One fact holds clear: exclusion of Japan does not invalidate the hypothesis. Whether one wishes to use a one-tailed or two-tailed test of significance, one.
Private sector sector shrinkage shrinkage and and the the growth growth of industrialized industrialized Private economies: Reply* Reply* economies: MICHAEL L. MARLOW
u.s. Department Department of of Treasury, Treasury, Washington, Washington, DC DC 20220 U.S.
1. Introduction In my my 1986 paper on private private sector tested the the hypothesis hypothesis that, that, In sector shrinkage, 1986 paper shrinkage,I tested ceteris paribus, economies economies with with large large and and growing growing government government sectors exert ininsectorsexert ceterisparibus, real economic economic growth. growth. The The data data used used in in my my paper paper were were pubpubverse influences influences on real verse OECD (1985) and contained contained various various measures measures of government government spendlished in OECD lished spend(1985)and over the and economic growth of 19 major industrialized economies over the ing industrialized economies 19 major ing and economic growth interval there exists that there existssuffisuffithe data data concluded concludedthat interval1960-1981. 1960-1981. My My analysis analysisof the withrelarelacient in support the hypothesis that economies economieswith evidencein cientempirical hypothesisthat supportof the empiricalevidence tively larger public sectors are associated with relatively poorer economic with economic associated sectors are relativelypoorer tively larger public growth and into question the wisdom·of wisdomof allowing call into that the the results results'clearly and that allowing 'clearlycall questionthe growth public 1986:152). sectors'(Marlow, at the the expense to grow sectorsto 152). (Marlow,1986: expenseof private privatesectors' grow at publicsectors than being II quoted 'Ratherthan observationthat, OECD's (1985: that, 'Rather widely being widely 14) observation (1985: 14) quoted OECD's staregarded and macroeconomic macroeconomicstaeconomicgrowth as aa major contributorto economic growthand regardedas majorcontributor on that the growth and financing of the public sector has, bility, the view the sector and that the the view has, on financing public growth bility, this 'widebalance, stifled growth now attracts widespread support.' Despite'this 'widebalance, stifled growthnow attractswidespreadsupport.'Despite crossspread' macroeconomiclevel, 'At the the macroeconomic noted that, level, preliminary that, 'At preliminarycrossview, II noted spread'view, into reveal revealan an inhavefailed failedto country the Secretariat Secretariathave undertakenby by the countrycomparisons comparisonsundertaken as and economic performance as verse between public sector size and economic size sector between verse relationship performance public relationship reflected in GDP growth rates ... 15). (1985: 15). ...,' ,' OECD (1985: growth rates and subsequent that, based on OECD's analysis Saunders (1987) argues OECD's based on Saunders(1987) argues that, subsequent analysis and Saunders(1986), elaboration and Saunders in Saunders Saunders(1985) resultsin the empirical elaborationof the (1986), (1985)and empiricalresults conclusionsapand that that 'my OECD's data are are correct correctand these data own analysis OECD'sown ap'my conclusions analysisof these that pear to be unwarrantedly strong.' Saunders (1988) alleges that my analysis Saunders to be my analysisof (1988)alleges unwarrantedlystrong.' pear real unconventional the the data contains numerous flaws; namely, my use of the unconventional real the data containsnumerousflaws; namely,my use time to the the time government resultsto 'extreme'sensitivity my results series, 'extreme' sensitivityof my expenditureseries, governmentexpenditure are hereare Theviews viewsexpressed comments.The ** II would valuablecomments. R. Mascaro Mascarofor for valuable liketo to thank thankAngelo wouldlike expressedhere AngeloR. the U.S. solely those the views of the author and do not necessarily reflect the views of the U.S. Department reflect do not and the author those Departmentof necessarily solely Treasury. Treasury.
errorin in including in the interval, the sample and the the exclusion exclusionof 'apinterval,my my error includingJapan Japanin sampleand 'apin propriate' control control factors factors in my regression equations. propriate' my regressionequations. In this In this paper, submitthe data to the data to further furtherscrutinity and conclude concludethat that the the paper,II submit scrutinityand resultthat that economic result economicgrowth and government size are are inversely relatedconcongrowthand governmentsize inverselyrelated tinues to to be be an an obvious obviousimplication tinues the data. data. implicationof the
2. Relative price price effect
The first The first issue issue concerns concernsthe the measurement measurementof the the scale scaleof aa government's government'sparparin ticipation in the economy. Saunders states that my use of the price-deflated the Saunders states that use the serticipation economy. my price-deflatedseries on on government in contrast ies 'stands in contrastto in this to all all other other work work in this governmentexpenditures expenditures'stands which adopts the more more conventional area, conventionalratios ratios of nominal nominalexpenditures area, which adopts the expendituresto nominalGDP.' nominal effect' which GDP.' The Thebasic basicissue issueconcerns concernsthe the 'relative 'relativeprice effect' whichrefers refers price the possibility to the that the growth the share of nominal government spending that the of the share nominalgovernmentspending possibility growth may differ from fromthe the growth the share shareof real realgovernment Thatis, may differ growthof the is, governmentspending. spending.That the real real index, the constant dolconstantdollar dollargovernment dolindex, GE(R) GE(R) == constant governmentspending/ spending/constant lar GDP, will differ differ from from the the nominal nominalindex, lar nominalgovernment GDP, will index, GE(N) GE(N) == nominal government nominal spending/nominal GDP, when the price indices of government when the indices and GDP, spending/ price governmentspending spendingand GDP differ. differ. GDP Saunders'comment commentis is curious Saunders' discussed in in Saunders (1986), there there curioussince, as discussed Saunders(1986), since, as is aa debate debateon this this issue; is e.g., see Beck (1976), Beck (1979), Beck (1985) see Beck Beck Beck and issue; (1976), (1979), (1985)and Heller(1981). In Heller In any event, it would appear that given the lack of a consensus it would that the lack a consensus (1981). any event, appear given the issue, on the price be appropriate ask what what difference differencethe the relative relativeprice issue, itit might might be appropriateto ask effect the estimation estimation of the the government government size economic growth growth rereeffect makes makes on the size - economic lation. lation. Tables Tables 1display estimations estimations of the the basic basic equations equations of my my 1986 paper: 1986paper: 1- 3 display RGDP RGDP RGDP
= ao a o + alGE alGE + e € = RGE + 4c/J = {3o + {3IGE 0 + 3IGE + (32 32RGE
(1) (1)
(2) (2)
where where RGDP RGDP = compound compound annual annual rate rate of growth growth of real real GDP GDP GE(R) GE(R) GE(N) GE(N) RGE(R) RGE(R) RGE(N) RGE(N) €,9 E,
c/J
= = ratio ratio of real real government government expenditure expenditure to GDP GDP in in initial initial periperiod of RGDP RGDP measurement; measurement; = = ratio ratio of nominal nominal government government expenditure expenditure to nominal nominal GDP GDP in initial initial period period of RGDP RGDP measurement; measurement; = = compound compound annual annual growth growth rate rate of GE(R); GE(R); = = compound compound annual annual growth growth rate rate of GE(N); GE(N); = = random random disturbance disturbance terms. terms.
I have have deleted deleted one set set of equations equations from from my my 1986 1986 period period - those those which which run run RGDP RGDP on only a constant constant and and RGE. RGE. These These were were routinely routinely poor poor fits fits and and do not
alter the the present present discussion. discussion. I have have also also deleted deleted the the equations equations dealing dealing with with soalter focus on the main main source source of Saunders' Saunders' criticriticial expenditures expenditures for for brevity brevity and and to focus cial cism. In In contrast contrast to Saunders Saunders (1987), (1987), I have have displayed displayed the regression regression results results for for cism. GE is highly highly statistically statistically (1) as as well well as as (2). (2). As can can be seen, seen, either either measure measure of GE (1) significant for for the the period period 1960-1970 and and 1960-1980. As in my my 1986 1986 paper, paper, significant the government government scale scale variable variable is generally generally not statistically statistically different different from from zero zero the over 1970-1980. Moreover, Moreover, as as shown shown in Saunders' Saunders' comment, comment, the use use of a over nominal-based government government scale scale variable variable GE(N) GE(N) results results in higher higher levels levels of nominal-based statistical significance significance and and relatively relatively large large increases increases in the the equations' equations' explanaexplanastatistical tory power. Given that my 1986 paper did find much support for the hypofind 1986 did not much for the Given that tory power. my paper hyposupport that the the rate rate of change change of government government expenditure expenditure exerts exerts a significant significant ininthesis that thesis economic growth, growth, the the 'relative 'relative price price effect' effect' issue issue would would appear appear to fluence on economic fluence offer no additional addi~ional information information on the the validity validity of that that hypothesis. hypothesis. ConseConseoffer quently, based the fact that there exists no consensus how measure that there exists consensus on how to measure based on the fact quently, government scale and the the fact fact that that Saunders' preferred measure measure actually actually Saunders' preferred scale and government provides stronger for my my hypothesis, hypothesis, I view view this this particular particular criticism criticism as as provides strongersupport supportfor curious. curious.
3. Excluding from the the sample sample Japan from ExcludingJapan The that Japan should not not be be included included with the the assertion assertionthat The next next criticism criticismdeals deals with Japanshould in the sample of observations an assertion that I interpret as implying that in the sample observations an assertionthat interpretas implyingthat some reason, Japan for us us as as economists. economists. For For some no relevant informationfor has no relevantinformation reason, Japan has as the OECD OECD should should be be treated treatedas Saunders memberof the believesthis this particular Saundersbelieves particularmember different Interestingly, Saunders offers no offers no Saunders membercountries.! differentfrom fromthe the other othermember countries.1Interestingly, a priori priori reasoning it would justify this fromthe the sample. wouldapthis deletion deletionfrom Rather,it apsample.Rather, reasoningto justify is aa that Japan justification arrived pear after Saunders Saundersobserved observedthat arrivedex post after that justification Japanis pearthat and major economic growth inverse relation relation between between economic the inverse contributorto the growth and major contributor government size. Without a doubt, Japan has a relatively high economic economic has a Without a size. doubt, relatively high Japan government growth low level level of government and aa correspondingly overthese these periods rateover government correspondinglylow periodsand growthrate valuable participation in the economy. I wonder inferany if economists could infer could I wonder economists in the any valuable economy. participation lessons not. Their Their deletionmay its deletion from this this country? lessons from may suggest suggestnot. country?Proponents Proponentsof its the argument 'contaminate'the that its its inclusion inclusionmight on the the fear fearthat be based basedon might 'contaminate' may be argumentmay testing of the economic growth-government size hypothesis. size the economic hypothesis. growth-government testing without Somewhat re-estimatedmy the criticism, aboutthe curiousabout Somewhatcurious criticism,II re-estimated equationswithout my equations Japan. Using both realand nominal-based measures of government measures nominal-based scale, II governmentscale, Japan. Using both real- and in the Tables re-estimated the lower lowerparts aredisplayed and they re-estimatedmy partsof Tables they are displayedin my equations equationsand andgoodness levelsand 13. Saunders the significance is correct on one one count: count:the correcton 1-3. Saundersis goodnessof significancelevels fit do fall result of aa this is the expected result is excluded. However, this fall when Japan Japan is to limit the amount of variation in the variables scheme that seeks variables being exexseeks limit amined. amined.
Table results) - dependent dependent 1960-1970 (regression 1. Government and economic economicgrowth, Governmentgrowth Table1. (regressionresults) growthand growth, 1960-1970 variable: Real GDP GDP growth 1960-1970 variable:Real growth1960-1970 Equation
Constant
GE(R) GE(R)
(1) (1)
12.91 12.91 (4.76) 14.21 14.21 (6.95) 13.11 13.11 (5.06) (5.06) 14.26 (6.53)
-23.97 -23.97 (2.87)
(2) (2) (3) (4) (4)
GE(N)
RGE(R) RGE(R)
RGE(N)
-31.69 -31.69 (4.46) - 23.91 (3.00) (3.00)
-0.40 -0.40 (1.54) (1.54) - 31.70 -31.70 (4.30)
-0.03 -0.03 (.09)
2
11 iR2
N
.32
16
.56
16 16
.39
16 16
.52
16
.06
15 15
.18
15 15
-.02 -.02
15 15
.11
15 15
Excluding Japan Japan Excluding (5)
(6) (7) (7) (8) (8)
8.06 (3.33) 9.76 (3.96) 8.34 (3.14) 9.55 (3.59)
- 9.82 -9.82 (1.34) - 16.85 -16.85 (2.01)
-10.52 -10.52 (1.33)
-0.07 -0.07 (0.32) - 16.57 -16.57 (1.89)
0.07 (.28)
One fact invalidate the hypothesis. One exclusionof Japan does not not invalidate the hypothesis. fact holds holds clear: clear:exclusion Japandoes Whether one one one wishes Whether wishesto use use aa one-tailed one-tailedor or two-tailed two-tailedtest test of significance, significance,one or both both of the the measures size continues be statistically differor measuresof government size to be differcontinues statistically government in conclusions ent from reached here and by by Saunent from zero. zero. Why Saunthe difference differencein conclusionsreached here and Why the ders? Several be correct. correct. One explanation may may be that that SaunSeveralexplantions ders? One explanation Saunmay be explantionsmay ders appears to prefer test the hypothesis using a two-tailed test. This may ders to test the a two-tailed test. This appears prefer hypothesisusing may be an point and should depend on how one formulates the alternative and on how be an arguable should one formulates the alternative arguablepoint depend in turn, some a priori priori considerations hypothesis, which, is related relatedto some considerations which,in turn, generally hypothesis, generallyis or previous previous empirical hypothesis in al = in (1) or evidence.If one one views the hypothesis viewsthe as H empiricalevidence. (1) as o: ca Ho: d then aa two-tailed o0 vs. appropriate. A one-tailed one-tailed test would vs. H two-tailedtest test is is appropriate. test would Ha: a: a l ;c 0, then at1 = : : be appropriate for the hypothesis H a = vs. H a < O. be a 0 for the vs. hypothesis appropriate o l Ho: Ha: a atl < 0. II originally part, may in part, chose aa one-tailed one-tailedtest, some of his his which, in test, which, originallychose may explain explainsome questioning levels of significance. rationale behind behind aboutmy statementof levels Therationale my statement questioningabout significance.The are are negative and this choice this and choice was was my that the the coefficients on ai coefficients on my expectation negative expectationthat ai at best, that, at best, one could expect government size exert a neutral influence on one could size to exert a neutral influence that, expectgovernment On economic growth. On page 145, I discussed the possibility that governments economic discussed the that growth. page 145, I possibility governments could exert but that the public public sector could exert aa 'facilitating' role on on the the economy, that the sector 'facilitating'role economy, but growth over over this period was considerably above any threshold level that that might might this period was considerably above any thresholdlevel growth in economic growth be wedged in between government growth-as-facilitator be between of economic wedged governmentgrowth-as-facilitator growth and government that, and The observation observation that, government growth-as-impediment growth-as-impedimentto growth. growth. The based on Table paper, the level of the government's share share based Table11 of my 1986paper, the average the government's my 1986 averagelevel
GDP (GE(R)) (GE(R» had had risen risen from from .32 to .40 over over 1960-1980 appeared appeared to be of GDP support the construction my hypotheses. Based on ample evidence in of of the construction evidence my hypotheses.Based ample support Saunders' preferred preferred measure measure GE(N), GE(N), government's government's average average share share of the the econeconSaunders' over the the same same period. period. Moreover, Moreover, the the construction construction grew from from .28 .28 to .42 over omy grew one-tailed test test would would appear appear to be consistent consistent with with OECD's OEeD's statement, statement, of a one-tailed , . .. Rather Rather than being being widely regarded regarded as a major contributor to economic '... growth and and macroeconomic macroeconomic stability, stability, the the view view that that the the growth growth and and financing financing growth the public public sector sector has, has, on balance, balance, stifled stifled growth growth now now attracts attracts widespread widespread of the support' (p. (p. 14). 14). Moreover, Moreover, prior prior empirical empirical evidence evidence exists exists in Landau Landau (1983), (1983), support' Orzechowski (1984), and Weede (1984) that, varying degrees, public sector Weede to Orzechowski(1984),and (1984)that, varyingdegrees,publicsector size is is inversely inversely related related to economic economic growth. growth. While While less less than than conclusive, conclusive, these these size studies are are available available to Saunders Saunders and and were were listed listed in my my 1986 paper. Moreover, Moreover, studies 1986paper. and Eltis Eltis (1978), (1978), Smith Smith (1975), (1975), Cameron Cameron (1978), (1978), Saunders (1986) (1986) cites cites Bacon Bacon and Saunders Marsden (1983) and Gould (1983) as, at least, suggestive of an inverse relation Marsden(1983)and Gould (1983)as, at least, suggestive an inverserelation between government government size and economic economic growth. growth. Therefore, Therefore, contrary contrary to his his asassize and between sertion, my results results are are not not inconsistent inconsistent 'with 'with most most of the the other other empirical empirical work work sertion,my the relationship relationship between between the the size and growth growth of government government in in industrialindustrialsize and on the It would appear appropriate for Saunders to provide the imimized economies.' for Saunders ized economies.'It would appearappropriate providethe plied long long and and impressive impressive litany litany of empirical empirical studies the opposite opposite of studiesshowing showingthe plied my work before one can accept his assertion that my 1986 paper is inconsistent assertion that 1986 is his inconsistent work before one can my paper accept my 2 In In addition, with Saunders issue.2 work on on this this issue. with the the majority addition, Saunders empiricalwork majority of empirical might providing studies relation between positive relation between that estimate estimate aa positive studies that consider providing might consider government before one one might concludethat that aa two-tailed two-tailedtest test and growth size and mightconclude governmentsize growthbefore is hypothetical construct. construct. the appropriate is the appropriatehypothetical the choice vs. After out that that the choiceof aa one-tailed one-tailedtest test vs. all this, turnsout After having said all this, itit turns havingsaid not critical to the acceptance of my hypothesis, with or two-tailed test is with or to the two-tailed test is not critical acceptance my hypothesis, without measureof governSaunderspreferred inclusionof Japan. withoutthe the inclusion governUsing Saunders preferredmeasure Japan.Using arestatisticalment the coefficients coefficientson on GE(N) statisticalsize GE(N) and two-tailed two-tailedtests, mentsize tests, the GE(N)are GE(N)and and 1960-1980 at at the ly 1960-1970 and the intervals1960-1970 the intervals from zero zero during differentfrom duringthe ly different respective levels lists realof .10 and .05, or better. Curiously, Saunders lists only realSaunders and or better. .10 levels .05, Curiously, only respective In in his his comment on my derived commenton scale variables, derivedgovernment variables,GE(R), my paper. paper. In GE(R), in governmentscale to the the inclusion inclusionof Japan. conclusion, sensitiveto not extremely are not resultsare Japan. extremelysensitive conclusion,my my results
4. time period 4. Choice Choice of time period
time period chosen to the the time The sensitiveto areextremely resultsare that my The criticism criticismthat periodchosen extremelysensitive my results stems results presented in Saunders (1985: 11). He He interprets in Saunders results from empirical stems from (1985: 11). interprets presented empirical his are and government that economic growth his 1985 1985 paper as suggesting growth and government growth growth are paper as suggesting that inversely orare in orand not 1975-1981. 1975-1981. Three Three points related over over 1960-1973 1960-1973 and points are inversely related der. his empirical includes aa mismisas discussed discussed in the section below, his der. One, as empirical model includes specification drawn from from his his inferences drawn casts considerable doubt on the inferences specification which casts
Table 1970-1980(regression and economic economicgrowth, Governmentgrowth Table2. Government results)- dependent dependent (regressionresults) growth,1970-1980 growthand variable: 1970-1980 variable:Real Real GDP GDP growth growth1970-1980
Equation Equation
Constant Constant
GE(R) GE(R)
(1) (1)
5.05 5.05 (3.82) (3.82) 5.05 5.05 (3.82) (3.82) 5.27 5.27 (3.94) (3.94) 5.40 5.40 (3.62) (3.62)
-4.89 -4.89 (1.29) (1.29)
(2) (2) (3) (3) (4) (4)
GE(N) GE(N)
RGE(R) RGE(R)
RGE(N) RGE(N)
-4.89 -4.89 (1.29) (1.29) -0.34 -0.34 (1.03) (1.03)
-4.06 -4.06 (1.04) (1.04)
- 0.09 -0.09 (.56) (.56)
-5.39 -5.39 (1.35) (1.35)
R2 R2
N
.04 .04
16 16
.04 .04
16 16
.05 .05
16 16
.00 .00
16 16
-.06 -.06
15 15
-.06 -.06
15 15
.00 .00
15 15
-.06 -.06
15 15
Excluding Japan Japan Excluding
(5) (5) (6) (6) (7) (7) (8) (8)
3.90 3.90 (2.29) (2.29) 3.90 3.90 (2.29) (2.29) 3.84 3.84 (2.33) (2.33) 4.15 4.15 (2.41) (2.41)
-1.76 -1.76 (.37) (.37) -1.76 -1.76 (.37) (.37)
0.30 0.30 (.06) (.06)
-0.46 -0.46 (1.37) (1.37) -1.57 -1.57 (.33) (.33)
-0.17 -0.17 (1.02) (1.02)
Table dependent and economic 1960-1980(regression Table3. Government economicgrowth, Governmentgrowth growthand growth,1960-1980 results)- dependent (regressionresults) variable: GDP growth variable:Real Real GDP 1960-1980 growth1960-1980
Equation Equation (1) (1) (2) (2) (3) (3) (4) (4)
Constant Constant 9.95 (5.47) (5.47) 10.56 10.56 (7.47) (7.47) 10.07 10.07 (5.46) (5.46) 10.94 10.94 (6.46) (6.46)
GE(R) GE(R)
GE(N) GE(N)
RGE(R) RGE(R)
RGE(N) RGE(N)
- 17.61 -17.61 (3.14) (3.14)
- 22.07 -22.07 (4.49) (4.49) - 17.16 -17.16 (3.02) (3.02)
-0.27 -0.27 (.83) (.83)
- 22.74 -22.74 (4.30) (4.30)
-0.09 -0.09 (.44) (.44)
R2 R2
N
.37
16
.56 .56
16 16
.36 .36
16 16
.53 .53
16 16
.12 .12
15 15
.23 .23
15 15
.05 .05
15 15
.19 .19
15 15
Excluding Japan Japan Excluding
(5) (5) (6) (6) (7) (7) (8) (8)
7.23 7.23 (3.92) (3.92) 8.33 8.33 (4.39) (4.39) 7.30 7.30 (3.68) (3.68) 8.75 8.75 (4.23) (4.23)
-9.69 -9.69 (1.73) (1.73)
- 14.64 -14.64 (2.27) (2.27) --9.78 9.78 (1.67) (1.67)
--0.04 0.04 (.14) (.14) -15.27 -15.27 (2.28) (2.28)
-0.13 -0.13 (.60) (.60)
statistical coefficient on government statisticaltests. tests. Two, the coefficient is negative size is and sigTwo, the governmentsize negativeand sigon the basis of a two-tailed test at the .10 level and, nificant, over 1975-1981, a over the basis two-tailed at test the .10 level nificant, and, on the basis of aa one-tailed the .05 level. Three, my regression results the basis at the .05 level. one-tailedtest, test, at Three,my regressionresults for for the in Table the interval interval1960-1980, in Table 3, show statistical statisticalsignificance for the the 3, show significancefor that the period of the 1970s is senentire period. Even entireperiod. Evenif Saunders' that the Saunders'suggestion the 1970s is sensuggestion period sitive to the sitive is the 1973 1973 oil shock and simultaneous declinein in ecqnomic shock and simultaneousdecline growthis ecqnomicgrowth true, only be be of aa transitory transitory nature. these complications should only nature.As displayed true, these complicationsshould displayed in my period 1960-1980, these in estimationsof the the period these short-run short-rundisruptions, my estimations disruptions,if present, do be fundamental long-run implications implications of do not fundamentalto the the long-run not appear present, appear to be government size on economic growth. Consequently, it is difficult for me to it size is economic difficult for me government growth. Consequently, take this take this comment commentseriously. seriously.
5. Misspecification my equation Misspecificationof my equation Saunders ... none the results aregrounded in any Saundersargues that '... none of the resultsare modelof the the arguesthat' groundedin any model growth process, however naive this may be. By excluding other factors which naive this be. however other factors which growthprocess, may By excluding those influence economic may be be wrongly interpreted to those theirimpact influence economicgrowth, wronglyinterpreted growth,their impactmay in the explanatory variables which are In other other words, the analysis.' are included included in variableswhich words, analysis.' In explanatory in part, part, from Saunders from omitted Saundersasserts assertsthat that my conclusionsresult, omitted variable variable result, in my conclusions bias. To formulation of the process, Saunders Saunders ininthe growth bias. To correct correctmy 'naive' formulation my 'naive' growthprocess, fixed capital forcludes forthe share non-residentialgross fixed cludestwo two new new variables: variables:the shareof non-residential gross capital time period period and and the civilian mation in GDP relevanttime the share the relevant shareof civilian mationin GDP averaged overthe averagedover period. The extent of the employment in the initial year year of each each period. in agriculture the initial The extent the employmentin agriculturein economic behind their inclusion is only 'capture the impact is to the their inclusion behind economic reasoning only 'capture impact of reasoning and investment on economic growth and the possibilities for increased the increased economic investmentexpenditure possibilitiesfor growth expenditure growth potential in with larger in economies economieswith sectors.' largeragricultural agriculturalsectors.' growthpotential expected efIt little simple theory to show show the the expected macroeconomictheory It only takes aa little only takes simplemacroeconomic in the equation. Standard fect variablein the equation. Standardtreatment treatment investmentvariable fect of including an investment includingan G is is of the between capital K and and government size G the relation relationbetween governmentsize capitalK K K = f(O) f(G) + iJ.I.
(3) (3)
See Aschauer which postulates that Aschauer out capital-formation. crowdsout that government whichpostulates capital-formation.See governmentcrowds the (1987) for a recent paper, with supporting empirical evidence, the effects on effects with a for recent evidence, empirical supporting paper, (1987) = With of public public expenditure on private capital accumulation. With Y = output and accumulation. output and expenditure privatecapital form aa production production relation the form relationof the
+
Y = f( K) f(K)+
+
E E
(4) (4)
it follows that estimation estimationof it follows that Y = f(K, Y G) + l'7 f(K, G)
(5) (5)
may Y is to G. find the the coefficient on G G to to be be zero zero since since Y is inversely relatedto coefficienton G. may find inverselyrelated In are true, then estimating is incorrect incorrectsince In other otherwords, and (4) since words,if (3) true, then (3) and (4) are estimating(5) (5) is the the coefficient coefficient on on G G is is calculated calculatedafter after the the correlation correlationbetween betweenK K and and G G is is taken if G exerts all influence on Y via K, then after regreson via taken account accountof, 'and, G exerts all influence then after and, K, regression sion of Y on on K K and and G, to exert exerton on thereis is no no remaining influenceleft left for for G G to G, there remaininginfluence In this Y. In then this case, Y. case, then Y Y == f(G) + 6()
(6) (6)
is the true form of (3) and (4), is the true reduced reducedform not (5). resultthat that G G == 0 Saunders'result (3) and (4), not (5). Saunders' after (5) is consistent with my work when (3) and (4) are true. after estimating is with work when consistent and are true. estimating(5) my (3) (4) The above The abovedemonstrates demonstratesthat thateconomic economictheory does not not provide the basis basisfor for theorydoes providethe Saunders'findings Saunders' that government size does does not not influence economicgrowth influenceeconomic findingsthat governmentsize growth once one one controls once may controlsfor for the the effects effects of capital formationon on Y. Y. While Whilesome some may capitalformation be tempted be to 'throw' as many variables into an equation as one has data, 'throw' as variables into an as one has the tempted data, the many equation apply a 'little' theory as not include applied researcher must be careful researcher must be to a 'little' careful so as to not include applied apply theory inappropriate variables. Inclusion variablesalong with appropriate Inclusionof inapinappropriatevariables along with appropriatevariables. inappropriate variables variables could believe that that economic growth is not relatrelatcouldlead leadone one to believe economicgrowth propriate ed in fact, ed to government when, in fact, an appropriately specified specified equation equation size when, an appropriately governmentsize would demonstrate demonstratethe the inverse inverserelation. relation. would course, if a larger larger governgovernInterestingly, Saunders reports '. Saunders(1985) '..... Of course, Interestingly, (1985) reports ment ment sector has crowded-out crowded-out private private investment investment and and thus thus lowered lowered the the investinvestsectorhas ment public expenditure expenditure on economic economic ment share, the overall overall negative negative impact impact of public share, the Y) ititsomewhatgreater growth will will be be somewhat greater than than implied implied by by the the coefficient coefficient on (G growth (G// Y) self.' Moreover, Moreover, Saunders cites Smith (1975), Cameron (1978), and Saunders(1985) cites Smith Cameron and (1985) (1975), (1978), Marsden empirical evidence evidence that, that, given given a positive positive Marsden (1983) (1983) as as providing providing some some empirical relation an inverse inverse relation relation beberelation between between investment investment and and economic economic growth, growth, '... ' ... an tween tween the the public public expenditure expenditure share share and and economic economic growth growth was was implied' implied' (p. (p. 11). 11). It is a curious curious matter matter as as to why why the the specification specification problem problem was was hinted hinted at at in earliearlier er work work but but not in the the recent recent comment comment on my paper. paper.
6. Conclusion Conclusion Further Further tests tests and and thoughts thoughts on the the OECD OECD data data lead lead me me to conclude conclude that, that, if anyanything, my 1986 1986 paper paper underestimated underestimated the the magnitude magnitude of the the inverse inverse relation relation bebething, tween tween economic economic growth growth and and government government size. size. If one takes takes the nominal-based nominal-based measure government scale, scale, as as advised advised by Saunders, Saunders, the the significance significance levels, levels, measure of government
coefficient coefficient magnitudes and goodness fits improve improve over over what found with with what I found magnitudesand goodnessof fits initial investigation. my I would suggest that Saunders reconsider his relucwould that Saunders reconsider his relucmy initial investigation. suggest tance that the believe that public sector unrelated to economic economic tance to believe the size size of the the public sector is is unrelated in OEeD period. growth in OECD countries countriesover over this this time time period. growth One additional One relevant to the policy debate additionalthought relevant the current currentpolicy debateconconthought appears appears major industrialcerning budget deficits and economic performance within the deficits and economic within the industrialcerningbudget performance major ized economies.The ized economies. displayed here paper sugThe empirical workdisplayed in my hereand and in 1986paper empiricalwork my 1986 sugseriousproblems gests serious problems associated various proposals proposals urging urging governgovernwith the associatedwith the various gests ments raisetaxes policy. Elsewhere, ments to raise taxes and/or and/or 'ease' 'ease' fiscal fiscal policy. have suggested Elsewhere,II have suggested that availableempirical that available empirical evidence implies that plans to increase taxes as as a way evidenceimplies that plans increasetaxes way out of budget budget deficits deficits are are plans plans that carry the potential for raising government that carrythe potentialfor raisinggovernment 3 Coupled with spending possibly· future the evidence and possibly evidence future deficits deficits as as well. well.3 Coupled with the spending and we should presented here, potential of tax shouldalso also recognize the potential tax increases increasesto raise raise here, we presented recognizethe the level level of government the participation in a country and, accordingly, exert inin a inexert governmentparticipation countryand, accordingly, verse influences performance as verse influenceson its its future futureeconomic economicperformance as well. well. As in As suggested suggestedin my paper, the 1986paper, the empirical evidencemay the following While my 1986 empiricalevidence may suggest suggestthe irony:While followingirony: political participants participants may cravelarger and larger non-marketresource resourceallocaallocapolitical may crave largerand largernon-market tions, their future ability to satisfy that craving may very well be severely their future that well be contions, ability satisfy cravingmay very severelyconstrained by the appetite. strainedby the satisfaction satisfactionof that that same same appetite. Notes Notes 1. It is interesting 1. potential insights insights that that It is that Saunders Saunders(1986: concludesthe the opposite. 'The potential (1986:58) interestingthat 58) concludes opposite. 'The might comparative crossfollow from crossfrom further furtherstudy the Japanese within a comparative might follow study of the Japaneseexperience experiencewithin country proceed frameworkwould wouldappear considerable.This This is morefruitful fruitfulway is surely countryframework surelyaa more appearconsiderable. way to proceed than to discard discardthe by treating as a statistical than the informational informationalcontents contentsof the the Japanese case by it as statistical Japanesecase treatingit outlierwhich whichdistorts outlier distorts"more "more reliable" reliable"relationships.' relationships.' 2. The Theinterested 2. and Russek Russek (1987) useful discussion discussion of the interestedreader readercould couldsee see Barth, Keleherand for a useful the Barth,Keleher (1987)for available economic growth empirical evieviavailableevidence on the size - economic relation.Recent Recentempirical evidenceon the government governmentsize growthrelation. inverse relation between government economic growth growth dence denceon on the the consistency an inverse sizeand andeconomic relationbetween consistencyof an governmentsize is contained contained in Barth and (1987). within withindifferent in Barth and Bradley differentequation equationspecifications Bradley(1987). specificationsis and 3. Marlow discusses the 3. Marlow(1987) on the the tax-spend the available availableempirical evidenceon tax-spendhypothesis (1987) discusses hypothesisand empiricalevidence budget constraint argues constrainttype the hypothesis evenif one one does does not not accept the budget that, even arguesthat, acceptthe type hypothesis, hypothesis hypothesis,the that expenditures will legislated tax increases can can not not be be reasonably reasonably rejected. rejected. that willalso withlegislated alsorise risealong taxincreases expenditures alongwith
References References in the the United UnitedStates States1953Aschauer, Net private private investment and public public expenditure expenditure in investmentand Aschauer,D.A. (1987). (1987).Net 1984. Bank of Chicago. 1984. Occasional OccasionalPaper FederalReserve ReserveBank SM-87-10, Federal Chicago. PaperSM-87-10, London:MacMacBacon, Britain's economic problem: Too producers. London: Toofew producers. W. (1978). and Eltis, economicproblem: Bacon,R., R., and Eltis, W. (1978).Britain's millan. millan. spending on Barth, .R., and on economic economicactivity. andBradley, M.D. (1987). Theimpact activity. Barth,JJ.R., Bradley,M.D. (1987).The impactofgovernment of governmentspending Unpublished manuscript. Unpublishedmanuscript.
Barth, scale of of government government and and economic economic F.S. (1987). and Russek, The scale Russek, F.S. Barth, J.R., J.R., Keleher, Keleher,R., R., and (1987). The manuscript. activity. activity.Unpublished Unpublishedmanuscript. National Tax Journal 29: 29: Beck, public sector: evidence. National TaxJournal M. (1976). Theexpanding sector:Some Somecontrary Beck,M. contraryevidence. (1976).The expandingpublic 15-21. Beck, M. growth: A real perspective. Public Public Finance Finance 34: 34: 313-356. M. (1979). real perspective. Publicsector sectorgrowth: Beck, (1979).Public Beck, M. (1985). Public and resource allocation. Public Public Finance Finance 30: M. (1985). relativeprices, resourceallocation. 30: Publicexpenditure, Beck, prices,and expenditure,relative 17-34. Cameron, D.R. expansion of the the public public economy: economy: A comparative comparative analysis. analysis. American American D.R. (1978). Theexpansion Cameron, (1978).The Political Science Review 72: 72: 1243-61. Political ScienceReview 1243-61. Gould, F. (1983). The development of public public expenditures in Western Western industrialized industrialized countries: countries: A F. (1983). Thedevelopment Gould, expendituresin comparative Public Finance Finance 38: 38: 38-69. comparativeanalysis. analysis.Public Heller, in the P.S. (1981). trendsin the shares sharesof nominal nominaland and real realgovernment Heller,P.S. (1981).Diverging Divergingtrends governmentexpenditures expenditures National Tax Journal 34: in policy. National in GDP: GDP: Implications TaxJournal for policy. 34: 61-74. Implicationsfor Landau, and economic Governmentexpenditure economic growth: Landau, D. (1983). (1983). Government expenditureand growth:A cross-country cross-countrystudy. study. 49: 783-792. Southern Economic Journal Journal 49: SouthernEconomic PubMarlow, M.L. (1986). Privatesector sectorshrinkage and the the growth industrializedeconomies. economies.PubMarlow,M.L. (1986).Private shrinkageand growthof industrialized lic lic Choice 49: 143-154. Choice49: 143-154. Marlow, spend hypothesis: Empirical evidence M.L. (1987). Taxand and spend and implications. evidenceand Marlow,M.L. hypothesis:Empirical (1987). Tax implications.UnpubUnpublished paper. Washington, lished paper. DC. Washington,DC. K. (1983). Links between growth: Some Some empirical World Marsden, betweentaxes taxesand and economic economicgrowth: evidence.World Marsden,K. (1983).Links empiricalevidence. Bank Staff Bank Staff Working DC. 605, Washington, WorkingPaper, Paper, No. 605, Washington,DC. OECD. Problems of growth and OECD. OECD. (1985). Socialexpenditure 1960-1990: Problems and control. control. Paris: Paris:OECD. (1985).Social expenditure1960-1990: of growth W. (1984). Orzechowski, perspectives on policy and growth. EconomEconomInternationalperspectives on tax tax policy andeconomic economicgrowth. Orzechowski,W. (1984).International ic Outlook. Outlook.Washington, DC: US US Chamber Chamberof Commerce. Commerce. Washington,DC: performance in Journal Saunders, Publicexpenditure and economic in OECD economicperformance OECDcountries. countries.Journal Saunders,P. (1985). (1985).Public expenditureand of Public Policy Policy 5: 5: 1-21. of Public Saunders, public sector Whatcan we learn can we learnfrom frominternational internationalcomparisons sectorsize and size and Saunders,P. (1986). (1986).What comparisonsof public economic performance? European European Sociological Review 2: economicperformance? 2: 52-60. SociologicalReview Saunders, Privatesector sectorshrinkage and the the growth industrializedeconomies: economies:ComComSaunders,P. (1988). (1988).Private shrinkageand growthof industrialized Public Choice ment. ment. Public Choice 58: 58: 277-284. Smith, performance. National National Westminister Bank Publicconsumption and economic economicperformance. Westminister Bank Smith,D. (1975). (1975).Public consumptionand Review 17-30. Review E. (1984). Weede, E. socialism in in rent-seeking sociand ideological Weede, socialism,and (1984).Democracy, Democracy,creeping creepingsocialism, ideologicalsocialism rent-seekingsociPublic Choice eties. eties. Public 44: 349-365. Choice44: