Great Recession and Income Inequality: a State-level Analysis

Transcription

1 Journal of Economics, Race, and Policy (2018) 1: ORIGINAL ARTICLE Great Recession and Income Inequality: a State-level Analysis Mehmet E. Yaya 1 Received: 6 February 2018 /Revised: 20 July 2018 /Accepted: 23 July 2018 /Published online: 7 August 2018 # Springer Nature Switzerland AG 2018 Abstract This paper analyzes the impact of the Great Recession on the income inequalities of racial and ethnic groups, namely whites, blacks, Hispanics, and Asians, in the USA. As the US economy fell into a deep recession during the late 2000s, the unemployment rate skyrocketed, the stock markets crashed, and incomes significantly declined. Using the American Community Survey from 2005 to 2016, this paper presents novel results that suggest the Great Recession not only increased the overall income inequality in the USA but also within- and between-group income inequalities among racial and ethnic groups. Furthermore, the impact of the recession on inequality is not uniform across these racial and ethnic groups. More specifically, during the Great Recession, inequality among blacks, Hispanics, and Asians has significantly increased, while whites experienced only a moderate increase in income inequality. The findings of this work are both timely and relevant, especially in terms of public policy. Policymakers cognizant of problems associated with within- and between-group inequalities might concentrate on policies alleviating the impact of the recession, especially those geared toward racial and ethnic minorities. Keywords Great Recession. Income inequality. Race. Ethnic minorities Introduction * Mehmet E. Yaya myaya@emich.edu 1 Eastern Michigan University, Ypsilanti, MI 48197, USA The US economy, infamous for its high income inequality among the developed economies, was rocked with a deep recession in the late 2000s, known as the Great Recession. The recession started in the housing market with sharp declines in housing prices; then, it spilled into the financial sector with the collapse of the Lehman Brothers in late 2008 due to declines in the values of mortgage-backed securities. Eventually, the effects of the recession were felt in the entire economy including the labor markets, industrial production, and international trade. The National Bureau of Economic Research (NBER) announced the official beginning of the recession as December 2008 after three-quarters of negative GDP growth. During the Great Recession, the unemployment rate rapidly increased from 5 to 10%. Furthermore, more than 5 million Americans lost their health insurance due to declining incomes and employment Holahan (2011), and a quarter of the families lost three-quarters of their wealth during the Great Recession. These losses disproportionally affected the lower-income, lesseducated, and minority families Pfeffer et al. (2013). The Great Recession finally concluded at the end of Some recently published articles have analyzed the consequences of recessions (see Pfeffer et al. 2013; Meyer and Sullivan 2013; Hellebrandt 2014; Piketty and Saez 2013; Salgado et al. 2014). From the labor market impact to the mental health, these papers showed the adverse effects of recessions. However, only a few published articles discussed the impact of economic downturns on income inequality. One possible reason for the lack of extensive research on the true impact of recessions on inequality might be the complex nature of income inequality. More specifically, the changes in inequality are multifaceted, and inequality is a persistent process. Furthermore, no single factor can solely explain the majority of the changes in income inequality, especially in the short run. However, it is natural to think that inequality, social problems, and demographical changes such as discrimination, segregation, poverty, immigration, and access to education and health are closely related to each other. This is especially true when a recession has non-uniform effects on vulnerable populations such as different racial and ethnic groups in the economy. In addition to the issues pertaining to the Great Recession, the US economy has been facing challenges associated with race and ethnicity, including but not limited to these minorities earnings, education, health, and income inequalities. It should be noted that the US population is very heterogeneous, with

2 J Econ Race Policy (2018) 1: blacks and Asians as the largest two minorities in the US population at approximately 12.6 and 4.8% of the US population, respectively, according to the 2010 US Census. Furthermore, Hispanics, the largest ethnic minority, constitute approximately 16.3% of the US population. These racial and ethnic groups in the USA have varying levels of earnings, poverty rates, educational attainments, health, and income inequality characteristics. Asians are repeatedly shown to earn the most (family income of Asians was $77,900 in 2014), while non-white Hispanics and blacks earn the least (both groups earn approximately $43,000). Furthermore, Asians have the lowest poverty rate, highest educational attainment (53% of Asians completed a college degree), and highest income inequality compared to blacks and Hispanics. In comparison, blacks tend to have highest poverty rates (26% of blacks live in poverty) while Hispanics have the lowest educational attainment (only 15% of Hispanics complete a college degree). 1 Furthermore, inequality in the USA has been increasing since the 1970s. Orrenius and Zavodny (2013) argued that Hispanics face higher inequality than non-hispanic whites but lower income inequality than non-hispanic blacks, based on the 2010 Current Population Survey. Their findings suggest that, with their relatively high poverty rate, Hispanics are at the bottom of the income distribution. Hoover and Yaya (2010, 2011) showed substantial differences in inequality among the racial and ethnic groups in the USA. Their results suggest that inequality is the lowest among Hispanics, followed by blacks and whites. 2 As important as the distribution of income and income inequality, the literature truly lacks a comprehensive analysis of the impact of the Great Recession on these issues. The needed comprehensive analysis would provide a better understanding of the impact of the Great Recession, facilitating a path to a public policy discussion regarding how to alleviate the adverse effects of recessions, especially in the context of within- and between-group income inequalities among the racial and ethnic groups. Income inequality is an indisputable measure of social mobility and available economic opportunities to individuals in a society; hence, the discussion of public policies toward understanding and alleviating income inequality has utmost importance. Furthermore, a comprehensive analysis of income inequality should incorporate the vulnerable populations who have not been fully examined in the literature. More specifically, females, immigrants, young adults, less-educated individuals, and racial and ethnic groups are some examples of populations that need to be studied for a complete understanding of these social issues. Unfortunately, few studies have examined the impact of recessions on overall income inequality, and almost none of them has focused on 1 For a comprehensive analysis of racial and ethnic groups earnings, poverty, education, and health in the USA, see (Pew Research Center 2016) andspe- cifically for Asians, see Segal et al. (2002). 2 For a general overview of income inequality in the USA, see Gottschalk and Danziger (2005). vulnerable populations. Furthermore, most of the earlier studies used overall income inequality. In the context of vulnerable populations, specifically racial and ethnic groups, it is possible to decompose the income inequality into within- and betweengroup inequalities, which demonstrate the share of inequality that can be attributed to the income inequality within each group as well as between one group and the rest of the population. Again, the decomposition of income inequality within and between vulnerable populations has not been explored in the literature, particularly during and after economic downturns, business cycles, and recessions. This paper fills the gap in the literature with the impact of the Great Recession on income inequalities of one of the vulnerable populations, i.e., racial and ethnic groups. More specifically, this paper analyzes the impact of the Great Recession on within- and between-group inequalities based on the Gini Coefficient and state-level average personal incomes. Income inequalities within- and between-groups at the state level over a 12-year span are analyzed before and after the Great Recession using the pooled regression approach with robust standard errors. The research questions that are being analyzed are fourfold: (1) Has the Great Recession increased the overall income inequality in the USA? (2) Is the impact of the Great Recession on income inequality uniform among the racial and ethnic groups? (3) Has the Great Recession changed the within- and between-group inequalities among whites, blacks, Asians, and Hispanics? and (4) Is the impact of the Great Recession on income inequality short-lived, or is there a permanent effect of the Great Recession on income inequality in the USA? The rest of the paper is organized as follows: the BLiterature review^ section presents a brief literature review. The BData^ section presents the data and the descriptive statistics of the variables. The BMethodology^ section introduces the econometric model, which estimates the effect of the Great Recession on overall and within- and between-group income inequalities among racial and ethnic groups across 50 states and the District of Columbia over a period of 12 years. The BResults^ section presents the results from the econometric models. The BConclusion^ section concludes. Literature review The Great Recession has quickly drawn attention in the literature, related to issues pertinent to labor markets, retirement age and investment, mental health, income, and wealth inequality (Pfeffer et al. 2013; Meyer and Sullivan 2013; Hellebrandt 2014; Piketty and Saez 2013; Salgado et al. 2014; Cho and Newhouse 2013; Couch and Fairlie 2010). Bucknor (2015) analyzed the labor market outcomes for whites and blacks especially during the Great Recession. She showed that whites with college degrees have higher employment rates than blacks, due to the fact that blacks were

3 114 J Econ Race Policy (2018) 1: disproportionately affected by the disemployment effects of the recession. She concluded that black men were among the worst hit during the Great Recession and recovery. The Great Recession also impacted the mental health of Americans, as the foreclosures in the housing market led to more frequently reported depression among adults (Cagney et al. 2014). Furthermore, McFall (2011) found that recession-related declines in labor market earnings and stock market losses increased the retirement age by 2.5 months. Finally, Argento et al. (2015) showed that people tapped into their retirement savings more during the Great Recession. Cynamon and Fazzari (2016) tied the slow economic recovery after the Great Recession to higher income inequality. The authors hypothesized that the rise of inequality had become a drag on expenditure, which in return slowed the economic recovery in the post-recession period. On the other hand, Fisher et al. (2013) showed that the Gini coefficient increased in the time period that coincided with the beginning of the Great Recession. Furthermore, Pfeffer et al. (2013) showed that wealth inequality has increased by 10% during the Great Recession. Meyer and Sullivan (2013)analyzedthe income and consumption inequality and found that these inequalities rose significantly during the Great Recession. In their working paper, Masterson et al. (2017) used the Levy Institute Measure of Economic Well-Being (LIMEW) to analyze the impact of changes in income and wealth on racial economic inequality, especially during the Great Recession. The authors showed that the racial inequality in the USA remains a function on within-group rather than betweengroup inequality, even during the Great Recession. Finally, Kochhar and Fry (2014) examined wealth inequality with the comparison of wealth of whites, blacks, and Hispanics during and after the Great Recession. Data from the Federal Reserve s Survey of Consumer Finances suggests that the wealth inequality between whites and blacks as well as the whites and Hispanics increased during the Great Recession. Unfortunately, the report by Kochhar and Fry (2014) neither delves into the decomposition of income inequality nor analyzes the potential long-run permanent effects of the Great Recession on racial and ethnic inequality in the USA. It is also important to examine the fundamental determinants of income inequality, since the main research question of the paper is to test whether the biggest economic downturn since the Great Depression, known as the Great Recession, has any impact on income inequality. The determinants of income inequality have been a subject of interest since the seminal work of Kuznets (1955), who examined the relationship between economic growth and income inequality across nations. Kuznets (1955) asserted that income inequality follows an inverted U-shape, known as Kuznets Curve. Low income inequality is observed at low levels of economic development, and inequality tends to increase with economic development up until a threshold level. With further economic development, inequality eventually decreases. A decade later, Aigner and Heins (1967) tested the Kuznets Curve using the states in the USA using average income, education, age, the share of whites, and the share of urban population across the states as main determinants of income inequality. Their results suggest that average income, age, education, and race (the share of whites) across states are the significant determinants of inequality. More specifically, Aigner and Heins (1967) showed that the states with higher incomes and higher shares of whites exhibit higher inequalities. Finally, they presented that the states with older populations have more equally distributed income due to the fact that older populations have a higher skill rigidity. The Kuznets Curve has been tested empirically over the years, repeatedly suggesting that there exists an inverted U-shape relationship between economic development and inequality (Rossi et al. 2001; Mollick 2012). Furthermore, Zandvakili (1999) deserves special attention in this context. He examined not only the potential determinants of income inequality measured with generalized entropy but also decomposed the inequality into within- and between-group measures. The fundamental factors that Zandvakili considered were race, age, education, and marital status. He argued that race, in conjunction with education and age, are the important factors determining the income inequality among female household heads. In terms of the decomposed inequality, he found that between-group income inequality increased from the 1960s to 1990s, and any improvement in the income inequality for female households is observed within each group that the author analyzed. As seen above, few studies indicate that inequality generally increases during periods of down- turn. However, other than Masterson et al. (2017), not many studies introduce race into the complex relationship among inequality and recessions. In fact, there are papers that have analyzed the impact of economic downturns on income inequality (see Hoover et al. 2009), but they do not incorporate race into their analysis. Yet, there are other papers that analyze the determinants of the racial and ethnic groups inequalities, such as Zandvakili (1999), Orrenius and Zavodny (2013), or Hoover and Yaya (2010, 2011), but these papers did not specifically examine the economic downturns. Consequently, the literature truly lacks a comprehensive analysis of the intersection of race, income inequality, and recessions. This paper endeavors to fill the gap in the literature with an extensive examination of income inequality before and after the Great Recession within and between racial and ethnic groups in the USA. Data The American Community Survey (ACS) is a nationally representative annual survey that provides information on income,

4 J Econ Race Policy (2018) 1: demographics, as well as the state of residence of respondents. The survey does not follow the same individual over time; however the large number of respondents in the survey makes ACS ideal to pursue a state-level analysis over time. For the purpose of the current analysis, the following variables from the 2005 to 2016 surveys were utilized: personal income, education, age, gender, and employment and immigration status of respondents to generate annual state-level averages. The time frame selected for this study conveniently provides data from 5 years prior and 5 years after the Great Recession, which lasted approximately 2 years. The recession indicator is a binary variable, taking the value of 1 if the year is 2010 or 2011, and zero otherwise. The sample is restricted to adults aged 18 to 65. The income variable is the sum of all personal incomes from various sources, including wages, salaries, self-employment income, investment income, and social security income. 3 The education variable in ACS is converted into an indicator, taking a value of one if a respondent has a college degree or above and zero otherwise. The age, gender, and immigrant status variables are self-explanatory. The employment status variable from ACS is used to generate the state-level unemployment rates for each racial and ethnic group by dividing the number of respondents indicating being employed by the number of people in the labor force. The selection of the explanatory variables is mainly based on the previous literature. These explanatory variables are frequently used by researchers as the main determinants of income inequality. Average state income can be used as a proxy for economic development as presented in Kuznets Curve which argues that states with higher income should have higher income inequality up to a threshold income level. Age has also been used as a determinant of inequality as in the works of Aigner and Heins (1967), Zandvakili (1999), and Hoover and Yaya (2010, 2011). States with older populations are expected to have equally distributed income due to skill rigidity, as proposed by Aigner and Heins (1967). Gender is also expected to play a role in inequality determination. Employment of females has shown to have a disequalizing effect on income inequality due to the types of jobs females hold as well as the higher share of females working in parttime jobs; therefore, states with larger shares of females are expected to have more equally distributed incomes. Immigration increases the heterogeneity of the US population, since immigrant demographics are vastly different than the US population. Therefore, states with larger shares of immigrants are expected to have higher income inequality. Education measures the share of state population with a college degree. 3 In few instances, the respondents reported negative income due to possible losses in investment or self-employment. These negative values were converted to zero. The number of negative income responses is very limited. Furthermore, in the literature, the respondents with negative income values are either removed (see Zandvakili (1999)) or converted into zero (see Hoover and Yaya (2010, 2011). It should also be noted that the personal income variable in ACS is top coded to maintain the anonymity of the respondents. As college degree holders increases in the population, the income distribution stretches due to the fact that education brings more opportunities to people with more education. Therefore, states with larger shares of college degree holders are expected to have higher income inequality. Finally, the higher unemployment rate compresses the income distribution. As more people fall into unemployment, states with higher unemployment are expected to have higher income inequality as in the works of Hoover and Yaya (2010, 2011). Using personal income, initially the Gini coefficient has been computed for each racial and ethnic group for each of the 50 states plus the District of Columbia over 12 years from 2005 to The Gini coefficient is a commonly used measure of inequality that takes a value between zero and one. The higher the coefficient, the more unequal the income distribution is. The Gini coefficient is popular in the literature, since it is sensitive to the changes in income at the bottom and top of the distribution. More specifically, the changes in the Gini coefficient are more pronounced when the share of income earned by the poor (or rich) changes. In comparison, other inequality metrics such as relative mean deviation or standard deviation are mainly concerned with dispersion around the average income. Figure 1 depicts the Gini coefficient for the entire sample and racial/ethnic groups between 2005 and The gray shaded area is the period of the Great Recession for the years 2010 and Although NBER announced the beginning of the recession as early as December 2007 and the official end date being the second quarter of 2009, the changes in income inequality generally lag behind the business cycles due to its persistent nature. Furthermore, the respondents of ACS were asked about their income in the previous year, making 2010 and 2011 the ideal years for the recession variable in the dataset. (Meyer and Sullivan 2013) used the same time period (2010 and 2011) within their study. The figure is consistent with the existing literature in terms of the increasing trend in the overall inequality observed in the USA. This overall increase in income inequality can be attributed to the overall decreases in income as well as non-uniform labor market outcomes among the vulnerable populations such as racial and ethnic groups during the recession. More clearly, incomes of racial and ethnic minorities dropped much faster than those of whites (Pfeffer et al. 2013); hence the income distribution in the USA stretched, leading a higher income inequality. Figure 1 clearly shows Asians have the highest Gini coefficient among the racial and ethnic groups, while Hispanics have the lowest. This finding is consistent with the literature, showing Asians exhibiting the most dispersed income distribution. There might be several explanations for this dispersion. One, it is possible that the incomes of Asians are drawn more from the tails of the income distribution, i.e., high incomes as well as low incomes, than other racial and ethnic groups. Furthermore, many Asians reside in large metropolitan areas with varying

5 116 J Econ Race Policy (2018) 1: Fig. 1 Gini Coefficient by racial and ethnic groups: Source: Author s own estimations based on the American Community Survey (ACS) for The estimates are based on nominal personal income. The total number of observations is approximately 22.8 million respondents over 12 years. Approximately 18 million respondents were white, 2.5 million were black, 3.0 million were Hispanic, and 1.4 million were Asian. levels of income. Finally, Asians have the highest average education, possibly opening doors to well-paying job opportunities in industries that might not be available to other racial or ethnic groups. On the other hand, Hispanics are shown to have the lowest dispersion of income mostly attributable to their low income coupled with the low average education among Hispanics. Again, the tight distribution of income suggests that the incomes of Hispanics are heavily clustered at the lower tail of the income distribution. There are sharp increases in inequality for the racial and ethnic groups, i.e., whites, blacks, Asians, and Hispanics, during the recession years. This result is consistent with and supports the findings of Pfeffer et al. (2013) and Meyer and Sullivan (2013). Inequality among whites follows the entire sample closely, and this is not a surprise, given the high share of this group in the entire sample. On the other hand, inequalities among blacks and Asians exhibit much sharper increases during the recession compared to whites and Hispanics. This suggests that blacks and Asians are subject to the adverse effects of the recession more than others, possibly due to factors specific to each of the groups, such as the types of professions these people work, their education level, and poverty rates. Furthermore, the Gini coefficients started at very close figures for blacks and Hispanics, but they diverged significantly in the post-recession period of 5 years. The same argument can be made for Asians and whites. These observations lead to the assertion that the recession has fundamentally changed the income inequality characteristics of racial and ethnic groups. The Gini coefficients are fairly flat for whites and Hispanics, whereas they tend to increase further for blacks and Asians within 5 years after the Great Recession. One possible explanation for the observed changes in inequality might be related to the labor market effects of the Great Recession. It is clear that the income inequality in the USA has not reverted back to prerecession levels; moreover, some racial and ethnic groups, especially blacks and Asians, have experienced further deterioration, possibly due to their inability to return to employment or earning levels in the post-recession period. Pyatt (1976) suggested that it is possible to decompose the Gini coefficient into three components; namely within-group, between-group, and overlap. The within-group inequality component shows the share of inequality due to income inequality that arises from the dispersion of income among the poor and rich within each group that make up the population. On the other hand, the between-group inequality can be obtained if every income in every subgroup were replaced by each relevant subgroup mean (Lambert and Richard Aronson 1993). The remaining portion of the Gini coefficient can be named residual or overlap. (Mookherjee and Shorrocks 1982) discussed the residual and the interaction effect, which they argued impossible to interpret precisely, except this residual is needed to maintain identity. Figure 2 shows the within-group inequality based on Pyatt (1976) decomposition of the Gini coefficient. Within-group

6 J Econ Race Policy (2018) 1: Fig. 2 Pyatt s within-group inequality by racial and ethnic groups: Source: Author s own estimations based on the American Community Survey (ACS) for The estimates are based on nominal personal income. The total number of observations is approximately 22.8 million respondents over 12 years. Approximately 18 million respondents were white, 2.5 million were black, 3.0 million were Hispanic, and 1.4 million were Asian. inequality explains approximately two thirds of the overall income inequality. Within-group income inequality is lowest among whites, and the other racial and ethnic groups have much higher within-group inequalities. Hispanics have the lowest within-group inequality, while Asians have the highest inequality. These findings are also consistent with the literature (Masterson et al. 2017). Unlike the overall income inequality shown in Fig. 1, within-group inequalities for all racial and ethnic groups are fairly flat, with only moderate changes from 2005 to In fact, within-group income inequality slightly declined in the pre-recession period for minorities, while it was constant for whites. The gray shaded areas again represent the period of recession. A noticeable increase during the recession is observed in within-group inequality in all racial/ethnic groups, and the increases among Asians and Hispanics are more pronounced, possibly due to heterogeneity of these racial and ethnic groups in terms of their performances in their respective labor markets. More specifically, it is clear that Hispanics sort themselves in professions that are mainly located at the bottom of the income distribution as suggested by their average income. These professions, such as those in the construction industry, might be heavily hit by the recession which leads to more dispersion of income within Hispanics. The same argument can be made for Asians, but this time, it should be noted that Asians earn the highest income on average, and recession again might hit the Asian income earners but this time at the top of the distribution, leading to more unequally distributed income within Asians. Between-group inequality based on Pyatt (1976)decomposition of the Gini coefficient measures how much of the overall inequality can be attributed to the inequality between these racial groups. For example, the between-group inequality for whites measures the inequality between whites and nonwhites. It should be noted here again, the between-group income inequality based on Pyatt s decomposition can be obtained if every income in every subgroup is replaced by each relevant subgroup mean. In other words, the between-group inequality of whites vs. non-whites can be calculated by replacing the incomes of blacks, Asians, and Hispanics with the average income of whites. Once the replacement is completed, the Gini coefficient can be calculated as usual. Figure 3 exhibits these between-group inequalities, showing visible spikes among whites vs. non-whites and blacks vs. non-blacks during the recession. However, the between-group inequalities remain fairly flat for Asians vs. Non-Asians and Hispanics vs. Non-Hispanics. The spike in between-group inequality for whites vs. non-whites during recession is expected. Masterson et al. (2017), Pfeffer et al. (2013), Meyer and Sullivan (2013) have all shown that the Great Recession had a non-uniform impact on whites and minorities. The dispersion of income between whites and non-whites increased

7 118 J Econ Race Policy (2018) 1: Fig. 3 Pyatt s between-group inequality by racial and ethnic groups: Source: Author s own estimations based on the American Community Survey (ACS) for The estimates are based on nominal personal income. The total number of observations is approximately 22.8 million respondents over 12 years. Approximately 18 million respondents were white, 2.5 million were black, 3.0 million were Hispanic, and 1.4 million were Asian. during the Great Recession, possibly due to minorities disproportional exposure to the adverse effects of the recession, such as declines in income. The same argument can be made for blacks, since Fig. 3 exhibits a similar spike during the recession. The between-group income inequality for blacks and non-blacks has significantly increased. It should be noted for both groups, i.e., whites and blacks, the between-group inequality in the post- recession period reverts back to approximate pre-recession levels within 5 years of the Great Recession. On the other hand, the between-group inequality for Asians vs. non-asians as well as Hispanics vs. non- Hispanics does not seem to respond to the Great Recession. Both of these between-group inequalities have upward trends for the 12 years analyzed. It is also noteworthy that betweengroup inequality for Asians is significantly lower than the other racial and ethnic groups. Table 1 shows the descriptive statistics for both dependent variables, the Gini coefficient, and explanatory variables, i.e., the state average values of income, education, age, share of females, unemployment rate, and share of immigrants. The total number of observations is 2448 for a strictly balanced panel of 50 states plus the District of Columbia over 12 years among four racial and ethnic groups. Table 1 panel A shows the mean, standard deviation, and the extreme values of the Gini coefficients. There are three Gini coefficients, which are the overall Gini coefficient, and the decomposed forms, i.e., Table 1 Descriptive statistics Panel A: dependent variables Obs. Mean Std. Dev. Min. Max. Gini coefficient Pyatt s between-group inequality Pyatt s within-group inequality Panel B: explanatory variables Obs. Mean Std. Dev. Min. Max. Income 2448 $31, $ $ $88, Age Female share Immigration share Unemployment rate College share

8 J Econ Race Policy (2018) 1: within- and between-group inequalities introduced by Pyatt (1976). The mean value of the overall Gini coefficient among the racial and ethnic groups is The minimum Gini coefficient of 0.32 is observed among Hispanics in South Dakota in 2006 and the maximum value of 0.74 among Asians in Vermont in When the Gini coefficient is decomposed, the between-group inequality constitutes a much smaller share compared to within- group inequality. There are racial and ethnic groups in few states, and values are for relatively small groups in their respective states and years. For instance, between-group inequality is zero for Asians in Kansas in 2006, Hispanics in West Virginia in 2005, and blacks in New Mexico in In comparison, between-group inequality for whites, which is a sizable group in each state and year, never takes a value of zero. Table 1 panel B shows the descriptive statistics of the explanatory variables. The average income is $31,505 with a standard deviation of $9874. The average age is years, and the female share is approximately Moreover, the immigration share on average is 0.34, with a minimum of zero observed in states with rather sparse populations. For instance, Wyoming and Montana are such states. In addition, Mississippi, Arkansas, Louisiana, and West Virginia have very low immigrant shares in the dataset. Finally, the average unemployment rate and the share of population with college degrees are 0.08 and 0.10, respectively. The unemployment rate ranges between zero and 0.28, and the lowest unemployment rate in the dataset is observed in a few sparsely populated states among blacks, Hispanics, and Asians. Likewise, the share of the population that has college degree is zero in some states with a small number of respondents. For example, Hispanics in 2006 in North Dakota have 0% of respondents with a college degree. Methodology Figures 2 and 3 indicate that the impact of the Great Recession is not uniform across the racial and ethnic groups in the USA. These findings are, however, only descriptive; an econometric analysis is required to test the hypothesis of the non-uniform effect of the Great Recession on inequality. This paper analyzes the impact of the Great Recession on the within- and between-group inequalities among the racial and ethnic groups using state-level data and the multivariate pooled ordinary least squares (OLS) regression model. The methodology is based on an estimation model utilized by (Hoover and Yaya 2010, 2011), who presented the factors contributing to income inequality at the state level. Using a similar panel estimation technique but incorporating the Great Recession as a possible explanatory variable, the following three models are estimated: Gini i;s;t ¼ α 0 þ α 1 Recession t þ α 2 Income i;s;t þ α 3 Education i;s;t þ α 4 Age i;s;t þ α 5 Sex i;s;t þ α 6 Unemployment i;s;t þ α 7 Immigrant Share i;s;t þ γ s þ δ t þ σ i þ Within Inequality i;s;t ¼ α 0 þ α 1 Recession t þ α 2 Income i;s;t þ α 3 Education i;s;t þ α 4 Age i;s;t þα 5 Sex i;s;t þ α 6 Unemployment i;s;t þ α 7 Immigrant Share i;s;t þγ s þ δ t þ σ i þ ð1þ ð2þ Between Inequality i;s;t ¼ α 0 þ α 1 Recession t þ α 2 Income i;s;t þ α 3 Education i;s;t þ α 4 Age i;s;t þα 5 Sex i;s;t þ α 6 Unemployment i;s;t ð3þ þ α 7 Immigrant Share i;s;t þγ s þ δ t þ σ i þ Equations 1, 2, and 3 analyze the overall, within-, and between-group Gini coefficients, respectively. The subscripts i, s, and t indicate the race (whites, blacks, Asians, and Hispanics), states (50 states and the District of Columbia), and time (from 2005 to 2016). Furthermore, the models include state and racial/ethnic group fixed effects as well as time fixed effects. These models are estimated for the pooled sample, including all racial and ethnic groups, also separately for each of the racial and ethnic groups. The models use state-level averages on income, education, age, female share, unemployment rate, and immigrant share as explanatory variables. As it is explained above, the selection of the explanatory variables is mainly based on the previous literature. Following the original Kuznets 1955) work, states with higher income is expected to have higher income inequalities, so the expected sign for income is positive. Moreover, states with older populations are expected to have more equally distributed income due to skill rigidity proposed by Aigner and Heins (1967); therefore, the expected sign for age is negative. States with larger share of females is expected to have a less equally distributed income; hence, larger share of females is expected to increase the income inequality. In addition, the states with a larger share of immigrants are expected to have a higher income inequality, so the expected sign for immigrant share is positive. As the college degree holders increase in the population, the income distribution stretches due to the fact that education brings more opportunities to people with more education. Therefore, states with a larger share of college degree holders are expected to have a higher income inequality. Finally, the higher unemployment rate compresses the income distribution. States with higher unemployment is expected to have higher income inequality; hence, positive coefficient is expected for unemployment rate in these regression models. Finally, the econometric models include an indicator, i.e., recession, which takes a value of one during the Great Recession, and zero otherwise. The Great Recession has increased the unemployment rate, decreased the earnings; hence, the hypothesis is that the recession should have a statistically significant and positive impact on

9 120 J Econ Race Policy (2018) 1: Table 2 Regression results for the Gini coefficient Dependent: Gini coefficient All Whites Blacks Asians Hispanic Recession (0.0034)*** (0.0023)*** (0.0073)*** (0.0069)*** (0.0058)*** Income (0.0031)*** (0.0037)*** (0.0104)** (0.0058)*** (0.0114)*** Age (0.0012)*** (0.0014)*** (0.0019)*** (0.0028)** Female share Immigrant share (0.0549)* Unemployment rate (0.0454)*** (0.0492)** (0.0718)** (0.1119)*** (0.0855)*** College share Constant (0.0466)*** (0.0666)*** (0.0998)*** (0.0881)*** (0.1227)*** Time trends Yes Yes Yes Yes Yes State fixed effects Yes Yes Yes Yes Yes Racial/ethnic fixed effects Yes No No No No R N Dependent variable is Gini coefficient. Robust standard errors are in parenthesis *p < 0.1; **p < 0.05; ***p <0.01 inequality in the USA, suggesting that the Great Recession should increase overall, within-, and between-group inequalities among racial and ethnic groups. Results Table 2 shows the results from the estimations using Eq. 1.The dependent variable is the overall Gini coefficient in all columns. The first column presents the results for the ordinary least squares (OLS) with robust standard errors for the entire sample. The model includes state fixed effects, racial and ethnic group fixed effects, as well as time fixed effects. Number of observations for the full sample is 2448 (50 states plus DC over 12 years for 4 racial and ethnic groups). The average impact of the recession on the Gini coefficient is positive and statistically significant. The results suggest that the Great Recession has increased the overall Gini coefficient in the USA by Given the fact that the Census Bureau reported the Gini coefficient to be hovering around 0.46 late in 2000s, the recession has increased the income inequality by approximately 5%. 4 Furthermore, the coefficients for income, age, and 4 historical-income-inequality.html. unemployment rate are all statistically significant. More specifically, high inequality is associated with high income and unemployment rates, but the reverse is true for age. The coefficients are in line with the expectations and consistent with the earlier literature (Hoover and Yaya 2010, 2011). It should be noted that female, immigrant, and college degree holder shares in the population do not have any significant impact on inequality. Although not reported for brevity, the racial and ethnic group fixed effects are all statistically significant, suggesting that the recession did not have a uniform impact on the overall inequality among the racial and ethnic groups. In the next four columns, the same model is estimated on each racial and ethnic group individually, namely whites, blacks, Asians, and Hispanics. In each of these regressions, racial and ethnic fixed effects are excluded. The results for whites are similar to column one, with the same explanatory variables being statistically significant. Notably, the major difference is that the magnitude of the effect of the recession on the inequality of whites is half of what has been shown in column one. For whites, the Great Recession has increased the Gini coefficient but only moderately with a magnitude of The Great Recession has increased the inequality for whites but at a much smaller magnitude compared to the average impact of the recession. It is possible to assert that whites did not experience the adverse effects of the recession in terms of

10 J Econ Race Policy (2018) 1: inequality as much as the racial and ethnic minorities. It is also notable that age has a positive impact on the inequality for whites, suggesting that aging leads to higher inequality among whites. Column three presents the results for blacks. The Great Recession has increased the inequality for blacks by , which is three times that of the whites and higher than the average impact estimated in column 1 which is The recession has increased the income inequality of blacks the most. Furthermore, income and unemployment are associated with higher inequality while age is with lower inequality among blacks, consistent with the previous literature. The results for Asians are presented in column 4. The impact of the recession is approximately and is still positive and statistically significant, slightly above the average effect. It is also worth to mention that age does not seem to affect the inequality among Asians. Finally, the impact of the recession on inequality among Hispanics is provided in column 5. The impact of the Great Recession on inequality for Hispanics is the lowest, after whites. This is expected since Hispanics have the lowest Gini coefficient as shown in Fig. 1, which can be explained with the group s low income, low education, and high share of immigrants. For Hispanics, the share of immigrants is also significant, suggesting that in states with higher share of immigrants Hispanics have lower income inequality. It should be noted that in the last two to three decades, the south and southwestern states where Hispanics predominantly reside have been receiving an influx of immigrants with low education and they fill low-paying positions which leads to even lower income inequality due to higher concentration of population stacking at the lower tail of the distribution. As presented above, Pyatt (1976) decomposed the Gini coefficient into three parts: within-group, between-group, and overlap. The within-group inequality component shows the share of inequality that arises from the dispersion of income among the poor and rich within each group that makes up the population. On the other hand, the between-group inequality can be obtained as follows: hypothetically replace every income in every subgroup by each relevant subgroup s mean income then calculate the Gini coefficient as usual. Finally, residual or overlap is the remaining portion of the Gini coefficient, which can be considered as the interaction effect and it is impossible to interpret precisely, except this residual is needed to maintain identity. In Table 3, the Gini coefficient is replaced with the decomposed within-group income inequality measures, as shown in Eq. 2. AsseeninFig.2, the majority of the overall income inequality can be explained by the within-group inequality. The results are presented in the same manner as in Table 2, with the dependent variable being the only difference. Table 3 Regression results for Pyatt s within-group income inequality Dependent: Pyatt s within inequality All Whites Blacks Asian Hispanics Recession (0.0040)** (0.0019)* (0.0015)*** (0.0017)*** Income (0.0015)** Age (0.0008)** Female share (0.0260)*** (0.1065)* Immigrant share (0.0113)*** (0.0771)* Unemployment rate (0.0352)*** College share Constant (0.0322)*** (0.0740)*** (0.0239)*** (0.0149)*** (0.0228)*** Time trends Yes Yes Yes Yes Yes State fixed effects Yes Yes Yes Yes Yes Racial/ethnic fixed effects Yes No No No No R N Dependent Variable is within-group inequality. Robust standard errors are in parenthesis *p <0.1;**p <0.05;***p <0.01

11 122 J Econ Race Policy (2018) 1: In column one, the entire dataset with all four racial and ethnic groups is examined. The results for each racial and ethnic group are presented in the same order, i.e., whites, blacks, Asians, and Hispanics, in the following four columns. The results are slightly more mixed than the overall income inequality results that are presented in Table 2. In Table 3 column 1, the average impact of the recession on within-group inequality is shown to be , much smaller than the overall Gini coefficient presented in Table 2 column 1. Interestingly, income is not statistically significant, suggesting that within-group income inequality is independent of income. When all racial and ethnic groups are pooled, the results suggest that age, female and immigrant shares, and unemployment rate are the fundamental factors affecting the within-group inequality. More specifically, higher age, female share, and unemployment rate decreases income inequality. On the other hand, higher immigrant share increases the within-group inequality. When whites are examined in Table 3 column 2, the recession does not seem to impact the within-group inequality of whites. Furthermore, neither income nor age or unemployment can explain the changes in the within-group inequality of whites. The only two factors that seem to explain the changes in the within-group inequality for whites are the female and immigrant shares. The results suggest that whites who reside in states with higher shares of females and immigrants have higher within-group inequality. The results for whites call for further research in order to understand the true dynamics of withingroup inequality. There might be some possible explanations for fairly stable within-group inequality among whites. First, as the majority race in the population, any change in the withingroup inequality might require several unobservable factors such as attitudes toward minorities, discrimination, or role of diversity to change, which is not easy to capture in a regression model. Alternatively, these observables might be fairly stable, captured in the constant term of the model. Finally, two statistically significant factors, i.e., the female and immigrant shares might actually capture the majority of the variation in the within-group inequality among whites. Figure 2 which depicts the within-group inequality for whites support the results in Table 3 column 2. In the figure, the within-group inequality seems to be fairly flat over the 12 years. For non-whites, i.e., blacks, Asians, and Hispanics, the within-group income inequalities increased during the Great Recession. The largest impact is observed within Asians, with the impact being approximately (Table 3, column 4), while blacks and Hispanics experienced lower increases in within-group inequalities. The increases in within-group Gini coefficients due to the Great Recession are and for blacks and Hispanics, Table 4 Regression results for Pyatt s between-group income inequality Dependent: Pyatt s between inequality All White Black Asian Hispanic Recession (0.0023)*** (0.0019)*** (0.0010)*** (0.0006)*** Income (0.0020)** (0.0026)*** (0.0008)*** (0.0008)*** Age (0.0016)*** (0.0001)** Female share (0.0143)*** (0.0052)** Immigrant share (0.0068)*** (0.0853)*** Unemployment rate (0.0206)*** College share (0.0180)** Constant (0.0066)*** (0.0063)*** Time trends Yes Yes Yes Yes Yes State fixed effects Yes Yes Yes Yes Yes Racial/ethnic fixed effects Yes No No No No R N Dependent Variable is between-group inequality. Robust standard errors are in parenthesis *p <0.1;**p <0.05;***p <0.01