research paper 700-800 words


Hispanic Population Growth and Rural Income Inequality

Emilio A. Parrado, University of Pennsylvania William A. Kandel, Economic Research Service, USDA

We analyze the relationship between Hispanic population growth and changes in U.S. rural income inequality from 1990 through 2000. Applying comparative approaches used for urban areas we disentangle Hispanic population growth's contribution to inequality by comparing and statistically modeling changes in the family income Gini coefficient across four rural county types; established Hispanic, rapidly growing Hispanic, rapidly growing non-Hispanic, and slow-growth or declining counties. Results support perspectives that stress growing social heterogeneity for understand- ing the contribution of minority population growth to inequality, including changes in human capital and industrial restructuring. We find remarkably similar inequality growth across rapidly growing Hispanic and non-Hispanic counties. This suggests that growing rural inequality stems largely from economic expansion and population growth rather than changing Hispanic composition.


Rapid growth ofthe U.S. Hispanic population, especially through immigration, has generated public concern and academic debate about its contribution to in- come inequality (Reed 2001; Borjas et al. 1997; Lerman 1999). Two general perspectives inform this discussion. The first argues that relatively less educated and unskilled Hispanic immigrants contribute to income inequality because their low levels of human capital restrict their earning power (Borjas 1999). The sec- ond perspective argues that structural forces fueling the demand for low-skilled, low-paying jobs are the main cause of income inequality, and the increase in low- skilled immigration simply reflects those forces {Piore 1979).

Empirical trends underlying the debate are clear. From 1990 to 2000 the Hispanic population in the United States increased more than 50 percent and in 2003 officially surpassed non-Hispanic blacks as the nation's largest minority group. Much of this growth resulted from immigration. More than 7 million Latin-American migrants entered the United States during the 1990s, almost dou- bling the number of foreign-born Hispanics in a single decade. Rapid Hispanic population growth also coincided with rising income inequality. U.S. Census estimates indicate that from 1989 through 1999 the Gini coefficient measuring income inequality among U.S. families increased 7 percent, from .401 to .429 (U.S. Census Bureau 2006). Combined with the 10 percent increase during the

Financial support was provided through a cooperative agreement ($43-5AEU-4-80l24) with the Economic Research Service, USDA. Views expressed do not reflect the opinions of the Economic Research Service or the U.S. Department of Agriculture. Direct correspondence to Emilio Parrado, Department of Sociology, University of Pennsylvania, 3718 Locust Walk, Philadelphia, PA 19104- 6299. E-mail: [email protected],edu.

" TlH Unlterslly at North Csrolini Ptau Social Forces 86(31 1421-1450. March 7ain

1422 • Social Forces BB{3)

1980s, this trend of growing inequality represents a reversal of both consistent declines in income inequality during the first four decades ofthe 20'*" century and stable levels of low-income inequality registered from 1940 to 1970.

Several studies have attempted to disentangle the connection between Hispanic population growth-especially through immigration-and income inequality (Altonji and Card 1991; Borjas et al. 1996; Bradbury 1996; Chevan and Stokes 2000; Peti 2006). The majority of these analyses, however, have been conducted either at the national level or only for urban areas. Studies of mechanisms under girding rural inequality are relatively rare despite scholarly recognition that the forces fueling income inequality might differ in rural and urban areas (Domazlicky 2005; Kuznets 1955; McLiughlin 2002). Moreover, with few exceptions (Albrecht et al. 2005; McLaughün 2002) the handful of studies with a specific focus on rural inequality has not paid explicit attention to the impact of changing minority composition and immigration on inequality.

However, understanding this association is increasingly relevant. Data from the 2000 U.S. Census revealed unanticipated and dramatic Hispanic popula- tion growth in all U.S. regions but particularly in new rural destinations in the Southeast and Midwest. In fact, in the 1990-2000 decade, the rate of non-met- ro Hispanic population growth exceeded that of metro counties (Kandel and Cromartie 2004) highlighting the need to expand the geographic focus of most previous ethnic and immigration research. AJthough often overlooked in current discussions of immigration and economic trends, non-metropolitan counties are not trivial entities, encompassing roughly 77 percent of all U.S. territory and 17 percent of its population. The rapid influx of minority groups in non-traditional destinations has considerable economic and policy ramifications for rural areas.

Accordingly, this research examines the relationship between Hispanic popula- tion growth and changes in family income inequality in non-metropolitan coun- ties from 1990 to 2000. Our analytical approach builds on urban-based research linking inequality across local labor markets to differences in the relative supply of immigrants and minorities (Card 2005; Friedberg and Hunt 1995). This approach disentangles the contribution of minority population growth to inequality by generating counterfactual comparisons of inequality trends across labor markets with different population trajectories.

While cities function readily as delineated local labor markets by allowing researchers to compare individual cases (e.g.. Card 1990), analysis of more sparsely populated rural areas benefits from grouping areas according to their changing population compositions. Our approach is to create a county-level typology that distinguishes established Hispanic, rapidly growing and recently settled Hispanic, rapidly growing non-Hispanic, and slow-growth or declining populations. We convey trends captured in our typology by mapping several county type distribu- tions and qualitatively describing illustrative cases in each group. The quantita- tive analysis that follows compares inequality patterns across these four

Hispanic Population Growth and Rural Inequality • 1423

types and models such changes according to labor force characteristics, industria] change and additional sources of heterogeneity. Overall, our comparative ap- proach reveals remarkably similar inequality trends across Rapid Growth Hispanic and non-Hispanic rural counties, suggesting that increases in rural inequality are largely the product of economic expansion and population growth and not of increases in Hispanic composition per se.

Recent Rural Hispanic Population Growth

U.S. Census estimates reported in Table 1 show that the non-Hispanic non-metro population' was relatively stagnant in the 1980s, with a growth rate well below metro areas. Hiis pattern changed significantly during the 1990s, when non- metro population grow revived to more than 8 percent and rivaled the growth in metro areas. This increase in population growth was even more dramatic among Hispanics, whose non-metro numbers grew 27 percent during the 1980s (rela- tive to 56 percent in metro areas) and 67 percent during the 1990s, surpassing the 57 percent growth registered in urban areas and accounting for more than 25 percent of all non-metropolitan population growth (Kandel and Cromartie 2004; Kirschneretal. 2006).

These general trends mask considerable variation across regions. Particularly striking was the growth of Hispanic popuiations outside traditional destination ar- eas in the Southwest, where the majority of rural Hispanics have resided since the turn of the century Media reports have illustrated dramatic examples of Hispanic influx in places such as Dalton, Georgia; Storm Lake, Iowa; and Siler Gity, North Carolina, and a growing body of ethnographic research documents the mixed reception Hispanics typically receive in small communities with little experience or few public programs to assist foreign-born newcomers (Gozdziak and Martin 2005; Griffith 1995; Kandel and Parrado 2004, 2005; Zúñiga and Hernández- León 2005; Massey 2008).

Table 1 illustrates these changes. From 1990 to 2000, nonmetropolitan Hispanics increased 35 percent in the Southwest," 113 percent in the Midwest, 81 percent in the West, and a staggering 204 percent in the South. Most of the growth and dispersion of the Hispanic population was driven by the foreign born. In the South during the 1990s, for example, the native Hispanic population in the South grew 38 percent compared to an astounding 211 percent among the foreign born, explaining over 82 percent of total Hispanic population growth in the region. The majority of tJie foreign bom (60 percent) entered the United States after 1990, and 40 percent entered between 1995 and 2000. A similar pattern occurred in the Midwest, where the growth rates of native and foreign born Hispanics were 53 and 206 percent, respectively. The growth of the foreign population explains the majority of the growth (61 percent) of the Hispanic population in the Midwest with 46 percent entering the United States after 1990. Ihe main implication for our purposes is that understanding the role of Hispanic

1424 • Social Forces 88{3)

Table 1: Hispanic and Non-Hispanic Population Growth by Metropolitan Status and Region

Total Population

Non-Hispanics Metro

1980 1990 2000

Nonmetro 1980 1990 2000

Hispanics Metro

1980 1990 2000

Nonmetro 1980 1990 2000

163,899 177,360 193,133

48,038 48,995 52,983

13,111 20,452 32,130

1,498 1,902 3,176

% Chanqe

8.2 8.9

2.0 8.1

56.0 57.1

27.0 66.9


.5 2.4

4.0 4.1

44.0 38.9

71.2 72.3

% Growth Midwest

1.8 6.1

-2.3 4.4

36.0 78.0

25.8 113.2

across U.S. South

14.3 14.3

2.8 9.5

73.5 93.1

1.2 204.0

Reqions West

16.9 19.0

8.1 16.7

62.2 129.6

54.8 81.1


16.0 9.4

6.8 13.0


59.9 50.1

26.8 35.3

Source: U.S. Census, SF1 files, 1980-2000 Note: Regions are census regions, except for the Southwest which borrows from the West and the South and consists of Arizona, California, Colorado, New Mexico and Texas.

population growth on economic outcomes overlaps with understanding the eifect of immigration which has been the primary contributor to the growing Hispanic population in new areas of destination and the leading force affecting social het- erogeneity. The rapidity and magnitude of these changes highlights the need to evaluate their impact on inequality in rural America.

Hispanic Population Growth and Social Heterogeneity

Much empirical scholarship has analyzed the forces fueling inequality. Because we consider the particular phenomenon of Hispanic population growth within a mostly non-Hispanic rural population, our analysis builds on elaborations that stress the role of social heterogeneity to explain cross-sectional variation in inequality and their evolution over time. These articulations began with Kuznets' (1955) work and later received expanded treatment by Nielsen and colleagues (Nielsen 1994; Nielsen and Alderson 1997; Nielsen and Alderson 2001; Möller et al. forthcoming). A central tenet of this perspective is that inequality is "generated by social heterogeneity related to a specific stage ofthe development process."(Nielsen 1994:655) Rather than expecting monotonie growth or decline in inequality with economic development, this perspective emphasizes transitional aspects of growth that enlarge economic disparities. The

Híspante Population Growth and Rural Inequality • 1425

main expectation is that any dimension associated with economic growth that is unevenly distributed across individuals and affects their income prospects will translate into inequaHty (Nielsen and Alderson 1997).

The classical example of a transitional mechanism affecting inequality is seaor dualism. According to Kuznets (1955), population shifts between a traditional and a modern sector of the economy produce an inverted U-shaped relationship between development and inequality. Specifically, shifting a society's labor Force from a traditional agricultural sector with low productivity and wages to a modern sector with high productivity and wages is expected to affect the evolution of income inequality over time. Low levels of inequality prevalent among agricultural societies are predicted to increase as employment shifts to the higher paying modern sector during early stages of development. Inequality is expected to peak at intermediate levels of development and ultimately decrease at some advanced stage when substan- tial portions ot the population are employed in the modern sector.

The predicted Kuznetsian pattern of declining inequality at advanced levels of economic development was supported in studies of the evolution of inequality across U.S. counties (Nielsen and Alderson 1997). However, the reversal in the evolution of income inequality in recent decades highlights the importance of other sources of heterogeneity in affecting inequality trends, including educa- tional composition, industrial structure and demographic change (Nielsen and Alderson 1997). Given our focus on the role of growing Hispanic representation, we concentrate on transitional mechanisms that could account for the role of minority population growth on income inequality across rural areas. In general the connection between inequality and heterogeneity suggests that rapid population growth, especially of low-skilled minority groups, should contribute to inequality (Chenery et al 1974; McNicoll 1984). The mechanisms, however, are diverse and depend on the forces attracting people to rural areas.

Two general sources of heterogeneity are central to our analysis (Chevan and Stokes 2000; Katz and Murphy 1992). The first, often associated with supply- side explanations that link Hispanic population growth with income inequality, is growing heterogeneity in population composition, specifically human capital endowments, foreign-born status and family structure. The general expectation is that Hispanic population growth expands the supply of low-skilled workers, altering the composition ofthe labor force and thereby fostering inequality. This is particularly so for human capital endowments. Among rapidly growing rural counties the sudden influx of immigrant Hispanics expands the lower end ofthe educational distribution, potentially contributing to inequahty. Borjas (1999), for example, finds that between 1980 and 1995 immigration increased the number of high school dropouts by 21 percent, a period during which dropouts' wages fell 11 percent relative to more educated workers. At the other end ofthe educational distribution, rural counties attracting highly skilled professionals could also expe- rience growing inequality (McLaughlin 2002).

1426 • Social Forces S8(2)

In addition, the Hispanics moving to new rural destinations during the past two decades tend to be primarily of immigrant origin (Kandel and Cromartie 2004). Foreign-born status often embodies several detrimental labor market characteristics, including less English fluency, less U.S. work experience, and, frequently, unauthorized status. The increasing heterogeneity resulting from the growing relative representation of foreign-born groups in rural counties is also expected to contribute to inequality.

The final supply-side process affecting inequality associated with minority population growth relates to changes in demographic composition, particularly age and family structure. In general, the life-cycle profile of earnings implies that the relative growth ofthe population at older ages should contribute to income inequality due to relatively higher earnings exhibited among older working age groups (Formby et al. 1989). TKis effect, however, might be mitigated in rapidly growing Hispanic counties because Hispanics, especially immigrants, are younger than the general population. Hispanic population growth is also likely to alter the family structure of receiving counties, including the proportion oi female-headed households and patterns of female labor force participation. Disadvantages affect- ing female-headed households imply that their growing representation should also contribute to income inequality (Snyder and McLaughlin 2004). At the same time, female labor force participation represents an important source of family income that compensates for individual income disparities and thus reduces inequality (Cancian and Reed 1998). Female employment, however, is lower in Hispanic families, reducing its potential to abate inequality in rapidly growing Hispanic counties compared to other counties.

The second general source of heterogeneity, often associated with demand-side explanations linking Hispanic population growth and income inequality, is growing heterogeneity in the sectoral composition of employment and industrial restructur- ing. The main expectation is that the changing industrial composition of rural coun- ties will contribute to inequality above and beyond changes in the socioeconomic characteristics of the labor force. Numerous studies have linked industrial restruc- turing to growing inequality (Harrison and Bluestone 1990; Morris et al. 1994; Neilsen and Alderson 1997; Chevan and Stokes 2000; McLaughlin 2002). These studies stress that the shifts in employment away from manufacturing and toward services increases income inequality. The main expectation here is that manufactur- ing employment reduces inequality due to its relatively high level of productivity' and unionization, which provide the means and incentives for corporations to pay high wages even to low-skilled workers. This was in fact the case for immigrants and mi- norities during the early 20''' century, when manufacturing employment facilitated access into the middle class. We expect the decline in manufacturing to reduce the number of good jobs available to low-skilled workers and thus to increase inequality.

While much of this literature emphasizes the dichotomy between manufacturing and service sectors, applying the industrial restructuring framework to rural areas

Hispanic Population Growth and Rural Inequality • 1427

requires special attention to industries attracting low-skilled Hispanics, including foreign-born workers to non-metropolitan counties. This comprises changes not only in manufacturing and services, but also in agriculture and construction. Given the emphasis on the characteristics of jobs available within these sectors, including constraints on wage increases and union proteaion, we expect that increases in agricultural and construction employment, as well as service industry employment, would contribute to inequality in rural areas, relative to manufacturing employment.

The literature documents other socioeconomic sources of heterogeneity affecting inequality (Barro 2000; McLaughlin 2002; Martin 2006) including growth ofthe non-Hispanic black population, unemployment, median family income, labor force size and prevalence of full time employment. Key among these factors is family income because it direcdy ties with Kuznets' argued link between economic growth and income inequality. Although we control for such mechanisms in our statistical analysis, they are not the focus of our study Rather, we emphasize the role of chang- ing supply and demand conditions, because they directly address the unresolved is- sue of whether it is Hispanic population growth per se or broader economic changes that account for inequality trends in rapidly growing Hispanic destinations.

Analytic Strategy: Caunterfactual Comparisons and Rural County Inequality

One challenge for our analysis is the difficulty of separating the unique impact of Hispanic population growth from broader processes of socioeconomic change in rural counties. One approach is to include a measure of the changing Hispanic composition in a county and estimate its effect on income inequality. This option is unlikely to capture the dramatic changes of new rural Hispanic destination areas be- cause the number of emerging rural Hispanic counties is not large, and effects might not be captured with a continuous measure of change in Hispanic representation across all rural U.S. counties. Moreover, such an analysis does not provide straight- forward comparisons across rural areas experiencing dissimilar population trends.

An alternative approach, to identify counties with different labor market struc- tures and relate trends in inequality to their particular population trajectories, has been applied extensively in metropolitan studies ofthe impact of immigration on natives' wages (Card 1990, 2005; Friedberg and Hunt 1995). This analyti- cal strategy compares the local wage structure across cities with large and small influxes of immigrants. This approach is particularly advantageous in our case because it generates cleat counterfactuals that permit comparison of changes in inequality across rural areas that have recently received a large influx of Hispanics against those that did not. it also allows for a deeper and broader understanding ofthe transformations occurring in rural areas by distinguishing demographic and economic patterns associated with changes in minority composition. In addition, it can be used to geographically locate different local labor markets with particular population trajectories and thereby identify the regional concentration of the economic and social transformations affecting rural areas.

1428 • Social Forces 88[2)

While cities function naturally as local labor markets, facilitating individual case comparisons, rural area comparisons are less straightforward. Our approach is to construct a rural county typology that distinguishes distinct population trajectories from 1990 to 2000. The typology combines three factors: Hispanic share of 1990 county population, the change in county Hispanic share from 1990 to 2000, and total 1990-2000 county population change. Together, these factors produce four county types: (1. counties with established Hispanic populations, (2. counties with rapid Hispanic population growth in regions with little previ- ous Hispanic presence, (3. counties that grew rapidly but do not have sizeable Hispanic populations, and (4. demographically stagnant counties.' Comparing inequality across these county types, especially between Rapid Growth Hispanic and Rapid Growth Non-Hispanic, is our main objective.

Data and Methods

Data for this analysis come from the 1990 and 2000 U.S. Census SF3 files. The unit of analysis is the non-metropolitan county as defined in 2003 (see footnote 1). The dependent variable is the change in Gini coefficient for family income from 1990 to 2000. We focus on family rather than individual inequality because it more directly reflects overall population well-being, especially for women and children.

Sociologists have long debated appropriate statistical approaches for analyzing socioeconomic change when the dependent variable is measured at two points in time, and in particular how best to account for omitted variables in panel data designs. Two methods most commonly proposed are: the lagged-regressor variable method, in which the dependent variable measured at time 2 (}^) is regressed on the dependent variable measured at time 1 ( Y¡) and additional covariares (A); and the difference score method, in which the time 1 score is subtracted from the time 2 score {Y,'Y) and then regressed on X. It is important to note that the regression of Y^'Y on both Y^ and A" is computationally equivalent to the regressor variable method and produces the same results (see Allison 1990; Werts and Linn 1970).

The literature Rirther distinguishes pure-difference from semi-difFerence models depending on how the predictor variables (A) are specified. In the pure-difference model, the change score {Y^-Y) is regressed on the difference score of the indepen- dent variables [X^-X). In the semi-difference model, the change score is regressed on the level of the predictor variables at time 1 {X^ which more closely resembles the lagged-regressor variable method in which predictors are measured at time 1.

Both for methodological and theoretical reasons Allison (1990) and Firebaugh and Beck (1994) support the pure-difference approach. Methodologically, the pure- difference model eliminates the potential bias due to omitted variables. Unmeasured enduring traits of individual counties are removed when differentiating both the dependent and independent variables because constant effects get cancelled out. Theoretically, model choice depends on the causal connection expected between

Hispanic Population Growth and Rural Inequality * 1429

the variables. The ¡agged-regressor variable model assumes a temporal ordering from Y^ to X^ to Y^ which is not appropriate in two-wave pane! designs in which X is measured contemporaneously with Kat both time points (Allison 1990).

In addition, Firebaugh and Beck (1994) argue that the theoretical argument for expecting rhe level of the independent variables at time 1 to affect change in the dependent variable is unclear. They show that a semi-diiFerence model also implies that a change in Kis caused by a change in A*and Kbut during the previ- ous interval. A semi-difference model assumes that the effect of a change in A'lies dormant for a period before affecting Y. The extent and rationale for the dormancy period is usually unclear. Especially iithe interval in a pure difference model can be extended beyond the expected dormancy period, regressing difference scores on change in the levels ofthe independent variables more appropriately estimates the causal connection between YandX'm two panel designs (Allison, 1990).

Accordingly, we specify a pure difference model which predicts change in income inequality between 1990 and 2000 according to four types of independent variables: rural county types (7), changing population composition (P), changing industrial composition (I), and other sources of heterogeneity (0). Our working equation is:

where the dependent variable [Y^^-Y^^ corresponds to the arithmetic difference between the 2000 and 1990 Gini Concentration Ratios computed for family income by county. We estimate our models using OLS techniques.

To assess the robustness of our results to model specification, we tested ad- ditional models, including a lagged-regressor model where Y vizs regressed on K^̂ and the change in independent variables, as well as a complete semi-difference model in which K^-K^was regressed on the change between 1990 and 2000 and the level of independent variables in 1990. While the magnitude ofthe coefficients differs, results do not change our substantive findings, especially the role of county types and changing socioeconomic conditions on rural county inequality.'*

Independent Variables

Rural County Types

Table 2 summarizes the criteria used for the construction of our county typology. Established Hispanic Counties were at least a 10 percent Hispanic in 1990. Rapid Growth Hispanic Counties were less than a 10 percent Hispanic in 1990 and saw their percent Hispanic increase by more than 2.5 percentage points between 1990 and 2000. Rapid Growth Non-Hispanic were less than 10 percent Hispanic in 1990, saw their percent Hispanic increase by less than 2.5 percentage points between 1990 and 2000, but experienced overall population growth exceeding 17 percent. Lastly, Slow Growth and Decline Counties were less than 10 percent Hispanic in 1990, saw their percent Hispanic increase by less than 2.5 percentage

1430 • Socia/Forces 5Z[Z)

Table 2: Criteria for Non-metro County Typology % Hispariic Change in % % Growth In Total

Composition Hispanic Population County Type 1990 1990-2000 1990-2000 Established Hispanic counties ^ 10 Rapid growth Hispanic counties < 10 ^2 .5 Rapid growth non-Hispanic counties < 10 < 2.5 ^ 17 Slow qrowth & decline counties* < 10 < 2.5 < 17

'Refers to the total county population, including any Hispanic population

points between 1990 and 2000, and experienced overall population growth of less than 17 percent. The cut-off points for Hispanic composition in 1990, dif- ference in percent Hispanic (1990-2000), and population growth (1990-2000) correspond to the variable means plus one half their standard deviations.

Changing Population Composition: Education, Foreign-born Status and Family Structure

To account for the impact of heterogeneity in population composition on Inequality, our statistical model includes measures of change in educational composition, foreign-born representation, and demographic and family change in rural counties. We include two variables measuring 1990-2000 change in the share ofthe population with less than a high school education and at least a four-year college degree. We expect these variables to be positively associ- ated with growing inequality, lelative to expansion at the intermediate levels of education, since they directly measure increased heterogeneity in educational endowments. Moreover, given relatively low average education levels among Hispanics, we expect growth of the less educated population to mediate the impact of Hispanic population growth on inequality. In addition, we include a measure of counties' 1990-2000 change in foreign-born composition. Because immigrants typically occupy lower labor market positions, we expect a …