Racial diversity and aggregate productivity in U.S. industries: 1980-2000.

• Author: Sparber, Chad

1. Introduction

   The United States grew increasingly diverse between 1980 and 2000 as whites declined from 83% to 73% of the labor force. Many states are reconsidering their affirmative action policies while the federal government reevaluates immigration and citizenship statutes. Unfortunately, we do not yet fully understand the aggregate effects of diversity. International empirical analyses often argue that racial (or ethnic) heterogeneity hinders macroeconomic growth and development by fostering conflict, corruption, political instability, and competitive rent seeking. Many top U.S. business executives reject these views, however, and claim that individuals from varied backgrounds bring unique approaches to problem solving and complement each other in production. Several academic analyses echo these sentiments. This paper adds to the debate on the costs and benefits of diversity by building upon the traditional economic growth and development framework to measure racial diversity's net effect on productivity within sectors of the United States.

   My methodology advances the existing literature by addressing two limitations of prior studies. First, although cross-country evidence for the costs of diversity is clear, the United States may be especially effective in managing diversity. This paper determines whether the U.S. experience differs from international accounts. Second, typical international growth analyses measure aggregate outcomes for countries but fail to address differences in marginal effects that could exist across sectors of the economy. I depart from this strategy by analyzing industrial performance across U.S. states. To begin, I measure racial diversity by using decennial census data from the Integrated Public Use Microdata Series (IPUMS) (1) to construct fractionalization indices for state-industry cells between 1980 and 2000. I then employ two-stage weighted least squares regressions with fixed effects to estimate diversity's industry-specific influence on average wages paid to workers--a proxy for labor productivity in the absence of observed output per worker measures.

   Unqualified statements regarding the costs and merits of diversity are unwarranted, as racial heterogeneity increases productivity within most, but not all, U.S. industries. These results are robust to methodologies controlling for individual-level heterogeneity, state-specific trends, industry-specific trends, and supply factors. Prior academic research, anecdotes from political events, and the sector-specific coefficients in this paper provide intuition about the industrial characteristics that determine whether diversity leads to economic gains or losses. I use the U.S. Department of Labor's O*NET Consortium data (2) to show that industries heavily reliant upon creative decision making, problem solving, and customer service benefit from diversity. In contrast, sectors characterized by high levels of group work or teamwork experience losses, which suggests that heterogeneity encumbers common action. The results, therefore, reconcile competing literatures by recognizing both the costs and the benefits of diversity.

2. Motivation

   Do racially diverse environments generate net economic gains or losses? International growth economists typically find evidence for the latter. Easterly and Levine's (1997, p. 1239) seminal investigation argues that "A movement from complete heterogeneity to complete homogeneity is associated with an income increase of 3.8 times." To them, diversity increases polarization, facilitates competitive rent seeking between groups, and promotes growth-reducing political policies. (3) A number of social scientists have argued that diversity also conflicts with common action, and that heterogeneous societies tend to oppose wealth redistribution and public goods provision. (4) Others find evidence that diversity escalates corruption and political instability. (5)

   Many top U.S. executives, in contrast, argue that a racially diverse workforce bolsters productivity. In the late 1990s, the University of Michigan began defending the constitutionality of its affirmative action admissions policies in a series of well-publicized court cases. In October 2000, executives from 30 Fortune 500 companies united to file an amicus brief supporting the school. The brief contends that "A diverse group of individuals educated in a cross-cultural environment has the ability to facilitate unique and creative approaches to problem-solving arising from the integration of different perspectives" and that "such individuals are better able to develop products and services that appeal to a variety of consumers and to market offerings in ways that appeal to these consumers." (6) These anecdotal claims do find support in academic literature. In psychology, Campbell (1960, p. 391) argues, "persons who have been thoroughly exposed to two or more cultures seem to have an advantage in the range of hypotheses they are apt to consider, and through this means, in the frequency of creative innovation." Simonton (1999, pp. 207, 213), also citing Campbell, further maintains that "creativity is favored by an intellect that has been enriched with diverse experiences and perspectives.... It is as if the mere exposure to different lifestyles and divergent values enables individuals to expand the range and originality of their ideational variations." (7)

   Empirical, qualitative, and anecdotal evidence suggests that an ideal theoretical model of diversity and productivity should reconcile the costs and benefits of diversity within a single production function. Assume an economy is composed of [L.sub.i] workers from groups i = 1 ..., N. The number of groups (N) is fixed, and group i is composed of a fraction of the labor force equal to [[theta].sub.i]. A generic function C([L.sub.1], [L.sub.2], ..., [L.sub.N]) measures the fraction of output remaining after firms (and society) suffer costs associated with employing individuals from disparate groups. These costs might reflect communication and conflict costs, diversion from productive activity, and/or disputes over proper public goods provision. (8) C([L.sub.1], [L.sub.2], ..., [L.sub.N]) is homogenous of degree 0 so that only the relative size of groups (i.e., [[theta].sub.1], [[theta].sub.2], ...,[[theta].sub.N]) will affect output. Costs are maximized when each group maintains an identical share of the labor force; costs decrease as the labor force share of one group approaches unity.

   Suppose, however, that workers of different types also supply unique skills so that a diverse workforce provides a greater range of perspectives, ideas, and problem-solving techniques. Workers from varied groups will complement each other in production. A "love of labor variety" term can account for these complementarities, as can a nested constant elasticity of substitution specification allowing imperfect substitutability across labor groups. (9) Let G([L.sub.1], [L.sub.2], ..., [L.sub.N]) represent a homogenous of degree 1 function identifying potential gains. An equal labor force share across groups will maximize production complementarities; gains are negligible if a single group composes nearly all of the labor force.

   Equation 1 combines these cost and complementarity terms into a single production function where output (Y) experiences constant returns to capital (K) and labor ([L.sub.1], [L.sub.2], ..., [L.sub.N]), augmented by an exogenous technology parameter (A). Equation 2 represents average labor productivity (output per worker, y) as a function of capital per worker (k) and each group's proportion of the labor force ([[theta].sub.1], [[theta].sub.2], ..., [[theta].sub.N]). An increase in diversity refers to a mix of labor force shares that comes closer to an equal representation of each group in production.

   Y = F(K, [L.sub.1], [L.sub.2], ..., [L.sub.N]) = A x [K.sup.1 - [alpha]] x C[([L.sub.1], [L.sub.2], ..., [L.sub.N]).sup.[alpha]] (1)

   [y = A x [k.sup.1 - [alpha]] x C([[theta].sub.1], [[theta].sub.2], ..., [[theta].sub.N]) x G([[theta].sub.1], [[theta].sub.2], ..., [[theta].sub.N]).sup.[alpha]] (2)

   Clearly, if diversity causes both costs and labor complementarities to increase, one must impose further assumptions on the nature of C(x) and G(x) to predict the theoretical effect on net productivity. Rather than explore these theoretical implications, I advocate empirical analysis to improve understanding. While the majority of prior work has assessed ethnic diversity and economic performance across countries, this paper prefers analysis of racial diversity in the United States as an alternative strategy.

   First, the costs and benefits of diversity could vary across national borders. Though major intercultural conflict or warfare is certainly a common occurrence for some countries, others could conceivably exploit complementarities in creativity and problem solving. Collier (2000, 2001) addresses this potential for cross-country variation in attitudes toward diversity by arguing that diversity is only detrimental to nondemocracies, since democracies establish better systems to manage cultural conflict. Alesina and La Ferrara (2005) prefer instead to focus on stages of economic development. To them, diversity negatively affects gross domestic product per capita growth in poor countries but not in their developed counterparts. A third alternative for avoiding the variation in attitudes toward diversity is to simply focus on a single country (e.g., the United States) since cultural sentiments, ethnic strife, and racial tolerance are likely to be more consistent across states or cities within a country than they are across national boundaries. This method also has the obvious advantage of providing insight into a specific country's experience with diversity. (10)

   Second, ambiguous ethnicity definitions confound analysis. Ethnicity classifies people according to cultural, linguistic, religious, or...