JEL CLASSIFICATION: H4, H5, M13 KEY WORDS: Public Capital, Public Investment, Small Business, Entrepreneurship

Transcription

1 Ths Draft: 1 October 003 Publshed: March 004, Atlantc Economc Journal 3:1 Publc Captal and Small Frm Performance Kevn Chrst* Rose-Hulman Insttute of Technology Rchard Green** Unversty of the Incarnate Word ABSTRACT: Prevous research on the role of publc captal (nfrastructure) n prvate producton has emphaszed potental complementartes between publc and prvate captal at an aggregate level. Presumably such effects, f they exst, arse from benefts enjoyed by ndvdual unts of producton. Because of the potental for them to be locaton-specfc or captal constraned, t s concevable that small busnesses may beneft dsproportonately from publc captal. Tests usng fnancal data for 871 small frms from 199 to 1996 ndcate a postve and statstcally sgnfcant elastcty between prvate labor productvty and the level of publc captal n the area where the frms are located. Such a postve elastcty provdes further evdence that publc and prvate captal are complementary nputs nto producton, and has mportant polcy mplcatons. JEL CLASSIFICATION: H4, H5, M13 KEY WORDS: Publc Captal, Publc Investment, Small Busness, Entrepreneurshp * Department of Humantes and Socal Scences, Rose-Hulman Insttute of Technology, 5500 Wabash Avenue, Terre Haute, IN Tel Fax: E-mal: kevn.chrst@rose-hulman.edu ** School of Busness and Appled Arts and Admnstraton, Unversty of Incarnate Word, UPO #318, 4301 Broadway, San Antono, Texas Tel E-mal: greenr@unverse.uwtx.edu Ths research has been supported through a grant from the Kauffman Center for Entrepreneural Leadershp, the support of whch the authors gratefully acknowledge.

2 1. Introducton Regons and states exhbt mportant dfferences when t comes to the apparent health of ther busness envronments for entrepreneural actvty. There are clearly dfferences among the states wth respect to small-busness starts and falures, and dstnctons among regons and states may extend to fnancal performance. Table 1 dsplays one set of statstcs for 1998, concernng the vablty of new and small busness actvty across the states. Rates of busness startups range from a low of.36 to a hgh of.84 starts per 1,000 resdents. Ratos of busness falures to busness starts exhbt smlar varablty, rangng from a low of 1.0 to a hgh of Such statstcs lead naturally to questons about what makes one regon or state more entrepreneural than another for new busness development and success. Ths paper nvestgates the potental role of publc captal, or nfrastructure, n creatng or sustanng an envronment conducve to the vablty and success of small busnesses. Specfcally, t nvestgates the possble nfluence of locaton n areas wth relatvely hgh amounts of publc captal on the fnancal performance of ndvdual frms. Whle not all small busnesses are entrepreneural enterprses n the sense n whch that phrase s often used, t s probably the case that most new busnesses begn small. Thus, even though entrepreneuralsm s often equated wth self-employment, not smallness of enterprse, the authors beleve that our focus on the fnancal performance of small busnesses represents a sutable proxy for entrepreneuralsm. The startng pont for the analyss n ths paper s the proposton that small busnesses may beneft dsproportonately from hgh levels or hgh qualty of nfrastructure. Small busnesses face many constrants that large frms do not, such as 1

3 restrcted access to captal and low levels of purchasng power n nput markets. Addtonally, t has long been recognzed that small frms are "locus locked,".e. small frms, unlke large corporatons, are economcally "ted" to the state and cty of the resdence of the entrepreneur (Schmenner, 198). Because of these (and other) fnancal and busness decson constrants, small frms may beneft comparatvely more than do large frms from free nputs such as publc nfrastructure. The polcy mplcaton of ths proposton s that publc captal nvestment n nfrastructure may contrbute to an enhanced entrepreneural envronment by makng the busness envronment more conducve to small busness success.. Publc Captal and Productvty The potental contrbuton of publc captal to regonal economc envronments has been the subject of study snce at least the md 1960s. Early dscussons of the topc dstngushed between socal publc captal (educatonal and publc health facltes) and economc publc captal (nfrastructure that ndrectly supports producton actvtes), and outlned the dfferental roles these categores of publc captal may play n contrbutng to regonal economc growth [Hansen, 1965]. Early research nto the effects of publc captal nvestment generally followed three lnes of nvestgaton: (1) the relatonshp between publc captal nvestment and economc development as measured by changes n populaton, employment, ncome or the number of frms; () the effect of such nvestment upon over-all productvty; and (3) the effect of such nvestment upon new busness starts [Bartk, 1991; Wasylenko, 1991; Munnell, 199; Fox and Murray, 1993; and Fsher, 1997].

4 There are at least three ways n whch publc captal nvestment could affect busness performance: (1) as an un-prced nput to producton, () as a reducton n the prce pad by busness for servces provded by publc captal nvestment, and (3) as a producton complement to prvate nputs that generates reductons n the user cost of those prvate nputs [Bartk,1991]. Some examples of these effects are obvous. For any ndvdual frm, publc hghways provde an un-prced nput; publc arports reduce the cost of ar transportaton; and publc educaton reduces the qualty-adjusted prce of labor. One study that specfcally addressed the possble lnk between publc captal nvestment and entrepreneural actvty nvestgated the relatonshp between hghway mles per square mle upon (1) small busness startups and () new busness plants, and reported postve, but not statstcally sgnfcant relatonshps [Bartk, 1989, 1991]. Specfc estmates of the aggregate cost savngs provded to prvate enterprse by publc captal nvestment can be both surprsng and controversal. One set of research, usng annual state-level data from 1970 to 1987 to study the effects of hghways and water and sewer systems upon manufacturng costs found that there were large, sgnfcant savngs for prvate enterprse [Morrson and Schwartz,1996]. In 198, for example, the margnal dollar of such nfrastructure n the South saved manufacturers an estmated 31 cents per year n operatng costs; whle n other regons, the analogous fgures ranged from 16 cents to 18 cents. In the early 1990s, research n ths area tended to focus on what came to be known as the "publc captal hypothess", wth specfc nvestgatons nto the role of publc nvestment on regonal and natonal productvty and output [Aschauer, 1989, 1997; Munnell,1990a]. Theoretcal and emprcal nqures were prompted by the rough 3

5 concdence of a slowdown n U.S. publc nfrastructure nvestment wth a declne n U.S. productvty growth n the late 1960s and 1970s [Berndt and Hansson, 1991; Hulten and Schwab, 199]. Generally, the hypothess argued that nvestment n nonmltary publc captal had both a drect and an ndrect nfluence on natonal output. The drect effect arses as publc captal, such as transportaton systems, provdes drect nputs to the producton process. The ndrect effect arses from the asserton that publc captal makes prvate captal more productve that the two knds of captal are complements n the producton process. If publc captal and prvate captal are complements, then expendtures on publc captal should be assocated wth hgher margnal productvty of prvate captal. By extenson, prvate frms should be able to reap benefts of publc nvestment, and realze hgher returns on ther own nvestment. Standard tests of the publc captal hypothess (or varants of t) typcally employ Bureau of Economc Analyss data on publc captal stocks, dsaggregated among the 48 contguous states [Costa, Ellson, and Martn, 1987; Munnell, 1990b]. Regardng functonal forms used n emprcal nvestgatons, prevous research nto potentally benefcal effects of publc captal on the prvate sector generally falls nto one of two categores reflectve of the dual nature of the frm s optmzaton problem. The frst category of research uses a producton functon approach to test whether publc and prvate captal are complements n the producton process [Aschauer, 1989, 1997; Munnell,1990a; and Tatom, 1991]. A second category of emprcal research uses cost functons to test whether publc captal reduces the costs of producton [Lynde and Rchmond, 199; and Morrson and Schwartz, 1996]. Ths research has generated a varety of estmates concernng the mportance of nfrastructure nvestment. No 4

6 researchers, to our knowledge, have concluded that nvestments n publc captal are detrmental. Some have concluded that the nfluence s neglgble. Most, however, have concluded that publc captal does have some benefcal effects at an aggregate level, partcularly for manufacturng. All of ths earler research nvestgates the nfluence of publc captal at an aggregate level, testng for nfluence at a natonal or regonal level. Presumably, however, f publc captal does generate benefcal producton effects, those effects must orgnate at a basc level of producton such as the ndvdual frm. Testng for ths potental nfluence s best done wth data for small frms, snce such frms are more lkely to undertake nvestments locally. 1 Thus, dfferences among the states n levels of publc captal mght be correlated wth dfferences n returns for small frms. 3. Publc Captal and Prvate Inputs To model the potental nfluence of publc captal on prvate producton, the analyss descrbed here begns wth a smple three-factor producton functon for frm output: ( N, K G ) Y = Af, (1) Where Y = the market value of frm 's output, A s a technology parameter, N and K are measures of the frm's labor and prvate captal nputs respectvely, and G represents a flow of servces from the publc sector. In adoptng ths econometrc approach, the 1 Large frms often undertake busness nvestment across state and nternatonal borders. An underlyng assumpton used here s that small frms are more lkely to undertake busness nvestment n a local area, specfcally, wthn one state. 5

7 authors focus solely on the role of publc captal as a free nput to the frm's producton process, and on the possble provdental effects of that nput on labor productvty. Ths paper does not focus drectly on the queston of whether publc captal reduces costs of producton. Ths choce of functonal form places ths paper wthn the realm of the frst category of emprcal research mentoned at the end of secton. Adopton of a generalzed Cobb-Douglas technology, and an assumpton that publc servces are n proporton to the publc captal stock allows restatement of the producton functon n logarthmc form: y = a0 + a1n + ak + a3g + ε () where the lower case represents the natural log of the varables, and g s some measure of publc captal stock relevant at the frm level. Assumpton of constant returns to scale allows expresson n factor-ntensve form, yeldng output per unt of labor nput: y ( k n ) + b ( g n ) + ε n = b + b1 0 (3) In ths form, the coeffcent b s nterpretable as an elastcty relatng output per unt of labor to the relatve ntensty of publc captal. If publc captal contrbutes to ndvdual frm performance va hgher output per labor, then the coeffcent b wll be postve. Alternatvely, assume that the producton functon exhbts constant returns to scale n the prvate nputs, but ncreasng returns to scale over all nputs, prvate and publc. If ths were the case, then n equaton () a + a 1, whle a + a + a 1, and the revsed factor ntensve form of equaton (3) s 1 = 1 3 > y ( k n ) + d g + ε n = d + d1 0 (4) Equatons (3) and (4) correspond to the equatons to be estmated: y n ( k n ) + b ( g n ) + αx + βz + ε = b0 + b1 (5) 6

8 y n ( k n ) + d g + αx + βz + ε = d0 + d1 (6) where y s the natural log of the frm's net sales, k s the natural log of the frm's total assets, n s the natural log of the number of people employed by the frm, g s the natural log of the publc captal stock per square mle n the state n whch the frm s located. X s a vector of state and ndustry dummy varables, and Z s a vector of frm-specfc dummy varables capturng the age and ownershp characterstcs of the frms. The ncluson of such frm characterstcs follows common conventon n entrepreneural research [Brd, 1989; Blanchflower and Oswald, 1998; Brock and Evans, 1986; Devne, 1994; Schller and Crewson, 1997; Wortman, 1987]. 4. Emprcal Analyss The emprcal analyss estmates the foregong models for 881 small frms durng the fve years from 199 to Frm level data on net sales, employment, and total assets, as well as frm characterstcs are derved from the Kauffman Fnancal Statements database. Whle most entrepreneural research equates entrepreneural actvty wth self employment, the authors use data from small busnesses as a proxy for entrepreneural actvty n order to contrbute somethng useful to the dscusson of condtons necessary for entrepreneural success. Presumably, f entrepreneurs are successful, they eventually ether grow nto a small busness or sell the busness concept to other, larger, successful frms. Further, not all self-employed people fll the tradtonal role of entrepreneur one who conceves of and brngs to fruton a new busness concept. For example, someone who cleans houses for others s only an entrepreneur n a lmted sense. Thus, when 7

9 selectng a metrc for entrepreneural actvty, small busness actvty may be just as useful a proxy as self employment. Data on state-level publc captal stocks are taken from the authors' own estmates of such stocks, and are assgned to ndvdual frms based on the state locaton of those frms. Thus, whle the unt of analyss s the frm, the key varable under nvestgaton s common to all frms wthn a state. Ths rases one potental caveat n nterpretaton of any results. Ths caveat s best explaned wth a hypothetcal example. Imagne two frms, dentcal n all respects, the frst of whch resdes n a "hgh" publc captal state, and second of whch s located n a "low" publc captal state. If the frst frm enjoys hgher labor productvty, when controllng for all other observable trats, can one assume that t s the nfluence publc captal that accounts for the hgher labor productvty? Not necessarly, because publc captal ntensty can vary greatly wthn a state. The authors cannot rule out the possblty that the frst frm was located n an area of the hgh publc captal state that actually enjoyed lttle access to that publc captal, and that the hgher labor productvty s actually attrbutable to some other unobserved nfluence. Heren resdes a lmtaton of ths analyss that could only be overcome wth a more fnte matchng up of publc captal ntensty and frm locaton. Matchng up frms wth ther physcal surroundngs as such a level of detal s beyond the scope of possblty for now. Another challenge n makng the theoretcal model operatonal s that the data on publc captal s an aggregate (state-level) varable, whle the other varables are frmspecfc. The publc captal nput, therefore, must be scaled down n some fashon so as to make t relevant at the frm level. One possblty s to measure t on a per capta bass. 8

10 In 199 the per capta publc captal stock for the Unted States was $10,80. Publc captal stocks, however, are probably drven by geography as well as populaton. Thus n 199 the state wth the hghest per capta publc captal stock was Wyomng at $0,093, and the state wth the lowest per capta publc captal stock was New Hampshre at $6,47. Snce ths study consders the nfluence of publc captal on the bass of physcal locaton, the authors have chosen nstead to use a varable that measures the relatve spatal ntensty of publc captal, specfcally publc captal per square mle, measured at the state level. In 199, the mean value of ths varable was $1,653,066. The state wth the hghest publc captal ntensty, as measured n ths manner, was New Jersey, and the state wth the lowest was Montana. Ths seems to be a more realstc approach to the ssue at hand than the use of per-capta data. The use of publc captal per square mle, however, ntroduces one other consderaton. As constructed, publc captal per square mle s hghly correlated wth populaton densty. Therefore, f the coeffcent estmate on ths varable s postve, one mght rghtfully ask whether t s the publc servces nput that contrbutes to hgher labor productvty, or whether there s somethng ntrnsc about operatng n a area of dense populaton. It s the authors' contenton that the state dummy varables adequately capture and control for state characterstcs such as populaton densty. When dong emprcal work on publc captal, the way n whch estmates of t are constructed are very mportant because publc captal bulds up and deprecates over very long tme perods. The publc captal stock at a gven pont n hstory s the culmnaton of a long hstorcal process. The Bureau of Economc Analyss has publshed aggregate 9

11 estmates of publc captal stocks (tangble wealth) for the U.S. from 195 to As brefly noted n Secton, dsaggregaton of such data nto a state-by-state panel s central to most nvestgatons of the regonal effects of publc captal on prvate producton. Generally, the dsaggregaton process nvolves complaton of state-level data on captal expendtures for long tme perods, deprecaton of dfferent classes of captal expendtures, and the "buldng up" of publc captal stock estmates for each state. For ths paper, the authors compled captal expendture data from 1957 forward, employed BEA deprecaton technques, and generated a state-level panel of the net stock of publc captal for 1970 to The authors then compared ther own aggregate captal stock estmate to that of the BEA, and found that for 1970 to 1996 the authors' estmates were never at varance wth the BEA total by more than %. Fnally, the authors compared the propertes of ther publc captal panel wth the state-by-state data prevously employed by other researchers, and were satsfed that the estmates of statelevel publc captal stocks employed here were n concert wth, and n some respects, superor to those employed n prevous research. Descrptve statstcs for key varables are provded n Table. The sample s lmted to frms wth at least fve employees, but fewer than 500. The average frm n the sample has 48 employees. Frms wth fewer than fve employees were excluded n an attempt to ensure that the sample truly was comprsed of frms n the tradtonal sense of a busness wth a payroll. Net sales were used nstead of net ncome as a proxy for the market value of output nstead of net ncome. The authors hypothesze that for small Detals concernng the constructon of ths new data set on publc captal are avalable from the authors upon request. 10

12 (and often prvately held) frms, net sales offers a better gauge of labor productvty than does net ncome. Total assets serve as a proxy for captal nputs. Publc captal per square mle s measured n current year dollars, and ranges from $63,000 (Montana, 199) to $1.7 mllon (New Jersey, 1996), wth a mean value of about $1.7 mllon (199). Generally, the data exhbt sgnfcant, and potentally meanngful, dfferences among the states n the level and growth rates of publc captal stocks. The medan growth rate of publc captal stock from 1970 to 1996 (not reported n Table ) was 3.8%, whle the mean growth rate was 3.3%. West Vrgna accumulated captal at the slowest pace,.1%, whle Nevada s publc captal stock grew at the fastest pace 4.86% annually. Over a 7-year perod, such dfferences n growth rate result n sgnfcant changes n the dstrbuton of publc captal across states. Of course such dfferences are naturally drven by geography and populaton trends. Measurng publc captal on a per square mle bass factors n the geographc element. Tables 3 presents OLS regresson results for equaton 5 for the full sample of frms, as well as for manufacturng and servce sectors. Generally, the results presented here are robust to random changes n sample composton and sze. Both the prvate and publc captal elastcty coeffcents are measured wth a hgh degree of precson and have the expected sgns. The prmary coeffcent of nterest, b, may be nterpreted as the elastcty of the output per worker to publc captal ntensty per worker. Specfcally, the dependant varable measures sales per employee, whle the key ndependent varable measures publc captal per square mle per employee. The 199 full sample coeffcent estmate for b of mples that a 1% change n the rato of publc captal per square 11

13 mle to a frm s employee level s assocated wth an 11 bass pont ncrease n the rato of net sales per employee. To put ths result nto perspectve, consder that n 199 the average frm n the sample had 48 employee, and sales of $4.698 mllon on assets of $.337 mllon, mplyng a sales to asset rato of.01, and $97,468 sales per employee. A 1% ncrease n the publc captal varable, assocated wth an 11 bass pont ncrease n sales per employee, would ncrease sales per employee to $97,581. Whle statstcally sgnfcant, the economc mportance of ths result may be open to queston. Nevertheless, t s mportant to recall that sgnfcant dfferences exst among the states wth regard to the spatal ntensty of publc captal. For example, Oho's publc captal per square mle s 74% greater than ts neghborng state to the west, Indana. These results mply that ncreases n the rato of publc captal per square mle are assocated wth hgher labor productvty (and, by extenson, hgher prvate captal productvty). Whle at the frm level, these dfferences may be slght, the dfferences among the states are sgnfcant enough to be meanngful for frms that are makng locaton decsons. There s a sharp dfference between ths coeffcent estmate for the manufacturng and servces subsets of the sample. The publc captal elastcty appears to be stronger for servce frms than for manufacturng frms. Servce frms are tradtonally less captal ntensve than are manufacturng frms. Such results are at odds wth fndngs of prevous research. One potental nterpretaton of these results s that busnesses for whch prvate captal nputs are relatvely less sgnfcant beneft dsproportonately from publc captal nputs. A potental polcy mplcaton of such a fndng would be that 1

14 publc captal plays a relatvely mportant role for areas seekng to attract and noursh a servce-orented economy. Regardng senstvty of these results to functonal form, Table 4 presents the analogous results for equaton 6, n whch the assumpton of constant returns to scale across all nputs s dropped. Here evdence of a strong relatonshp between publc captal and prvate output dsappears, suggestng that the results may be dependent on functonal form. The dfference n results between Tables 3 and 4 also may suggest that ncreasng returns to scale n publc and prvate nputs s an unreasonable hypothess at the ndvdual frm level. None of the frm-specfc age and ownershp varables exhbt any sgnfcant nfluence. More than anythng, such a lack of statstcal sgnfcance probably supports an effcent markets hypothess. At least for ths data, the age or tenure of frms or the characterstcs of ts ownershp do not seem to grant frms any specfc benefts or mpose any specfc barrers. 5. Summary The role of small busnesses n a market economy has been the subject of volumes of research, and the belef that small busnesses are pvotal to economc development has led to many polcy ntatves amed at assstng small busnesses. At the same tme, urban and regonal economsts have long pondered the factors that contrbute to a vbrant busness envronment. Ths paper brdges the gap between these two felds of study by nvestgatng the potental nfluence of publc captal usually thought of as nfrastructure on the vablty and success of small busnesses. 13

15 One obvous queston that proceeds from the analyss n secton 4 s what aggregate value would such ncreases n returns on prvate assets have f the analyss were extended to the economy as a whole? That queston s not part of our analyss, but has been addressed ndrectly by those who have studed the effect of publc captal usng aggregate producton functons. Such studes have usually found a postve effect. The authors' perspectve s from the other end of the mcroscope. If postve returns to publc captal accrue n the aggregate, then they must arse va uncountable small effects on frms throughout the economy. In ths respect, ths study s the frst of ts knd to test for frm-level effects. One would not have expected for these effects to be earth-shatterng n magntude. That they seem to exst at all lends further credence to the vew that publc captal has postve external effects on prvate busness, and suggest that these effects may beneft small frms n partcular. One mght extrapolate from these results that nvestment n publc captal should be ncreased on all fronts. Ths may not be so. Publc captal may have nonlnear effects on prvate productvty [Aschauer, 1997]. In other words, there may exst an optmum rato of publc to prvate captal, and by extenson to our paper, an optmal level of publc captal per square mle. Such an optmal level mght easly be expected to dffer across states because of ther dfferent geographc natures. One would not expect the same optmal level of publc captal per square mle to apply wth equal valdty to New Jersey and Montana. Ths s an area open to further study. Nevertheless, the results of ths paper do have some clear polcy mplcatons. The noton that publc spendng on nfrastructure does contrbute to an envronment conducve to the success of small busness provdes an mportant argument for 14

16 mantanng and extendng hgh qualty nfrastructure. Furthermore, t wll be mportant to keet ths potental relatonshp n mnd as the meanng of publc nfrastructure changes. Today s nfrastructure ntatves are as lkely to be the layng of fber optc networks as the constructon of hghways. Because t was constructed from hstorcal data on state captal outlays, the publc captal data used n ths study s heavly weghted wth the latter type of publc captal. There would seem to be extensve opportuntes to nvestgate the nfluence of new economy nfrastructure on prvate busness actvty. Such observatons pont the way for future research. 15

17 References Aschauer, Davd A. Is Publc Expendture Productve? Journal of Monetary Economcs 3 (1989), Aschauer, Davd A. Output and Employment Effects of Publc Captal. Workng Paper (Aprl 1997). Bartk, Tmothy J. (1991) Who Benefts from State and Local Economc Development Polces? Kalmazoo, MI: W.E. Upjohn Insttute. Berndt, Ernst R., and Bengt Hansson. Measurng the Contrbuton of Publc Infrastructure Captal n Sweden. NBER Workng Paper N. 384 (September 1991). Brd, Barbara J. Entrepreneural Behavor. Glenvew: Scott, Foresman and Company, Blanchflower, Davd G. and Andrew J. Oswald. What Makes an Entrepreneur? Journal of Labor Economcs 16:1 (1998), Brock, Wllam, and Davd Evans. The Economcs of Small Busness. New York: Holmes and Meer (1986). 16

18 Devne, Theresa J. Characterstcs of Self-employed Women n the Unted States. Monthly Labor Revew (March 1994), Costa, Jose de Slva, Rchard W. Ellson and Randolph C. Martn. Publc Captal, Regonal Output, and Development: Some Emprcal Evdence. Journal of Regonal Scence 7:3 (1987), Fsher, Ronald C. (1997) "The Effects of State and Local Publc Servces on Economc Development," New England Economc Revew. Boston: Mar/Apr: Fox, Wllam F. and Matthew N. Murray (1993) "State and Local Government Polces." In Davd Barkley, ed., Economc Adaptaton: Alternatves for Rural Amerca. Boulder, CO: Westvew Press, Inc. Hansen, Nles. "Unbalanced Growth and Regonal Development." Western Economc Journal 4 (1965), Hulten, Charles R., and Robert M. Schwab. Infrastructure Spendng: Where Do We Go From Here? Natonal Tax Journal 46 (199), Lynde, Catherne, and James Rchmond. "The Role of Publc Captal n Producton." The Revew of Economcs and Statstcs 74 (199),

19 Morrson, Catherne J. and Amy E. Schwartz. (1996) "State Infrastructure and Productve Performance." The Amercan Economc Revew; Nashvlle; Dec 86(5): Munnell, Alca H. (199) "Infrastructure Investment and Economc Growth." Journal of Economc Perspectves, vol. 6: Munnell, Alca H. Why Has Productvty Growth Declned? Productvty and Publc Investment. New England Economc Revew (January/February 1990), 3-. Munnell, Alca H. How Does Publc Infrastructure Affect Regonal Economc Performance? New England Economc Revew (September/October 1990), Schller, B. R., and P.E. Crewson. Entrepreneural Orgns: A Longtudnal Inqury. Economc Inqury 35 (1997), Schmenner, Roger W. Makng Busness Locaton Decsons. Englewood Clffs, NJ: Prentce-Hall (198). Tatom, John A. "Publc Captal and Prvate Sector Performance." The Federal Reserve Bank of St. Lous Revew 73 (May/June 1991),

20 Wasylenko, Mchael J. (1991) "Emprcal Evdence on Interregonal Busness Locaton Decsons and the Role of Fscal Incentves n Economc Development." In Henry Herzog and Alan Schlottmann, eds., Industry Locaton and Publc Polcy. Knoxvlle, TN. Unversty of Tennessee Press. Wortman, Max. Entrepreneurshp: An Integratng Typology and Evaluaton of the Emprcal Research n the Feld. Journal of Management 13: (1987),

21 Table 1 Busness Starts and Falures by State, 1998 Mean Standard Devaton Mnmum Maxmum Starts/Falures Rato Starts/Falures Rato* Starts per 1,000 resdents Starts per 1,000 resdents* Falures per 1,000 resdents Falures per 1,000 resdents* Source: Statstcal Abstract of the Unted States, * Trmmed statstcs, excludng top and bottom two states of the dstrbuton. Such statstcs are more meanngful because they exclude the nfluence of Delaware, a clear outler n terms of busness ncorporatons. 0

22 Table Descrptve Statstcs Standard Mean Medan Mnmum Maxmum Devaton Employees Net Sales (000s) 199 4,698 1, ,000 9, ,48, ,000 10, ,030, ,000 14, ,75, ,000 17, ,167, ,000 0,481 Total Assets (000s) 199, ,687 1,39 11,000 6, , ,464 13,133 94,61 6, ,84 946,15 18, ,000 7, ,044 1,079,470 15,561 04,000 8, ,33 1,097,137 7, ,000 13,069 Publc Captal / Sq. M. (000s) 199 1, ,55, , ,085, , ,607, , ,187, , ,79,791 Fnancal data and employment statstcs are based on frm level data (N = 871). Publc Captal per Square Mle statstcs are based on state-level data as descrbed n the Secton 3 (N = 48). 1

23 y n = b Table 3 OLS Regresson Results + b k n + b n Model: 0 1( ) ( ) g + αx + βz Full Sample N ε b 4.079** 0 (0.353) 4.389** (0.35) 4.496** (0.33) 4.633** (0.344) 4.610** (0.36) b 0.598** 1 (0.0) 0.591** (0.03) 0.581** (0.01) 0.583** (0.0) 0.555** (0.0) b 0.115** (0.00) 0.098** (0.00) 0.101** (0.018) 0.093** (0.018) 0.117** (0.019) R Manufacturng N b 3.183** 0 (0.64) 3.678** (0.58) 4.063** (0.536) 4.148** (0.547) 4.306** (0.539) b 0.736** 1 (0.041) 0.683** (0.039) 0.663** (0.036) 0.668** (0.036) 0.638** (0.034) b 0.105** (0.03) 0.086** (0.08) 0.069** (0.05) 0.051* (0.030) 0.06* (0.06) R Servces N b 3.756** 0 (0.896) 3.764** (0.90) 3.891** (0.917) 3.301** (0.91) 3.780** (1.05) b 0.497** 1 (0.047) 0.477** (0.048) 0.48** (0.050) 0.544** (0.05) 0.470** (0.055) b 0.3** (0.049) 0.39** (0.048) 0.19** (0.046) 0.11** (0.045) 0.57** (0.05) R Standard errors are provded n parentheses. ** Statstcally sgnfcant at a 1% level of confdence. * Statstcally sgnfcant at a 5% level of confdence.

24 y n Table 4 OLS Regresson Results = d + d k n + d Model: 0 1( ) g + αx + βz Full Sample N ε d 5.657** 0 (0.548) 5.59** (0.541) 5.685** (0.496) 5.935** (0.55) 6.163** (0.557) d 0.614** 1 (0.01) 0.607** (0.0) 0.595** (0.00) 0.597** (0.01) 0.564** (0.0) d (0.037) (0.036) (0.03) (0.033) (0.035) R Manufacturng N d 3.49** 0 (1.103) 3.75** (0.871) 3.780** (0.790) 3.994** (0.815) 3.898** (0.800) d 0.713** 1 (0.039) 0.670** (0.035) 0.660** (0.03) 0.650** (0.03) 0.65** (0.031) d (0.068) (0.058) (0.05) (0.053) (0.051) R Servces N d 5.011** 0 (1.306) 3.886** (1.38) 4.811** (1.307) 4.14** (1.31) 5.004** (1.485) d 0.505** 1 (0.044) 0.504** (0.046) 0.500** (0.047) 0.553** (0.048) 0.47** (0.051) d 0.08 (0.08) (0.081) (0.077) (0.075) 0.065** (0.086) R Standard errors are provded n parentheses. ** Statstcally sgnfcant at a 1% level of confdence. * Statstcally sgnfcant at a 5% level of confdence. 3