 Tennessee

37 17331

2.7

10.9

639 24 Maryland

37 15922

2.5

18.4

640 42 Penn

37 13957

2.2

6.2

641 55 Wisconsin

37 11393

1.8

8.5

642

37

9297

1.5

14.3

643 37 N. Carolina

37

8016

1.3

4

644 20 Kansas

37

7989

1.2

13.9

6 California

9 Connect.

159

Table C.33. 1993 Flow Values and Percentages Across States and Commodities (Continued) 645 25 Mass.

37

6951

1.1

6.9

646 51 Virginia

37

7313

1.1

7.3

647

4 Arizona

37

6212

1

10

648

1 Alabama

37

5892

0.9

7.3

649 27 Minnesota

37

5739

0.9

6

650 40 Oklahoma

37

6077

0.9

13.5

651 41 Oregon

37

6081

0.9

10.2

652

37

3848

0.6

7.3

653 19 Iowa

37

3528

0.6

5.3

654 22 Louisiana

37

3684

0.6

4.3

655 45 S. Carolina

37

4097

0.6

5.3

656

37

3114

0.5

5.8

657 28 Mississippi 37

3520

0.5

7.3

658 10 Delaware

37

2760

0.4

18.1

659 49 Utah

37

2697

0.4

8.5

660 31 Nebraska

37

1799

0.3

4.9

661 54 W. Virginia

37

1932

0.3

7.1

662 32 Nevada

37

1142

0.2

7.3

663 16 Idaho

37

327

0.1

2.2

664 23 Maine

37

712

0.1

3.8

665 33 New Hamp. 37

933

0.1

6.1

666 35 New Mexico 37

451

0.1

4.5

667 38 N. Dakota

37

465

0.1

6.3

668 44 Rhode Is

37

894

0.1

5.2

669 50 Vermont

37

400

0.1

5.1

670 30 Montana

37

203

0

2.6

671 46 S. Dakota

37

272

0

3.4

672 56 Wyoming

37

306

0

6

673 36 New York

38 44379

22.3

18.6

674

6 California

38 24754

12.4

4.2

675 34 N. Jersey

38 15273

7.7

6.3

676 17 Illinois

38 13299

6.7

4.4

677 25 Mass.

38

8597

4.3

8.5

678 48 Texas

38

8456

4.2

2.1

8 Colorado

5 Arkansas

160

Table C.33. 1993 Flow Values and Percentages Across States and Commodities (Continued) 679 12 Florida

38 7188

3.6

4.7

680 42 Penn

38 6969

3.5

3.1

681 13 Georgia

38 6716

3.4

3.4

682

8 Colorado

38 5109

2.6

9.7

683

9 Connect.

38 5010

2.5

7.7

684 39 Ohio

38 4680

2.3

1.6

685 27 Minnesota

38 4375

2.2

4.6

686 18 Indiana

38 3925

2

2.5

687 24 Maryland

38 3636

1.8

4.2

688 26 Michigan

38 3376

1.7

1.4

689 37 N. Carolina

38 3442

1.7

1.7

690 47 Tennessee

38 2977

1.5

1.9

691

38 2724

1.4

4.4

692 55 Wisconsin

38 2748

1.4

2.1

693 29 Missouri

38 2639

1.3

2.2

694 51 Virginia

38 2206

1.1

2.2

695 53 Washington 38 1927

1

1.7

4 Arizona

696 41 Oregon

38 1597

0.8

2.7

697 49 Utah

38 1450

0.7

4.6

698 31 Nebraska

38 1284

0.6

3.5

699 45 S. Carolina

38 1282

0.6

1.6

700

38

972

0.5

1.2

701 20 Kansas

38

935

0.5

1.6

702 40 Oklahoma

38 1008

0.5

2.2

703 44 Rhode Is

38

722

0.4

4.2

704

38

624

0.3

1.2

705 19 Iowa

38

694

0.3

1

706 21 Kentucky

38

667

0.3

0.7

707 22 Louisiana

38

564

0.3

0.7

708 33 New Hamp. 38

531

0.3

3.5

709 10 Delaware

38

370

0.2

2.4

710 28 Mississippi 38

433

0.2

0.9

711 35 New Mexico 38

397

0.2

3.9

712 54 W. Virginia

443

0.2

1.6

1 Alabama

5 Arkansas

38

161

Table C.33. 1993 Flow Values and Percentages Across States and Commodities (Continued) 713 23 Maine

38

115

0.1

0.6

714 32 Nevada

38

180

0.1

1.2

715 38 N. Dakota

38

142

0.1

1.9

716 46 S. Dakota

38

225

0.1

2.8

717 16 Idaho

38

51

0

0.4

718 30 Montana

38

91

0

1.2

719 50 Vermont

38

84

0

1.1

720 56 Wyoming

38

89

0

1.7

721

39 22430

13.6

3.8

722 48 Texas

39 12153

7.4

3

723 36 New York

39 11655

7.1

4.9

724 42 Penn

39 10529

6.4

4.7

725 17 Illinois

39

9238

5.6

3

726 34 N. Jersey

39

8492

5.2

3.5

727 12 Florida

39

6755

4.1

4.4

728 39 Ohio

39

6609

4

2.2

729 27 Minnesota

39

5956

3.6

6.3

730 25 Mass.

39

5378

3.3

5.3

731 13 Georgia

39

5231

3.2

2.7

732 53 Washington 39

4282

2.6

3.8

733 29 Missouri

39

4149

2.5

3.5

734

39

4001

2.4

7.6

735 18 Indiana

39

3649

2.2

2.3

736 47 Tennessee

39

3425

2.1

2.2

737 26 Michigan

39

3178

1.9

1.3

738 55 Wisconsin

39

3141

1.9

2.4

739

5 Arkansas

39

2946

1.8

5.5

740 44 Rhode Is

39

2852

1.7

16.6

741 21 Kentucky

39

2615

1.6

2.6

742 37 N. Carolina 39

2552

1.6

1.3

743

39

2382

1.4

3.7

744 49 Utah

39

2156

1.3

6.8

745 24 Maryland

39

1842

1.1

2.1

746 51 Virginia

39

1771

1.1

1.8

6 California

8 Colorado

9 Connect.

162

Table C.33. 1993 Flow Values and Percentages Across States and Commodities (Continued) 747 45 S. Carolina

39

1708

1

2.2

748 20 Kansas

39

1560

0.9

2.7

749

4 Arizona

39

1304

0.8

2.1

750 32 Nevada

39

1172

0.7

7.5

751

39

911

0.6

1.1

752 19 Iowa

39

1022

0.6

1.5

753 40 Oklahoma

39

1004

0.6

2.2

754 41 Oregon

39

995

0.6

1.7

755 31 Nebraska

39

748

0.5

2.1

756 22 Louisiana

39

595

0.4

0.7

757 28 Mississipi

39

611

0.4

1.3

758 54 W. Virginia

39

728

0.4

2.7

759 23 Maine

39

441

0.3

2.3

760 33 New Hamp. 39

576

0.3

3.8

761 10 Delaware

39

314

0.2

2.1

762 16 Idaho

39

252

0.2

1.7

763 35 New Mexico 39

329

0.2

3.3

764 46 S. Dakota

39

389

0.2

4.9

765 50 Vermont

39

385

0.2

4.9

766 30 Montana

39

167

0.1

2.2

767 38 N. Dakota

39

55

0

0.8

768 56 Wyoming

39

12

0

0.2

769

75 47183

10.6

7.9

770 37 N. Carolina

75 46309

10.4

22.9

771 34 N. Jersey

75 38589

8.7

16

772 47 Tennessee

75 39024

8.7

24.6

773 13 Georgia

75 36436

8.2

18.5

774 45 S. Carolina

75 22199

5

28.5

775 39 Ohio

75 19538

4.4

6.5

776 36 New York

75 17188

3.9

7.2

777 48 Texas

75 17441

3.9

4.3

778 42 Penn

75 16286

3.7

7.2

779

75 14138

3.2

22.8

75 11911

2.7

11.8

1 Alabama

6 California

4 Arizona

780 25 Mass.

163

Table C.33. 1993 Flow Values and Percentages Across States and Commodities (Continued) 781

1 Alabama

75 10179 2.3

12.6 10.4

782

51 Virginia

75 10417 2.3

783

18 Indiana

75

8829

2

5.5

784

55 Wisconsin

75

9134

2

6.8

785

12 Florida

75

7625 1.7

5

786

17 Illinois

75

7307 1.6

2.4

787

21 Kentucky

75

5920 1.3

6

788

29 Missouri

75

4990 1.1

4.2

789

23 Maine

75

4556

1

24.2

75

4187 0.9

6.4

790

9 Connect.

791

53 Washington

75

3808 0.9

3.3

792

5 Arkansas

75

3623 0.8

6.8

793

28 Mississippi

75

3787 0.8

7.8

794

19 Iowa

75

3292 0.7

4.9

795

20 Kansas

75

3172 0.7

5.5

796

24 Maryland

75

3318 0.7

3.8

797

26 Michigan

75

3039 0.7

1.3

798

32 Nevada

75

3285 0.7

21

799

41 Oregon

75

2712 0.6

4.6

800

10 Delaware

75

2210 0.5

14.5

801

27 Minnesota

75

2056 0.5

2.2

802

49 Utah

75

2264 0.5

7.2

803

44 Rhode Is

75

1611 0.4

9.4

804

8 Colorado

75

1259 0.3

2.4

805

22 Louisiana

75

1474 0.3

1.7

806

40 Oklahoma

75

1280 0.3

2.8

807

31 Nebraska

75

982 0.2

2.7

808

33 New Hamp.

75

1001 0.2

6.5

809

35 New Mexico

75

485 0.1

4.8

810

50 Vermont

75

483 0.1

6.1

811

54 W. Virginia

75

492 0.1

1.8

812

56 Wyoming

75

537 0.1

10.5

813

16 Idaho

75

122

0

0.8

814

30 Montana

75

81

0

1.1

164

Table C. 34. Three-Digit Breakdown of 1993 Commodity Flows

Com.

Value ($Million) Share

20

855597

1.000

201

162693

0.190

202

77194

0.090

203

162142

0.190

204

73322

0.086

205

40840

0.048

206

8944

0.010

207

26411

0.031

208

134932

0.158

209

131874

0.154

22

99532

1.000

221

10537

0.106

222

8806

0.088

223

337

0.003

224

12966

0.130

225

6609

0.066

227

13989

0.141

228

16893

0.170

229

14475

0.145

23

263205

1.000

231

62852

0.239

233

125404

0.476

235

1499

0.006

237

225

0.001

238

29895

0.114

239

25198

0.096

24

126657

1.000

241

15095

0.119

242

39862

0.315

243

47344

0.374

244

1273

0.010

249

16663

0.132

Com: Commodity

165

Table C. 34.Three-Digit Breakdown of 1993 Commodity Flows (Continued)

25

68705

1.000

251

54056

0.787

253

1379

0.020

254

6154

0.090

259

3005

0.044

26

194066

1.000

261

3108

0.016

262

59799

0.308

263

20920

0.108

264

60633

0.312

265

36267

0.187

266

2730

0.014

28

532909

1.000

281

96086

0.180

282

66372

0.125

283

140165

0.263

284

74197

0.139

285

21866

0.041

286

888

0.002

287

25851

0.049

289

40730

0.076

29

346712

1.000

291

325858

0.940

295

8733

0.025

299

5035

0.015

30

172527

1.000

301

20132

0.117

302

2641

0.015

303

0

0.000

304

5505

0.032

306

12096

0.070

307

102484

0.594

31

44092

1.000

311

2032

0.046

166

Table C. 34.Three-Digit Breakdown of 1993 Commodity Flows (Continued)

312

0

0.000

313

174

0.004

314

29557

0.670

315

138

0.003

316

2128

0.048

319

725

0.016

32

86142

1.000

321

6142

0.071

322

17656

0.205

324

6290

0.073

325

6364

0.074

326

6638

0.077

327

21714

0.252

328

1649

0.019

329

13183

0.153

33

227828

1.000

331

112303

0.493

332

16146

0.071

333

10088

0.044

335

48674

0.214

336

8583

0.038

339

9082

0.040

34

236993

1.000

341

11385

0.048

342

35423

0.149

343

15671

0.066

344

54264

0.229

345

17268

0.073

346

25969

0.110

348

7446

0.031

349

57937

0.244

35

406496

1.000

351

9356

0.023

352

27809

0.068

167

Table C. 34.Three-Digit Breakdown of 1993 Commodity Flows (Continued)

353

56330

0.139

354

30730

0.076

355

27446

0.068

356

47089

0.116

357

136199

0.335

358

23203

0.057

359

14557

0.036

36

408148

1.000

361

33156

0.081

362

36345

0.089

363

34350

0.084

364

31580

0.077

365

52119

0.128

366

38089

0.093

367

122815

0.301

369

20647

0.051

37

625681

1.000

371

455111

0.727

372

90883

0.145

373

12369

0.020

374

2780

0.004

375

2210

0.004

376

1091

0.002

379

10228

0.016

38

183310

1.000

381

14576

0.080

382

10281

0.056

383

3322

0.018

384

59331

0.324

385

4142

0.023

386

66392

0.362

387

4075

0.022

168

Table C. 34.Three-Digit Breakdown of 1993 Commodity Flows (Continued)

39

146934

1.000

391

15734

0.107

393

1945

0.013

394

35635

0.243

395

5880

0.040

396

7164

0.049

399

62816

0.428

169

APPENDIX D LIST OF US. CUSTOM DISTRICTS∗∗

District Code



District Name

01

PORTLAND, MAINE

02

ST. ALBANS, VERMONT

04

BOSTON, MASSACHUSETTS

05

PROVIDENCE, RHODE ISLAND

06

BRIDGEPORT, CONNECTICUT

07

OGDENSBURG, NEW YORK

09

BUFFALO, NEW YORK

10

NEW YORK CITY, NEW YORK

11

PHILADELPHIA, PA.

13

BALTIMORE, MARYLAND

14

NORFOLK, VIRGINIANIA

15

WILMINGTON, N. CAROLINA

16

CHARLESTON, S. CAROLINA

17

SAVANNAH, GEORGIA

18

TAMPA, FLORIDA

19

MOBILE, ALABAMA

20

NEW ORLEANS, LOUISIANA

21

PORT ARTHUR, TEXAS

23

LAREDO, TEXAS

24

EL PASO, TEXAS

25

SAN DIEGO, CALIFORNIA

26

NOGALES, ARIZONA

27

LOS ANGELES, CALIFORNIA

28

SAN FRANCISCO, CALIF.

29

COLUMBIA-SNAKE

30

SEATTLE, WASHINGTON

31

ANCHORAGE, ALASKA

32

HONOLULU, HAWAII

Bureau of the Census, 1993, US export and Imports of Merchandise on CD-ROM, Washington D.C.

170

33

GREAT FALLS, MONTANA

34

PEMBINA, NORTH DAKOTA

35

MINNEAPOLIS, MINNESOTA

36

DULUTH MINNESOTA

37

MILWAUKEE, WISCONSIN

38

DETROIT, MICHIGAN

39

CHICAGO, ILLINOIS

41

CLEVELAND, OHIO

45

ST. LOUIS, MISSOURI

49

SAN JUAN, PUERTO RICO

51

VIRGINIAN ISLANDS OF THE US

52

MIAMI, FLORDIA

53

HOUSTON/GALVESTON, TEXAS

54

WASHINGTON, D.C.

55

DALLAS/FORT WORTH, TEXAS

58

SAVANNAH/WILMINGTON

59

NORFOLK/MOBILE/CHARLESTON

60

VESSELS UNDER OWN POWER

70

LOW VALUE SHIPMENTS

80

MAIL SHIPMENTS

89

ELECTRICITY

90

UNDOCUMENTED EXPORTS TO CANADA

171

APPENDIX E DESCRIPTIONS OF DATABASES

E. 1. 1993 and 1997 Commodity Flow Surveys E. 1. 1. 1993 Commodity Flow Survey The 1993 Commodity Flow Survey (CFS) is produced by the U.S. Department of Commerce, Bureau of the Census, as part of the Economic Census, in partnership with the Bureau of Transportation Statistics of the U.S. Department of Transportation, and sponsored by the U.S. Department of Transportation. The CFS provides data on the movement of goods both for each of the 50 States and the District of Columbia, and for each of the 89 U.S. National Transportation Analysis Regions (NTARs). The NTARs may cross state boundaries, but are made of a set of whole counties. Thus, the structure of the data is suitable to conduct both state and NTAR level analyses. Shipments originating from businesses located in Puerto Rico and other territories, shipments traversing the U.S., and shipments from a foreign location to a U.S. location are not included in the CFS. Imported products shipments are included after they leave the importer’s domestic location for another location.

Export shipments are also

included until they reach the port of exit from the U.S. Shipments through a foreign country, with both the origin and destination in the U.S., are included. However, in the calculation of the mileages for these types of shipments, the foreign segment is not included. Movements of commodities in the CFS are generated by establishments in manufacturing, mining, wholesale, selected retail, service, and auxiliary activities. Sectors such as agriculture, government, and some other retail, that may have significant shipments, are not included. Also, shipments from “Oil and Gas Extraction” are not included since most of the establishments in this sector have undeliverable mailing 172

addresses.

However, agricultural products from initial processing centers onward are

included. The CFS includes 11 data tables, and the same table layout is used for all 11 tables. The table layout includes 20 variables. However, in each of the tables, only specific variables are

recorded.

The Origin variable indicates the place where the commodity flow originates, while the Destgeo variable is the destination of the flows. Both variables are recorded with the same codes. When origin and destination are states, the code is the two-digit census code, (50 states, and Washington D. C.). However, for the NTARs the codes are three-digit (89 NTARs). The Commodity distributed

by

variable

establishments.

represents the items that are produced, sold or They

are

classified

according

to

the

Standard

Transportation Commodity Classification (STCC) system. The data presented in the CFS are aggregated to the two and three-digit levels. A three-digit STCC list is given in Appendix A. The Modeseq and Mode variables are defined together. Both of them uniquely define one mode or a modal combination. For example, the Modeseq

value 12 and the

Mode value 02 together define private truck. Four different levels of mode aggregation are defined. The first is the sum of all modes. The second defines only single modes, multiple modes, and other and unknown modes. The third level sums modal totals: parcel, U.S. postal service, or courier total; truck total; air total; rail total; inland water total; deep sea water total; pipeline total; and other and unknown modes total. The fourth level is the most disaggregated level and includes the following: parcel, U.S. postal service, or courier; private truck; for-hire truck; air; rail; inland water; Great Lakes; deep sea water; pipeline; private truck and for-hire truck; truck and air; truck and water; truck and pipeline; rail and water; inland water and Great Lakes; inland water and deep sea; other and unknown modes. The Strat variable defines either a distance stratum or a weight stratum. In each case 10 strata are defined.

This variable is defined only for tables 3, 4, 10 and 11. In

tables 3 and 11, it denotes distance, while it denotes weight in tables 4 and 10. The other variables are the measurement variables: Value is the value of the commodities shipped in $ million. This is the net selling value, f.o.b. plant, exclusive of freight charges and excise taxes; Weight is the weight of the commodities shipped in million (short) tons (one short ton is 2000 pounds); Wgtdis is the weight-distance of the 173

commodities shipped in million ton-miles; and Avgdis defines average miles per shipment. It should be noted that in several tables distance shipped is given by Table E. 1. Variables included in the CFS Variable Name ORIGIN

Type

Len

Dec

Description

C

3

0

Geographic code for state/NTAR of origin

COMMODITY

C

3

0

Commodity classification code

MODESEQ

C

2

0

Mode of transportation sequencing code

MODE

C

2

0

Mode of transportation code

DESTGEO

C

3

0

Geographic code for state/NTAR of destination

STRAT

C

2

0

Weight or distance stratum code

VALUE

N

9

0

Value of commodities shipped (in $ 1,000,000)

VALUEF

C

1

0

Flag for above field

VALUEPCT

N

5

1

% of total value of commodities shipped

VALUEPCTF

C

1

0

Flag for above field

WEIGHT

N

9

0

Weight of commodities shipped (in 1,000,000 tons)

WEIGHTF

C

1

0

Flag for above field

WEIGHTPCT

N

5

1

% of total weight of commodities shipped

WEIGHTPCTF

C

1

0

Flag for above field

WGTDIS

N

9

0

Weight-distance of commodities shipped(in1,000,000ton-miles)

WGTDISF

C

1

0

Flag for above field

WGTDISPCT

N

5

1

% of total weight-distance of commodities shipped

WGTDISPCTF

C

1

0

Flag for above field

AVGDIS

N

9

0

Average miles per shipment

AVGDISF

C

1

0

Flag for above field

intervals (as the strata mentioned above), and these distances are estimated along the Great Circle between origin and destination. However, except for these cases, mileage calculations were made by Oak Ridge National Laboratories. Valuef, Weightf, Wgtdisf, and Avgdisf denote the flags for above-mentioned variables. Two types of flag are used: (D) denotes figures that were withheld to avoid disclosing data for individual companies; and (S) denotes data that do not meet publication standards due to high sampling variability. Valuepct,

Weightpct,

and

Wgtdispct

define

percentage

distribution

of

measurement variables. The universe of percentage distribution changes according to table specifications. For example in Table 1, the percentage distribution is computed according to transportation modes for each variable and for each origin, while Table 2 defines % distributions for the Wgtdis variable only. Finally, the Valuepctf, Weightpctf, and Wgtdispctf variables denote flags for the above-defined percentage variables, and they are defined in the same manner as the other 174

flags.The CFS includes 11 data tables reporting data with different levels of aggregation and data organizations. Since the study directly uses Table 5, and Table 9 in the 1993 CFS, only these tables are explained below. Table E. 2. Tables in the Commodity Flow Survey Table Title Table_01. dbf

Table Content Mode of transportation for all origins

Table_02. dbf

Total modal activity for all origins

Table_03. dbf

Mode of transportation and distance shipped for all origins

Table_04. dbf

Mode of transportation and shipment size for all origins

Table_05. dbf

Commodity for all origins

Table_06. dbf

Commodity and mode of transportation for all origins

Table_07. dbf

All destinations for all origins

Table_08. dbf

All destinations and modes of transportation for all origins

Table_09. dbf

All destinations and commodities for all origins

Table_10. dbf

Commodity and shipment size for all origins

Table_11. dbf

Commodity and distance shipped for all origins

Table 5 is titled “Commodity for All Origins”. In this table, the values, weights, weight-distances and average distances of commodity shipments (at the two-digit STCC level) going out from each origin (both state and NTAR) are reported. The modes, destinations, and percentages of measurement variables are not reported. In this table, it is specifically possible to monitor the value, weight, weight-distance, and average-distance of, say, forestry products going out from Kansas. There are 4,726 records in Table 5: 1,700 states, and 3,026 NTARs. Table 9 is titled

“All Destinations and Commodities for All Origins”. In Table 9,

Table 7 is further disaggregated by two-digit STCC commodity groups. The values, weights, and weight-distances of two-digit commodity groups going out from each origin to every other destination are provided.

From this table, it is possible to read off the value,

weight, and weight-distance of forestry products going out from Kansas to Alabama. In Table 9, there are 356,014 records: 86,700 states, and 269,314 NTARs. E. 1. 2. 1997 Commodity Flow Survey The 1997 CFS is the second and latest survey in its series. The general data structure, variables, industrial coverage, and geographic coverage have not been changed drastically in the 1997 CFS. For this reason, the content of each table is not analyzed and reported here in detail as for the 1993 CFS. However, there are some notable differences in the 1997 CFS. 175

First of all, the sample size of the survey decreased to 100,000 establishments in 1997 from 200,000 in 1993, out of a universe of about 800,000 establishments. In the 1993 CFS, data are reported by states and by NTARs. However, in 1997, in order to obtain a certain degree of coordination with other Census Bureau statistics, the data are reported for three major geographical bases: Census Regions and Divisions, States, and Major Metropolitan Areas and the Remainder of States. Since we are planning to conduct the research at the state level, this change will not affect our study. Although the industry coverage has not been changed in 1997, the commodity classification system has been changed. The 1993 CFS uses the Standard Transportation Commodity

Classification

(STCC)

system

to

report

data,

whereas

the

Standard

Classification of Transported Goods (SCTG) code is used in 1997. The American Association of Railroads developed the STCC for rate assignments by the Interstate Commerce Commission (ICC) in the 1960’s. The SCTG coding system was created by the US and Canadian Governmental Agencies, and it is based on the Harmonized System (HS) of commodity classification in order to address worldwide logistics requirements.

Again,

since this study will be conducted at the two-digit commodity aggregation level for the two periods (1993 and 1997), this change should not create a significant problem. A list of the SCTG codes at the two-digit level is presented in Appendix B. E. 2. County Business Patterns The County Business Patterns (CBP) data are produced each year by The Bureau of the Census.

CBP uses three levels of geographic aggregation: the US, state, and

county. Data for agricultural production, railroads, much of government, and household employment are not included in CBP, as CBP data are derived from Social Security Administration files. CBP provide data on the total # of establishments, and # of establishments by employment size classes, mid-march total employment, total annual payroll, and total first quarter payroll. The universe of the database is all establishments with one or more paid employees.

176

Table E. 3. Variables in the US Data Files Variables

Type

Width

Description

USIND

C

2

US Summary Code

SICCODE(1)

C

4

SIC Code

TOTFLAG

C

1

Data Suppression Flag- Total

TOTTEMPM

N

8

Total Mid-March Employees

TOTPAYQ1

N

9

Total First Quarter Payroll ($1,000)

TOTPAY

N

10

Total Annual Payroll ($1,000)

TOTTEST

N

8

Total # of Establishments

FLAGX-Y*

C

1

DataSuppression Flag: Establishments Having X-Y Emp

EMPX-Y

N

8

Mid-March Employees: X-Y Employee Size Class

EMPWGX-Y

N

8

First Quarter Wages: X-Y Employee Size Class

EMPANX-Y

N

9

Annual Wages: X-Y Employee Size Class

ESTNUX-Y

N

7

Number of Establishments: X-Y Employee Size Class

X and Y denote lower and upper limits of employee size classes, and 9 size classes are defined for the US file. The intervals are as follows: 1-4; 5-9; 10-19; 20-49; 50-99; 100-249; 250-499; 500-999; 1000+.

Data are reported by 2, 3, and 4 digit SIC level for all geographic aggregations. The name of the variables and their definitions for county, state, and the US tables are presented in Tables 5, 6, and 7, respectively. Table E. 4. V ariables In States Data File. Variables FIPSTATE

Type C

Width 2

Description FIPS State Code

SICCODE1

C

4

SIC Code

TFLAG

C

1

Data Suppression Flag

TFLAG

C

1

Data Suppression Flag

TEMPMM

N

8

Total Mid-March Employees

TPAYQ1

N

8

Total First Quarter Payroll ($1,000)

TANPAY

N

9

Total Annual Payroll ($1,000)

TESTAB

N

6

Total Number of Establishments

EMPXQY*

N

6

Mid-March Employees: X-Y Employee Size Class

FQWGXQY

N

7

First Quarter Wages: X-Y Employee Size Class

AWGXQY

N

8

Annual Wages: X-Y Employee Size Class

ESTNOXQY

N

6

Number of Establishments: X-Y Employee Size Class

FLAGX

C

1

DataSuppression Flag: Establishments Having X-Y Emp.

STATE1B

C

2

Census State Code

X and Y denote lower and upper limits of employee size classes, and 9 size classes are defined for the state files. The intervals are as follows: 1-4; 5-9; 10-19; 20-49; 50-99; 100-249; 250-499; 500-999; 1000+.

177

Table E. 5. Variables In County Data Files. Variables FIPSTATE2

Type C

Width 2

Description State Code

FIPSCTY2

C

3

County Code

SICCODE2

C

4

SIC Code

TFLAG

C

1

Data Suppression Flag

TEMPMM

N

12

Total Mid-March Employees

TPAYQ1

N

12

Total First Quarter Payroll ($1,000)

TANPAY

N

12

Total Annual Payroll ($1,000)

TESTAB

N

6

Total Number of Establishments

CTYEMPL1

N

6

# of Establishments/Employment Size Class 1-4

CTYEMPL2

N

6

# of Establishments/Employment Size Class 5-9

CTYEMPL3

N

6

# of Establishments/Employment Size Class 10-19

CTYEMPL4

N

6

# of Establishments/Employment Size Class 20-49

CTYEMPL5

N

6

# of Establishments/Employment Size Class 50-99

CTYEMPL6

N

6

# of Establishments/Employment Size Class 100-249

CTYEMPL7

N

6

# of Establishments/Employment Size Class 250-499

CTYEMPL8

N

6

# of Establishments/Employment Size Class 500-999

CTYEMPL9

N

6

# of Establishments/Employment Size Class 1000+

CTYEMPL10

N

6

# of Establishments/Employment Size Class10001499

CTYEMPL11

N

6

# of Establishments/Employment Size Class 15002499

CTYEMPL12

N

6

# of Establishments/Employment Size Class 25004999

CTYEMPL13

N

6

# of Establishments/Employment Size Class 5000+

STATE2

N

2

Census State Code

COUNTY2

C

3

Census County Code

As mentioned in Chapter 4, some sectoral employment data are missing in the CBP because of data disclosure problems, and are replaced in the following manner. While there are missing observations at the two-digit level, one-digit level state employment data are all available. Summing up two-digit employment for each state up to one digit level, and

subtracting this sum from the exact one-digit employment, we obtain the total value

of all missing observation. For each one-digit sector, the total missing value is apportioned over the “missing” two-digit sectors using CBP data for earlier or later years. For example, say, in 1993, in state A, sector 20 and 25 employments are suppressed. State A’s given manufacturing employment in 1993 is say

2500. When we

sum given sub-sectors employments, we get 2300. Obviously, sum of the sector 20 and 25 employment is 200 in this case. The question is then how to distribute this sum among these sectors. We search previous or next years database where sectors 20 and 25 are not missing and They are 225 and 75 respectively, say in 1992. Their relative percentage is 0.75 and 0.25 in this case, and we distribute our 200 with these percentages. 178

E. 3. Census of Manufactures Manufacturing

is

defined

as

“the

mechanical

or

chemical

transformation

of

substances or materials into new products. The assembly of component parts of products also is considered to be manufacturing if the resulting product is neither a structure nor other fixed improvement. These activities are usually carried on plants, factories, or mills that characteristically use power-driven machines and materials-handling equipment”* The employees.

universe The

is

all

manufacturing

establishments

are

establishments

classified

in

with

one

manufacturing

or

more

Standard

paid

Industrial

Classification Codes 2011 through 3999. Data contents and aggregation levels vary significantly across files. However, in general, the files contain data on the number of establishments, employment, payroll, value of shipments, value added, capital expenditures, and other statistics for the establishments engaged in manufacturing. The census of manufactures is produced every 5 years, for years ending in 2 and 7. Data for the intervening years are collected in the Annual Survey of Manufactures (ASM). The 1992 Census of Manufactures is published in three report series: the Industry Series, the Geographic Area Series, and the Subject Series. The Final Industry Series report data for the US with two exceptions: File MC92F2, “Industry Statistics by State”, and MC92F6B, “Product Statistics by State”.

Auxiliary

establishments are not included in these records. The File MC92F2, “The Industry Statistics by State” has 11,453 records and may be useful for our study. Data are aggregated by 4 digit SICC, and by state. Geographic Area Series includes both operating establishments and auxiliary establishments data records. Explanations of the data files included in the Geographic Area Series, and their attributes, are presented in Table E. 6. The Subject Series include “the Final General Summary”, and “Manufactures’ Shipments

to

Characteristics

Federal of

Government

Manufacturing

and

Agencies” Wholesale

files.

Additionally,

Establishments

that

“The Export”

Selected file

is

produced for the Analytical Report Series. Only some files of the “Final General Summary” database may be relevant to our study since they are presented at the state level. Explanations of these files and their attributes are presented in Table E. 7.

*

MC92GS.TXT, 1992, in 1992 Economic Census CD-ROM/MC92, Bureau of The Census, Washington D.C..

179

Table E. 6. Geographic Area Series Table Name MC92A1

Title

SICC

Geography

# of Rcrds 1,243

Years

Employment Statistics for States and Metro Areas

All industries

States, MSAs, PMSAs

and

MC92A2

General Statistics for States and Metro Areas

All Industries

States, MSAs, PMSA

and

500

92, 87, 82, 77

MC92A3A

Summary Statistics for States

All Industries

States

52

92

MC92A3B

Assets and Related Statistics for States

All Industries

States

52

92

MC92A4

General Statistics for States, Metro Areas, Counties& Places

2, 3, and 4 Digit

States, MSA, PMSA, Cnt&Plc.

86, 986

92

MC92A9

Establishments by Employment Size for States and Counties

2 Digit

US, States, Counties

69,410

92

92, 87, 82, 77

Table E. 7. Final General Summary Database Relevant Files Table Name MC92SF2

Title

SICC

Geography

Industry Statistics

2, 3 digits

States

# Rcrds 5,886

MC92SA1

Historical Employment

All industries

States

208

92, 87, 82,77

MC92SA2

Historical Statistics for Operating and Auxiliary Establishments

All Industries

States

208

92, 87, 82, 77

MC92SA3A

Detailed Statistics

All Industries

States

52

92

MC92SA3B

Assets

All Industries

States

52

92

MC92SA4

Metropolitan Areas

All Industries

States, MSA. PMSA

861

92

MC92SA5

Employment Size

2 digit

States

1144

92

MC92SM4

Auxiliary Establishments

All Industries

States

52

92

MC92L4

Location Manufacturing

4 digit

States, Cnts, Places

357,475

92

of

180

of

Years 92