How local are IT services markets: Proximity and IT ... - CiteSeerX

How local are IT services markets: Proximity and IT outsourcing

Ashish Arora, Chris Forman1 Carnegie Mellon University [email protected], [email protected] March 2007

Abstract We examine the question of which services are tradable within a concrete setting: the outsourcing of IT services across a broad cross-section of establishments in the US. If markets for IT services are local, then we should expect increases in local supply should increase the likelihood of outsourcing by lowering the cost of outsourcing. If markets are not local then local supply will not affect outsourcing demand. We analyze the outsourcing decisions of a large sample of 99,775 establishments in 2002 and 2004, for two types of IT services: programming and design and hosting. Programming and design projects require communication of detailed user requirements whereas hosting requires less coordination between client and service provider than programming and design. Our empirical results bear out this intuition: The probability of outsourcing programming and design is increasing in the local supply of outsourcing, and this sensitivity to local supply conditions has been increasing over time. This suggests there is some nontradable or “local” component to programming and design services that cannot be easily removed. In contrast, the decision to outsource hosting is sensitive to local supply only for firms for which network uptime and security concerns are particularly acute.

1

Corresponding author. Authors are listed in alphabetical order. We thank the associate editor and two anonymous referees for very helpful comments. We also thank the track co-chairs (Eric Clemons, Rajiv Dewan, and Robert Kauffman) and three anonymous reviewers for the minitrack on Competitive Strategy, Economics, and IS for the 40th Annual Hawaii International Conference on System Sciences. We thank the Alfred P. Sloan Foundation for financial support, and Harte Hanks Market Intelligence for providing essential data. All errors are our own.

1. Introduction The outsourcing and offshoring of services in the US is an important and growing phenomenon that has recently attracted widespread attention. At present, widely varying projections of the number of jobs “at risk” have been presented, mostly by consulting firms [31]. Ultimately, these estimates turn on the extent to which IT services are tradable.2 There have been two prevailing views on the tradability of IT services. One view emphasizes the role of information technology in reducing the costs of performing services at a distance. Under this view, IT reduces the costs of coordinating economic activity over long distances. Proponents of this view argue that all services are potentially tradable. A second view argues that humans work best in physical proximity to one another, and that some face-to-face interaction is required for the execution of many types of services. Proponents of this view argue that offshoring is fraught with hidden costs arising from inexperienced services personnel, differences in language and culture, and time differences between vendor and client site [29]. Though a great deal of case study work has examined offshore project decisions and governance in a variety of situations (e.g., [10]), this is ultimately a question not of what is possible but rather what is predominant. In this paper, we examine the extent to which markets for purchased IT services are local. If some elements of IT services delivery must be supplied locally, then suppliers must be located near customers, and so are not tradable. By contrast, if the markets for IT services are not local, then suppliers need not be collocated, and so IT services are tradable. Our focus is on the tradability of purchased IT services: we do not consider the tradability of services that are performed within the boundaries of the firm. That is, our results do not inform which services can be offshored within a firm. However, at present the vast majority of offshored services are also outsourced, and captive offshoring is usually restricted to large firms that have IT as a central part of their value creation strategy [34]. We examine the IT outsourcing decisions of a large cross-section of establishments in the US

2

In this paper we follow the international trade literature and in particular Jensen and Kletzer (2005) in using the label ‘tradable services’ to refer to those that can be conducted at a distance.

1

during two years: 2002 and 2004. We investigate the extent to which the outsourcing decision depends upon the local supply of outsourcing firms. Our major hypothesis is that if some components of IT services need to be delivered locally, then increases in local supply should increase the likelihood of outsourcing by lowering the cost of outsourcing. We extend this analysis by examining how this relationship has changed over time, how it differs across IT tasks, and how it is moderated by the characteristics of the establishment. We examine the decisions of firms to outsource two types of IT services: programming and design and hosting. Programming and design refers to the decision to outsource programming tasks or the planning and designing of information systems. To be successful, these outsourcing projects typically require communication of detailed user requirements. Hosting involves management and operation of computer and data processing services for the client, as well as hosting of Internet services and web servers. After an initial set-up period, the requirements of such hosting services will be relatively static and will require less coordination between client and service provider than programming and design. As a preview to our results, we show that: 1. The probability of outsourcing programming and design is increasing the local supply of outsourcing. 2. The sensitivity to local supply conditions appears to have increased over time. This suggests that there is some irreducible non-tradable or “local” component to programming and design services that cannot easily be removed through advancement in software development practices or through improvements in communications capabilities engendered by advancements in IT. 3. Outsourcing of hosting services is less sensitive to variance in local supply than is the decision to outsource programming and design, though the sensitivity of the hosting decision has also increased over time. This increase in sensitivity appears to be concentrated among those firms for which network uptime and security concerns are particularly acute. 2. Related Literature 2

This paper is related to three areas of prior research. First, we contribute to recent work that has examined which types of service work are most effectively conducted offshore. Second, we advance work in the IT outsourcing and nascent IT offshoring literatures. Third, we contribute to recent work that seeks to understand the geographic dispersion in the location of high technology industries. We view our research as building upon recent attempts to understand which services are tradable across a broad cross-section of the economy. Jensen and Kletzer [22] examine which services are tradable by examining geographic concentration in economic activity. Tradable industries will be geographically concentrated to take advantage of economies of scale and favorable location factors. By contrast, nontradable industries must locate where demand is and thus must have a geographic distribution that is similar to that of overall economic activity. Our approach is complementary: If a service is easily tradable, demand decisions will not be sensitive to whether the service is locally available (or the extent of its availability). Ono [33] examines manufacturing firm decisions to outsource white-collar services. She examines how the outsourcing decision varies with potential demand as proxied by total population and a demand shifter. Like Ono [33], we examine how the decision to outsource services depends upon local market conditions, however our analysis focuses on identifying which IT services are tradable and we focus on a broader cross-section of industries. We also explicitly model local supply, and treat it as endogenous in the sense of potentially depending upon aggregate local demand. We also contribute to recent field research conducted in other industries that has examined the operational risks of outsourcing services that require intensive coordination or transfer of tacit knowledge between buyer and supplier (e.g., [3], [12], [19]). In contrast to this research that relies on small samples or case studies, we provide systematic evidence on which IT services can most easily be offshored using a broad cross-section of industries in the U.S. economy. Prior research on the IS outsourcing decision has often focused on variation in establishment- or firm-level factors. These could be economic explanations such as the desire of firms to obtain cost advantages through economies of scale or scope (e.g., [2], [25]) or the role of transaction costs on the 3

sourcing decision [2]. Other work has focused on political factors [24] or on strategic responses to institutional influences [1]. To our knowledge, no prior work has addressed our primary question: the conditions under which purchased IT services need to be conducted locally. 3 We also contribute to recent research that has examined the spatial distribution of economic activity in high-technology industries ([7], [23]). Much of this prior literature demonstrates that technology-intensive industries concentrate for one of three reasons: thicker labor markets, the availability of complementary resources, or knowledge spillovers [28]. In our research we provide one explanation for the geographic dispersion in software production in the U.S. 3. Framework and Hypotheses for the Decision to Outsource We examine whether the decision to outsource is increasing in the local supply of outsourcing firms. Prior work in media richness theory (e.g., [15], [16]) suggests that transmission of information that is equivocal, uncertain, or complex is better accomplished through the use of richer media such as face-toface communications. Software development tasks typically require frequent small adjustments among co-workers, particularly in early stages of software development [32]. In other words, insofar as IT outsourcing tasks require transmission of information that is complex or equivocal, or require frequent adjustments among developers, the theory suggests that the costs of outsourcing will be lower when providers of outsourcing services are nearby. It is well established in other nontradable services markets that prices are declining in the number of entrants, so long as labor supply is sufficiently elastic [9]. Thus, since the price of local IT services is decreasing in the number of local suppliers, we expect the likelihood of outsourcing to be increasing in the supply of local IT services firms. Hypothesis 1: Outsourcing is increasing in the supply of local firms, other things equal. We also examine how the sensitivity to local supply depends upon the outsourcing task. Case study research on globally distributed software research suggests that activities that are difficult to 3

We also advance recent work that has investigated the factors influencing the performance of IT outsourcing engagements (e.g., Gopal et al 2003; Gopal and Sivaramakrishnan 2006). In contrast to prior work which has focused on characteristics of the contract or the engagement, we examine how proximity to outsourcing firms influences the perceived value of outsourcing.

4

conduct at a distance include those that require customer interaction, customer proximity, and deep domain knowledge, while those that are easier to conduct at a distance include activities like coding and maintenance, or infrastructure maintenance [11]. In our study we examine the outsourcing of two sets of activities: programming and design and hosting. The decision to outsource programming and design tasks, due to the requirement of transmitting equivocal information (particularly at early stages of the software development lifecycle) and to the need for frequent coordination between supplier and customer, will be more sensitive to changes in local supply than will the decision to outsource hosting services. Hypothesis 2: The decision to outsource programming and design will be more sensitive to variations in local supply than will the decision to outsource hosting. A long line of information systems research has suggested that IT reduces communication and coordination costs within and between firms (e.g., [26], [13], [17], [18]). To examine how IT influences the importance of proximity to suppliers, we examine how the relationship between the decision to outsource and local supply changes over time. If technological improvements in IT are decreasing the coordination costs of sourcing at a distance, then the sensitivity of the outsourcing decision to changes in local supply will decline over time. However, it is also possible that sensitivity to local supply will increase over time. For example, increases in concerns over information security may require increasing proximity for hosting providers if security incidents require immediacy in communications between clients and service providers. Though we state our hypothesis in terms of the sensitivity of local supply increasing over time, we treat it as an empirical question as to which hypothesis is supported by the data. Hypothesis 3: The sensitivity of outsourcing to local supply is increasing over time. In our last set of hypotheses, we examine how the sensitivity to local supply varies based upon the characteristics of the firm. In particular, we examine the effects of variation in firms’ IT infrastructure environment on the relationship between local supply and the decision to outsource hosting. We have

5

conducted a similar set of analyses for programming and design, but we focus on the results for hosting services for the sake of brevity. We examine five classes of application infrastructure within a firm: enterprise applications, functional applications, network applications, security applications, and other applications. For hosting services, one major benefit of proximity is the immediacy of replies in face-to-face communications. This will be particularly valuable for firms with heavy investments in network and security applications. For firms with large investments in network infrastructure, reliability of the network is particularly important and network downtime must be minimized. Network problems, when they occur, must be handled quickly. Firms with large investments in security applications will be particularly concerned with information security issues. When security issues arise, they must be handled quickly, and web servers and firewalls may need to be accessed physically.4 That is, in both cases, immediacy of communication between service provider and client is particularly important. This suggests that proximity to suppliers will be relatively more important for firms with greater investments in security and network infrastructure software. Hypothesis 4a: The relationship between outsourcing hosting and local supply is stronger for firms with greater investments in security software. Hypothesis 4b: The relationship between outsourcing hosting and local supply is stronger for firms with greater investments in network infrastructure software.

4. Framework and Econometric Model We motivate our econometric model with a simple theoretical framework that examines an establishment’s decision to staff IT projects with internal or external IT staff. Establishments face the following maximization problem:

4

This is vividly illustrated in the HBS case study, The iPremier Company: Denial of Service Attack, HBS 9-601114 [6], in which an online retailer under denial of service attack was unable to access the hosting facility remotely and an IT staffer had to physically visit the hosting site to address the issue.

6

max π ( x1 , x2 , z ) − w1 x1 − w2 x2 x1 , x2

where x1 and x2 represent the number of external or internal IT employees hired (respectively), w1 and w2 represent the wage or cost of hiring an additional external or internal worker, and z represents other characteristics that condition profits (and hence, demand for outsourcing). The function π (⋅) represents the value of IT projects. To decide upon the optimal level of outsourcing and IT employment, we take first order conditions:

d π ( x1 , x2 , z ) − w1 = 0 dx1 d π ( x1 , x2 , z ) − w2 = 0 dx2 leading to the following optimal levels of outsourcing and internal IT employment: x1 = f ( w1 , x2 , z )

(1a)

x2 = f ( w2 , x1 , z )

(1b)

The focus of our analysis will be on the optimal level of outsourcing x1. To econometrically estimate the outsourcing decision embedded in these first order conditions in (1a), we must make a number of additional assumptions. First, we assume that the cost of hiring external workers is a linear function of local supply, w1 = g (os) + η , where os represents local supply and η is the error term. This assumption is motivated by the arguments presented prior to hypothesis 1: (1) The provision of some IT services is most effective and less costly when it is performed in close proximity to clients and (2) If markets for goods and services are local, then their price is declining in local supply [9]. While we recognize that wages may depend on factors other than local supply, these other factors will be included in the error term η . Indeed, to the degree that markets are not local, local supply will not matter for the outsourcing decision. Thus, this specification merely makes concrete our hypotheses. We do not observe the true quantity of employees outsourced, but only a binary variable indicating whether outsourcing had been used. Thus, the number of outsourcing employees hired will be a latent variable x1* . Thus, the decision we observe for establishment i will be

7

x1*i = f (osi + ηi , x2i , zi ,ν i ) . Assuming that x1i* is linear in parameters gives us x1*i = α osi + β x2i + γ zi + ηi + ν i

(2)

We assume that ηi and ν i are distributed normally and we estimate a linear probability model of the decision to outsource. Our major interest is in testing whether α > 0 , that is, whether the decision to outsource is increasing in local supply. Of course, as (2) is a cross-sectional regression, one may be concerned that osi may be correlated with unobserved location-specific factors ηi that increase the likelihood of outsourcing. For example, outsourcing firms may prefer to locate in places with a more highly skilled workforce, which may also lower the costs of outsourcing. In this case, estimates of α will be inconsistent. Further, x2 may be correlated with unobservables that increase the value of outsourcing at an establishment. To address these issues, we use instrumental variable (IV) techniques, as described later.

5. Data

The data we use come from the Harte Hanks Market Intelligence CI Technology database (hereafter CI database). The CI database contains establishment-level data on (1) establishment characteristics, such as number of employees, industry and location; (2) use of technology hardware and software, such as computers, networking equipment, printers and other office equipment; and (3) use of outsourcing. Harte Hanks collects this information to resell as a tool for the marketing divisions at technology companies. We obtained data from the CI database for two years: 2002 and 2004. Interview teams survey establishments throughout the calendar year; our sample contains the most current information as of December 2002 and 2004 respectively. We focus on the establishment rather than the firm as the unit of analysis because establishmentlevel data will enable us to more precisely measure the impact of changes in local supply on the costs of outsourcing. Most software investment decisions in our data are made at the establishment level. For

8

instance, 80% of the establishments in our sample that responded to the question stated that IT investment decisions were made at the establishment rather than the firm level. Moreover, we control for multiestablishment firms in our estimation. Our sample from the CI database contains all commercial establishments with over 100 employees: there were 91,129 such establishments in 2002 and 89,776 in 2004. We dropped establishments which did not answer questions on outsourcing or had other crucial data missing, which resulted in samples of 52,191 establishments in 2002 and 47,584 in 2004. In general, excluded establishments were smaller and had fewer investments in IT. Thus, establishments with relatively heavy investments in IT are over-represented in our sample. However, since these are more likely to be able to reach out beyond local suppliers, our estimates are likely lower bounds on the true effects. 5.1 Identifying Decisions to Outsource

Our dependent variable is x1i, the extent of outsourcing by establishment i. This variable x1i is latent. We observe only discrete choices: whether or not the establishment chooses to outsource a particular service or not, with the observed decision assuming a value of either one or zero, respectively. Harte Hanks tracks 20 separate binary measures of outsourcing services that an establishment may use. We aggregate these 20 different outsourcing services into two categories that have similar production technologies. These two categories will comprise the dependent variables for our baseline model. Later we report on robustness checks in which we explore alternative classifications. The first variable measures an establishment’s decision to outsource programming or design services. An establishment is considered to have outsourced programming and design if it answers yes to

outsourcing any of the following services: application design; contract programming; application development; package software implementation; or Internet/web application development. The second variable measures an establishment’s decision to outsource hosting services. An establishment is considered to have outsourced hosting services if it answers yes to outsourcing any of the following services: Internet/web server hosting; Internet routers; web site management; Internet firewalls; LAN client/server; LAN network management; or LAN maintenance. 9

5.2 Independent Variables

Summary statistics on the independent variables are included in Table 1. Measures of local supply osi are calculated from County Business Patterns data. Our measure of the local supply of programming and design establishments is equal to the log of one plus the number establishments in North American Industry Classification System (NAICS) codes 541511 and 541512.5 Our measure of the local supply of hosting establishments is equal to the log of one plus the total number of establishments involved in computer facilities management, hosting, and data processing services (NAICS 541513, 514210, and 518210).6 We use two different measures of the change in internal IT services ( x2i ), depending upon the measure of outsourcing. When x1i measures outsourcing of programming and design, then x2i measures changes in the number of programmers at the establishment over the prior two years.7 When x1i measures outsourcing of hosting services, then x2i measures changes in the number of non-PC servers at the establishment. Controls: We include as additional controls in our regressions industry fixed effects (three-digit NAICS dummies), the log of establishment employment, a dummy indicating that the establishment comes from a multi-establishment firm, the number of PCs per employee and the number of non-PC servers per employee. 5.3 Instruments 5

According to the Census bureau, NAICS code 541511 “comprises establishments primarily engaged in writing, modifying, testing, and supporting software to meet the needs of a particular customer.” NAICS code 541512 “comprises establishments primarily engaged in planning and designing computer systems that integrate computer hardware, software, and communication technologies.” 6 NAICS 541513 “comprises establishments primarily engaged in providing on-site management and operation of clients' computer systems and/or data processing facilities.” NAICS 514210 and 518210 both refer to data processing, hosting, and related services. The NAICS underwent a revision in 2002, so NAICS 514210 was relabeled 518210. NAICS 518210 “comprises establishments primarily engaged in providing infrastructure for hosting or data processing services. These establishments may provide specialized hosting activities, such as web hosting, streaming services or application hosting, provide application service provisioning, or may provide general time-share mainframe facilities to clients.” 7 For example, for our 2002 analyses, we examine the change in programmers between 2000 and 2002. In fact, number of programmers is measured as a categorical variable with the following ranges: 1-4; 5-9; 10-24; 25-49; 5099; 100-249; 250-499; 500-999; and 1000 or more. We use the midpoint of each interval and then calculate the change between 2000 and 2002.

10

We use six sets of instruments for osi . Changes in each of these variables will proxy for geographic variation in the local demand for IT services. By capturing the determinants of local IT services demand, they will influence entry of IT services providers and be correlated with the local supply of programming and design, and hosting services. However, these variables capture characteristics of the establishments in a geographic location that will influence the demand for outsourcing; since we control for similar characteristics at the establishment level, it is unlikely that these variables will be correlated with unobservables influencing establishment outsourcing demand. We use the log of county employment from CBP to measure the aggregate potential market size in the county. When calculating county employment—as in all of the independent variables in this section— we exclude establishments in the three-digit NAICS industry 541 (Professional, Scientific, and Technical Services).8 As is well known, industries differ substantially both in the extent of their IT use and in their geographic distribution [18]. Locations with a higher percentage of IT-intensive firms may have greater average demand for IT outsourcing services. We use three sets of measures to capture differences in the IT intensity of firms across counties in our sample. The IT-intensity index captures how differences in industry mix will affect the demand for

outsourcing. It is calculated by first identifying each industry’s use of IT services as a fraction of total inputs using BEA Input-Output Benchmark Tables for 1997. To calculate IT intensity for county l, these fractions are weighted by industry employment. Thus, for each county l and industry m, IT − INTENSITYl = ∑ m

(industry m spending on IT services) (total county l employment in industry m) (industry m total spending on inputs) (total county l employment)

Next we control for the percentage of IT establishments – establishment in the county that are in industries that are involved in the production of IT. We follow the classification developed by the Department of Commerce as described by Cooke [14].

8

NAICS 541 includes 541511-541513 plus the “other category” 541519. We exclude the entire three-digit category because in our calculation of IT-intensity it was impossible for us to identify the relevant six-digit industry in the BEA input-output tables.

11

Last, we control for the percentage of establishments in different industry groupings: Manufacturing (NAICS 321-339); Wholesale and Retail Trade (NAICS 421-454); Finance, Insurance, and Real Estate (NAICS 521-525); Information Processing (NAICS 511-514 and 551); and Other Services (NAICS 561-814).9 We next examine how the size of establishments in a location affects the entry of suppliers. As noted by earlier studies [2], large firms may achieve economies of scale without the use of IT services firms. Thus, controlling for market size, the supply of IT services firms will decrease as the size of establishments increase relative to the industry average. To measure how variation in the establishment size distribution affects industry employment, we construct the following index of establishment size:

(total employment in county l and industry m) (total establishments in county l and industry m) (total county l employment in industry m) ESTSIZEl = ∑ (total employment in industry m in US) (total county l employment) m (total establishments in industry m in US) . Entry of outsourcing firms may also be influenced by cost differences across locations. One thing that may influence the costs of outsourcing firms may be the availability of highly skilled workers. To proxy for this, we also include the log of total enrollment from post-secondary colleges and universities in the county, obtain from Barron’s Educational Series. We also instrument for x2i since it may be correlated with establishment-specific unobservables that influence the likelihood of outsourcing. As an instrument for changes in the number of programmers, we calculate the change in programmers in other firms in the same 2-digit NAICS industry in other locations that the firm has an establishment. The instruments pick up factor changes in industry specific demand (but not location specific demand) and should be correlated with an establishment’s change in programmers but not with the propensity of the establishment to outsource, conditional on its industry. We instrument for changes in the number of servers using this variable plus changes in the number of 9

The excluded category includes Mining (211-213), Utilities (221), and Construction (233-235, and Transportation and Warehousing (481-493).

12

servers in other firms in the same 2-digit NAICS industry in other locations that the firm has an establishment. We include both variables as instruments.

6. Results

We begin by examining the geographic variation in the supply of IT services firms. Figures 1 and 2 show the number of programming and design and hosting establishments across the counties of the contiguous 48 states. Figure 1 shows that programming and design establishments are widely dispersed across the U.S. 62.4% of counties have at least one programming and design establishment in them, and the distribution of establishments broadly follows the density of economic activity in the U.S. Large cities and their surrounding communities have a large supply of such establishments, while isolated counties in the Midwest and West usually do not. The (Pearson) correlation coefficient between county employment and number of programming and design establishments is 0.8930, while the Spearman rank correlation is 0.8302. Figure 2 shows that while the supply of hosting establishments is also greater in locations with more IT outsourcing demand, the number of hosting establishments appears to be more concentrated in a relatively small number of counties. Only 46.7% of counties have at least one hosting and management establishment in them. The Pearson correlation coefficient between number of hosting and management establishments and total county employment is 0.9462; however the Spearman rank coefficient is 0.7635. The high Pearson correlation is high partly because of the large number of hosting establishments in large cities. Moreover, the map shows that although hosting establishments are generally located in the largest cities, many counties in the surrounding metropolitan areas have few or no hosting establishments. 6.1 Baseline Results

We first examine the relationship between local supply and the likelihood of outsourcing in 2002. Table 2 displays our 2002 baseline results for the association between local supply and the outsourcing of programming, design, and hosting services. Columns (1) and (2) show results without instrumental variables; columns (3) and (4) show the results of instrumenting for local supply but not for changes in 13

internal programmers and servers; and columns (5) and (6) show the full specification with instruments for local supply and changes in internal programmers and servers. Our first stage regressions show that our instruments explain a substantial fraction of the variation in our endogenous supply variables. The F-statistic test that the all of the coefficient estimates for the instruments are equal to zero for the supply of programming and design regression is 52561.48 and the comparable F-statistic for the hosting and management regression is 43103.52; both of these statistics reject the null at the one percent level of significance.10 However, our instruments for changes in programmers and servers are weak: The null hypothesis that all of the coefficient estimates for the instruments are equal to zero cannot be rejected for either programming and design (F-statistic 0.74; pvalue 0.7031) or for hosting (F-statistic 0.5777; p-value 0.87). However, the estimates for the regressions including and excluding the change in internal services (x2) are very similar, as a comparison of columns 3 and 4 with columns 5 and 6 shows. As a result, we will focus on the results for instrumenting supply only (columns 3 and 4). However, all results on the relationship between outsourcing and local supply are robust to the inclusion of the x2 variables, both with and without the use of instruments. The results show that increases in the local supply of programming and design establishments increases the likelihood of outsourcing those services, while increases in the local supply of hosting establishments does not increase the likelihood of outsourcing hosting. This is true regardless of the extent to which instrumental variables are used. Columns (1), (3), and (5) show that increases in the local supply of programming and design firms have a statistically significant impact (at the 1% level) on the decision to outsource those services. The results in column (3) imply that a one standard deviation increase in the log of local programming and design establishments increases the probability of outsourcing programming and design by 1.0 percentage points; this is significant when compared to a mean probability of outsourcing of 26.0%. In contrast, increases in the local supply of hosting establishments have no statistically significant impact on the decision to outsource hosting services. Thus, Table 2 provides support for hypothesis 1 for programming and design services, and also provides 10

These statistics are from columns 5 and 6; F-statistics from columns 3 and 4 are similar.

14

support for hypothesis 2 (programming and design outsourcing is more sensitive to variations in local supply than is hosting outsourcing). The R-squared values are small in all of these models, ranging from 0.0193 to 0.0228. The reason is that our dependent variable is a binary rather than a continuous measure of the extent of outsourcing. As has been shown by previous authors, it is not uncommon to have R-squared values as low as these in limited dependent variable models [5]. Columns (7) and (8) estimate a simultaneous equations model of the decision to outsource programming and design and hosting services. In this model, we use an aggregate measure of local supply that combines programming, design, and hosting establishments, and then instrument for this using the same set of variables as those in columns (3) and (4). This model allows us to statistically test for a difference in the coefficient estimates across the two models. The results are qualitatively similar to those in columns (3) and (4); in particular, they show that the coefficient estimate for the supply of programming and design is greater than that for hosting at the 5% significance level. In Table 3 we estimate the same set of models using 2004 data to examine how the importance of proximity to local supply has been changing over time. There is no evidence in our data that the importance of proximity is declining over time, at least over this short time period. If anything, proximity to local suppliers has become more important to the decision to source externally. In other words, our results support hypothesis 3. All of the coefficients for local supply variables are larger in Table 3 than the comparable estimates in Table 2. The results in column (3) suggest local programming and design supply has a statistically significant impact on the decision to outsource programming and design (at the 1% level) and that a one standard deviation increase in the log of local programming and design establishments increases the likelihood of outsourcing those services by 1.3 percentage points or by 6.3%: this compares to a 1.0 percentage point increase for an identical increase in supply in 2002. Column (4) also shows that in 2004 the decision to outsource hosting services was sensitive to variations in local supply; this is in contrast to our 2002 results which showed that local supply played little role on the decision to outsource hosting. Increases in local hosting establishments has a statistically 15

significant impact on the decision to outsource hosting (at the 1% level), however the association is somewhat weaker than that for programming and design. A one standard deviation increase in the number of hosting establishments is associated with an increase in the likelihood of hosting of 0.9 percentage points or about 4.0%. In columns (7) and (8) we show the results of a simultaneous equations model of the joint decision to outsource these two sets of activities; these results confirm that the coefficient estimate for programming and design is larger than that for hosting at the 10% level. Thus, despite greater sensitivity to local supply in 2004, the decision to outsource hosting remains less sensitive to local supply conditions than the decision to outsource programming and design. 6.2 Robustness Checks

We conducted a variety of robustness checks; the results of all of these are shown in Table 4. We examined the robustness of our results to alternative measures of our dependent variables: in particular, we re-estimated the model separating programming from design and separating Internet hosting from hosting that excluded Internet applications. Since the design phase of the systems development lifecycle may require greater face-to-face communication than programming to establish specifications for the IT project, we may expect greater sensitivity to local supply. This is exactly what we find. We conducted other robustness checks as well. We reran our models where we summed IT services rather than using a binary measure of outsourcing. We also re-estimated the model using the log of total employment instead of the number of establishments as our measure of local supply. The results remained qualitatively similar. 6.3 IT Infrastructure and the Importance of Proximity

In this section we examine the extent to which our data support Hypotheses 4a and 4b. We classified a firm’s application portfolio into five categories: security applications, functional applications, enterprise applications, network applications, and other applications. This was based upon prior work by Bresnahan and Greenstein [8], who classified applications in the CI database into scientific and manufacturing software (which we group together and label functional applications); standard business applications software and database software (which we group together and label enterprise applications); communications and networking software (which we label network applications); and system software 16

and utilities (which we include in our other category). It is also related to recent work by McAfee [30], who groups IT into enterprise IT, network IT, and function IT. We add security applications as a separate category because of the potential importance of vendor proximity for clients for whom security is a major concern. We generate a dummy variable for each of these five categories.11 Using these variables, we estimate the following model: x1*i = α osi + β x2i + ∑ δ j osi × app j + γ zi + ηi + ν i j∈C

where appj indicates a dummy variable for the jth application, where j ∈ C ≡ {Security, Enterprise, Network, Functional} . Table 7 shows the results of including these variables in our outsourcing regression and interacting them with our measures of local supply. To eliminate potential biases engendered by unobserved differences across establishments that do and do not report their use of software applications, this sample includes only establishments that have reported software applications and as a result the sample size is lower than that in earlier tables (to 35,082 and 29,664 in 2002 and 2004, respectively). Columns (1) and (2) show the baseline results for hosting in 2002 and 2004. The results in column (1) show that for hosting outsourcing, the interaction between local supply and enterprise applications is positive and significant at the one percent level in 2002. However, the direct 11

In brief, our dummy variable for security applications was equal to one if and only if the establishment reported the use of security/encryption software. Our dummy variable for functional applications was one when the establishment reported one of the following applications: CAD/CAM/CAE; customer/salesforce management; data analysis & simulation; office automation; spreadsheet; suites; engineering; imaging; project management; and science/research. Our dummy variable for enterprise applications was equal to one when the establishment reported use of one of the following applications: accounting; business intelligence; database management; e-commerce software; ERP; general business software; MRP; payroll; personnel management; TP monitors; customer service; decision support; financial management; inventory; marketing; order administration; personnel; sales/marketing; web site database; data warehouse; and knowledge management. Our dummy variable for network applications was equal to one when the establishment reported one of the following: application server software; electronic mail; groupware software; web browser; web design tools; web development tools; homepage; intranet; communication control; network control; terminal emulation; and web server software. Our dummy variable for other application was equal to one when the establishment reported one of the following: application development; interactive programming; program testing/debugging; programming utilities; backup & recovery; file management; industry specific software; job scheduling; messaging middleware; micro link; performance measurement; report generation/management; store management; system utilities; system/software management; education; government; help desk; resource management. We also experimented with using the percentage of applications in each of these five categories, however the mass of the distribution is concentrated at zero and one for each of these applications. These results are qualitatively similar to the ones presented in the paper.

17

effect of local supply is negative and significant, so that the combined direct effect and interaction term is never significantly positively associated with an increase in the outsourcing probability. Because of the weak relationship between outsourcing supply and demand for hosting in 2002, we interpret these results as our inability to measure jointly the effects of variance in the direct effect and interaction term. In 2004, the direct effect of local supply on hosting is approximately zero and statistically insignificant, but the coefficient estimate of the interaction term with of network applications is positive (0.0068) and statistically significant at the 5% level, while that with security applications (0.0057) is marginally significant with a p-value of 10.2%. The sum of the parameter estimates for the direct effect of local supply with the interaction term for network applications is significant at the 5% level, though the sum of the parameter estimate for local supply and the interaction term for security applications is not significant at conventional levels. We interpret these results as supportive for Hypothesis 4a and weakly supportive for hypothesis 4b. More broadly, these analyses suggest that hosting’s increase in sensitivity to local supply is driven primarily by establishments for whom network uptime and security is a priority. Columns 3 and 4 show the results of a robustness check that uses an alternative classification of applications.12 The results remain qualitatively the same. 7. Discussion and Conclusions 7.1 Findings

In this paper we have examined the geographic variation in supply and the decision to outsource two types of outsourcing services: programming and design and hosting. Differences in the characteristics of these services and the manner in which they are supplied has led to substantial differences in their geographic dispersion and, in turn, their tradability. Table 6 lists our primary findings. We find some support for all of our major hypotheses. For one class of services, programming and design, establishment decisions to outsource are significantly influenced by the magnitude of local supply. These results suggest that markets for

12

In particular, the following enterprise applications are reclassified as functional applications: accounting, payroll, personnel management, marketing, and inventory.

18

programming and design services are at least partly local. In contrast, micro-level outsourcing decisions for hosting services are less sensitive to the characteristics of the local market. Our results showed that the decision to outsource hosting services was not sensitive at all to changes in local supply in 2002. We did find evidence on the effects of local supply in 2004; however this appeared to be most important for firms for which network uptime and information security were particular concerns. A long line of information systems research has suggested that IT reduces communication and coordination costs ([26], [13]). Some authors have argued that the coordination benefits of IT will lead to increases in offshoring [35]. We find no evidence that IT has significantly eroded the importance of proximity. Despite substantial progress in information technology, the importance of proximity has not decreased between 2004 and 2002, and remains substantial. Moreover, those establishments that had the deepest investments in networking technologies that would reduce communication costs were also the ones for which proximity was most important. 7.2 Implications

Going back to March and Simon [27], scholars have understood the challenges of software development in a distributed environment. These challenges have led to the emergence of a literature on globally distributed software development. However, heretofore there has been relatively little systematic empirical evidence on the conditions under which IT services are tradable. Our research contributes to the existing literature by demonstrating how proximity to users is also an important determinant of the costs of outsourcing. Moreover, in contrast to prior work which has focused on case studies, we demonstrate this in a broader environment than has previously been possible. Our work also contributes to the outsourcing literature. In contrast to much of the literature on the decision to outsource, which focuses on firm characteristics that lead to variance in the costs or benefits to outsourcing [25], or the influence of political or institutional factors within the firm or between contracting parties ([24], [1]), we focus on the role of proximity to suppliers. Our research also contributes to understanding the economic geography of the software industry in the US. Prior work has demonstrated that the US software industry is less concentrated that other 19

technology industries such as computing hardware [23]. Our research shows that this pattern may be explained in part by the need for proximity to users. In short, this work shows that there may be some irreducible component to software development and system design that cannot take place at a distance from users. This suggests that it may be some time before certain types of IT activities can be economically offshored; at the very least, it is consistent with the finding that offshore outsourcing firms tend also to have a substantial domestic presence in the United States. More speculatively, it suggests that some complex programming and design tasks are not easily partitioned. This has implications for the location of inventive activity in software development, potentially implying that some difficult design and development tasks cannot be produced away from the United States, where software development activity is centered. Arora, Forman, and Yoon [4] provide further evidence on this latter conjecture in their recent study on the globalization of software R&D activity. These results have implications for understanding how trends in outsourcing and offshoring will influence US employment growth. However, in contrast to prior work which has answered this question by examining the concentration of broad three-digit industries [22], we examine the tradability of narrower industry activities based upon the nature of the activity and systematic evidence on demand. While some tasks in programming and design can undoubtedly be conducted at a distance, these results suggest that providers of such services must maintain some local presence. Our research also provides guidance to managers on which IT services tasks can be most easily conducted at a distance, and suggests environments in which offshoring may be particularly problematic. In particular, our results suggest that hosting offshoring is particularly problematic in environments with complex network infrastructure, and when security concerns are an issue. 7.3 Limitations and Future Research

Our focus is on outsourcing rather than offshoring. The two are related but distinct. Offshoring implies that the activity takes place offshore, but may be carried out by the firm itself or its foreign subsidiaries. Outsourcing implies that the activity is carried out by another firm, be it nearby or offshore. 20

It is possible that there are subtle interactions between need for proximity and contracting across firm boundaries so that the potential for offshoring may be greater than that implied by our results for programming and design services. However, the large number of programmers stationed by such vendors near their customers (witness the ongoing uproar about the use of H1-B visas by IT firms) supports our findings that there is a significant need for proximity in some (though not all) aspects of software design and development. It is possible that this need may be satisfied by foreign programmers being moved to be close to the clients; the fact remains that the activity takes place locally. Nonetheless, further work is needed to understand whether the need for proximity in internal offshoring transactions is the same as that for outsourced offshoring. Another limitation to our study is that in our baseline analysis the dependent variable is binary: we observe only the extensive margin of whether firms choose to outsource. Moreover, by necessity we aggregate different classes of outsourcing services into two broad categories. If firms have a distribution of IT projects that vary in the extent to which they can be conducted at a distance, our method will systematically undervalue the importance of proximity to the outsourcing decision. Though we conducted a robustness check that showed the impact of proximity on number of service types outsourced, we are unable to measure variations in the intensive margin of outsourcing within a service category that are due to differences in proximity. Further work that uses the dollar value of IT outsourcing would be better able to quantify the value of proximity. Current work in offshoring [10] has suggested that some firms are systematically more capable in offshoring than others. Are these less capable firms systematically more reliant on proximity to suppliers? What makes a firm more or less capable? Further work is also needed to understand which firms and outsourcing tasks most require proximity. Our paper presents a methodology for identifying tradable and non-tradable services that can be useful outside of an IT setting. Use of this method could be useful in identifying which positions are most at risk for being moved to alternate locations. Moreover, this method could also be useful for identifying whether the set of positions at risk is changing over time, due to improvements in outsourcing practices, 21

technological change in IT that may reduce the coordination costs associated with distance [18], or some other reason.

22

References 1. Ang, S., and Cummings, L.L. Strategic Response to Institutional Influences on Information Systems Outsourcing. Organization Science, 8, 3 (May-June 1997), 235-256. 2. Ang, S., and Straub, D. Production and Transaction Economies and IS Outsourcing: A Study of the U.S. Banking Industry. MIS Quarterly, 22, 4 (December 1998), 535-552. 3. Aron, R., and Liu, Y. Determinants of Operational Risk in Global Sourcing of Financial Services: Evidence from Field Research. In S.M. Collins and L. Brainard (eds.), Brookings Trade Forum 2005: Offshoring White-Collar Work. Washington, DC: Brookings Institution Press, 2005, pp. 373-398. 4. Arora, A.; Forman, C.; and Yoon, J. Software. In J. Macher and D. C. Mowery (eds.) Globalization of Research. Washington, DC: National Academy of Sciences Press (tentative title), 2007. 5. Athey, S. and Stern, S. The impact of information technology on emergency health care outcomes. RAND Journal of Economics, 33, 3 (Autumn 2002), 399-432. 6. Austin, R. D. The iPremier Company (A): Denial of Service Attack. Harvard Business School Case 9-601-114. 7. Beardsell, M., and Henderson, V. Spatial evolution of the computer industry in the USA. European Economic Review, 43, 2 (February 1999), 431-456. 8. Bresnahan, T., and Greenstein, S. Technical Progress in Computing and in the Uses of Computers. Brookings Papers on Economic Activity, Microeconomics (1996), 1-78. 9. Bresnahan, T. and Reiss, P. Entry and Competition in Concentrated Markets. Journal of Political Economy, 99, 5 (October 1991), 977-1009. 10. Carmel, E., and Agarwal, R. The maturation of offshore sourcing of information technology work. MIS Quarterly Executive 1, 2 (June 2002), 65-76. 11. Carmel, E., and Tija, P. Offshoring Information Technology: Sourcing and Outsourcing to a Global Workforce. Cambridge: Cambridge University Press, 2005. 12. Clemons, E. and Aron, R. Maximize Your Outsourcing Benefits Through Competitive Arbitrage. Working Paper, Wharton School of Business, University of Pennsylvania, 2004. 13. Clemons, E.; Reddi, S. P.; and Row, M. The impact of information technology on the organization of production: The “move to the middle” hypothesis. Journal of Management Information Systems 10, 2 (Fall 1993), 9-35. 14. Cooke, S. Information technology workers in the digital economy. In Digital Economy 2003, Chapter II. Washington, DC: Department of Commerce. www.esa.doc.gov/DigitalEconomy2003.cfm. 15. Daft, R.L., and Lengel, R.H. Information richness: A new approach to managerial behavior and organization design. In B. Staw and L. Cummings (eds.) Research in Organizational Behavior Greenwich, CT: JAI Press, 1986, pp. 191-233. 16. Daft, R.L.; Lengel, R.H. ; and Trevino, L. Message Equivocality, Media Selection, and Manager 23

Performance. MIS Quarterly 11, 3 (September 1987), 355-366. 17. Forman, C. The Corporate Digital Divide: Determinants of Internet Adoption. Management Science 51, 4 (April 2005), 641-654. 18. Forman, C.; Goldfarb, A.; and Greenstein, S. How Did Location Affect Adoption of the Commercial Internet? Global Village v. Urban Leadership. Journal of Urban Economics 58 (November 2005), 389-420. 19. Gopal, A.; Mukhopadhyay, T.; and Krishnan, M.S. The Role of Software Processes and Communication in Offshore Software Development. Communications of the ACM 45, 4 (April 2002), 193-200. 20. Gopal, A.; Sivaramakrishnan, K.; Krishnan, M.S.; and Mukhopadhyay, T. Contracts in Offshore Software Development: An Empirical Analysis. Management Science 49, 12 (December 2003), 16711683. 21. Gopal, A., and Sivaramakrishnan, K. On Vendor Preferences for Contract Types in Offshore Projects: The Case of Fixed Price versus Time and Materials Contracts. Working Paper, Smith School of Business, University of Maryland, 2006. 22. Jensen, J. B., and Kletzer, L. G. Tradable Services: Understanding the Scope and Impact of Services Offshoring. In S.M. Collins and L. Brainard (eds.), Brookings Trade Forum 2005: Offshoring White-Collar Work. Washington, DC: Brookings Institution Press, 2005, pp. 75-116. 23. Kauffman, R. and Kumar, A. A Combined Scale-and-Scope Theory of IT Industry Cluster Growth. Working Paper, Carlson School of Management, University of Minnesota, 2006. 24. Lacity, M. C., and Hirschheim, R. Information Systems Outsourcing: Myths, Metaphors, and Realities. New York: John Wiley & Sons, 1993. 25. Loh, L., and Venkatraman, N. Determinants of Information Technology Outsourcing: A CrossSectional Analysis. Journal of Management Information Systems 9,1 (Summer 1992), 7-24. 26. Malone, T.; Yates, J.; and Benjamin, R. Electronic Markets and Electronic Hierarchies. Communications of the ACM. 30, 6 (June 1987), 484-497. 27. March, J. G., and Simon, H.A. Organizations. New York: Wiley, 1958. 28. Marshall, A. Principles of Economics. London: Macmillan, 1920. 29. Matloff, N. Offshoring: What Can Go Wrong? IT Professional 7, 4 (July | August 2005), 39-45. 30. McAfee, A. Mastering the Three Worlds of Information Technology. Harvard Business Review Reprint R0611J (2006), 1-9. 31. McCarthy, J. C. 3.3 Million US Services Jobs To Go Offshore. TechStrategyTM Research, Forrester Research (November), 2002. 32. Mintzberg, H. Structure in Fives: Designing Effective Organizations. Englewood Cliffs, NJ: Prentice Hall, 1993. 24

33. Ono, Y. Outsourcing Business Service and the Scope of Local Markets. Working Paper WP 200109, Federal Reserve Bank of Chicago, 2001. 34. Sood, R. IT, Software, and Services: Outsourcing & Offshoring. Austin: AiAiYo Books, 2005. 35. Whitaker, J.; Mithas, S.; Kumar, S.; and Krishnan, M.S. Antecedents and Performance Outcomes of Onshore and Offshore Business Process Outsourcing. Working Paper, Ross School of Business, University of Michigan, 2006.

25

Table 1: Descriptive Statistics for Establishment Outsourcing Analysis 2002 Data Outsource Programming & Design Outsourcing Hosting Ex Internet Log(Local Programming & Design Establishments) Log(Local Hosting Establishments) Change in Programmers Change in Servers Change in Programmers Instrument Change in Servers Instrument Log of County Employment IT Intensity Index Establishment Size Index Log University Enrollment Percent of Establishments in Manufacturing Percent of Establishments in Whlse/Retail Trade Percent of Establishments in FIRE Percent of Establishments in Info Processing Percent of Establishments in Other Services Percent of Establishments in IT-Producing Log Establishment Employment Multi-Establishment Dummy PCs per Employee Non PCS per Employee 2004 Data Outsource Programming & Design Outsourcing Hosting Log(Local Programming & Design Establishments) Log(Local Hosting Establishments) Change in Programmers Change in Servers Change in Programmers Instrument Change in Servers Instrument Log of County Employment IT Intensity Index Establishment Size Index Log University Enrollment Percent of Establishments in Manufacturing Percent of Establishments in Whlse/Retail Trade Percent of Establishments in FIRE Percent of Establishments in Info Processing Percent of Establishments in Other Services Percent of Establishments in IT-Producing Log Establishment Employment Multi-Establishment Dummy PCs per Employee Non PCS per Employee

Mean

Std. Dev.

Minimum

Maximum

Number Obs

0.260409 0.3224502 4.5142 2.863149 0.837357 1.088138 0.314018 0.242966 12.03868 0.007193 1.520034 8.953789 0.157896 0.183636 0.071733 0.057296 0.373022 0.047099 5.567426 0.428963 0.529886 0.009858

0.438862 0.4674187 2.19376 1.751509 28.81435 70.45643 5.764956 9.864905 1.729237 0.003516 1.246224 3.53963 0.107231 0.035955 0.034512 0.030596 0.067172 0.033799 0.808246 0.494933 4.510383 0.07051

0 0 0 0 -500 -10000 -346.655 -883.463 3.496508 0 0.479992 0 0 0.016416 0 0 0.092736 0 4.60517 0 0 0

1 1 7.765145 5.774552 500 6000 250 410.5 15.16671 0.086475 36.02635 13.1765 0.956435 1 0.319149 0.347092 1 0.470837 12.76769 1 1001.05 8.033334

52191 52191 52191 52191 52191 52191 52191 52191 52191 52191 52191 52191 52191 52191 52191 52191 52191 52191 52191 52191 52191 52191

.1969359 .2303295 4.554013 3.298404 .0804052 .2619368 .056879 .2000097 11.99981 .0135147 1.524583 8.984475 .1438124 .1833131 .0757714 .0581592 .3866645 .0216583 5.562991 .486466 .5534679 .0068324

.3976877 .4210482 2.182343 1.849864 27.32528 63.2404 5.361489 6.758896 1.724756 .1106074 1.170749 3.506497 .1016654 .0365394 .0356119 .0333942 .0672548 .0216625 .8210779 .4998221 .8117137 .0657178

0 0 0 0 -500 -6000 -250 -633.8888 3.970292 0 .2685202 0 0 .0123518 .0049407 0 .0677201 0 4.60517 0 0 0

1 1 7.821242 6.329721 500 5996 372 270 15.14559 1.917011 40.29307 13.1765 1 1 .3773585 .370019 1 .460733 10.91509 1 100 6.08

47584 47584 47584 47584 47584 47584 47584 47584 47584 47584 47584 47584 47584 47584 47584 47584 47584 47584 47584 47584 47584 47584

26

Table 2: Analysis of Establishment Outsourcing Decision (2002)

Log(Local Supply) § Change in Programmers Change in Servers Log Establishment Employment Multi-Establishment Dummy PCs per Employee Non PC Servers per Employee Constant Observations R-squared

No Instruments Outsource Programming Outsource & Design Hosting (1) (2) 0.0046 0.0012 (0.0009)** (0.0013) 0.0004 ... (0.0001)** ... ... 0.0001 ... (0.0000)** 0.0286 0.0011 (0.0026)** (0.0027) -0.0942 -0.0808 (0.0041)** (0.0044)** 0.0022 0.0006 (0.0013)+ (0.0008) 0.135 0.1065 (0.0495)** (0.0288)** 0.1183 0.3167 (0.0395)** (0.0409)** 52191 52191 0.0235 0.0195

Instrument for Local Supply Outsource Programming Outsource & Design Hosting (3) (4) 0.0045 0.0018 (0.0010)** (0.0013) ... ... ... ... ... ... ... ... 0.0291 0.0011 (0.0026)** (0.0027) -0.0943 -0.0809 (0.0041)** (0.0044)** 0.0022 0.0006 (0.0013)+ (0.0008) 0.1417 0.1094 (0.0509)** (0.0288)** 0.1161 0.3156 (0.0395)** (0.0409)** 52191 52191 0.0228 0.0193

Instrument for Local Supply and Change in Programmers/Servers Outsource Programming Outsource & Design Hosting (5) (6) 0.0039 0.002 (0.0011)** (0.0014) 0.0055 ... (0.0032)+ ... ... 0.0021 ... (0.0024) 0.0228 0.0006 (0.0049)** (0.0048) -0.0925 -0.0806 (0.0044)** (0.0045)** 0.0018 0.0004 (0.0010)+ (0.0007) 0.0507 0.0424 (0.0657) (0.0945) 0.1536 0.3171 (0.0463)** (0.0465)** 52191 52191 ... ...

Simultaneous Equation Model (Instrument for Local Supply) Outsource Programming Outsource & Design Hosting (7) (8) 0.0045 0.0012 (0.0010)** (0.0010) … ... … ... ... … ... … 0.0291 0.0010 (0.0026)** (0.0028) -0.0943 -0.0809 (0.0042)** (0.0045)** 0.0022 0.0006 (0.0004)** (0.0005) 0.1418 0.1095 (0.0272)** (0.0290) 0.1155 0.3147 (0.0385)** (0.0411)** 52191 52191 ... ...

Standard errors are in parentheses. All regressions include dummy variables for three-digits NAICS. § In columns (1), (3), and (5) local supply is equal to the sum of programming and design establishments (NAICS 541511 and 541512); in columns (2), (4), and (6) it is the number of local hosting establishments (NAICS 541513 and 514210); in columns (7) and (8) it is equal to the sum of establishments in the four aforementioned categories. +significant at 90% confidence level. *significant at 95% confidence level. **significant at 99% confidence level.

27

Table 3: Analysis of Establishment Outsourcing Decision (2004)

Log(Local Supply)§ Change in Programmers Change in Servers Log Establishment Employment Multi-Establishment Dummy PCs per Employee Non PC Servers per Employee Constant Observations R-squared

No Instruments Outsource Programming Outsource & Design Hosting (1) (2) 0.0058 0.0045 (0.0010)** (0.0012)** 0.0002 … (0.0001)** … … 0.0001 … (0.0000) 0.0275 0.0113 (0.0027)** (0.0027)** -0.0676 -0.0542 (0.0039)** (0.0041)** 0.0329 0.0173 (0.0132)* (0.0080)* 0.1059 0.0782 (0.0543)+ (0.0503) 0.0059 0.17 (0.0369) (0.0410)** 47584 47584 0.0192 0.0114

Instrument for Local Supply Outsource Programming Outsource & Design Hosting (3) (4) 0.0061 0.0050 (0.0010)** (0.0012)** … … … … … … … … 0.0277 0.0113 (0.0027)** (0.0027)** -0.0677 -0.0542 (0.0039)** (0.0041)** 0.033 0.0172 (0.0133)* (0.0079)* 0.107 0.09 (0.0545)* (0.0495)+ 0.004 0.1692 (0.0369) (0.0410)** 47584 47584 0.0189 0.0113

Instrument for Local Supply and Change in Programmers/Servers Outsource Programming Outsource & Design Hosting (5) (6) 0.0060 0.0050 (0.0011)** (0.0012)** -0.0019 … (0.0046) … … 0.0009 … (0.0026) 0.0289 0.0113 (0.0040)** (0.0028)** -0.0684 -0.055 (0.0043)** (0.0047)** 0.0338 0.018 (0.0142)* (0.0087)* 0.116 -0.1098 (0.0617)+ (0.5954) -0.0022 0.1695 (0.0400) (0.0412)** 47584 47584 … …

Simultaneous Equation Model (Instrument for Local Supply) Outsource Programming Outsource & Design Hosting (7) (8) 0.0061 0.0041 (0.0009)** (0.0010)** … … … … … … … … 0.0276 0.0113 (0.0024)** (0.0026)** -0.0678 -0.0542 (0.0039)** (0.0041)** 0.0330 0.0171 (0.0024)** (0.0025)** 0.1070 0.0900 (0.0280)** (0.0298)** 0.0028 0.1659 (0.0376) (0.0400)** 47584 47584 … …

Standard errors are in parentheses. All regressions include dummy variables for three-digits NAICS. § In columns (1), (3), and (5) local supply is equal to the sum of programming and design establishments (NAICS 541511 and 541512) ; in columns (2), (4), and (6) it is the number of local hosting establishments (NAICS 541513 and 518210); in columns (7) and (8) it is equal to the sum of establishments in the four aforementioned categories. +significant at 90% confidence level. *significant at 95% confidence level. **significant at 99% confidence level.

28

Table 4: Analysis of Establishment Outsourcing Decision, Robustness to Alternative Modeling Assumptions (2004) Splitting Programming and Design

Log(Local Supply)§ Log Establishment Employment Multi-Establishment Dummy PCs per Employee Non PC Servers per Employee Constant Observations R-squared

Outsource Programming (1) 0.0018 (0.0009)* 0.0180 (0.0023)** -0.0810 (0.0036)** 0.0010 (0.0011) 0.1213 (0.0429)** 0.1315 (0.0364)** 52191 0.0190

Outsource Design (2) 0.0048 (0.0008)** 0.0175 (0.0020)** -0.0386 (0.0031)** 0.0014 (0.0005)** 0.0505 (0.0240)* 0.0372 (0.0309) 52191 0.0112

Instrument for Local Supply Outsource Outsource Hosting ex Internet Internet Hosting (3) (4) 0.0024 0.0042 (0.0012)* (0.0010)** 0.011 0.0008 (0.0021)** (0.0021) 0.0067 -0.0699 (0.0032)* (0.0032)** 0.0031 0.0185 (0.0025) (0.0077)* 0.0915 0.0165 (0.0326)** (0.0355) 0.0686 0.1214 (0.0329)* (0.0320)** 47584 47584 0.0067 0.0183

Use of Employment for Local Supply Outsource Programming Outsource & Design Hosting (5) (6) 0.0051 0.0033 (0.0009)** (0.0008)** 0.0276 0.0112 (0.0027)** (0.0027)** -0.0678 -0.0543 (0.0039)** (0.0041)** 0.0329 0.0172 (0.0133)* (0.0079)* 0.1068 0.0899 (0.0544)* (0.0496)+ -0.0009 0.166 (0.0369) (0.0410)** 47584 47584 0.0188 0.0113

Dependent Variable is Number of Services Outsourced Outsource Programming Outsource & Design Hosting (7) (8) 0.0107 0.0135 (0.0022)** (0.0032)** 0.0678 0.019 (0.0063)** (0.0073)** -0.1217 -0.1427 (0.0084)** (0.0105)** 0.0663 0.0394 (0.0272)* (0.0182)* 0.2016 0.1211 (0.1286) (0.1026) -0.0419 0.2191 (0.0968) (0.0752)** 47584 47584 0.0154 0.0112

Standard errors are in parentheses. All regressions use instruments for local supply, but do not instrument for change in the number of programmers or servers. All regressions include dummy variables for three-digits NAICS. Dependent variable in columns (7) and (8) is number of services outsourced. § In column (1) local supply equal to the number of programming establishments (NAICS 541511); in column (2) it is equal to the number of design establishments (NAICS 541512); in column (3) is equal to the number of hosting and facilities management providers (NAICS 541513); in column (4) it is equal to the number of hosting providers (off site) (NAICS 518210); in column (5) it is equal to the total programming and design employment (NAICS 541511 and 541512); in column (6) it is total hosting employment (NAICS 541513 and 518210); in column (7) local supply is equal to the sum of programming and design establishments (NAICS 541511 and 541512); in column (8) it is the number of local hosting establishments (NAICS 541513 and 518210). +significant at 90% confidence level. *significant at 95% confidence level. **significant at 99% confidence level.

29

Table 5: Analysis of Establishment Outsourcing Decision with Interactions

Log(Local Hosting Establishments) Security Applications Present Enterprise Applications Present Network Applications Present Functional Applications Present Log(Local Supply) X Security Applications Present Log(Local Supply) X Enterprise Applications Present Log(Local Supply) X Network Applications Present Log(Local Supply) X Functional Applications Present Log Establishment Employment Multi-Establishment Dummy PCs per Employee Non PC Servers per Employee Constant Observations R-squared

Outsourcing Hosting 2002 (1) -0.0055 (0.0030)+ 0.0543 (0.0177)** 0.0351 (0.0109)** -0.0191 (0.0105)+ 0.0082 (0.0269) 0.0041 (0.0051) 0.0068 (0.0033)* 0.0035 (0.0032) -0.012 (0.0079) -0.0227 (0.0035)** -0.0671 (0.0056)** -0.0002 (0.0005) 0.0198 (0.0318) 0.552 (0.0553)** 35082 0.0243

Outsourcing Hosting 2004 (2) -0.0003 (0.0031) 0.0112 (0.0132) 0.0266 (0.0123)* -0.0203 (0.0115)+ 0.0755 (0.0580) 0.0057 (0.0035) 0.0026 (0.0033) 0.0068 (0.0031)* 0.0007 (0.0154) -0.0086 (0.0035)* -0.0361 (0.0057)** -0.0042 (0.0040) 0.0558 (0.0329)+ 0.3347 (0.0593)** 29664 0.0161

Outsourcing Hosting 2002 (Alt Classification) (3) -0.0049 (0.0029)+ 0.0529 (0.0177)** 0.0349 (0.0109)** -0.0194 (0.0105)+ 0.0319 (0.0141)* 0.0037 (0.0051) 0.0072 (0.0033)* 0.0030 (0.0032) -0.0059 (0.0043) -0.0243 (0.0035)** -0.0676 (0.0056)** -0.0002 (0.0004) 0.0101 (0.0318) 0.5585 (0.0553)** 38082 0.0248

Outsourcing Hosting 2004 (Alt Classification) (4) -0.0003 (0.0031) 0.0112 (0.0132) 0.0265 (0.0123)* -0.0203 (0.0115)+ 0.0786 (0.0579) 0.0057 (0.0035) 0.0026 (0.0033) 0.0068 (0.0031)* -0.0010 (0.0153) -0.0086 (0.0035)* -0.0361 (0.0056)** -0.0041 (0.0040) 0.0560 (0.0329)+ 0.3346 (0.0593)** 29664 0.0189

Standard errors are in parentheses. All regressions using instruments for local supply, but do not for change in the number of programmers or servers. All regressions include dummy variables for three-digits NAICS. In all columns local supply is equal to the number of local hosting and management establishments (NAICS 541513 and 518210). +significant at 90% confidence level. *significant at 95% confidence level. **significant at 99% confidence level.

30

Table 6: Summary of Hypotheses and Results Hypothesis Hypothesis 1: Outsourcing is increasing in the supply of local firms, other things equal.

Prediction α >0

Hypothesis 2: The decision to outsource programming and design will be more sensitive to variations in local supply than will the decision to outsource hosting. Hypothesis 3: The sensitivity of outsourcing to local supply is increasing over time. Hypothesis 4a: The relationship between outsourcing hosting and local supply is stronger for firms with greater investments in security software. Hypothesis 4b: The relationship between outsourcing hosting and local supply is stronger for firms with greater investments in network infrastructure software.

α PROGRAMMING > α HOSTING

Result Supported, particularly for programming and design Supported

α 2004 > α 2002

Supported

HO STING δ SECURITY >0

Supported in 2004 data.

HO STING δ NETWO RK > 0

Supported in 2004 data.

31

Figure 1: Geographic Distribution of Programming and Design Establishments in the U.S. (2003)

[0.00,0.00] (0.00,1.00] (1.00,7.00] (7.00,2492.00]

Notes: Light Green (0 Establishments); Medium Light Green (1 Establishment); Medium Dark Green (2-7 Establishments); Dark Green (over 7 Establishments)

Figure 2: Geographic Distribution of Hosting Establishments in the U.S. (2003)

[0.00,0.99] (0.99,1.99] (1.99,11.99] (11.99,560.00]

Notes: Light Green (0 Establishments); Medium Light Green (1 Establishment); Medium Dark Green (2-11 Establishments); Dark Green (Over 11 Establishments)

32