Journal of Agricultural and Applied Economics, 32,3(December 2000):565-57202000 Southern Agricultural Economics Association

Social Capital and Economic Growth: ACounty-Level Analysis

Anil Rupasingha, Stephan J. Goetz, and David Freshwater

ABSTRACT

The effect of social capital on economic growth is examined using linear regression anal-ysis and U.S. county-level data. Results reveal that social capital has a statistically signif-icant, independent positive effect on the rate of per-capita income growth.

Key Words: economic growth, social capital, U.S. counties.

The growing literature on social capital hasadded a new dimension to studies of economicgrowth, as mainstream economists start to an-alyze the roles of culture and social ties asfactors of production similar to physical andhuman capital. In particular, economists in-creasingly recognize that interactions amongpeople and their feelings for each other haveclear and predictable economic implications(Robison).

Despite this growing recognition, the effectof social capital on economic growth at thelevel of U.S. counties has not been formallyand systematically analyzed. In this paper theconventional B arro-type empirical growthmodel is expanded to test for the independenteffects of social capital or civic engagementon economic growth. Our results suggest thatthe level of social capital had a statistically

Rupasingha is a post-doctoral research associate atTVA Rural Studies in the Department of AgriculturalEconomics at the University of Kentucky. Goetz isprofessor of Agricultural Economics at Penn StateUni-versity and Director of the Northeast Regional Centerfor Rural Development. Freshwater is professor of Ag-ricultural Economics at University of Kentucky. Cor-respondence: Anil Rupasingha, 400 Agr. Eng. Bldg.,University of Kentucky, Lexington, KY 40546-0276.E-mail: anil@rural.org.

The authors thank two reviewers for their com-ments.

significant, independent positive effect on therate of per-capita income growth in U.S. coun-ties between 1990 and 1996. At the same time,excluding the social capital vector does notmaterially change the estimate of the conver-gence parameter or the effect of human capitalon growth.

Literature and Conceptual Background

Social capital in its present form was firstidentified by Jane Jacobs, Pierre Bourdieu andJean-Claude Passeron, and Glenn Loury, buthas since been developed most extensively byJames Coleman and Robert Putnam. 1 Interestin the concept of social capitrd was renewedwith the publication of Putnam’s Making De-mocracy Work.

Putnam uses the term social capital to referto features of social organization, such as net-works, norms and tmst, which facilitate co-ordination and cooperation for mutual gain.Coleman defines social capital as “a varietyof different entities, with two elements incommon: they all consist of some aspect ofsocial structure, and they facilitate certain ac-tions of actors—whether personal or corporate

1See Woolcock on the history of the definition ofsocial capital.

566 Journal of Agricultural and Applied Economics, December 2000

actors—within the structure” (p. 598). Wool-cock describes social capital as “. . . a broadterm encompassing the norms and networksfacilitating collective action for mutual bene-fit. ” In general, social capital includes the in-stitutions, the relationships, the attitudes, val-ues and beliefs that govern interactions amongpeople and contribute to economic and socialdevelopment (World Bank).z

Robison, Schmid and Siles address thequestion of whether social capital has features(such as durability, decay and service poten-tial) that are similar to features of physicalcapital. They also present results of a surveyadministered at a conference on social capitalthat yielded rankings of various definitions ofsocial capital. No single definition of social

capital was universally agreed to by confer-ence participants.

Putnam’s primary focus is on civic engage-ment—participation in voluntary associations,along with activities such as voting. He arguesthat civic engagement builds social capital byfostering personal interaction. Repeated inter-action in turn facilitates communication andamplifies information about the trustworthi-ness and cooperation of others, which reducestransaction costs associated with economic ex-change. Robison and Schmid suggest that themain cause of economic and social failures islack of social capital, in other words a lack ofcaring, goodwill, loyalty, sense of belonging,sense of community, or social closeness. Fu-kuyama claims that the presence or absence ofsocial capital or “trust” in a society is a keydeterminant of economic success. Tivo recentsurveys in the Journal of Economic Literature

emphasize the importance of social capital inthe economic performance of a country or re-gion (Temple; Collier and Gunning).

A major economic effect of social capitalis that it reduces information and transactioncosts. When transaction costs and the costs ofgathering and disseminating information arereduced, less risk is involved and more ex-change takes place, thus enlarging the scope

2WorldBankSocial Capitalweb-sitecontainsoth-er key referencesto social capital.URL: http://www.worldbank.orglpovertylscapitall

of transactions and interactions. Conversely, alack of social capital results in demands formore external controls, such as tougher lawenforcement and security systems, monitoringand enforcement (Flora and Flora). Schmidand Robison explore how social capital be-come a part of capital inputs in a productionprocess. They describe how interpersonal trustcan be a substitute for physical inputs of po-lice surveillance and legal service (p. 59).

Another contribution of social capital isthat it affects the supply of certain publicgoods. The provision of public goods is sub-ject to free riding or shirking if most users donot participate in joint actions to make the pro-vision of public good a success. In these sit-uations conventional theories of collective ac-tion have concluded that individuals willresort to strategic behavior by refusing to con-

tribute toward the public good in order to ob-tain a benefit far greater than the cost theyhave to pay. When social capital is present,externalities are internalized, which has the ef-fect of eliminating or reducing the free riderproblem and the misuse of public goods whileat the same time increasing investments inpublic goods.

Social capital also has several negative as-pects. While “good” social capital may facil-itate collective action and promote economicdevelopment, “bad” social capital can stiflecooperation and economic development.Good-old-boy (girl) clubs and gangs are ex-amples of bad social capital. Levi suggeststhat it may have been better for individuals tobowl alone in certain cases (for example, Tim-othy McVeigh and other co-conspirators in theOklahoma City bombing were members of abowling league). Strong ties within groups canlead to the exclusion of others, who are there-fore unable to benefit from group membership.For example some ethnic groups dominatecertain occupations and industries and mer-chants in some countries refuse to deal withtraders of a different ethnic background, andnewcomers are often excluded from stronglytied industrial groups (Portes and Landolt).

Olson argued that some groups deliberatelysuppress economic growth by securing a dis-proportionate share of national resources or in-

Rupasingha, Goetz, and Freshwater: Social Capital and Economic Growth 567

hibiting individual economic advancement byplacing heavy personal obligations on mem-bers, which prevents them from participatingin broader social networks. He points out thatin societies that permit free trade and free or-ganization, coalitions will form around mar-ketable goods and services. These coalitionsmay engage in rent-seeking activities, some-time at the expense of society in general. Forexample, producer co-operatives form to pro-tect members’ assets and, if possible, to boostprofits by raising prices. Labor unions orga-nize to bargain for wages. Doctors and law-yers join professional organizations to protecttheir own interests.

Only a small number of empirical studiesdirectly relate social capital variables to eco-nomic growth. Robison and Siles empiricallytested the relationship between changes in so-cial capital indicator variables and the leveland disparity of household income in the Unit-ed States and found that changes in social cap-ital have a significant effect on the disparityand level of household income. Temple andJohnson found that various social capital in-dices perform well in predicting economicgrowth across countries. Using social capitalindicators from the World Values Survey,Knack and Keefer determined that trust andcivic norms are stronger in nations with higherand more equal incomes. They also found thatmembership in formal organizations—Put-nam’s measure of social capital—is not asso-ciated with improved economic performance.Narayan and Pritchett found that for a sampleof Tanzanian villages membership levels invarious associations were positively correlatedwith per-capita income. Helliwell and Putnamfound significant evidence that per-capita GDPconvergence is faster, and equilibrium levelsof income are higher, in regions with higherlevels of social capital. They used data fromnorthern and southern Italy and three measuresof social capital (the extent of civic commu-nity, institutional performance, and citizen sat-isfaction) to test their hypotheses.

In this paper we estimate a Barre-typegrowth model to analyze the effects of socialcapital or civic engagement on economicgrowth in the U.S. at the county level. The

impact of social capital on expanding econom-ic activity has not been empirically substanti-ated at the county level. One advantage of us-ing county-level data is that they contain fewerstructural differences than cross-country data,which is a criticism raised by Miller. In addi-tion, certain measures are available whichwould be difficult to collect across differentcountries, and the US data tend to be collectedin a consistent manner. Related recent studiesof income convergence at the country levelinclude Levine and Renelt, Grammy and As-sane, Murthy and Chien, and Greasley andOxley, who use time series data from selectedOECD countries. Goetz and Hu also estimatea county-level economic growth model, but donot include social capital variables among theregressors.

We test the general hypothesis that socialcapital, while not overriding the effects ofconventional measures of growth, has additiveeffects on economic growth. We examine theimpact of the density of various types of as-sociations, economic agents’ financial supportfor charitable organizations and participationin elections, as well as crime rates—as mea-sures of social capital—on the rate of econom-ic growth. We realize that endogeneity is a po-tential concern but we have dealt with it asbest as we could (given the availability of dataonly in certain years) by including startingconditions at a point in time that precedes thegrowth period.

Specification

To determine the impact of social capital onincome growth we specify an economicgrowth model using change in county-levelper-capita personal income across the U.S. asthe dependent variable. In previous studies,per-capita income (output) growth has beenmodeled as a function of initial per-capita in-come, initial human capital stock, and a set ofcontrol and related variables. An extended ver-sion of a general economic growth model canbe written as in Barro and Sala-I-Martin, p.421:

(1) Dy, = F(Y,-[, H,.,, SC,-,, R,-l),

568 Journal of Agricultural and Applied Economics, December 2000

where Dy< is per-capita income growth, Y,.linitial per-capita income, H,-, initial humancapital stock, SC,-, initial social capital stock,and R,-, a set of control variables. The empir-ical specification of the growth model closelyfollows that in Goetz and Hu (p. 356), withthe addition of our measures of social capitalstocks:

(2) Dy, = aO + Q, Y,., + cqH,.T + @SC,-T

+ ~R,_T + U,,

where Dy, = (Y, – Yt.,)/Y,.,, Y~and y~-, arepersonal income per capita in 1990 and 1996,~ = 6, a, @ and r are parameters to be esti-mated, and u is a random disturbance. Edu-cational attainment serves as a proxy for hu-man capital stocks, as in Goetz and Hu: H,.,

is the percent of population 25 years or olderin 1990 with a bachelor’s, graduate, or a pro-fessional degree.

A major obstacle often cited in studies ofsocial capital is the empirical measurement ofthe variables in vector SC,-,. We use CountyBusiness Patterns and USA Counties datafrom the Commerce Department to constructindicators of associational activity, a crime in-dex, and a voter participation rate, and datafrom the National Center for Charitable Sta-tistics (NCCS) to measure county-level publicsupport for charitable organizations.

Associational Activity

The measure of principal interest is the num-ber of the following establishments in eachcounty: (a) bowling centers, (b) public golfcourses, (c) membership sports and recreationclubs, (d) civic and social associations, (e) re-ligious organizations, (~) labor organizations,(g) business associations, (h) professional or-ganizations, (i) political organizations. Ourmeasure of the density of associational activity(All Groups) in counties is the total number ofestablishments per 10,000 persons in 1990.We also test the possibility indicated in Knackand Keefer, that rent-seeking organizationshave a negative effect on economic perfor-mance, by dividing the above organizations

into “Olsonian” and “Putnam-esque” groups.Groups (a) through (e) are “Puntam-esque”(P-Groups) or those groups that are unlikelyto be rent-seeking, but which involve socialinteraction that can promote trust and coop-eration. Groups ~) through (i) are consideredto be rent-seeking organizations in the senseof Olson (0-Groups).

“Perverse” Social Capital (or Negative

Social Capital)

Criminal activity has been interpreted as a

symptom of deteriorating social capital or lackof social capital (e.g., Rubio). Low crime ratesin a county may indicate a more cohesivecommunity. The modified crime index (crimeindex and arson) for 1990 from the USACounty data set is used to measure “perverse”social capital (Crime).

Charitable Giving and Voting

These two variables complete the vector of so-cial capital indicators (Knack). County-levelpublic support data for charitable organiza-tions (Charity) are obtained from the NationalCenter for Charitable Statistics (NCSS). TheNCCS Core files are based on the IRS’s An-nual Return transaction files. The Charity var-

iable had 230 missing values which were re-placed by the sample mean value for theanalysis. Variable voting is the percent of eli-gible voters participating in the federal elec-tion, averaged over 1988 and 1992 (county-level data on participation in local electionsare not available on the CD-Rem).

On the basis of Goetz and Hu’s study, thefollowing variables are included in vector R ascontrols and related influences. Urban denotesmetropolitan counties and Rural is countieshaving places with no more than 2,500 resi-dents. These capture the effects of populationagglomeration in a county on growth; non-ur-ban/non-rural counties are the excluded cate-gory. Highway indicates whether (=1) or not(=0) the county has an interstate highway ac-cess ramp, to measure basic physical infra-structure and the degree of isolation of a coun-ty. Public highway spending (Hiwyexp)

Rupasingha, Goetz, and Freshwater: Social Capital and Economic Growth 569

proxies for public investment. Variable Corp-

tax is the percent of state taxes that are col-lected from corporations. A higher value is hy-pothesized to discourage new firm locationsand, therefore, to depress economic growth. Astate-level right-to-work law indicator variable(O, 1) is included to measure the strength oflabor, independently of item (~) (Labor orga-nizations) above. Seven other indicator vari-ables are included to capture regional effects:New England, Mideast, Plains, Southeast,Southwest, Rocky Mountain, and Far West(the East/North central region or Great Lakesregion is the excluded category).

Empirical Results

Table 1 reports ordinary least squares (OLS)estimates of equation (2), using data from3,040 U.S. counties, corrected for heteroske-dasticity.3 The growth period covers 1990through 1996. In Model 1 all social capitalindicators are excluded. According to Model2, the association activity variable, All Groups,

exerts a statistically significant, positive effecton economic growth. This indicates that as-sociation activity influenced county-level eco-nomic performance in the 1990s to a signifi-cant degree and independently of the otherfactors affecting economic growth. As hypoth-esized, Crime is negative and significant, in-dicating that a high crime rate in a county hasa negative impact on its economic growth. Theeffect of charitable giving on growth, on theother hand, is positive but not statistically dif-ferent from zero. Voting exhibits a non-lineareffect, leading first to an increase and then adecrease in income growth. The maximumvalue of this quadratic function is reached ata voter participation rate of 50.5 percent. Thisresult therefore confirms that participation inelections affects economic growth, but the ef-fect is subject to diminishing returns. Param-eter estimates for all of the other regressorsare statistically different from zero and, withthe exception of corporate taxes and the right-

3The Breusch–Paganchi-squared statisticshowsthatheteroskedasticityis presentin the dataset.

to-work law variable, each has the expectedsign.

A comparison of Models 1 and 2 showsthat the coefficient estimates on starting in-come (al) and human capital (a2) do not differmaterially, so that the specification bias asso-ciated with excluding social capital variablesin this regression model and this particulardata set is relatively small. Even so, a Waldtest of the hypothesis that the coefficients ofall the social capital variables (All Groups,

Crime, Charity, and Voting) are jointly zero isrejected at the 1-percent level, indicating thatthis vector needs to be included in the regres-sion.

We next test for the possibly negative ef-fects on economic growth of rent-seeking or-ganizations (P-Groups) and positive effects oforganizations that promote trust and coopera-tion (O-Groups). The results in Model 3 in-dicate that both types of organizations havepositive and statistically significant effects oneconomic growth in US counties. Thus wefind no support for Olson’s argument that or-ganizations with redistributive goals encour-age anti-growth rent-seeking. In contrast, rent-seeking organizations in this data set have apositive and significant effect on economicgrowth. These results do not change when weinclude only labor organizations in P-Groups,

or delete religious organizations from the O-Groups.

The positive effect of social capital on eco-nomic growth may be more pronounced inpoorer counties, if social capital is more pro-ductive in situations where local cooperationsubstitutes for the direct provision of servicesby government or profit-seeking firms. To testthis hypothesis—and explore potential inter-actions between social capital and initial in-come levels—we estimated auxiliary regres-sions with selected interaction terms. Onlyresults for the interacted variables and inter-action terms are reported here.

If the above hypothesis is valid, the inter-action terms for “good” social capital vari-ables (All Groups and Voting) with initial in-come will be negative, suggesting thatdecreases in income enhance the positive ef-fects of these social capital variables on eco-

Journal of Agricultural and Applied Economics, December 2000

*

Rupasingha, Goetz., and Freshwater: Social Capital and Economic Growth 571

nomic growth, while the interaction term be-tween “bad” social capital (Critne) and initialincome will be positive (and significant). Inthis case, a decline in income increases thenegative effect of the “bad” social capital oneconomic growth. The estimated coefficientsin this case

Dy, =

Dy, =

Dyl =

are:

0.0014A11 Groups – O.13-04Y,-,

(0.62) (5.04)

+ 0.87 -07A11 Groups* Y,_,

(0.67)

0.76 Voting + O.11-05Y,-T

(4.07) (0.82)

– 0.22 -04 Voting*Y,.,

(2.77)

–0.28-05Crime – 13-04Y,-,

(4.52) (9.60)

+ O.12-09 Crime*Y,_T

(3.65)

where t-statistics are in parentheses. The in-teraction term All Groups*Y,-T is positive butnot statistically significant. The negative andsignificant coefficient on Voting* Yt_7 and thepositive and significant coefficient onCrime* Y,-, tend to provide support for our hy-pothesis.

Conclusion

The results presented here suggest that socialcapital or civic engagement is an important in-dependent determinant of economic growth inU.S. counties. We find significant evidencethat per-capita income grows more rapidly incounties with high levels of social capital,measured using the density of membership or-ganizations, crime rate, charitable giving andvoter participation. These results also suggestthat future studies of economic growth shouldincorporate measures of social capital to avoidspecification bias, although this type of biasmay be relatively small.

References

Barre, R.J. and X. Sala-I-Martin. Economic

Coleman, J.S. “Social Capital in the Creation ofHuman Capital.” American Journal of Sociol-

ogy 94(1988 ):S95–S 120.

Collier, R and J.W. Gunning. “Explaining AfricanEconomic Performance.” Journal of Economic

Literature 37(March 1999) :64-1 11.Flora, C.B. and J.L. Flora. The National Rural De-

velopment Partnership and Entrepreneurial So-

cial Infrastructure: Building a Sustainable

Communip of Interest. North Central RegionalCenter for Rural Development, 1996.

Fukuyama, 1? Trust: The Social Virtues & the Cre-

ation of Prosperity. New York The Free Press,1995,

Goetz, S.J. and D. Hu. “Economic Growth and Hu-man Capital Accumulation: Simultaneity andExpanded Convergence Tests. ” Economic Let-

ters 5 1(1996):355–362.

Gramrny, AR. and D. Assane. “New Evidence onthe Effect of Human Capital on EconomicGrowth.” Applied Economics Letters 4(1996):

121–124.

Greasley, D. and L. Oxley. “Time-Series BasedTests of the Convergence Hypothesis: SomePositive Results.” Economics Letters 56(1997):

143-147.

Helliwell, J. and R.D. Putnam. “Economic Growthand Social CapitaJ in Italy, ” Eastern Economic

Journal 21( Summer 1995):295–307.Knack, S. “Civic Norms, Social Sanctions and Vot-

er Turnout. ” RationaliQ and Society 4(April

1992):133–156.

Knack, S. and I? Keefer. “Does Social Capital havean Economic Payoff? A Cross-Country Inves-ligation. ” Quarterly Journal of Economics

112(November 1997): 125 1–1288.

Levi, M. “Social and Unsocial Capital: A ReviewEssay of Robert Putnam’s Making DemocracyWork.” Politics and SocieV 24(March 1996):45–55.

Levine, R. and D. Renelt. “A Sensitivity Analysisof Cross-Country Growth Regressions. ” Amer-

ican Economic Review 82(1 992):942–963.

Miller, S.M. “A Note on Cross-Country GrowthRegressions. ” Applied Economics 28( 1996):

1019-1026.

Murthy, V.N,R. and 1.S. Chien. “The Empirics ofEconomic Growth for OECD Countries: SomeNew Findings. ” Economics Letters 55(1997):

425–429.

Narayan, D. and L. Pritchett, 1997, “Cents and so-ciability: Household income and social capitalin rural Tanzania,” World Bank Policy Research

Growth. New York McGraw-Hill, 1995. Working Paper No. 1796.

572 JoumalofAgricultural and Applied Economics, December 2000

Olson, M. The Rise and Decline of Nations. Con-necticut: Yale University Press, 1982.

Portes, A.and I? Landolt. “TheDownsideo fSocialCapital, ” The American Prospect 26(May–Jtme

1996):18–21,

Putnam, R.D. Making Democracy Work: Civic Tra-

ditions in Modern Italy. Princeton: PrincetonUniversity Press, 1993.

Robison, L.J. “In Search of Social Capital in Eco-nomics. ” Staff paper No. 96-102, Departmentof Agricultural Economics, Michigan State Uni-versity, December 1996.

Robison, L.J. and M.E. Siles. “Social Capital andHousehold Income Distributions in the UnitedStates: 1980, 1990. ” Journal of Socio-Econom-

iCS 28(1999):43–93.Robison, L.J. and A.A. Schmid. “Can Agriculture

Prosper Without Increased Social Capital. ”Choices (Fourth Quarter 1994):29–3 1.

Robison, L.J., A.A. Schmid and M.E. Siles. “Is So-

cial Capital Really Capital?” Dept. Agric.Econ., Michigan State University, E. Lansing,MI, Staff Paper No. 99-21, June 1999; 24 pp.

Rubio, M. “Perverse Social Capital: Some Evi-dence from Columbia.” Journal of EconomicIssues 3 l(September 1997):805–16.

Schmid, A.A. and L.J. Robison. “Application ofSocial Capital Theory. ” Journal of Agriculturaland Applied Economics 27(July 1995):59–66.

Temple, J. “The New Growth Evidence. ” Journalof Economic Literature 37(March 1999):112–156.

Woolcock, M. “Social Capital and Economic De-velopmen~ Toward a Theoretical Synthesis andPolicy Framework. ” Theory and Society27(1998):151–208.

World Bank, “The Initiative on Defining, Monitor-ing and Measuring Social Capital: Overviewand Program Description.” Working Paper No.1, Social Capital Initiative, the World Bank,April 1998.