Abstract
The United States experienced the Great Recession between 2007 and 2009 and many cities, suburbs, and communities are still suffering from its legacy. Yet, little is known about whether or not the crisis led to increases in student mobility, particularly for minority households. This study analyzed data from a large urban district located in San Bernardino, California, to examine the interplay of school mobility and race and class during the economic downturn. Using a logit model, findings show that race and class together interacted in ways that made Black students particularly vulnerable to school changes. Research and policy implications are discussed.
The United States experienced the Great Recession between 2007 and 2009 and many American cities, suburbs, and communities are still suffering from its legacy. The housing crisis and economic recession that ensued disrupted the residential stability of many families through a wave of foreclosures and unemployment. During this time, Black residents were particularly vulnerable as recent research suggests predominantly Black neighborhoods having been the most negatively impacted by the housing crisis (Hall, Crowder, & Spring, 2015; Rugh & Massey, 2010). Not only did more Blacks, proportionally, lose jobs, but also those losses were more likely to force Black residents to move. Similarly, more Black homeowners, proportionally, entered foreclosure and were more likely to end up moving than foreclosed households of all other races (Stoll, 2013). Between 2007 and 2009, Blacks were more than 70% more likely to lose their homes to foreclosure than Whites (Bocian, Li, & Ernst, 2010). Black homeowners were made even more vulnerable because they were 30% more likely to receive high-cost subprime loans relative to White subprime borrowers with similar credit scores. One concern is that their children may have had to change schools as a result of a residential move, particularly since school mobility of students in primary and secondary urban schools has been found to be associated with a wide range of negative academic outcomes, including graduation rates, and academic achievement outcomes, to name a few (Hanushek, Kain, & Rivkin, 2004; Rumberger & Larson, 1998; Temple & Reynolds, 2000; Voight, Giraldo-García, & Shinn, 2017). Involuntary moves from either a foreclosure or a parental job displacement particularly in a context of relatively high poverty, characteristic of many urban settings, are likely even more harmful to students as choices may be limited by the urgency of the move (Ellen, Madar, & Weselcouch, 2015; Scanlon & Devine, 2001).
Although many recent studies have argued that the U.S. housing bust and subsequent foreclosure crisis was a racialized process (Been, Ellen, & Madar, 2009; Fisher, 2009; Hyra, Squires, Renner, & Kirk, 2013; Wyly, Moos, & Hammel, 2012), there is an ongoing debate about which matters more in American life, race or class (Valant & Newark, 2016). The purpose of this study is threefold: (a) To contribute to the debate on the importance of race versus class in predicting life outcomes; (b) to take a step toward filling the gaps in our understanding of how the housing market crash, that was particularly severe for Black households, impacted Black student mobility in schools; and (c) to introduce the issue to public policy discourse.
The present article is part of a larger study examining the impact of the Great Recession in an area at the epicenter of the housing market bust (Mordechay, 2017a). This study applies a logit model to administrative data from a large urban school district in California’s San Bernardino County, a region on the suburban fringe of Los Angeles, in an effort to understand the extent to which The Great Recession affected the mobility of Black urban students. Geographically, the district is located in what was at one point considered one of the foreclosure and unemployment capitals of the nation (Pfeiffer, 2012). In addition to the region’s unique exposure to the housing crisis, prior to the crash, the suburban area’s housing market allowed for exceptionally high rates of homeownership among minorities, in particular, Black families (Molina, 2016). With this in mind, the analysis in this article is guided by the following three research questions:
This study applies an intersectional frame of analysis for understanding the ways in which class, race, and place factors comingle to produce inequalities in public education (Kozol, 1992). Although race continues to play an important role in educational inequalities in the United States, SES and other demographic variables are also critical factors in explaining disparities in educational achievement, particularly among Black urban students (Storer, Mienko, Chang, & Kang, 2012). This article asserts that class, race, and place are intricately bound to one another and that a singular focus on any of these factors is insufficient in explaining educational outcomes. This analysis is an attempt to move beyond the traditional lenses of race or SES as the primary drivers of educational inequalities, particularly on the experiences of Black students, because no other racial group has been subjected to the same degree of race, class, and spatial inequality (Massey & Denton, 1993; Mordechay & Orfield, 2017; Orfield, Frankenberg, & Siegel-Hawley, 2010). As a point of clarification, the term “urban” is used to refer to students, schools, and communities situated within the context of a large and dense metro area characterized by high concentrations of poverty, racial, ethnic, linguistic, and social class diversity (Milner & Lomotey, 2014).
This article begins with a brief discussion of the context of the study area and why San Bernardino County provides an appropriate case study for this analysis. This is followed by a review of the literature on the long-standing debate of the importance of race versus class in American society. Next, the racial and class dynamics of the housing boom in the years leading up to The Great Recession are discussed, followed by a review of previous research on the relationship between school mobility and a variety of educational outcomes. After describing the data and methodology, analyses and findings are presented. The article concludes with a discussion of key findings, possible explanations, suggestions for future research, and policy recommendations.
Context of Study Area
Located in Los Angeles’s eastern fringe, San Bernardino County provides an exemplary case study for examining the effects of the housing market crash on the school mobility of Black children for several notable reasons: The county led the country in population growth and economic development throughout the 1990s and 2000s (P. Johnson, Reed, & Hayes, 2008; Mordechay, 2014), generating an enormous housing boom to the regional economy before collapsing at the end of 2007. Racially, it is extraordinarily diverse by all measures and was at one point the fastest growing population for Blacks in the western United States and was the fastest growing Black suburban region nationwide (Tobar, 1990). Between 1980 and 2007, there was an estimated net migration of about 130,000 Black residents into the region, many of which were middle class, pursuing homeownership in suburban communities not far from the entertainment capital of Los Angeles. In fact, between 1970 and 2010, the region’s Black population increased by a factor of eight (Mordechay, 2014), with some scholars characterizing the heavy migration from Los Angeles to the Inland Empire as “Black flight” (J. H. Johnson & Roseman, 1990). In addition to the two decades of rapid population and economic growth in the Inland Empire, much of the migrating Black population was more affluent than the region’s once majority White population (Gabriel & Painter, 2003).
By 2008, the decades housing bubble had burst, plunging the region into economic peril. In 2009, the region was ranked as having the third highest rate of foreclosures of any metropolitan area in the nation, and the second highest unemployment rate (Mordechay, 2011). At its peak in 2010, San Bernardino County had an unemployment rate of 14% (Mordechay, 2011), and a foreclosure rate of over 5% (U.S. Census Bureau, 2014). In 2013, median household income across the Inland Empire County was just over $52,000, down 18% since 2007, the beginning of the housing crisis. For Black families, the decline was even more stark, decreasing by over 25% during the same time period (U.S. Census Bureau, 2014).
The present study uses internal administrative data from one of California’s largest urban high school districts located in San Bernardino County to examine what happened to Black students as a result of the housing market collapse, where since the beginning of the recession in 2007 to 2013, the percentage of students on the free and reduced lunch program more than doubled from 25% to 56%, an increase of 125% (CA Department of Education, 2015). This study has important implications for policy makers and educators for at least two reasons: (a) Little is known about how the foreclosure crisis played out for public school children, particularly for minority families that were especially vulnerable in housing market leading up to the boom (Mordechay, 2017b) and (b) disruptive school moves are linked to a multitude of children’s academic problems, such as grade retention, school completion, and a lack of interpersonal skills (Scanlon & Devine, 2001).
The results from this study indicate that after controlling for a variety of covariates including SES, Black students experienced substantially higher mobility rates during the beginning of the Great Recession (2008) compared with other racial groups. In addition, the probability of being mobile for middle-class 1 Blacks in the sample was significantly higher than for the poorest Blacks in the data. Together, the findings suggest a complex interaction between race and class for Black students as “middle class” Blacks in the sample were particularly vulnerable to school changes during the recession.
Existing Research
The Race or Class Debate
The tension between a race-based account of inequality versus the class-based perspective dates back to at least W. J. Wilson’s (1978) thesis in The Declining Significance of Race. W. J. Wilson (1978) theorized that when analyzing contemporary life chances among Blacks, the diverging experiences were no longer primarily determined by race as it had been historically. Rather, it was the effect of large-scale economic changes that grew increasingly more important in understanding Black occupational status and job placement. These shifts in economic and political structures, W. J. Wilson (1978) contended, had rendered class more significant than race.
More contemporary analyses have also suggested that income and class have begun to overshadow race when predicting opportunity in American society (Iceland, Sharpe, & Steinmetz, 2005; Logan, Stults, & Farley, 2004; St. John & Clymer, 2000). Some scholars suggest that racial differences controlling for class are decreasing while class differences controlling for race are increasing in America (Reardon, 2013). Much of the research suggesting that class now supersedes race as a predictor of life chances points to the 1980s as a turning point (Bailey & Dynarski, 2011; Reardon, 2013). For example, using nearly 70 years of data from the U.S. Census, Bailey and Dynarski (2011) find that the asymmetry between rich and poor students in college completion has grown by about 50% since the late 1980s. In addition, Reardon’s (2013) findings also suggest the growing importance of class by examining the standardized test score gap between the children of the poor families and the children from affluent families, suggesting that the “income achievement gap” grew by as much as 40% and is now more than 50% larger than the Black-White racial achievement gap, a stark reversal of the trend 50 years ago.
A consensus is emerging that views SES and the effects of poverty as important factors in explaining racial differences in educational achievement (Thames, Karimian, & Steiner, 2016; Brooks-Gunn & Duncan, 1997; Brooks-Gunn, Klebanov, & Liaw, 1995; Orfield & Lee, 2005).
Yet, other scholars suggest a historical link to racism that is still very much related to opportunity in American society, regardless of class or SES (Massey & Denton, 1993; Briggs, 2005; Oliver & Shapiro, 1995). Educationally, in several studies of the Black-White test score gap, the race effect retained statistical significance after covariates were incorporated, including SES (Fryer & Levitt, 2006; Kao & Thompson, 2003). Other studies documenting the unique experiences of middle-class Black families suggest a distinct race effect as they tend to live in neighborhoods with significantly lower median incomes, higher poverty rates, and a higher incidence of other related problems than do the middle class from other races (Adelman, 2004; Adelman, 2005; Alba, Logan, & Stults, 2000; Pattillo-McCoy, 1999). There is also evidence that equity losses during the Great Recession were particularly severe for minority households and that subprime lending during the housing boom was significantly more prevalent in minority neighborhoods (Gould Ellen & Dastrup, 2012).
Racial Dynamics of the Housing Boom and Bust
In the years leading up to The Great Recession, Latinos and African American homeowners, in particular, were significantly more likely than Whites to hold high-interest subprime loans, even after controlling income and credit scores (Faber, 2013). From the perspective of the borrower, the main feature distinguishing between subprime and prime loans is that subprime loans actively price the loan based on the risk associated with the borrower. In theory, the interest rate on the loan depends mostly on the income to debt ratio and credit scores (Keys, Mukherjee, Seru, & Vig, 2010). However, several scholars have documented that the racial disparities in subprime lending throughout the 1990s and early 2000s were largely a result of lending discrimination and that up to half of the subprime borrowers could have qualified for less costly prime mortgages (Barwick, 2010). In addition, a federal report from HUD (U.S. Department of Housing and Urban Development) found that subprime loans were more than 5 times greater in predominately African American neighborhoods than White ones (Fishbein & Bunce, 2000).
There is also evidence that the disproportionate concentration of subprime mortgages in minority-populated neighborhoods was facilitated by broader patterns of residential racial segregation, resulting in entire communities becoming vulnerable to predatory lending and the subsequent collapse of the housing market (Hyra et al., 2013; Rugh & Massey, 2010). Furthermore, there is also evidence that repossessed homes since the end of the recession have been overwhelmingly clustered in the same majority–minority neighborhoods (Ellen et al., 2015), leaving these same minority communities vulnerable to the prolonged impacts of unstable housing even after the recession. In addition to the racial disparities in the housing market, minorities were also hit hardest by employment layoffs when the recession began in 2007 (Hoynes, Miller, & Schaller, 2012) and were in a weaker position than White families to weather the financial burden of these layoffs. In California, for example, between 2007 and 2009, the Recession years, Blacks had higher underemployment rates compared with all other races irrespective of their educational attainment levels (Mordechay, 2011). Given Black families’ precarious position in the housing market leading up to The Great Recession and the group’s vulnerability to employment layoffs during the recession, there is reason to believe that Black children in schools might have been particularly susceptible to school changes. No study to date has specifically examined Black children from various social classes exposure to school changes during the largest economic downturn since the Great Depression in the 1930s.
The Harms of Changing Schools
A preponderance of empirical evidence supports the theoretical connection between school mobility and urban youths’ education outcomes. Research with nationally representative samples of youth has found that experiencing residential mobility decreases the likelihood that students will complete high school (Metzger, Fowler, Anderson, & Lindsay, 2015; Rumberger, 2003; Swanson & Schneider, 1999). Mobility poses significant problems not just for the students who are mobile but also the glut of negative spillover effects in their schools, including their nonmobile peers and teachers (Fleming, Harachi, Catalano, Haggerty, & Abbott, 2001; Gasper, DeLuca, & Estacion, 2010; Kerbow, 1996; Parke & Kanyongo, 2012; Rumberger, 2003). There is evidence, for example, that in the schools with many mobile students, nonmobile students show signs of lower academic achievement and higher dropout rates (South, Haynie, & Bose, 2007). For teachers, high levels of student mobility can disrupt orderly teaching and curriculum development, lower teacher morale, and increase administrative stress. These effects suggest serious externalities resulting from student mobility (Rumberger, 2003; Rumberger, Larson, Ream, & Palardy, 1999; Lash & Kirkpatrick, 1990). Students attending high-mobility schools are exposed to considerably less robust curriculums than those attending schools with infrequent mobility and these effects are possibly more harmful to low-income and minority students, potentially increasing racial and socioeconomic academic achievement gaps among students (Hanushek et al., 2004; Kerbow, 1996).
Although some types of school moves can have positive effects, the majority are associated with a range of negative outcomes including lower test scores in math and reading, grade retention, lower self-confidence, discipline issues in schools, dropping out, and even drug abuse later into adulthood, to name to just a few (Gasper, DeLuca, & Estacion, 2010; Parke & Kanyongo, 2012; Reynolds, Chen, & Herbers, 2009; Rumberger et al., 1999; Swanson & Schneider, 1999; Wood, Halfon, Scarlata, Newacheck, & Nessim, 1993). Despite the myriad of ways in which school changes could adversely impact students, there has been little research on whether or not the housing crisis of 2008 lead to increases in student mobility (see Mordechay, 2017b, for further discussion). Other scholars have documented the consequences of the housing crisis, namely, its effects on wealth (e.g., Pfeffer, Danziger, & Schoeni, 2013), physical and mental health (Houle, 2014), neighborhood crime (Ellen, Lacoe, & Sharygin, 2013), and community instability (Hall, Crowder, & Springer, 2015). Few have examined the recession’s impact on student mobility, particularly for Black families, who were the most vulnerable demographic group during the housing crisis (Bocian et al., 2010).
Although this work has been a critical contribution to our understanding of what happens to mobile school-aged children, the consequences of housing foreclosure and its racialized patterns, we still lack basic knowledge of how the housing crisis disrupted schooling for the most vulnerable students, namely, African Americans. It is likely that children, particularly those certain minority groups, may have had to change schools as a result of a foreclosure and move to a school whose quality differs from their original school. This analysis will fill this gap in the literature regarding Black children’s schooling, drawing on evidence from a large urban school district in San Bernardino County. Information about which areas, and which demographic groups have been most affected by the housing crisis is crucial to the development of policy responses that seek to stabilize vulnerable communities and mediating the individual and community-level effects in the event of future housing crises.
Data and Method
Sample
This article draws from administrative data from a large high school district (Grades 9-12) located in San Bernardino County, California. The district includes a total of 18 schools, with student enrollment across each school varying between 500 and 3,400, with substantial variability by race/ethnicity and SES (see Tables 1 and 2). Data are available at the student and school level, and incorporates a variety of covariates (e.g., race/ethnicity, grade point average [GPA], parent education). The descriptive analysis is based on 7 years of available data (2006 to 2013), where each year of the data includes between 23,000 and 25,000 observations (students). What follows the descriptive analysis is a more robust analysis using logistic regression to examine variability in student mobility in 2008 (n = 22,949), the year in which the descriptive analysis showed the steepest increase in student mobility. To determine whether school mobility in 2008 was associated with race by different levels of SES, I include an interaction term in the model, followed by plotting the predicted probability of students being mobile by race and SES, holding all other variables in the model constant. Addressing the confounding of race and SES is critical since much of the social science research on social and economic disparities has focused on either SES or race, often controlling for one factor when testing the effect of the other (Chen, Martin, & Matthews, 2006; Williams & Collins, 1995). However, as SES and race are closely intertwined, with members of many minority groups, on average, being lower in SES, some scholars have suggested that researchers ought to be testing for interaction effects between race and SES (Lacy, 2007; Pettigrew, 1981; Williams, 1999).
Student Demographics Overtime.
Student SES Overtime.
Note. SES = socioeconomic status; LSES = low socioeconomic status; LMSES = low-mid socioeconomic status; HMSES = high-mid socioeconomic status; HSES = high socioeconomic status.
Furthermore, the total number of students included in the sample between 2006 and 2013 is approximately 182,000. Table 1 below presents the demographic distribution of students, including race and a proxy for students’ socioeconomic background: the highest level of education either of their parents has attained. Although this is an imperfect method for characterizing SES, other researchers have used parent education as a proxy for SES (Ananat, Gassman-Pines, & Gibson-Davis, 2011; Stanfiel, 1973). In addition, past studies have suggested family income to be highly correlated with parents’ educational attainment and that education is the most stable predictor of SES (Harwell & LeBeau, 2010; Stanfiel, 1973; Williams & Collins, 1995). Finally, although SES can be measured in several ways (e.g., educational attainment, occupation, income), with each having different implications, given that minorities do not receive the same financial gains for equivalent years of education as do Whites (V. Wilson & Rogers, 2016), education may be a better indicator of distribution of SES across racial groups than income.
With this in mind, throughout the analysis, I will be referring to students from families whose parents have less than a high school diploma as the LSES (low SES) group (n = 38,243). For those students whose parents have a high school diploma, I will be referring to as LMSES (low-mid SES) group (n = 39,889). For those whose parents have gone to college but have no degree, I will be referring to as HMSES (high-mid SES) group (n = 52,479). Finally, for those students whose parents have college degrees, I will be referring to as the HSES (high SES) group (n = 51,545). Descriptive data for the sample are shown in Tables 1 and 2.
Outcome Variable
The outcome variable of interest to this study is school mobility at the student level in 2008. The year 2008 was selected as the outcome year because the descriptive analysis showed the starkest increase in student mobility during that year (see Table 3). Student mobility is treated as a binary variable and was coded as follows: 0 = did not move in the 2008 school year, 1 = moved at least once in the 2008 school year. A student is considered mobile if they meet at least one of the following four criteria: (a) they left school (truant), (b) moved to a verified public school in California, (c) moved to a verified U.S. school (outside California), and (d) moved to a foreign county.
Student Mobility Over Time.
Predictor Variables
The primary independent variables of interest are year, race/ethnicity, SES status, English Language Learner (ELL) status and Special Education status, and grade (9-12). The measures were created using student-level demographic data. Student race/ethnicity included 19 classifications that were collapsed into five classifications for the final analysis, which were coded categorically as follows: Latino = 1; White = 2; Black = 3; Asian = 4 (reference group); and Other = 5. ELL statuses included nine classifications and were collapsed into two. Special Education status included six classifications that were also collapsed into two. ELL status and Special Education status variables were each dummy coded, with non-ELL and non-Special Education students used as a reference group. Parents’ education level (a proxy for SES status) was equivalent to either the father’s or the mother’s highest level of education. Parents’ level of education was composed of six distinct levels of educational attainment. The six levels were collapsed into four categories of SES for the final analysis which were coded as follows: those with less than a high school education (employed as a reference group) = 1; high school graduates = 2; some college = 3; and college graduates = 4.
Data Analysis Method
The empirical approach to this analysis involves a logistic regression with cluster-robust standard errors. Logistic regression is a variant of regression. Similar to multiple regression, logistic models can be used in studies with continuous or categorical independent variables. Regression is used when the dependent variable is continuous, whereas a logistic regression is fitting for a dependent variable that is dichotomous in nature (e.g., mobile vs. nonmobile). Different from linear regression, logistic regression requires the data neither to be normally distributed nor have linear relationships between the independent and the dependent variables (Hair, Black, Babin, Anderson, & Tatham, 1998).
Clustered robust standard errors are used because the traditional standard error estimates for logistic regression models based on maximum likelihood from independent observations is not accurate for data sets with a clustered structure (students nested within schools), as observations in the same clusters tend to have similar characteristics and are highly correlated with one another other. Therefore, robust standard error estimates are needed to take into account the intra-cluster correlation where the assumption of independence is violated (Huber, 1967). In addition to Huber (1967), Cameron, Gelbach, and Miller (2008) have developed a practitioner’s guide that describes the procedures for getting accurate statistical inference, a fundamental component of which is obtaining accurate standard errors with cluster corrections. Clustering data structure is common in educational research, and adjusting for the clustering is therefore appropriate for this analysis given that the data nested within groups.
Furthermore, the logistic model for this analysis took the following form:
where
P is the predicated probability that the outcome is present;
Y is a binary dependent variable;
X1 through Xp are distinct independent variables; and
For the logistic regression model, the coefficients are exponentiated β (= eβ) to be interpreted as odds ratios with corresponding 95% confidence intervals. The odds ratio is the ratio of the probability that the event of interest occurs (mobility) to the probability that it does not (non-mobility). It is a more intuitive measure of effect size and is commonly used in the social science and medical research (Bland & Altman, 2000). In addition, the logistic model tested for interaction effects for all student race/ethnicity and SES variables to address the second research question. All statistically significant interactions were reported. Because interpretational difficulties can be challenging in nonlinear models (logistic regression) in the case of interaction effects (Peng, So, Stage, & St. John, 2014), I computed the change in predicted probabilities of being mobile for each race group by different SES classifications.
I began the following section by employing descriptive statistics to describe the basic features of both the outcome and predictor variables. This is followed by a more robust analysis of the 2008 “peak effect” (see “Descriptive Analysis” section for more information). The outcome variable of interest is Student Mobility. The predictor variables of interest are (a) race, (b) SES, (c) language status, (d) special education status, (e) GPA, (f) grade (9-12), and (g) year (2006-2013).
Descriptive Analysis
The descriptive analyses reveal that overall, student mobility rates increased substantially between 2007 and 2008, the beginning of The Great Recession. The increase from 3.6% to 6.0% between 2007 and 2008 was a statistically significant difference (p < .001). After an initial increase in student mobility rates in 2008, the rate begins to decrease in 2009 through 2013 (see Table 3). With this in mind, I will be referring to 2008 as the “peak effect” of the recession as defined by student mobility changes in the descriptive analysis.
The descriptive analysis for each student subgroup reveals a similar trend. Namely that in 2008, each of the different racial and socioeconomic subgroups experienced an uptick in mobility rates between 2007 and 2008 (Figures 1 and 2). Although each of the racial and ethnic subgroups displayed an increase between 2007 and 2008, the increases varied widely by student subgroup. Black students appear to have experienced the most substantial increase in mobility among all the race groups, rising from 3.8% in 2007 to 7.9% in 2008. This was followed by Latinos, which saw an increase during the same time period from 4.1% to 6.2%. The lowest increases were among Whites and Asians (Figure 1).
Student mobility by race, 2006-2013.
Student mobility by SES, 2006-2013.
Perhaps not surprisingly, for the different SES groups, between 2007 and 2008, the lowest SES groups (LSES) saw the steepest increases from all SES groups, a rise from 4.3% to 7.4%. Conversely, the highest SES group (HSES 4) saw the least severe change during this time, increasing from 2.9% in 2007 to 4.7% in 2008. Substantial increases were also observed with the samples Special Education population, rising from 3.3% in 2007 to 10.3% in 2008. The ELL subgroup saw an increase from 5.9% to 10.4% during the same time period.
Although each of the subgroups discussed above showed a similar pattern between 2007 and 2008, the peak-effect, the magnitude of the increase varied quite dramatically, with the highest increases found among traditionally marginalized subgroups. The unequal distributions further highlight the highly racialized dynamics of the Great Recession (2007-2009) that left whole communities vulnerable to the residential instability that accompanied the collapse of the housing market. To further test these effects, I employed a more robust model with controls that could help explain the extent of these patterns where a number of compelling findings emerged.
Results of Adjusted Logistic Regression Model
Table 4 displays the results from the logistic model. After controlling for various student background characteristics, including SES, race/ethnicity, ELL status and Special Education status, grade (9-12), and academic achievement (GPA), the following evidence emerges: Namely that compared with Latinos (the reference group), Black students were most likely to have moved in 2008, controlling for all variables in the model. Each of the racial and ethnic groups was more likely than Latinos to be mobile in 2008, with Blacks being the only group that was statistical significant (p < .001). The odds ratios and their corresponding 95% confidence intervals from the logistic model are reported in the model below (Table 4). Black students were 1.54 times as likely to move schools in 2008 than Latino students. Perhaps counter-intuitively, the model could not distinguish any statistically significant differences between the four SES groups. In fact, students from the highest SES group were slightly more likely to move in 2008 than other students although the differences were not statistically meaningful. At least one possible explanation is that higher income families were more likely to experience residential instability from a house loss whereas lower SES families were more likely to be part of the rental housing market.
Logistic Regression With Cluster Robust Standard Errors for Variables Predicting School Mobility in 2008 (n = 22,949), Controlling for Sociodemographic and Academic Variables.
Note. Latino, low SES, and Grade 12 are the reference categories. Standard error is of OR and is adjusted for 15 clusters of schools. OR = odds ratio; CI = confidence interval; SES = socioeconomic status; GPA = grade point average.
p < .05. **p < .01. ***p < .001.
After testing for main effects, a race by SES interaction term was entered into the model with mobility in 2008 as the outcome variable to test for ethnic group differences in mobility across the different levels of SES, finding the overall interaction to be statistically significant at p < .001, with a chi-square statistic of 87.69 (df = 28). Model coefficients were then used to generate predicted probabilities associated with each category of SES and race. This allowed us to see the probability of being mobile in 2008 for each category of SES and race, holding all other measures in the model constant. Figure 3 presents these data in a plot indicating the predicted probability.
Race and SES interaction.
The sample of Black students had a similar SES distribution as the White students, yet the group’s mobility rates were substantially higher in 2008, the middle of the Great Recession (see Table 1). This is a surprising finding considering the research indicating the predictive power of SES on a plethora of social and educational outcomes (Duncan & Magnuson, 2005; Hackman & Farah, 2009). The observed pattern with the Black student population could at least in part be explained by what some scholars have suggested, namely that middle-class status for Blacks is often more turbulent than that of other races (Conley, 1999; Pattillo-McCoy, 2008).
Next, significant interactions were explored using simple-effects analysis for each race group at different SES levels to test if they are statistically different from one another. The findings indicate that for Latino, Asian, and White students in the sample, the probability of being mobile is stable for each category of SES, as the interaction between race and SES did not enter as significant with respects to these groups. However, the data for the Black student population does reveal a different pattern. The interaction between race and SES entered as significant with respect to the Blacks at the mid-low SES (p < .001) level and the high SES (p < .01) level. Some possible explanations are discussed in the following section.
Discussion
The U.S. housing collapse of 2007-2009 was one of the most profound residential disasters of the past century, pushing millions of American households into foreclosure and many more into financial calamity. Yet, the burden of the crisis was not evenly distributed, with minority households much more likely than White households to experience either a job loss or foreclosure. With these racial disparities in mind, this article aimed to provide an understanding of the relationship between the Great Recession and student mobility for Black children in a school district located in one of the most impacted areas. This study extended existing research on the most recent housing crisis (2007-2009) on the mobility of Black students by focusing on an area both at the epicenter of the housing crash, and of “middle class” Black homeownership by assessing the relationship between race and class for Black students and their exposure to school mobility. Although other studies have examined the effects of the economic downturns on households (Ananat et al., 2011; Goldin, 1999) and students in schools (Mordechay, 2017a), few studies to date have examined the effects of The Great Recession on school changes, and none have examined the experiences of Black children from varying social class backgrounds.
The results of the descriptive analysis indicate that, in 2008, mobility rates spiked for all subgroups, with the starkest increase seen among Black students. Upon closer inspection of the “peak effect” in 2008, the logit model indicated that Black students in the sample were much more likely to be mobility than any other racial group (p < .001), controlling for SES and a variety of other covariates, suggesting that for Blacks, race was a stronger predictor of mobility than SES. In addition, the interaction effects of race and SES entered as significant with respect to the Blacks at the low-mid SES (p < .001) level and Black students at the high SES (p < .01) level. Although this study is unable to describe the causal mechanisms whereby housing crises impacted student mobility for this sample, the most striking finding from this study was the particularly stark rates of mobility for Black students from low-mid SES and high SES levels.
The pattern observed for Black students in 2008 might be explained by the group’s precarious position in the region’s housing market. For example, Black neighborhoods in the suburban region, where the school district is located, had a higher rate of foreclosures than non-Black areas in the same region and it is likely that Black families were relying more heavily on subprime loans, particularly those either on the periphery of the middle class or even “solidly” middle class (Molina, 2016). This is a trend that has been found in other metropolitan areas with high rates of homeownership among Blacks, even after controlling for major indicators such as income, occupation, and credit (Calem, Hershaff, & Wachter, 2004; Meyer & Pence, 2008). If so, this phenomenon of a “racialized home buying market,” where prospective and present Black homeowners are particularly vulnerable, could explain the counter-intuitive interaction effects found in the model.
There are a number of other possible explanations for the evident trends observed in 2008 for low middle and high SES Black students. For decades, scholars have maintained that Blacks who successfully enter the ranks of the so-called middle class face a number of unique challenges to maintaining their so-called middle-classness (Isaacs, 2007; Pattillo-Mccoy, 2005). Newly middle-class Black families are particularly susceptible to backsliding to their low-income origins, in part, because they trail other races on many of the other indicators of what sociologists have suggested constitute the “middle class,” such as income, wealth, homeownership, and educational attainment measures (Lopez-Calva & Ortiz-Juarez, 2014). Therefore, since considerable numbers of Black families are first- or second-generation middle class (Pettigrew, 1981; Wheary, 2006), many of these families have less of a buffer to absorb an exogenous shock to the household (e.g., job loss or foreclosure). In fact, one study suggests that Black adults are 4 times more likely than their middle-class White counterparts to have grown up in poverty (Heflin & Pattillo, 2006). Therefore, it is likely that these unique challenges make Blacks particularly vulnerable to financial shocks either on the individual or household level.
Implications for Policy and Practice
Public policy can play a critical role in stabilizing vulnerable neighborhoods and mediating the individual- and community-level effects in the event of future economic crises. We need to ensure that mortgage and lending policies and fair housing policies are enforced and strengthened, especially to protect minorities who are especially vulnerable in the housing market. The housing market and lending discrimination clearly contribute to growing racial disparities in the Recession and its aftermath. To more effectively enforce fair housing laws already in place, the federal policy should support more paired-testing programs nationwide. Paired-testing investigations, also known as auditing, happens when equally qualified auditors from different race and ethnic groups posing as homebuyers or renters approach housing providers, such as real estate and rental agents, and mortgage lenders, to inquire about the availability of the same or similar housing units or other housing-related services (Fix & Struyk, 1993). Such investigations with greater regularity might reduce incidents of discrimination and therefore might abate vulnerable subgroup’s precarious position in the housing market. In the effort to improve educational outcomes for urban students, ensuring housing stability for racial and ethnic minorities is essential.
At the school level, there is evidence to suggest that schools can make a difference by facilitating mobile student’s involvement in civic engagement activities such as leadership in school clubs or groups and involvement in afterschool programs (Voight et al., 2017). Intentionally focusing these opportunities at students known to be experiencing residential or school instability may build their resilience and buffer against the negative effects of residential or school mobility. In addition, schools with high levels of mobility could effectively extend counseling services to students and families (Nakagawa, Stafford, Fisher, & Matthews, 2002).
Strengths, Limitations, and Future Research Directions
As with all research, this study has both strengths and limitations. First, the data used for the project are noteworthy because it is relatively large with substantial subgroup variation, making it possible to identify shared and different patterns that cut across cases. In addition, conceptually, it is important to flush out the confluence of SES and race/ethnicity, each of which is a social classification that is highly correlated with many educational and life outcomes. Because these categories tend to overlap, many studies inadvertently confound the influence of one or more of the categories with each other (LaVeist, 2005). Although not without limitations, this research design makes a deliberate effort to disentangle race/ethnicity from socioeconomic status.
Also, although these data are well suited for disentangling changes in school outcome from family characteristics, they do not control for all potential confounding variables. In spite of this, this article does control for what social scientists have suggested are highly influential variables predictive of a variety of social and economic outcomes (Chen & Paterson, 2006; Duncan & Magnuson, 2005). Moreover, any generalizing from these findings to other regions and economic recessions at different times should be done with caution. There should be no presumption that the trends and effects found in this study hold for other metros, states, or countries. San Bernardino is a unique region that leading up to the 2007-2009 housing crisis was one of the fastest growing metropolitan areas in the county with a spectacular housing boom and subsequent bust. In addition, race effects can be highly locally dependent (McCall, 2001). Finally, the present study included only one measure of SES, parental education. Although this indicator is quite stable over time, it does not tell us about the impact of other SES characteristics such as wealth and occupational status to name just a few.
With this in mind, future research might also consider the experiences of minority children in other metro areas who are particularly vulnerable to the housing market collapse. For example, in the Miami, Atlanta, and New York City housing markets, foreclosures during the recession were disproportionately found in Black and Latino neighborhoods, even after controlling for income (Ellen et al., 2013).
Footnotes
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) received no financial support for the research, authorship, and/or publication of this article.