Advertisement

Journal of Economics, Race, and Policy

, Volume 1, Issue 2–3, pp 126–141 | Cite as

Racial Differentials in the Wealth Effects of the Financial Crisis and Great Recession

  • Ryan Compton
  • Daniel Giedeman
  • Leslie Muller
Original Article
  • 252 Downloads

Abstract

The financial crisis of 2007–2009 was arguably the most severe financial crisis in American history and the subsequent Great Recession was the worst economic downturn in the USA since the Great Depression. In this paper, we analyze data from the Panel Survey of Income Dynamics (PSID) to examine the effects of the crisis and recession on the wealth of White and Black families using graphical, cross section, and panel empirical models. While other studies have measured the short-term effects of the crisis and recession on American household wealth, we are able to look at longer-term wealth effects by incorporating data from the recently released 2015 wave of the PSID. Our results indicate that the negative consequences of the economic downturn on Black families’ wealth were severe and longer-lasting than for White families.

Keywords

Great recession Inequality Race Wealth 

JEL Classifications

D31 I31 J15 

Introduction

The financial crisis of 2007–09 and subsequent Great Recession have spawned a large body of work on the causes and consequences of the crisis and associated economic downturn.1 Because more than a decade has passed since the beginning of the financial crisis, we are now able to investigate longer-term effects of the crisis and Great Recession on wealth in the USA. Particularly, we focus on how the economic downturn may have had a differential outcome on Black and White families’ assets and wealth accumulation. Our analysis comes at a time when economic inequality is at the top of the economic research agenda, and so understanding the role the crisis and recession had on recent Black and White wealth accumulation will prove useful as researchers continue to study the sources of racial inequality in America. In order to examine this topic, we use nine waves (1999–2015) of family-level data from the Panel Study of Income Dynamics (PSID), a bi-annual longitudinal survey fielded by the University of Michigan. With these data, we employ graphical and econometric analysis to investigate the experience of Black and White families in terms of wealth and asset holdings over the period.

Financial crisis aside the reasons behind the Black/White wealth gap is an area of significant study (c.f. Killewald 2013; Maroto 2016; Killewald et al. 2017).2 Possible explanations for the wealth gap include, among others, the fact that compared to Black households, White households tend to have higher income, education, and employment (Barsky et al. 2002; Bricker et al. 2014; Campbell and Kaufman 2006), hold more investments and have greater household equity (DeVaney et al. 2007), receive larger private gifts and inheritances (Gittleman and Wolff 2004; McKernan et al. 2014a), have higher financial literacy (Lusardi 2008) and trust in the financial sector (Choudhury 2002), and are less likely to be single parent households or provide financial assistance to extended family and friends (Chiteji and Hamilton 2002; O’Brien 2012).3 We consider our work on the financial crisis as contributing to this larger discussion on racial differences.

The literature on the impact of the financial crisis and Great Recession on wealth, asset accumulation, and debt is large and draws in researchers from across multiple disciplines, including sociologists, demographers, and economists. While we save the discussion of this literature for the next section, at this point, we highlight that the financial crisis and Great Recession negatively affected the wealth of American households, with these effects generally being found to have been relatively larger for Black households than White ones (Pfeffer et al. 2013; McKernan et al. 2014c). Our study adds to the existing literature on several margins. First, while many previous studies have looked at the immediate aftermath of the financial crisis on wealth, our study takes a longer view, using data from 1999 through the 2015 wave of the PSID, which is the longest post-recession sample in the literature to our knowledge and allows us to further track the economy’s (and wealth’s) recovery beyond earlier studies. Further, while many of the studies in this literature rely on a pre-post descriptive methodology (this descriptive method looks at changes in wealth over the course of the recession or more commonly compares wealth differences before and after the downturn) for their analysis, we employ various econometric techniques such as logistic and fixed effects models to empirically examine the effect of the crisis and downturn on wealth. The use of these different techniques together will allow us to see if there is a consistent story emerging across the techniques. Finally, we not only consider total wealth, but also how different asset components were affected.

Previewing our results, we find evidence that the wealth of Black families has diverged from that of White families in the wake of the financial crisis and recession. This is partially because average Black family wealth had still not recovered to its pre-crisis levels even as late as 2015, and partially because Black families have not benefitted from more recent gains in various financial assets, such as stocks, nearly as much as White families have benefitted. Additionally, Black families are significantly more likely than White families to have experienced long-term wealth decreases as a result of the crisis and economic downturn.

The rest of the paper proceeds as follows. Section “Related Literature” provides a discussion of the related literature on the effects of the financial crisis and Great Recession, while section “Data and Graphical Evidence” outlines our data and provides graphical evidence of differences in Black and White wealth and asset accumulation over time. Section “Regression Models and Results” presents our econometric models and results, while section “Conclusion” concludes.

Related Literature

The financial crisis and resulting recession was obviously a generational economic event and has resulted in significant study. Researchers have investigated the effect of the recession on a wide range of outcomes including, among others, mothers’ health (Currie et al. 2015), mental health (Houle 2014), political participation (Huyser et al. 2017) as well as wealth and income in a range of countries such as Germany (Grabka 2015), Italy (D’Ambrosio and Rohde 2014), Spain (Amuedo-Dorantes and Borra 2017), and, of course, the USA.

Not surprisingly, the literature on the Great Recession and its wealth effects in the USA is considerable. In terms of the major papers in this area, much of the early work relied on the pre-post descriptive approach discussed earlier. Examples of these include works by Bricker et al. (2011), Wolff (2012), Bosworth (2012), Choi (2013), and Shapiro et al. (2013). The general finding from these studies is that wealth declined sharply for most American households over the course of the downturn, with African American and Hispanic households experiencing the largest relative declines. This finding is in line with Oliver and Shapiro (2006) which suggested that recessions and economic restructuring have a greater negative impact on Black earnings and assets than on White earnings and assets.

More recent work has moved towards more formal econometric modeling in order to examine the recession’s effect on wealth. For example, Pfeffer et al. (2013) use the 2007 through 2011 waves of the PSID and a logistic model to estimate the likelihood of wealth loss and the magnitude of wealth loss over the period. They find that while all socioeconomic groups experienced losses as a result of the recession, Whites and Asians were less likely to have lost significant wealth than African American or Hispanic households, and White/Asian households lost roughly 14% less than non-White households, holding all else equal (Pfeffer et al. 2013).

McKernan et al. (2014b, 2014c) use the Survey of Consumer Finance (SCF) data and synthetic cohorts to determine the effect of the Great Recession on wealth relative to the counterfactual of what wealth and its associated trajectory would have looked like in the absence of the recession. This is analyzed for different generations as well as racial groups, and the authors find that the Great Recession reduced wealth by almost 30% for American households, with African American families losing a greater amount of their wealth than Hispanic and White families. Further, all major wealth components were found to be harmed by the recession (McKernan et al. 2014b).

Amuedo-Dorantes and Pozo (2015) examine how the downturn affected the assets, wealth, and retirement plans of older households in the USA, using the 2006 and 2010 Health and Retirement Study (HRS), with a specific focus on how the effect differed depending on whether the household head and spouse were American born or immigrants. The authors find that the wealth for non-immigrant households fell roughly 26% over 2006–2010, with immigrant households falling 24% and mixed households falling 38%. Interestingly, while Hispanic households experienced larger declines relative to White households, Black households were found to experience smaller losses. This result for Black households is counter to what most studies find in the literature, although one must keep in mind that this is a very specific sample as it examines only older households (55 years and over). The HRS also includes pension wealth in their total wealth measure, while the PSID and the SCF—the two most commonly used datasets for wealth measurement—do not.

More recently, Maroto (2016) uses the SCF to examine wealth differences across the distribution by race in the USA over the 1998–2013 period. She uses descriptive techniques as well as unconditional quantile regression and the Blinder-Oaxaca decomposition to examine the sources of the racial wealth gap. While this paper is more interested in the general reasons for racial differences in wealth and does not attempt to account for the Great Recession, the author points out that her data indicate that while non-Hispanic White families experienced larger overall declines in net worth (both mean and median) following 2008, Hispanic and non-Hispanic Black households experienced larger relative declines. Further, on the general question of racial differences, Maroto finds that non-Hispanic Black and Hispanic households tend to have lower net worth than non-Hispanic White families across the distribution, even after controlling for covariates. Further, the decomposition shows that for high wealth households, demographics, income, and education accounted for more of the wealth gap, while access to credit markets and homeownership mattered more for low wealth households (Maroto 2016).

Gorbachev et al. (2017) use data from the PSID and American Housing Survey to investigate the housing boom and subsequent housing bust associated with the financial crisis and Great Recession. The authors document extreme increases in Hispanic wealth relative to other ethnic groups during the boom, associated with predominately living in metropolitan areas, and subsequent wealth loss (relative to other groups) which the authors tie to a contraction of credit, especially for undocumented immigrants. The authors also document wealth losses for African Americans relative to White Americans over 2007–2011 due to the housing market collapse.

Lastly, Zhang and Feng (2017) use SCF data from 1989 to 2013 and employ descriptive as well as basic OLS and decomposition techniques to examine the wealth recovery by race post-Great Recession. Their findings suggest that Black, Hispanic, and lower-educated families have not seen their wealth recover at the rates other families have due to their heavy concentration of wealth in real estate, which has not recovered at the rate other asset classes have.

In sum, when looking at the literature, it is clear that the Great Recession resulted in large losses of wealth for most American families, with Black and Hispanic households generally bearing greater relative losses than White households. Whether descriptive analysis or formal econometric modeling is used, the results confirm that the Great Recession negatively affected Black and Hispanics more than Whites.

Our paper fits into, and adds to, this literature on a number of margins. First, like Pfeffer et al. (2013) and the recently published Zhang and Feng (2017), we take aspects of both the early and more recent literature by combining some basic pre-post descriptive analysis to provide a sense of how wealth and related assets have changed over time, especially in the periods during and since the downturn. We then also turn to formal modeling techniques in the form of logistic and fixed effects analysis to empirically investigate the long-term effects of the crisis and recession. Second, while many of the papers have considered the immediate effects of the recession with data that ends in the few years following the downturn, our use of the most recent 2015 wave of the PSID allows us to track the downturn’s effect more than half a decade later and thus be able to consider how long lasting the effects of the financial crisis and recession have been. The addition of the 2015 wave is especially important given that 2013 is the most recent wave (of both the PSID and SCF) that other studies use and the economy was in a much different place in its recovery by 2015. For example, by 2015, the economy was almost fully recovered, with a 2015 annual unemployment rate of 5.3% as compared to 7.4% in 2013 (BLS 2018). Third, our use of fixed effects models as well as logistic models allows us to use approaches seen in the literature together and determine whether a consistent story is seen across the approaches.

Data and Graphical Evidence

Data Description

We use data from the 1999–2015 waves of the Panel Study of Income Dynamics (PSID), a nationally representative longitudinal survey of over 9000 families (Panel Study of Income Dynamics 2017). Fielded by the University of Michigan, it is the longest running longitudinal survey in the world. It contains information from 1968 to the present and has followed original family members and their children’s families throughout time. Because of its longevity and richness in wealth data, the PSID has been used in numerous longitudinal studies on wealth (see, e.g., Gorbachev et al. 2017; Pfeffer et al. 2013). The first wealth module was fielded in 1984, and every 5 years thereafter, until 1999. After 1999, the core survey switched from annual to bi-annual and also included the wealth module in each wave. Hence, to cover the most years before the recession and still maintain a constant number of years between waves, we begin our analysis in 1999. Hence, our data cover 9 waves: 1999, 2001, 2003, 2005, 2007, 2009, 2011, 2013, and 2015. Our final sample consists of 28,851 observations of heads of household over this time frame, which equates to 3246 families in each wave.

Since wealth is typically reported as a family variable, our unit of observation is the head of household for a family. We use several criteria in choosing our sample for analysis. We first include only those families that remain in the sample for the full time period, to allow for consistency in the length of the pre and post crisis periods. Because our focus is specifically on Black and White families, we eliminate families that reported any other race for their head in any of the relevant years.4 Furthermore, because Hispanics as a group had much more extreme swings in housing wealth during this time period than did Black and White families (Gorbachev et al. 2017), we keep only Black and White respondents with non-Hispanic ethnicity. Finally, because our primary focus is to compare Black and White wealth over time, we further restrict the sample to only those families who had either a Black head over the full period or a White head throughout the panel.5

We take the net wealth variable directly from the PSID. It was created by adding together the following asset categories (as defined in the PSID), minus debt:
  • Annuities: annuities and holdings in Individual Retirement Accounts

  • Checking and savings: checking and savings accounts, money market funds, certificates of deposit, government savings bonds, and treasury bills

  • Farm and/or business: value of the farm and/or business

  • Home equity: value of the home minus the remaining mortgage

  • Other assets: bond funds, cash value in life insurance policies, valuable collections for investment purposes, and rights in a trust or estate

  • Other real estate: value of second home, land, rental real estate, and land contracts

  • Stocks: shares of stock in publicly held corporations, mutual funds, and investment trusts

Debt consists of money owed aside from the primary home mortgage. This includes debt from credit cards, student loans, medical or legal bills, money owed on farm, business or other real estate, and personal loans.

Summary Statistics

We show sample means for the full panel in Table 1. Our unit of analysis is the head of household; however, the measures for income and number of children are at the family level. Because the PSID originally oversampled low-income families, of which many were Black, our sample has a larger proportion of Black families (35%) than the US population.
Table 1

Summary statistics (mean values)

Variable

Full sample

Black

White

Age

50 (14)

48 (12)

52 (14)

Married

55% (50)

33% (47)

66% (47)

Family income

85,750 (115,694)

50,586 (44,204)

104,360 (135,773)

Number of children

0.74 (1.1)

0.94 (1.2)

0.64 (1.0)

Male

68% (46)

47% (50)

80% (40)

No high school degree

11% (31)

18% (38)

7% (25)

High school degree

35% (48)

43% (49)

32% (46)

Some college

25% (43)

26% (44)

24% (43)

College degree

29% (46)

13% (34)

38% (49)

Own a home

72% (45)

82% (39)

52% (50)

Black

35% (48)

White

65% (48)

Observations

28,630

9908

18,722

Source: means calculated from the 1999–2015 waves of the PSID

Standard deviations are in parenthesis

Family income is reported in 2015 dollars

Table 1 shows that in our sample, White families have higher family income, higher marriage rates, and a larger percentage of heads of household with college degrees. A larger percentage of women are heads of households in Black families than in White families. Furthermore, 82% of White families are home owners, versus 52% of Black families.

Wealth Measures 1999–2015: Long-term Effects of the Great Recession

Given the abundance of research showing the absolute differences between Black and White wealth, it is no surprise that in our data, total family net wealth in each year of our analysis shows a large gap between Blacks and Whites (Fig. 1). All values are in real 2015 dollars.
Fig. 1

Source: mean values calculated from the 1999–2015 waves of the PSID. Total wealth = (annuities + checking and savings + farm/business equity +. other assets + primary home equity + equity in other real estate + stocks) − debt. Values are in 2015 dollars

In 1999, the total family wealth of Whites was 6.3 times that of Black families (Table 2). This gap declines monotonically until we hit a low in 2005, where White wealth was 4.6 times that of Blacks. From 2007 on, the gap continued to grow in a linear fashion6 until 2015, when the White/Black wealth ratio reached its largest, at 7.7 to 1. One thing is clear: the gap in wealth between white and Black families widened after the Great Recession, and not just in the immediate after-effects. Six years after the official end to the recession, the gap between White and Black wealth was at its largest since 1999.7 Figure 2 shows the relative pattern of movement for overall wealth for both groups over time.
Table 2

Net family wealth

Year

Black

White

Ratio

1999

59,945 (336,121)

374,866 (1,503,990)

6.3

2001

68,875 (226,020)

396,331 (1,803,450)

5.8

2003

89,331 (498,084)

408,694 (1,409,971)

4.6

2005

108,952 (475,556)

495,901 (1,796,074)

4.6

2007

102,104 (275,450)

612,780 (2,236,774)

6.0

2009

83,254 (263,241)

568,180 (2940,568)

6.8

2011

67,385 (167,754)

513,057 (1,578,484)

7.6

2013

71,408 (203,076)

514,873 (1,504,953)

7.2

2015

78,840 (211,536)

604,975 (1,936,948)

7.7

Source: means calculated from the 1999–2015 waves of the PSID

Standard deviations are in parenthesis

Total wealth = (annuities + checking and savings + farm/business + other assets + primary home equity + other real estate + stocks) − debt

Values are in 2015 dollars

Fig. 2

Source: means calculated from the 1999–2015 waves of the PSID. Total wealth = (annuities + checking and savings + farm/business equity +. other assets + primary home equity + equity in other real estate + stocks) − debt. Values are in 2015 dollars

Breaking up total wealth into its different components can shed light on some of the drivers of this gap. Because Blacks have historically held a larger share of their wealth in housing than have Whites,8 examining the value of primary home equity is a good place to start.9 Table 3 shows the percentage of total wealth held in home equity. In every year of our panel, Blacks held a higher proportion of primary home equity in their wealth portfolios than did Whites. This proportion generally increased as time went by, from 35% in 1999 to 58% in 2011 and back down to 50% in 2015. White families had 24% of wealth in home equity in 1999, with this proportion climbing to 28% in 2011 and settling back at 25% in 2015. Why this large increase in the percentage of equity to total wealth after the recession? By examining the value of financial assets, the value of stocks and annuities particularly, we see that it was not that the value of housing equity for Blacks increased substantially, but rather the value of their financial assets decreased, thereby increasing their proportion of housing equity to total wealth.
Table 3

Home equity

Year

Black

White

Ratio

Black % of total wealth

White % of total wealth

1999

21,113 (41,291)

89,635 (145,168)

4.2

35

24

2001

26,365 (48,329)

107,209 (172,258)

4.1

38

27

2003

28,937 (51,149)

127,233 (197,331)

4.4

32

31

2005

41,842 (89,441)

158,541 (251,564)

3.8

38

32

2007

44,734 (80,384)

175,516 (281,806)

3.9

44

29

2009

38,541 (69,389)

148,656 (246,405)

3.9

46

26

2011

38,754 (68,982)

141,254 (271,230)

3.6

58

28

2013

37,139 (63,011)

137,954 (259,839)

3.7

52

27

2015

39,111 (69,843)

149,493 (266,149)

3.8

50

25

Source: means calculated from the 1999–2015 waves of the PSID

Standard deviations are in parenthesis. Values are in 2015 dollars

Figure 3 shows the value of home equity over the length of our panel and Fig. 4 shows the growth in home equity for both races. While Black and White home equity growth patterns are more similar over the 2005–2007 period, Black growth has more extreme highs and lows in the early part of the panel. During the housing boom, Black equity growth hit a high of 45%, while Whites saw a maximum growth in the mid-twenty percent. Black families also had a more rapid decline in equity, with growth becoming negative by the start of the financial crisis in 2007. Shapiro et al. (2013) found a similar pattern in Black and White housing wealth, with median wealth falling by 53 and 16% for Blacks and Whites, respectively, between 2005 and 2009.
Fig. 3

Source: mean values calculated from the 1999–2015 waves of the PSID. All values are in 2015 dollars

Fig. 4

Source: mean values calculated from the 1999–2015 waves of the PSID. All values are in 2015 dollars

Why would Black home equity rise so quickly, and then fall by such a significant amount, in such a short time period? A hallmark of the housing boom and subsequent bust was subprime mortgage lending. Black households disproportionally purchased these types of mortgages (Massey et al. 2016; Rugh et al. 2015), which initially allowed those who were not able to qualify for a traditional mortgage a chance to own a home, but were more likely to end in foreclosure (Hwang et al. 2015; Rugh et al. 2015).

In the years after 2011, Black home equity in our sample begins to grow again. However, as we saw earlier in Table 2, the Black/White gap in total wealth in 2015 was at its largest since 1999. Examining the trends in specific financial assets and total net wealth without home equity (Table 4) explains what is happening here. In Table 4, we see a much wider gap between Blacks and Whites in 2015 than existed in 1999 (Whites had 7.3 times more wealth in 1999, but 11.5 times more in 2015)—a gap that is even larger than that for total wealth with equity. Because this measure is primarily comprised of financial wealth, it follows that it was the increase in financial assets that Whites saw after the Great Recession—and Blacks did not—that widened the gap. We can see this trend when looking at the value of stocks (Fig. 5) over the time period. While White families’ stock values were almost twice as large in 2015 as they were in 1999, Black families’ stock values were virtually unchanged over this time period.
Table 4

Net wealth (without home equity)

Year

Black

White

Ratio

1999

38,832 (330,565)

285,231 (1,431,646)

7.3

2001

42,511 (214,522)

289,122 (1,708,006)

6.8

2003

60,394 (489,914)

281,460 (1,317,216)

4.7

2005

67,110 (455,417)

337,369 (1,656,912)

5.0

2007

57,369 (235,727)

437,264 (2,096,911)

7.6

2009

44,713 (241,399)

419,524 (2,856,106)

9.4

2011

28,631 (137,899)

371,802 (1,413,595)

13.0

2013

34,269 (174,457)

376,919 (1,340,812)

11.0

2015

39,730 (175,087)

455,482 (1,784,834)

11.5

Source: means calculated from the 1999–2015 waves of the PSID

Standard deviations are in parenthesis

Total wealth, no equity = (annuities + checking and savings + farm/business + other assets + other real estate + stocks) − debt

Values are in 2015 dollars

Fig. 5

Source: mean values calculated from the 1999–2015 waves of the PSID. All values are in 2015 dollars. Stock consists only of stock held in publically owned corporations

The findings that it was predominately financial wealth that is driving the widening gap 6 years after the Recession ended is not surprising, given the existing literature. Gutter and Fontes (2006) and Hanna et al. (2010) found that Whites are much more likely to own risky assets (e.g., stocks and business assets). Grable and Joo (1999) found that account ownership and doing business with financial institutions is positively correlated with good financial outcomes, but because Blacks have more limited access to banking institutions, they do not have as many opportunities to pursue these connections (Robert and Reither 2004).

Regression Models and Results

To more fully examine the effect the financial crisis and subsequent recession on the wealth of Black and White families, we employ two types of regression models. Our first approach employs a fixed effects model to study the influence of the financial crisis on the wealth of Black and White families. The second approach uses a logistic model to provide a complementary analysis of the likelihood that the crisis affected the wealth of Black families more severely, both in depth and in duration, than that of White families.

Approach 1: Inverse Hyperbolic Sine Transformation with Fixed Effects Model

Our first approach for examining the potential differential effects of the financial crisis and recession across racial groups uses a fixed effects model of the type often employed in studies of wealth. Specifically, our regression model is of the form:
$$ {\mathrm{wealth}}_{it}={\beta}_0+{\beta}_1\;{\mathrm{i}\mathrm{ncome}}_{it}+{\beta}_2\kern0.24em {\mathrm{postcrisis}}_{it}+\gamma \prime {X}_{it}+{c}_{\mathrm{i}}+{u}_{\mathrm{i}\mathrm{t}} $$
(1)
where at time t (t = 1999, 2001, … 2015), wealth is the measure of family wealth or the family’s holdings of the particular asset under study, income is the family’s income, postcrisis is a dummy variable that takes a value of 1 if the year is 200910 or later, and X represents a series of variables to control for whether the family consists of a married couple, the age and age-squared of the head of household, the number of children under the age of 18 in the family, and the state of residence.11 The term ci is a fixed family (head of family) effect and uit is the error term. It is important to note that we express the wealth variable as family wealth after applying the inverse hyperbolic sine transformation to the data. The use of this transformation has been seen recently in work by Thompson and Suarez (2015) as well as Zhang and Feng (2017). This transformation works similarly to a standard log transformation with respect to mitigating the effects of outliers (which is an issue with wealth data where extremely high wealth families skew the distribution) but unlike with a log transformation, the data transformed by the inverse hyperbolic sine function remains defined even when the original data values are zero or negative.12 Because many families experience negative net worth for at least some years, the use of the transformation allows us to include these families in our analysis. We estimate Eq. (1) by splitting our sample according to race so that we can compare the effect of the crisis on wealth on both Black and White families.

As mentioned above, we estimate the model using fixed effects, where ci accounts for individual heterogeneity in each family head. Wealth is accumulated savings, which can be affected by innate unobservable factors such as financial time horizon, patience, and degree of risk. These factors are likely correlated with at least some of the control variables, for example, patient individuals are more likely to complete more years of schooling, which in turn affects income, estimating the regression with fixed effects analysis controls for these factors, producing consistent estimates.13

We choose the additional control variables in the model based on economic theory and previous empirical findings in the literature. The Life Cycle Hypothesis, formulated by Ando and Modigliani (1963), motivates the inclusion of age in a model explaining wealth. Quite simply, in early adulthood, individuals do not accumulate much wealth, instead borrowing to finance education and home ownership. As they reach midlife and have greater income, they accumulate assets for retirement. Spend-down of these assets occurs as the individual reaches retirement age. Hence, we should expect to see an increase in wealth up to a point in midlife, at which wealth begins to decrease. Given the non-linearity of this relationship, we include both age and age-squared in our analysis.

A correlation between marriage and wealth is often found in empirical investigations (Ruel and Hauser 2013; Addo and Lichter 2013). Marriage may be positively correlated with wealth due to economies of scale in living arrangements and by individuals pooling of assets, although these effects may be trivial (Killewald et al. 2017). The number of children in a family also can affect wealth by increasing expenses, hence lowering wealth. This is particularly evident in families headed by single women. Grinstein-Weiss et al. (2008) show that single women with children have the lowest overall asset levels, while Ozawa and Lee (2008) find that these types of households accumulate less wealth overall.

The results from these sets of regressions can be found in Table 5 for Black families and in Table 6 for White families. For our purposes, we are most concerned with the postcrisis variable. For interpretation purposes, consider the column (1) of Table 5 where we have an estimate of − 1.556. What this means is that holding all else constant, the postcrisis period has a negative effect on wealth of roughly 78%.14 We also see a larger negative effect on Black families for overall family wealth, wealth without home equity, and checking and savings accounts. The effects on annuities and home equity are similar across the races, while the crisis more severely affected White families’ holdings of stocks and other assets. These latter findings result most likely because, in general, White families are likely to have larger holdings of stocks and other assets than Black families (and if a family was not holding any, say, stocks, the financial crisis would not have caused their holdings of stocks to decrease).
Table 5

Results from inverse hyperbolic sine fixed effects models for Black families

Variables

(1) Wealth

(2) Wealth w/o eq.

(3) Check and saving

(4) Stocks

(5)Other assets

(6)Annuities

(7) Equity

Family income

0.0163*** (0.00269)

0.0167*** (0.00306)

0.0164*** (0.00192)

0.00334*** (0.00119)

0.00257 (0.00163)

0.00380** (0.00177)

0.0130*** (0.00175)

Married

1.350*** (0.467)

0.141 (0.507)

0.226 (0.230)

− 0.109 (0.136)

0.0177 (0.170)

0.118 (0.203)

2.200*** (0.351)

Age

0.115* (0.0670)

− 0.0429 (0.0691)

− 0.0155 (0.0306)

− 0.0258 (0.0178)

− 0.0275 (0.0273)

0.0883*** (0.0234)

0.255*** (0.0417)

Agesq

− 0.000291 (0.000633)

0.00131** (0.000653)

0.000379 (0.000311)

0.000174 (0.000161)

6.00e-05 (0.000265)

− 0.000596** (0.000240)

− 0.00183*** (0.000411)

Children

0.0946 (0.0997)

0.196* (0.108)

0.0596 (0.0483)

0.00979 (0.0261)

− 0.0114 (0.0405)

− 0.0287 (0.0367)

0.0547 (0.0681)

Postcrisis

− 1.556*** (0.234)

− 1.606*** (0.275)

− 0.504*** (0.116)

− 0.173** (0.0748)

− 0.125 (0.106)

− 0.315*** (0.108)

− 0.380*** (0.130)

Constant

6.937*** (2.464)

12.24*** (3.003)

4.169*** (1.541)

2.833** (1.243)

3.702*** (0.926)

1.464 (1.683)

− 7.571* (4.186)

Observations

9908

9908

9908

9908

9908

9908

9908

R-squared

0.021

0.015

0.029

0.013

0.007

0.008

0.051

Number of clusters

1228

1228

1228

1228

1228

1228

1228

State dummies included in regressions but not reported

Family income is in $1000s

Huber-White-Sandwich robust standard errors clustered on families are in parentheses

***p < 0.01; **p < 0.05; *p < 0.1

Table 6

Results from inverse hyperbolic sine fixed effects models for White families

Variables

(1) Wealth

(2) Wealth w/o eq.

(3) Check and saving

(4) Stocks

(5) Other assets

(6) Annuities

(7) Equity

Family income

0.00145*** (0.000417)

0.00140*** (0.000445)

0.00168*** (0.000287)

0.00126*** (0.000323)

− 0.000307 (0.000338)

0.00136*** (0.000359)

0.00167*** (0.000386)

Married

0.945*** (0.323)

0.228 (0.339)

0.473*** (0.113)

0.338** (0.166)

0.139 (0.169)

0.715*** (0.197)

2.999*** (0.244)

Age

0.204*** (0.0450)

0.136*** (0.0499)

0.0903*** (0.0205)

0.0290 (0.0254)

− 0.0529** (0.0247)

0.249*** (0.0288)

0.371*** (0.0314)

Agesq

− 0.000867** (0.000382)

− 0.000630 (0.000421)

− 0.000638*** (0.000192)

− 0.000559** (0.000239)

0.000422* (0.000232)

− 0.00192*** (0.000269)

− 0.00264*** (0.000285)

Children

0.316*** (0.0914)

0.301*** (0.107)

0.0560 (0.0345)

0.126** (0.0534)

0.0588 (0.0478)

0.155** (0.0604)

0.337*** (0.0573)

Postcrisis

− 1.041*** (0.142)

− 0.669*** (0.171)

− 0.121* (0.0694)

− 0.403*** (0.101)

− 0.263** (0.104)

− 0.330*** (0.117)

− 0.506*** (0.0849)

Constant

2.683 (2.273)

4.900** (1.975)

6.475*** (1.134)

3.714*** (1.096)

4.796*** (0.932)

− 3.755*** (1.020)

− 7.542*** (2.021)

Observations

18,722

18,722

18,722

18,722

18,722

18,722

18,722

R-squared

0.019

0.011

0.015

0.019

0.007

0.015

0.099

Number of clusters

2346

2346

2346

2346

2346

2346

2346

State dummies included in regressions but not reported

Family income is in $1000s

Huber-White-Sandwich robust standard errors clustered on families are in parentheses

***p < 0.01; **p < 0.05; *p < 0.1

Approach 2: Logistic Regression Models

In addition to the previous methodology, we use a series of logistic regression models to determine the association that race has on the odds that a family will experience wealth losses of various degrees of severity and duration. The models are cross-sections and take the form:
$$ {\mathrm{wealthdecrease}}_i=\alpha +{\beta}_1\kern0.24em \mathrm{black}+\gamma \prime {X}_i+{\varepsilon}_i $$
(2)
where wealthdecrease is a dummy variable that takes a value of 1 if the relevant measure of wealth or specific types of assets decreased over three distinct time periods: 2007 to 2009, 2007 to 2011, and 2007 to 2015. The control variables in these regressions are the same as in the previous models with the exception that we now also include dummy variables for the sex and level of education of the head of household.15 For each of these specifications, we limit our analysis to families who held positive amounts of the wealth measure being examined in 2007. Additionally, for total family wealth, we run three further specifications where wealthdecrease takes a value of 1 if (a) family wealth went from positive to negative, (b) family wealth declined by more than 10%, or (c) family wealth declined by more than 25%. Again, we limit our sample to families who had positive amounts of family wealth in 2007 and we run the models over the time periods 2007 to 2009, 2007 to 2011, and 2007 to 2015.
Table 7 presents the odds ratios from logistic regressions for our various measures of wealth. The results indicate the likelihood that Black families experienced a decrease in wealth from 2007 to 2009 compared to White families. If the odds ratio is 1.0, the likelihood of a decrease in wealth is the same for Black and White families; if the odds ratio is significantly greater than 1.0, it indicates that Black families were more likely to have experienced a decrease in wealth, and if the odds ratio is significantly less than 1.0, it indicates that Black families were less likely to have experienced a decrease in wealth than White families. For interpretation purposes, consider the estimate for Black in column (3), which is 1.324 and statistically significant. This would be interpreted as Black families are 32% more likely to have lost value in their checking and savings accounts than White families. With respect to overall family wealth, it can be seen that Black families were not significantly more or less likely to have experienced an absolute decrease in wealth than White families. This finding should not be that surprising since most families, Black or White, experienced at least some decrease in wealth over the period 2007–2009.
Table 7

Logistic regressions for an absolute decrease in wealth from 2007 to 2009 for total wealth and various assets

Variables

(1) Wealth

(2) Wealth w/o eq.

(3) Check and saving

(4) Stocks

(5) Other assets

(6) Annuities

(7) Equity

Family income

0.998*** (0.000500)

0.998*** (0.000537)

0.998*** (0.000537)

0.999** (0.000628)

0.998** (0.00110)

1.000 (0.000529)

1.000 (0.000490)

Male

1.104 (0.151)

1.093 (0.154)

1.346** (0.189)

9.865*** (3.827)

3.138*** (1.099)

1.958** (0.522)

1.757*** (0.319)

Married

1.253* (0.163)

1.275* (0.171)

1.025 (0.135)

0.522* (0.188)

1.443 (0.487)

1.233 (0.304)

0.919 (0.156)

Age

1.018 (0.0248)

0.986 (0.0250)

0.976 (0.0236)

0.994 (0.0507)

1.017 (0.0535)

1.013 (0.0443)

0.995 (0.0286)

Agesq

1.000 (0.000212)

1.000 (0.000219)

1.000 (0.000211)

1.000 (0.000430)

1.000 (0.000451)

1.000 (0.000385)

1.000 (0.000250)

Children

1.027 (0.0470)

0.967 (0.0461)

0.997 (0.0465)

0.996 (0.106)

1.034 (0.121)

0.896 (0.0709)

1.031 (0.0552)

High school degree

0.826 (0.138)

0.896 (0.157)

0.678** (0.134)

1.091 (0.651)

0.858 (0.398)

1.628 (0.768)

1.215 (0.248)

Some college

0.955 (0.166)

1.090 (0.198)

0.554*** (0.112)

1.840 (1.104)

1.329 (0.629)

2.067 (0.981)

1.314 (0.278)

College degree

0.942 (0.166)

0.914 (0.169)

0.578*** (0.117)

1.429 (0.840)

0.898 (0.419)

2.004 (0.939)

1.507* (0.319)

Black

1.057 (0.116)

1.070 (0.121)

1.324** (0.148)

0.687 (0.208)

1.602* (0.433)

1.221 (0.264)

0.865 (0.112)

Constant

0.986 (0.916)

6.434* (6.905)

2.563 (2.379)

1.812 (3.689)

0.233 (0.390)

0.214 (0.334)

0.457 (0.496)

Observations

2795

2591

2603

805

664

1141

2260

Pseudo R-squared

0.0315

0.0249

0.0243

0.120

0.0918

0.0514

0.0692

Rank

59

59

59

50

47

51

57

Odds ratios are reported

State dummies included in regressions but not reported

Family income is in $1000s

Standard errors in exponentiated form are in parentheses

***p < 0.01; **p < 0.05; *p < 0.1

Table 8 presents results from the same regressions as in Table 7 with the exception that the odds ratios have been calculated for decreases in wealth from 2007 to 2011. We can now see that for most types of wealth, including total family wealth, Black families were significantly more likely to have experienced wealth decreases. Table 9 presents results from the same regressions, but this time, extending the time period for decreases in wealth from 2007 to 2015. These results are similar to those from 2007 to 2011, indicating that Black families were much more likely than White families to have experienced lasting long-run decreases in wealth as a result of the financial crisis and subsequent recession.
Table 8

Logistic regressions for an absolute decrease in wealth from 2007 to 2011 for total wealth and various assets

Variables

(1) Wealth

(2) Wealth w/o eq.

(3) Check and saving

(4) Stocks

(5) Other assets

(6) Annuities

(7) Equity

Family income

0.998*** (0.000603)

0.998*** (0.000657)

0.999*** (0.000555)

0.999* (0.000521)

0.999 (0.00152)

0.998*** (0.000748)

0.999* (0.000445)

Male

1.611*** (0.218)

1.956*** (0.276)

1.859*** (0.262)

12.65*** (4.456)

7.972*** (2.778)

3.537*** (0.873)

2.503*** (0.443)

Married

0.949 (0.125)

0.814 (0.111)

0.734** (0.0985)

0.682 (0.223)

1.118 (0.386)

0.877 (0.202)

0.805 (0.136)

Age

0.995 (0.0254)

0.983 (0.0260)

1.022 (0.0260)

1.030 (0.0531)

1.037 (0.0560)

0.978 (0.0408)

1.026 (0.0307)

Agesq

1.000 (0.000214)

1.000 (0.000220)

1.000 (0.000213)

1.000 (0.000415)

1.000 (0.000442)

1.000 (0.000347)

1.000 (0.000250)

Children

0.982 (0.0477)

0.957 (0.0485)

1.037 (0.0514)

0.899 (0.1000)

0.902 (0.111)

0.899 (0.0718)

1.053 (0.0590)

High school degree

0.653** (0.110)

0.570*** (0.103)

0.648** (0.128)

0.876 (0.468)

0.578 (0.260)

0.848 (0.360)

1.210 (0.246)

Some college

0.809 (0.141)

0.766 (0.143)

0.591*** (0.119)

1.748 (0.945)

0.737 (0.338)

1.094 (0.468)

1.424* (0.300)

College degree

0.713* (0.127)

0.533*** (0.101)

0.475*** (0.0962)

1.060 (0.558)

0.788 (0.362)

0.919 (0.389)

1.347 (0.283)

Black

1.246** (0.135)

1.450*** (0.161)

1.489*** (0.165)

1.199 (0.334)

2.538*** (0.672)

1.684*** (0.334)

0.972 (0.123)

Constant

5.777 (6.250)

8.901** (9.884)

0.978 (0.952)

0.599 (1.233)

0.180 (0.320)

1.631 (2.534)

0.186 (0.217)

Observations

2854

2652

2662

875

739

1225

2340

Pseudo R-squared

0.0408

0.0379

0.0366

0.164

0.179

0.0680

0.0713

Rank

59

58

60

49

48

53

56

Odds ratios are reported

State dummies included in regressions but not reported

Family income is in $1000s

Standard errors in exponentiated form are in parentheses

***p < 0.01; **p < 0.05; *p < 0.1

Table 9

Logistic regressions for an absolute decrease in wealth from 2007 to 2015 for total wealth and various assets

Variables

(1) Wealth

(2) Wealth w/o eq.

(3) Check and saving

(4) Stocks

(5) Other assets

(6) Annuities

(7) Equity

Family income

0.997*** (0.000651)

0.996*** (0.000722)

0.997*** (0.000629)

0.998** (0.000750)

1.000 (0.00128)

0.999** (0.000703)

1.000 (0.000514)

Male

2.182*** (0.282)

2.226*** (0.297)

2.126*** (0.285)

12.47*** (3.820)

11.18*** (3.764)

5.424*** (1.304)

3.162*** (0.518)

Married

0.889 (0.112)

0.977 (0.128)

0.780* (0.100)

1.122 (0.310)

1.055 (0.351)

0.997 (0.223)

0.658*** (0.103)

Age

0.956* (0.0255)

0.953* (0.0263)

0.985 (0.0264)

0.962 (0.0513)

1.014 (0.0574)

0.908** (0.0406)

0.986 (0.0309)

Agesq

1.000** (0.000209)

1.000** (0.000216)

1.000 (0.000211)

1.000 (0.000404)

1.000 (0.000434)

1.001*** (0.000348)

1.000 (0.000245)

Children

0.919 (0.0499)

0.970 (0.0556)

1.003 (0.0553)

0.942 (0.112)

1.004 (0.130)

0.970 (0.0852)

0.913 (0.0557)

High school degree

0.949 (0.153)

0.851 (0.145)

0.887 (0.165)

1.094 (0.545)

1.073 (0.431)

1.107 (0.434)

1.290 (0.254)

Some college

1.093 (0.182)

0.950 (0.167)

0.874 (0.167)

1.603 (0.811)

1.258 (0.514)

1.548 (0.612)

1.703*** (0.348)

College degree

0.863 (0.147)

0.749 (0.135)

0.923 (0.177)

1.332 (0.659)

1.305 (0.531)

0.841 (0.329)

1.405* (0.286)

Black

1.634*** (0.173)

1.720*** (0.187)

1.828*** (0.200)

1.154 (0.309)

2.098*** (0.515)

1.369* (0.261)

1.248* (0.154)

Constant

3.617 (3.704)

4.232 (4.486)

0.760 (0.791)

1.402 (2.906)

0.103 (0.261)

6.069 (10.09)

0.154 (0.200)

Observations

2944

2745

2730

976

841

1316

2432

Pseudo R-squared

0.0541

0.0525

0.0401

0.205

0.224

0.105

0.0781

Rank

58

58

56

49

48

51

56

Odds ratios are reported

State dummies included in regressions but not reported

Family income is in $1000s

Standard errors in exponentiated form are in parentheses

***p < 0.01; **p < 0.05; * p < 0.1

Although Tables 7, 8, and 9 provide evidence of decreases in wealth and holdings of assets, they are unable to provide evidence of the severity of the decreases in wealth or how this severity might differ across races. To address this issue, we conducted additional logistic regressions to examine the likelihood that Black families would (a) experience a decrease in wealth such that wealth became negative, (b) experience a decrease in overall wealth of more than 10%, and (c) experience a decrease in overall wealth of more than 25%.16

Table 10 provides the results for the period from 2007 to 2009. We see that Black families were more likely than White families to have experienced a move to negative wealth and/or to have had their wealth decrease by more than 25%. Table 11 extends the period of observation from 2007 to 2011. It can been seen that Black families were now significantly more likely to have experienced at least a 10% decrease in wealth as well as a 25% or greater decrease. When the analysis extends the period of study out to 2015 (Table 12), Black families were 50% more likely than White families to still have had negative wealth, a 70% probability of having experienced at least a 10% decrease in overall wealth, and were almost twice as likely as White families to have experienced at least a 25% decrease in overall wealth.17 These findings are consistent with those of Pfeffer et al. (2013) which employed a somewhat similar methodology.
Table 10

Changes in wealth from 2007 to 2009

Variables

(1) Negative

(2) > 10% decrease

(3) > 25% decrease

Family income

0.991*** (0.00210)

0.998*** (0.000525)

0.998*** (0.000636)

Male

0.822 (0.194)

1.241 (0.168)

1.350** (0.185)

Married

0.822 (0.213)

1.080 (0.139)

0.870 (0.114)

Age

1.072 (0.0710)

1.020 (0.0244)

0.960* (0.0237)

Agesq

0.999** (0.000668)

1.000 (0.000207)

1.001** (0.000214)

Children

1.038 (0.0806)

1.013 (0.0456)

1.123** (0.0520)

High school degree

1.144 (0.394)

0.856 (0.140)

0.893 (0.148)

Some college

1.481 (0.519)

1.042 (0.177)

1.029 (0.177)

College degree

1.806 (0.666)

0.925 (0.160)

0.767 (0.135)

Black

1.508** (0.310)

1.112 (0.120)

1.796*** (0.197)

Constant

0.197 (0.335)

0.419 (0.378)

0.420 (0.443)

Observations

2688

2792

2785

Pseudo R-squared

0.143

0.0312

0.0486

Rank

46

58

57

Odds ratios are reported

State dummies included in regressions but not reported

Family income is in $1000s

Standard errors in exponentiated form are in parentheses

***p < 0.01; **p < 0.05; *p < 0.1

Table 11

Changes in wealth from 2007 to 2011

Variables

(1) Negative

(2) > 10% decrease

(3) > 25% decrease

Family income

0.995*** (0.00167)

0.997*** (0.000632)

0.997*** (0.000769)

Male

0.555*** (0.124)

1.523*** (0.203)

1.764*** (0.237)

Married

0.944 (0.228)

0.946 (0.122)

0.771** (0.101)

Age

0.935 (0.0491)

0.964 (0.0243)

0.964 (0.0248)

Agesq

1.000 (0.000485)

1.000 (0.000210)

1.000* (0.000214)

Children

0.973 (0.0756)

0.979 (0.0473)

1.066 (0.0528)

High school degree

1.187 (0.391)

0.747* (0.121)

0.884 (0.144)

Some college

1.626 (0.541)

0.956 (0.161)

1.068 (0.182)

College degree

1.592 (0.559)

0.792 (0.137)

0.828 (0.145)

Black

1.255 (0.230)

1.310** (0.140)

1.920*** (0.208)

Constant

5.789 (8.652)

9.557** (9.564)

0.691 (0.701)

Observations

5.789

9.557**

0.691

Pseudo R-squared

(8.652)

(9.564)

(0.701)

Rank

49

59

57

Odds ratios are reported

State dummies included in regressions but not reported

Family income is in $1000s

Standard errors in exponentiated form are in parentheses

***p < 0.01; **p < 0.05; *p < 0.1

Table 12

Changes in wealth from 2007 to 2015

Variables

(1) Negative

(2) > 10% decrease

(3) > 25% decrease

Family income

0.988*** (0.00222)

0.997*** (0.000672)

0.996*** (0.000771)

Male

0.798 (0.174)

2.166*** (0.278)

2.069*** (0.267)

Married

0.808 (0.201)

0.775** (0.0965)

0.767** (0.0964)

Age

0.926 (0.0521)

0.942** (0.0251)

0.926*** (0.0251)

Agesq

1.000 (0.000475)

1.001*** (0.000209)

1.001*** (0.000212)

Children

0.949 (0.0913)

0.925 (0.0506)

0.921 (0.0523)

High school degree

0.991 (0.336)

0.989 (0.159)

0.882 (0.144)

Some college

2.072** (0.692)

1.170 (0.195)

1.068 (0.180)

College degree

2.186** (0.778)

0.951 (0.162)

0.912 (0.159)

Black

1.505** (0.288)

1.719*** (0.181)

1.934*** (0.208)

Constant

6.957 (13.86)

3.098 (3.143)

2.643 (2.871)

Observations

2823

2940

2934

Pseudo R-squared

0.155

0.0500

0.0563

Rank

47

57

57

Odds ratios are reported

State dummies included in regressions but not reported

Family income is in $1000s

Standard errors in exponentiated form are in parentheses

***p < 0.01; **p < 0.05; *p < 0.1

Taken together, these results from our various econometric models and periods of observation strongly suggest that the financial crisis of 2007–09 and subsequent Great Recession more severely affected the wealth of Black families than the wealth of White families. We had anticipated that one reason this situation may have occurred was because Black families lost more home equity than White families, but we do not find strong evidence that this is the case. It is likely that Black families did not experience as much of a rebound in overall wealth because they were less likely to have stock holdings and so they were unable to capture as many of the gains from the rising stock market as White families were able to do. Additionally, we hypothesize that Black families may have been more severely impacted during the Great Recession than were White families, especially with respect to experiencing negative conditions in the labor market. If this hypothesis is correct, it would likely have resulted in Black families drawing down their net worth or even going into debt during the period under study. These explanations are consistent with our findings.

Conclusion

In this paper, we tie together two areas of significant economic research interest, the financial crisis and Great Recession of the late 2000s with US racial wealth inequality. Since more than a decade has passed since the beginning of the Great Recession, we are now able to look at the longer run outcomes the downturn had on wealth across Black and White families as well as their experience recovering from that downturn.

This paper adds to the literature in a number of ways. Many of the papers on the effects of the recession examine wealth fluctuations in the short run. We are able to look at more long-term effects of the downturn by including data through 2015, 8 years after the financial crisis began, and a time horizon longer than any existing papers in this literature to date. We also break total wealth into its different components, so as to better understand what the specific driving factors are that widened the wealth gap. Our general findings indicate that Blacks not only experienced declines in total family wealth in the short term, but that these losses persisted into 2015, a point where significant economic recovery had occurred. These results are evident in both our descriptive and econometric analysis, and robust to logistic and fixed effects estimation techniques. These findings are also important as the more work that is done using different datasets and approaches, the greater our confidence when we continue to see similar results across studies. Our paper has a number of results that confirm what we see in the existing literature.

On that note, there are a number of additional findings from this study that are worth highlighting. Not only has the wealth gap widened in the years after the Great Recession, but it is at its largest since 1999. We tie this widening gap to differences in financial wealth ownership and the increase in financial asset values that White families experienced following the Great Recession, while Black families, with much lower holdings of financial wealth, did not. Further, our empirical work builds on the findings of the basic data discussion with strong evidence that the financial crisis, subsequent downturn, and recovery were more detrimental to Black family wealth than White family wealth, with effects continuing to last midway through this decade. These results support the general theme of the findings of the literature that Black households have fared worse than White households as a result of these events. The lasting effects found in this study highlight the need for further study of the economic and financial factors that are serving as barriers to Black wealth accumulation in the post-recession period.

Footnotes

  1. 1.

    A good source of the timeline of the US financial crisis is available at the St. Louis Federal Reserve site: https://www.stlouisfed.org/financial-crisis/full-timeline. The National Bureau of Economic Research dates the US recession as beginning in December 2007 and ending in June 2009. See: http://www.nber.org/cycles.html

  2. 2.

    While our focus is on wealth differences, there is also a related literature on income differences. Please see Gradin (2014), Hoover et al. (2018), and Nau and Soener (2017) for examples of the type of work being done in this area.

  3. 3.

    An excellent recent study using both descriptive as well as OLS and decomposition techniques to examine the full range of possible factors to explain the Black-White as well as White-Hispanic wealth gaps generally is in Thompson and Suarez (2015). The authors find the Black-White gap has been increasing recently and attribute a significant portion of the wealth gap to differences in asset holdings. These are driven largely by observable factors such as standard demographic factors, income, home ownership, and types of financial holdings. Interested readers are encouraged to read Thompson and Suarez (2015) for more.

  4. 4.

    The PSID allows up to four responses for race type of the head. For someone who listed more than one race, we assign that person only the race that aligns with his first choice.

  5. 5.

    There are 221 observations that had heads switch from Black to White at least once during the relevant time period.

  6. 6.

    From 2011 to 2013, the gap drops slightly from 7.6 to 7.2, but then rises to 7.7 in 2015

  7. 7.

    As we can see from the relatively large standard deviations in Table 2, wealth is highly skewed. Hence, we report median values in Table 13 in the Appendix. Using medians as our statistic of choice does not change generally what we observe over time: that the ratio between White and Black wealth has increased from the beginning of the recession (2007, ratio of 8.6) to our most current wave (2015, ratio of 10.8). Using medians also paints a more bleak picture of wealth inequality in the country, as Table 2 shows a ratio of 6 in 2007 that increases to 7.7 by 2015. Hence, given that using means does not change our overall results, and also because the components of wealth we focus on (real estate equity, stock) have a median of $0 in many of our waves (particularly for Blacks), we will report means in the remainder of the paper.

  8. 8.

    This is a consistent finding in the literature. See, for example, Shapiro et al. (2013) and Blau and Graham (1990)

  9. 9.

    The PSID also asks respondents about “other real estate,” apart from primary home equity. This category includes the value of a second home, land, rental real estate, and land contracts minus any debt owed. Because assets in this category were likely affected by the housing bust, we also measure “total real estate” wealth by adding primary home equity and equity in other real estate together. We report mean total real estate equity in Table 14 in the Appendix. Aside from the fact that the percentage of total wealth held in this asset is larger for both Blacks and Whites than for primary home equity only (Table 3), the trend over time for the White/Black ratio is similar. Furthermore, the percentage of total wealth in real estate for Whites remains everywhere lower than for Blacks in all years, which is also evident when comparing the mean value. This finding is most likely due to a lower participation in the financial markets for Blacks.

  10. 10.

    We choose 2009 as the beginning of the post-crisis period since by mid-2009, the worst of the financial crisis had passed (as measured by such crisis indicators as the TED Spread and A2/P2 Spread for Non-Financial Commercial Paper). Additionally, the Great Recession officially ended in June 2009. Because PSID interviews are conducted between March and November of survey years, there are potentially some interviews in 2009 that were conducted before financial conditions had completely recovered and the economy begun its expansion. We consider this potential overlap of periods to be of relatively minor concern.

  11. 11.

    A dummy variable is included for each state, excluding one. These serve to control for characteristics specific to each state (e.g., tax laws, regulations).

  12. 12.

    The inverse hyperbolic sine transformation is defined as log(yi + (yi2 + 1)1/2), and with the exception of very small values of y is approximately equal to log(2yi) or log(2) + log(yi). For interpretation purposes, we can treat the dependent variable which is transformed according to this transformation exactly as we would a standard logarithmic dependent variable. For those interested in more on this transformation, please see Burbidge et al. (1988), MacKinnon and Magee (1990), and Pence (2006). We would like to thank Frances Woolley for making us aware of this via her Worthwhile Canadian Initiative blog.

  13. 13.

    Random effects estimation is another option we consider. However, a Hausman test strongly rejects the use of random effects. Statistics from this test are available upon request.

  14. 14.

    Note that in the case of the Inverse Hyperbolic Sine transformation, we would interpret our results as we would a model where the dependent variable is natural logged. Hence, the coefficient estimate is interpreted based on (eβ − 1) ∙ 100 which in our case is (e−1.556 − 1) ∙ 100 =  − 78.9 which is interpreted as a  78.9% reduction in wealth.

  15. 15.

    Sex and level of education were captured in our previous model by the inclusion of fixed effects. Education serves as a predictor of wealth, although it can sometimes be difficult to disentangle its effect from that of income effects. Individuals with more education can earn more income, and in turn save and accumulate higher wealth (Gittleman and Wolff 2000). Killewald et al. (2017) suggest that, in fact, education may even be a proxy for past income streams.

  16. 16.

    As before, for all three cases, we limited the sample to include only families who had positive wealth in 2007.

  17. 17.

    We note that we also ran these regression models without the inclusion family income since the crisis and recession may have also affected wealth via a change in household income. The results from these additional models indicate that Black families were even more likely than White families to have experienced negative on effects on their wealth, suggesting that an income effect might also have played a role in the wealth differences we see between races.

Notes

Acknowledgements

We wish to thank Colin Jones for valuable research assistance and seminar participants at Lakehead University for helpful comments. The usual disclaimer applies.

References

  1. Addo FR, Lichter DT. Marriage, marital history, and black – white wealth differentials among older women. Fam Relat. 2013;75:342–62.  https://doi.org/10.1111/jomf.12007.CrossRefGoogle Scholar
  2. Amuedo-Dorantes C, Borra C. Emerging wealth disparities after the storm: evidence from Spain. Rev Econ Househ. 2017;  https://doi.org/10.1007/s11150-017-9363-3.CrossRefGoogle Scholar
  3. Amuedo-Dorantes C, Pozo S. The impact of the recession on the wealth of older immigrants and native household in the United States. IZA J Labor Policy. 2015;4(1):6. 1–27CrossRefGoogle Scholar
  4. Ando A, Modigliani F. The life cycle hypothesis of saving: aggregate implications and tests. Am Econ Rev. 1963;53(1):55–84.Google Scholar
  5. Barsky R, Bound J, Charles KK, Lupton JP. Accounting for the black-white wealth gap: a nonparametric approach. J Am Stat Assoc. 2002;97(459):663–73.CrossRefGoogle Scholar
  6. Blau F, Graham J. Black-white differences in wealth and asset composition. Q J Econ. 1990;105(2):321–39.CrossRefGoogle Scholar
  7. Bosworth B.. Economic consequences of the great recession: evidence from the panel study of income dynamics. Center for Retirement Research Working Papers; 2012. WP#2012-4. p. 1–26.Google Scholar
  8. Bricker J, Bucks B, Kennickell A, Mach T, Moore K. Drowning or weathering the storm? Changes in family finances from 2007 to 2009. NBER Working Paper; 2011. p. 16985.Google Scholar
  9. Bricker J, Dettling LJ, Henriques A, Hsu JW, Moore KB, Sabelhaus J, et al. Changes in US family finances from 2010 to 2013: evidence from the survey of consumer finances. Fed Reserv Bull. 2014;100(4):1–41.Google Scholar
  10. Burbidge JB, Magee L, Robb AL. Alternative transformations to handle extreme values of the dependent variable. J Am Stat Assoc. 1988;83(401):123–7.CrossRefGoogle Scholar
  11. Campbell LA, Kaufman RL. Racial differences in household wealth: beyond black and white. Res Soc Stratif Mobil. 2006;24(2):131–52.CrossRefGoogle Scholar
  12. Chiteji NS, Hamilton D. Family connections and the black-white wealth gap among middle-class families. Review of Black Political Economy. 2002;30(1):9–28.CrossRefGoogle Scholar
  13. Choi L. Household net worth and asset ownership among the economically vulnerable. Community Development Research Brief. Federal Reserve Bank of San Francisco; 2013.Google Scholar
  14. Choudhury S. Racial and ethnic differences in wealth and asset choices. Soc Secur Bull. 2002;64(4):1–15.Google Scholar
  15. Currie J, Duque V, Garfinkel I. The great recession and mothers’ health. Econ J. 2015;125:F311–46.CrossRefGoogle Scholar
  16. D’Ambrosio C, Rohde N. The distribution of economic insecurity: Italy and the US over the great depression. Rev Income Wealth. 2014;60(S1):S33–52.CrossRefGoogle Scholar
  17. DeVaney SA, Anong ST, Yang Y. Asset ownership by black and white families. Fin Counsel Plan. 2007;18(1):33–45.Google Scholar
  18. Gittleman M, Wolff EN. Racial wealth disparities: is the gap closing? Levy Economics Institute Working Paper No. 311; 2000. Google Scholar
  19. Gittleman M, Wolff EN. Racial differences in patterns of wealth accumulation. J Hum Resour. 2004;39(1):193–227.CrossRefGoogle Scholar
  20. Gorbachev O, O’Flaherty B, Sethi R. Ethnicity, immigration, and wealth fluctuations in the United States. Rev Income Wealth. 2017;  https://doi.org/10.1111/roiw.12317.
  21. Grabka MM. Income and wealth inequality after the financial crisis: the case of Germany. Empirica. 2015;42(2):371–90.CrossRefGoogle Scholar
  22. Grable JE, Joo S. Financial help-seeking behavior: theory and implications. Fin Counsel Plan. 1999;10(1):13–24.Google Scholar
  23. Gradin C. Race and income distribution: evidence from the USA, Brazil, and South Africa. Rev Dev Econ. 2014;18(1):73–92.CrossRefGoogle Scholar
  24. Grinstein-Weiss M, Yeo YH, Zhan M, Charles P. Asset holding and net worth among households with children: differences by household type. Child Youth Serv Rev. 2008;30(1):62–78.CrossRefGoogle Scholar
  25. Gutter MS, Fontes A. Racial differences in risky asset ownership: a two-stage model of the investment decision-making process. Fin Counsel Plan. 2006;17(2):64–78.Google Scholar
  26. Hanna SD, Wang C, Yuh Y. Racial/ethnic differences in high return investment ownership: a decomposition analysis. Fin Counsel Plan. 2010;21(2):44–59.Google Scholar
  27. Hoover GA, Compton RA, Giedeman DC. More on the impact of economic freedom on the black-white income gap. Public Finance Rev. 2018;46(2):205–23.CrossRefGoogle Scholar
  28. Houle J. Mental health in the foreclosure crisis. Soc Sci Med. 2014;118:1–8.CrossRefGoogle Scholar
  29. Huyser KR, Perez JM, Ybarra VD, Hellwege JM, Sanchez L. Differential influence of the great recession on political participation among race and ethnic groups. Soc Sci Q. 2017;  https://doi.org/10.1111/ssqu.12399.CrossRefGoogle Scholar
  30. Hwang J, Hankinson M, Brown KS. Racial and spatial targeting: segregation and subprime lending within and across metropolitan areas. Soc Forces. 2015;93(3):1081–108.CrossRefGoogle Scholar
  31. Killewald A. Return to being black, living in the red: a race gap in wealth that goes beyond social origins. Demography. 2013;50(4):1177–95.CrossRefGoogle Scholar
  32. Killewald A, Pfeffer FT, Schachner JN. Wealth inequality and accumulation. Annu Rev Sociol. 2017;43:379–404.CrossRefGoogle Scholar
  33. Lusardi A. Financial literacy: an essential tool for informed consumer choice. Working Paper, Dartmouth College; 2008. Downloaded November 13, 2017.Google Scholar
  34. MacKinnon JG, Magee L. Transforming the dependent variable in regression models. Int Econ Rev. 1990;31(2):315–39.CrossRefGoogle Scholar
  35. Maroto M. Growing further apart: racial and ethnic inequality in household wealth across the distribution. Sociol Sci. 2016;3:801–24.CrossRefGoogle Scholar
  36. Massey DS, Rugh JS, Steil JP, Albright L. Riding the stagecoach to hell: a qualitative analysis of racial discrimination in mortgage lending. City Community. 2016;15(2):118–36.CrossRefGoogle Scholar
  37. McKernan S-M, Ratcliffe C, Steuerle E, Zhang S. Do racial disparities in private transfers help explain the racial wealth gap? New evidence from longitudinal data. Demography. 2014a;51(3):949–74.CrossRefGoogle Scholar
  38. McKernan S-M, Ratcliffe C, Steuerle E, Zhang S. Disparities in wealth accumulation and loss from the great recession and beyond. Am Econ Rev. 2014b;104(5):240–4.CrossRefGoogle Scholar
  39. McKernan S-M, Ratcliffe C, Steuerle E, Zhang S. Impact of the great recession and beyond: disparities in wealth building by generation and beyond. Urban Institute Working Paper; 2014c. p. 1–21.Google Scholar
  40. Nau M, Soener M. Income precarity and the financial crisis. Socio Econ Rev. 2017;  https://doi.org/10.1093/ser/mwx020.
  41. O’Brien RL. Depleting capital? Race, wealth and informal financial assistance. Soc Forces. 2012;91(2):375–96.CrossRefGoogle Scholar
  42. Oliver ML, Shapiro TM. Black wealth, white wealth : a new perspective on racial inequality. 2nd ed. New York: Routledge; 2006.Google Scholar
  43. Ozawa MN, Lee Y. The effect of children on the income status of female-headed households: an intercountry comparison, vol. 22. New York: Oxford University Press; 2008. p. 357–73.Google Scholar
  44. Panel Study of Income Dynamics. Public use dataset. Produced and distributed by the Survey Research Center. Ann Arbor: Institute for Social Research, University of Michigan; 2017.Google Scholar
  45. Pence KM. The role of wealth transformations: an application to estimating the effect of tax incentives on saving. Contrib Econ Ana Pol. 2006;5(1):20.Google Scholar
  46. Pfeffer FT, Danziger S, Schoeni RF. Wealth disparities before and after the great recession. Ann Am Acad Pol Soc Sci. 2013;650(1):98–123.CrossRefGoogle Scholar
  47. Robert SA, Reither EN. A multilevel analysis of race, community disadvantage, and body mass index among adults in the U.S. Soc Sci Med. 2004;59(12):2421–34.CrossRefGoogle Scholar
  48. Ruel E, Hauser R. Explaining the gender wealth gap. Demography. 2013;50(4):1155–76.CrossRefGoogle Scholar
  49. Rugh JS, Albright L, Massey DS. Race, space, and cumulative disadvantage: a case study of the subprime lending collapse. Soc Probl. 2015;62(2):186–218.CrossRefGoogle Scholar
  50. Shapiro T, Meschede T, Osoro S. The roots of the widening racial wealth gap: explaining the black-white economic divide. IASP Research and Policy Brief; 2013. p. 1–8.Google Scholar
  51. Thompson JP, Suarez GA. Exploring the racial wealth gap using the survey of consumer finances. Finance and Economics Discussion Series 2015–076. Washington: Board of Governors of the Federal Reserve System; 2015.  https://doi.org/10.17016/FEDS.2015.076.CrossRefGoogle Scholar
  52. United States, Department of Labor, Bureau of Labor Statistics (BLS). Labor force statistics from the current population survey. Retrieved June 26, 2018 from: https://www.bls.gov/cps/cpsaat01.htm.
  53. Wolff EN. The asset price meltdown and the wealth of the middle class. NBER Working Paper; 2012. p. 18559.Google Scholar
  54. Zhang S, Feng S. Understanding the unequal post-recession wealth recovery for American families. B E J Econ Anal Policy. 2017;17(4):20160229.  https://doi.org/10.1515/bejeap-2016-0229. ISSN (Online) 1935-1682CrossRefGoogle Scholar

Copyright information

© Springer Nature Switzerland AG 2018

Authors and Affiliations

  1. 1.University of ManitobaWinnipegCanada
  2. 2.Grand Valley State UniversityGrand RapidsUSA

Personalised recommendations