FINAL PAPER: BY CAROLA ESCOBAR
A COMPARATIVE ANALYSIS IN TERMS OF POVERTY IN THE U.S.: THE CASE OF LOWER MISSISSIPPI AND TEXAS BORDERLAND
ABSTRACT
This paper replicates a study presented by Slack et. al. 2009,which is based on a comparative analysis in a county-level poverty distribution across family types in two areas, Texas Borderland and the Lower Mississippi Delta. Using county-level figures from 1990 and 2000 U.S. Census Data, an Ordinary Least Squares Model was estimated with family type as the dependant variable. The dimensions of aggregate poverty level defined independent variables as employment, human capital, population structure, and non-metro residencies. We demonstrated as well as Slack et. al. did, the factors which contribute to poverty levels, with the exception of employment, are different between single mother’s families and married couples. We find that in areas like the Delta,which are characterized as rural and having a high migration rate,there is a tendency to suffer higher rates of poverty, especially in married couples.We alsofindthat Slack et. al. should not have estimatedthe pooled models but should have estimated the states separately. The results of these estimations show substantial differences from the pooled results in Slack et al.
A COMPARATIVE ANALYSIS IN TERMS OF POVERTY IN THE U.S.: THE CASE OF LOWER MISSISSIPPI AND TEXAS BORDERLAND
INTRODUCTION
The United Statesis a country characterized by continuous progress in economy, science, technology, and culture. Despite these improvements there are two regions that have lagged behind: the Texas Borderland and Lower Mississippi Delta. These two regions have been characterized by a consistent level of poverty. The majority of the population in these areas is constituted by racial/ minorities, blacks in the Delta and Hispanics in the Borderland, which is considered one of the factors that contribute to the increasedrate of poverty in these places.
A study performed by Slack et al 2009 highlights the problem of poverty in the Texas Borderland and Lower Mississippi, which is the base for our investigation. Slack et. al address the problem of poverty in terms of family type comparisons for the differences between regions. Family structure is their main point of their comparative analysis because the unequal economic differences between single – female headed families and families headed by married couples.
In our study, we provide a comparative analysis between the Borderland and Lower Mississippi, following the same procedures that Slack et al did for their study. In terms of observing similarities and differences in the patterns and aggregate mechanisms that power poverty in these two regions within family structure, we provide an alternative vision of looking at poverty in the U.S.
Regional Poverty
Past studies have pointed out the relationship between the place and space, and how this relationship causes differences between populations (Slack et. al 2009). In thelast two decades spatial inequality still persists due to resources differentials, but these are mostly due to income differentials related to labor scarcity (Kanbur and Rapoport 2004). This labor scarcity causes the movement of different types of population between states, counties, and even sometimes countries. Today, the U.S. is characterized by a level of consistent poverty and spatial inequality (Tickamyer, 2005).
There is also a relationship between the place and poverty. For instance, there are areas in U.S, where poverty is constant and persistent, such as Native American reservations includingAppalachia, the Delta, and the Borderland (Slack et. al 2009). According to Tickamyer, among these areas are counties where the poverty is persistent. These areas had different economic, cultural, andracial- ethnic histories.
The areas commonly characterized by constant poverty are located near the rural South, West U.S., and the Appalachia. These areas were “characterized by high concentrations of racial/ethnic minorities”(Slack et al 2009). The areas were also characterized by high rates of unemployment, no investment or infrastructure development, no services, an inefficient political system, and low levels of capital (Tickamyer, 2005).
The Borderland, which is an area where Latin and Germanic cultures interact, is a mixture and hybridization of Anglo and Mexican cultures. “A border subculture, whose practitioners call themselves fronterizos has apparently formed” (Jordan, 1994). Usually, in the agriculture business workers of Mexican descent have a low labor wage (Snipp 1996). The low wage labor increases the chance of being poor and cuts opportunities to obtain a better life. Also, the incapacity of moving labor and capital is the main characteristic between the borderland and the U.S interior. (Mora, and Davila 2006). According to Slack et al 2009, it has recently been established that “labor intensive plants known as maquilladoras” which are located on the Mexican side of the border, contributes to the low labor wage for Mexican workers in agribusiness. For Hispanic workers, there are border wage differentials comparing earnings between border and interior workers. In addition, earnings for Latin workers are lower because of “relative low average levels of human capital” (Mora, Davila, and Mollick 2007). In the last forty years, the Mexican and U.S. governments have tried to improve the conditions for workers in the Borderland. Different policies have been implemented like the Bracero Program (1942-1964), Industrialization Program (1965 to present), and North American Free Trade Agreement (1994 to present); however, the region still does not demonstrate much improvement in terms of economy (Slack et. al 2009).
The Delta is a region which is characterized by extreme poverty and discrimination amongst black and white. The success of local elites in ejecting external forces from gaining any advance in social, economic, or political means is the main characteristic of the region. These elites dominate local politics, having the power to discourage “external investment in the area to maintain their positions of power and privilege” (Tickamyer, 2005). The Delta is the place for fertile soil which is located in the Mississippi River Floodplain. Also, the region has the feature of cotton production, and plantation agriculture (Slack. et al 2009). In addition, this is an area considered as the “Black Belt”, which means “a southern sub region of counties with greater – than – average concentrations of African American residents” (Allen-Smith, Wimberley, and Morris, 2000). In absence of the sufficient infrastructure such as road systems and rails, and limitations of educational and economies opportunities; millions of blacks left the South, going to the Northern cities and the West. This “out- migration” started between the 1900s and 1970s (Slack et. al 2009). A consistent poverty is the main feature for the Borderland and Delta regions. Factors like race, rurality, and family structure contribute to increases in the undervelopment in these places in the U.S.
The family structure is an important factor that reinforces the tendency of poverty. The feminization of poverty is a reality that faces developed and undeveloped countries. According to Thibos et al 2007, “a feminized poverty is the rate of poverty among children, who disproportionately reside in female- headed households”. Despite the evolution of women among the labor force between 1960 and 1970, women are still part of a large proportion of the population. Poverty rates vary by family structure, because single female headed families have a higher risk of being poor, no matter their place of residence (Lichter and McLaughlin 1995, McLaughlin and Sachs 1988). Also, race is an important factor between the feminization of poverty, for black and Hispanic women it is harder to get out poverty. For Hispanic women who face economic hardship, a lack of viable employment opportunities, language barriers, and even sometimes legal status problems, the chances of obtaining a better life are low (Thibos, Loucks, and Martin 2007). Having a correlation between poverty and family structure, a welfare reform expresses the promotion of marriage and encouraging the population to maintain the status of “two parents headed families” (H.R. 3734). Marriage could be considered an alternative to obtaining a better job for improving the condition of the women, especially those who have disadvantages to obtain a joband language barriers, marriage is an option to expand economic barriers (Licher, Graefe, and Brown 2003). According to Hoffman & Averett, variation in the family structure has implications over the “income dynamics, especially poverty and welfare receipt”. Also, Hoffan & Averett present statistics which reveal for 2005; women have crossed the barrier of married. For the first time, the Census Bureau show “that less than half of women age 15 and older were living without a husband”. Compared with 1950, 35 percent were in this category. In addition in 2006, 120 million women were classified as aged 15 and older. Within this figure, 59.5 million, or under half were married; which means with the spouse present, or spouse absent. Then this group was divided by those who are separated, but not divorced. A twenty six percent of the 120 million of women never married, also 13 percent were actually separated or divorced. The rest, 9.4 percent is classified as widows.
Factors that Affect Poverty
According to Slack et al 2009, there are aggregate mechanisms that influence poverty. These mechanisms are classified in four dimensions: population structure, employment structure, human capital, and non-metro residence.
Population Structure
The theory has demonstrated that the structure of the places are “strongly correlated” to poverty levels (Slack et al 2009). Also, net migration is a factor to be considered powerful for the economy of any place. Age, is another important factor in determining poverty (Cotter 2002: Rupasingha and Goetz 2007). Immigration is a factor that contributes to poverty in a negative way.
Employment Structure
Research has demonstrated that people who work are negatively correlated to poverty. Also, people who work in sectors such as finance, insurance, and real state (FIRE), manufacturing are negatively related to poverty because they provide a better stability in terms of earnings ( Slack et. al 2009, Cotter 2002; Rupasingha and Goetz 2007). In contrast, people working in the agriculture sector have a positive correlation with poverty (Albrecht, Albrecht, and Albrecht 2000).
Human Capital
Past research has demonstrated that education is an important factor in determining poverty. Education levels are lower in rural areas (Mosley and Miller 2004). Also, English speaking ability is an important factor for the immigrant population(Davila and Mora 2000).
Non-metro Residence
Usually, poverty is related to rural areas, which are constantly compared with urban areas. A large part of the population lives in rural areas, which poverty rates are higher and constant (Adamas and Duncan 1992: Summers et. al 1993).
DATA AND METHODOLOGY
Our investigation is based on the data provided by the U.S. Census Files from 1990 and 2000. This data was distributed on a county-level in two areas; the same data Slack et al. 2009 used in their study. The first area, Borderland (Texas), stretches across Rio GrandeRiver from El Paso until Brownsville and includes 41 counties, where Hispanic people are the major population group in this area by 2000, representing 80, 2 percent of the total population in the year 2000. The second area to be studied was delineated by the Lower Mississippi Delta Development Commission 1.The Delta counties were restricted to three States: Mississippi, Arkansas, and Louisiana. This area includes counties along the Mississippi River from northeastern Arkansas to New Orleans, has and includes 133 counties. According to Slack et al., by the year 2000, 35 percent of the population in the Delta was Black, and Blacks constituted the majority population in 30 of the 133 counties surveyed. With the help of a map (figure 1, 2000 U.S. Census Summary Files) we followed the same line distribution as described by Slack et al to identify the counties to be studied in those two regions. This estimation procedure resulted in 41 counties for the Borderland (Texas) and 137 counties for the Delta Region. This included four more counties than were included in the Slack et al study but they were chosen according to the same criteria so those 4 counties were not dropped from this investigation.
Using the data from the 1990 and 2000 U.S. Census Summary Files, the dependent and independent variables were defined. The dependent variable is drawn from the 2000 U.S. Census Files, which is the percentage of families with children whose total income in 1999 was lower than the official poverty thresholds (Slack et. al 2009). “In 2000, the average poverty threshold for a family of four was $17,603; for a family of nine or more persons the threshold was $35,060 or for an unrelated individual aged 65 or more was $8,259” (US Department of Labor Bureau of Labor Statistics, March 2002). This set of family poverty definitionswas used classify families with children which were headed by married couples, singles females, and single males.
The independent variables were taken from the 1990 U.S. Census Summary Files. These variables were based on four dimensions of aggregate–level poverty predictors: employment, human capital, population structure, and non-metro residencies (Slack et. al. 2009).Employment by “industrial mix” is a very important factor in understanding poverty. For this reason is county–level employment structure was indentified as the percentage of employed people in manufacturing, agricultural, and FIRE sectors. Literature has also shown that the population structure of an area in this study is related to poverty rates. The variables associated with county–level population structure are: percentage of population under 15 years of age, net migration, percentage of the population which is foreign-born, and the percentage of the population in the predominate ethnic minority in the area of study (Slack et. al. 2009).
Human capital levels are also important predictors of county-level poverty. The variables indentified as “county–level human capital” are: percentage of people who do not speak English well or at all and the percentage of population aged 25 years and older with a high school or above education.
The degree to which an area is rural or metropolitan reveals important information about the specific areas. In general it is expected that rural areas will have poverty rates that are higher than in urban areas. The strategy followed by Slack et al., was to examine “descriptive statistics” comparing county- level poverty by family type in the Delta (Arkansas, Louisiana, and Mississippi to the Borderland (Texas)). Then, Slack et al. estimated an Ordinary Least Squares model, applying a lagged panel design. This design provided a way to connect family type poverty with a county-level employment structure, human capital, population structure, and non-metro residence through the two regions. In addition, this technique relates variables in different times. Slack et al. restricted the sample in their regression to counties with at least 30 married couples or single female-headed households for each county. Counties not meeting these criteria were eliminated from the study. This restriction required the removal of three counties for the Borderland. Because of the small number of counties in the Borderland, Slack et al. made estimations for “regionally pooled models”, which implied an addition of a dummy variable for the Borderland region.
Y = Percentemploy + Percentmanuf + PercentFIRE + Netmigra
+ Percentunage15 + Percentforeingb + Percentminor + Percentlessths
+ Borderland + Nonmet + Borderland*Nonmet
Where:
Percentemploy: percentage of population employed
Percentmanuf : percentage of population in the manufacturing sector
PercentFIRE : percentage of population in FIRE
Netmigra : Net Migration
Percentunage15 : percentage of population under age 15
Percentforeingb : percentage of population foreign born
Percentminor : percentage of population which is considered minority ( blacks in
Delta, Hispanics in the Borderland)
Percentlessths : percent of population with a less than high degree
Borderland : Borderland =1; Delta = 0
Nonmet : Non – metro = 1; Metropolitan = 0
After estimating the pooled regression model, the next step was testing models to measuring their accuracy. These tests included multicollinearity and spatial correlation. Multicollinearity problems forced Slack et al. to drop one the variables “English language ability.” Also, spatial autocorrelation was diagnostics. Such type of autocorrelation is presented when studies using geographically unit of analysis. Is a correlation of a variable with itself through space (Lembo, Jr Cornell). It can results from spatial errors or spatial lag (Slack et al. 2009). The results indicated that models did not present spatial problems.
Following the Slack et al. procedure, counties at least 30 married couples and single, female-headed families were identified and other counties were excluded. This cutoff resulted in three counties less for the Borderland (Kenedy, McMullen, and Terrel) being excluded with 38 counties remaining in the sample. The final sample thus included 137 counties for to the Delta Region and 39 counties for the Borderland. Ordinary least squares (OLS) were employed, using a lagged panel design which relates family type-specific poverty to county-level employment structure, human capital, population structure, and non-metro residence across these two areas.
We examined the models for models heteroskedasticity (using Whites Test) and for multicollinearity (using variance inflation factors). The model related to the Borderland (Texas)was found to have collinearity between the variables of Migration and English language ability, which forced us to remove these variables. Then, we added a dummy variable for the Borderland (Texas). Collinearity was also found to be a problem between the variables representing the percentage of the population that is Foreign born, and the variable representing English ability. These two variables were removed from the model in order to have a good model.
Since, following Slack et al., a pooled sample was being used; the sample was tested for structural change between counties from each state. The “Chow Test” was applied in our case in order to determine if the coefficients of our model are equal in separate subsamples (Davidson and MacKinnon 2004). We split our data in two groups, for example one group comprising three states such as Texas, Louisiana , Mississippi and the other by one state, in this case Arkansas. The process was repeated successively for the married headed – families, and the single headed- families.The test indicated that most of the coefficients were significantly different by state, with an exemption of Arkansas for married headed families, and for Texas in the case of single female – headed families. An OLS model was then estimated for each State without dummies, and variables like Foreign born, and English ability. This new model revealed no problems of heteroskedasticity or multicollinearity and the results from this model were our final results.