Annotated Bibliography (APA Format)

Wealth, Race, and Place

Author(s): Brian L. Levy

Source: Demography , February 2022, Vol. 59, No. 1 (February 2022), pp. 293-320

Published by: Duke University Press on behalf of the Population Association of America

Stable URL: https://www.jstor.org/stable/10.2307/48687237

JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact [email protected]. Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at https://about.jstor.org/terms

Population Association of America and Duke University Press are collaborating with JSTOR to digitize, preserve and extend access to Demography

This content downloaded from �����������63.236.101.18 on Sun, 21 Jan 2024 14:41:17 +00:00������������

All use subject to https://about.jstor.org/terms

Demography (2022) 59(1):293–320 DOI 10.1215/00703370-9710284 © 2022 The Author This is an open access arti cle dis trib uted under the terms of a Creative Commons license (CC BY-NC-ND 4.0).

ELECTRONIC SUPPLEMENTARY MATERIAL The online ver sion of this arti cle (https: / /doi .org /10.1215/00703370 -9710284) con tains sup ple men tary mate rial.

Published online: 18 January 2022

Wealth, Race, and Place: How Neighborhood (Dis)advan tage From Emerging to Middle Adulthood Affects Wealth Inequality and the Racial Wealth Gap

Brian L. Levy

ABSTRACT Do neigh bor hood con di tions affect wealth accu mu la tion? This study uses the National Longitudinal Survey of Youth 1979 cohort and a coun ter fac tual esti ma tion strat egy to ana lyze the effect of prolonged expo sure to neigh bor hood (dis)advan tage from emerg ing adult hood through mid dle adult hood. Neighborhoods have siz­ able, plau si bly causal effects on wealth, but these effects vary sig nif cantly by race/ eth nic ity and homeownership. White homeowners receive the larg est pay off to reduc­ tions in neigh bor hood dis ad van tage. Black adults, regard less of homeownership, are dou bly dis ad van taged in the neigh bor hood–wealth rela tion ship. They live in more­ dis ad van taged neigh bor hoods and receive lit tle return to reduc tions in neigh bor hood dis ad van tage. Findings indi cate that disparities in neigh bor hood (dis)advan tage fg ure prom i nently in wealth inequal ity and the racial wealth gap.

KEYWORDS Neighborhood effects • Wealth • Inequality • Race

Introduction

Wealth is a key mea sure of well­being and pre dic tor of life chances in the United States (Spilerman 2000). It plays an impor tant role in edu ca tional, labor mar ket, and health out comes (Killewald et al. 2017) and serves as both a safety net in eco nomic down turns and a means for upward mobil ity (Shapiro 2006). Wealth is also one of the most unequally dis trib uted resources and a prominent fea ture of U.S. racial inequal­ ity. In 2016, median house hold wealth of Whites was 10 times that of Blacks and 8 times that of His pan ics (Dettling et al. 2017). Wealth’s mobil ity­gen er at ing and safety net func tions make it crit i cal to racial strat i f ca tion (Shapiro 2006), and wealth is a cen tral deter mi nant of racial disparities in edu ca tional attain ment and wel fare receipt (Conley 1999, 2001). Thus, many con sider wealth the “sine qua non indi ca tor of mate rial wellbeing” (Oli ver and Shapiro 2006:203).

I argue that neigh bor hoods are an overlooked driver of wealth inequal ity. For many, homes are a key source of wealth (Shapiro 2006), and home val ues are closely related to the neigh bor hoods in which they are located (Galster et al. 2008). Neigh­ borhoods also affect edu ca tional attain ment, employ ment, income, and other fac tors

This content downloaded from �����������63.236.101.18 on Sun, 21 Jan 2024 14:41:17 +00:00������������

All use subject to https://about.jstor.org/terms

294 B. L. Levy

that impact wealth (Chetty et al. 2016; Vartanian and Buck 2005; Wodtke et al. 2011). Despite neigh bor hoods’ the o ret i cal impor tance, I am aware of no rig or ous causal anal y sis of neigh bor hood effects on wealth. Whereas meso­level char ac ter is tics, such as neigh bor hoods, are underexamined, research on macro­ and micro­level causes of wealth inequal ity is much more com mon (Keister 2005; Keister and Moller 2000). Still, our knowl edge of the sources of wealth inequal ity remains lim ited (Pfeffer and Schoeni 2016), and most ana ly ses of the racial wealth gap leave a siz able por­ tion unex plained (e.g., Campbell and Kaufman 2006; Herring and Henderson 2016; Maroto 2016; Oli ver and Shapiro 1995).

This study ana lyzes how neigh bor hood (dis)advan tage in adult hood relates to wealth at age 50 and helps explain the racial wealth gap. It makes three con tri bu tions to research on neigh bor hood effects, racial inequal ity, and wealth inequal ity. First, it identifes neigh bor hoods as an impor tant fea ture of wealth inequal ity. Second, it reveals two ways that neigh bor hoods con trib ute to the racial wealth gap: through (1) large disparities in neigh bor hood dis ad van tage (ND) and (2) Whites dis pa rately beneft ting from reduc tions in ND. Third, it responds to the recent call (Killewald et al. 2017) for research on the wealth of groups besides Blacks and Whites. Beyond these con tri bu tions, this study advances research on neigh bor hood effects by focus­ ing on an understudied period of the life course (adult hood), ana lyz ing het ero ge neous effects, and using coun ter fac tual meth ods with a con tin u ous treat ment.

Literature Review

Wealth Inequality in America

Wealth inequal ity in the United States is extreme and wid en ing, eclips ing even the lev els seen dur ing the Roaring Twenties (Piketty 2013/2014; Saez and Zucman 2016). The richest 1% now own 40% of wealth (Saez and Zucman 2016), and the share of house holds with no or neg a tive wealth is ris ing (Keister and Moller 2000; Pfeffer and Schoeni 2016). The mid dle of the dis tri bu tion also shows diver gence, with house hold wealth grow ing faster at the 75th per cen tile than at the median or the 25th per cen tile (Pfeffer and Schoeni 2016).

Research on wealth inequal ity has iden ti fed two broad types of deter mi nants: struc tural (macro­level) fac tors, includ ing the hous ing and stock mar kets, asset and tax pol i cies, and rac ism; and indi vid ual or fam ily (micro­level) driv ers, includ ing age, fam ily struc ture, edu ca tion, income, and inher i tances (Keister and Moller 2000). Research on the United States and 15 other high­income countries has found that income and inher i tances are the stron gest pre dic tors (Semyonov and Lewin­Epstein 2013). Several reviews offer fur ther insight into wealth inequal ity (Keister 2005; Keister and Moller 2000; Killewald et al. 2017), but nota bly absent from research on the causes of wealth accu mu la tion are meso­level fac tors, such as neigh bor hoods.

As is the case with over all wealth inequal ity, the racial wealth gap is sub stan tial and grow ing (Conley 2010; Oli ver and Shapiro 1995). The richest 100 U.S. house holds have as much wealth as all Blacks plus one third of His pan ics com bined (Collins and Hoxie 2015). The deter mi nants of the racial wealth gap vary from the deter mi nants of gen eral wealth accu mu la tion. Racial wealth inequal ity fol lows cen tu ries of rac ist

This content downloaded from �����������63.236.101.18 on Sun, 21 Jan 2024 14:41:17 +00:00������������

All use subject to https://about.jstor.org/terms

295Wealth, Race, and Place: Neighborhood (Dis)advan tage and Inequality

pub lic pol icy (Conley 2010; Oli ver and Shapiro 1995) that “sys tem at i cally prevented [Black Amer i cans] from accu mu lat ing prop erty” (Conley 1999:611). Homeowner­ ship is foun da tional for wealth accu mu la tion given that owning a home and dura tion of homeownership are pos i tively asso ci ated with wealth (Di et al. 2007; Turner and Leua 2009). Disparities in the rate and dura tion of homeownership can explain a large por tion of the racial wealth gap (Oli ver and Shapiro 1995; Shapiro 2006; Shapiro et al. 2013)—much more than income and edu ca tional attain ment explain (Sullivan et al. 2015). Still, non­White homeowners have lower equity and equity con di tional on socio eco nomic sta tus than do White homeowners (Killewald and Bryan 2016; Krivo and Kaufman 2004). Whites start with homes that have higher val ues, and their homes appre ci ate faster (Flippen 2004). One poten tial expla na tion for this is racial­ ized neigh bor hood access; non­Whites, par tic u larly Blacks, dis pro por tion ately reside in dis ad van taged neigh bor hoods (Massey and Denton 1993; Newman and Holupka 2016).

Most research incor po rat ing a range of indi vid ual­level var i ables can not fully explain the racial wealth gap (e.g., Campbell and Kaufman 2006; Herring and Henderson 2016; Maroto 2016; Oli ver and Shapiro 1995). A nota ble excep tion is Killewald and Bryan’s (2018) anal y sis of median racial wealth gaps at age 50. They con cluded that fam ily social ori gins explain about half of the median wealth gap, income and edu ca tion explain another quar ter, and homeownership and other house­ hold fac tors explain the fnal quar ter. Still, Maroto (2016) found that the racial wealth gap is large and dif f cult to explain at the top end of the wealth dis tri bu tion. Thus, expla na tions for the median gap may not trans late to the full wealth dis tri bu tion, and fur ther research on the gap is crit i cal (Killewald et al. 2017). For a new expla na tion, I turn to a key meso­level fea ture of fam i lies’ homes: the neigh bor hoods in which they sit.

Residential Segregation

The United States has a long his tory of racial res i den tial seg re ga tion. Documented back to the nineteenth cen tury (Du Bois 1899), seg re ga tion has waned only some­ what and remains a prob lem by con cen trat ing non­White, par tic u larly Black, Amer i cans in less­advan taged neigh bor hoods (Lee et al. 2014; Logan et al. 2015; Massey and Denton 1993). Contemporary seg re ga tion results from his tor i cal inequal­ ities (Sharkey 2013), ongo ing dis crim i na tion in mort gages and hous ing (Fischer and Lowe 2014; Pager and Shepherd 2008; Rugh and Massey 2010), and Whites’ pref er­ ence for neigh bor hoods with few non­White res i dents (Krysan et al. 2009).

Unlike racial seg re ga tion, class­based seg re ga tion emerged more recently as an impor tant con sid er ation. Income seg re ga tion increased from 1970 to 2012, with nota­ ble increases in the 1980s and 2000s (Jargowsky 1996; Reardon et al. 2018). Class­ based seg re ga tion is par tic u larly salient for racial and eth nic minor i ties; low­income Blacks and His pan ics are often seg re gated into the most dis ad van taged neigh bor­ hoods (Jargowsky 1996). Together, race/eth nic ity and class con sti tute the two key fea tures of con tem po rary neigh bor hood seg re ga tion (Lee et al. 2015).

These pat terns sug gest that neigh bor hoods could affect wealth. Rusk (2001) pos ited a “seg re ga tion tax,” with dis ad van taged groups receiv ing lower returns to

This content downloaded from �����������63.236.101.18 on Sun, 21 Jan 2024 14:41:17 +00:00������������

All use subject to https://about.jstor.org/terms

296 B. L. Levy

homeownership (see also Faber and Ellen 2016; Flippen 2004; Shapiro 2004). Denton (2001) fur ther spec u lated that the seg re ga tion tax is paid across the class dis tri bu tion. Still, this idea is not fully devel oped. Why might seg re gated neigh bor hoods reduce non­Whites’ wealth? Do highly seg re gated cit ies always have high wealth inequal ity? That is, does seg re ga tion lead to wealth disparities per se? Flippen (2010) found that met ro pol i tan racial seg re ga tion is asso ci ated with low rates of minor ity homeown­ ership, which is a poten tial mech a nism by which seg re ga tion could cause wealth disparities. Alternatively, do the char ac ter is tics of seg re gated neigh bor hoods drive wealth through dis pa rate access to advan taged, wealth­pro mot ing neigh bor hoods?

Neighborhood Effects

There are sev eral poten tial mech a nisms for neigh bor hood effects on wealth, which broadly fall into two groups: achieved sta tuses and hous ing. Considerable research has exam ined neigh bor hood effects on sta tus attain ment and behav ioral out comes. Neighborhood effects on edu ca tional and labor mar ket out comes are well established (Chetty et al. 2016; Sharkey and Faber 2014). Disadvantaged neigh bor hoods also increase the risk of crime and incar cer a tion (Hipp et al. 2010; Peterson and Krivo 2010), neg a tive behav ioral out comes (Sampson et al. 2002), and low lev els of health and well­being (Ludwig et al. 2012; Ross and Mirowsky 2001). Each of these out­ comes rep re sents a plau si ble path way for neigh bor hood effects on wealth. Achieved sta tuses seem espe cially likely to explain neigh bor hood effects on over all wealth inequal ity, whereas they may be less impor tant for neigh bor hood­based racial wealth disparities (Keister and Moller 2000; Semyonov and Lewin­Epstein 2013; Sullivan et al. 2015).

Neighborhood demo graph ics also cor re late with home val ues. Whites’ dis in cli­ na tion to move to low­income or non­White neigh bor hoods neg a tively affects home equity (Crowder and South 2008; Emerson et al. 2001; Galster et al. 2008; Krysan et al. 2009). Although both race (Anacker 2010; Coate and Schwester 2011) and class (Galster et al. 1999; Peng and Thibodeau 2013) are related to val ues, class is espe cially salient (Flippen 2004; Harris 1999). Because hous ing is a key source of wealth, the hous ing mar ket rep re sents a unique mech a nism for neigh bor hood effects on wealth—one not empha sized in most research on neigh bor hood effects.

Legacy and struc tural dis ad van tages in the hous ing mar ket imply that neigh bor­ hood effects oper at ing through hous ing may be salient for racial inequal ity. The his­ tory of redlining, block bust ing, and urban renewal (Faber 2020; Lipsitz and Oli ver 2010), cou pled with con tem po rary inequities in appraisal (Howell and Korver­Glenn 2021), mort gage lend ing (Korver­Glenn 2021; Stu art 2003), fore clo sure (Hall et al. 2015a, 2015b; Rugh et al. 2015), and sit ing of ame ni ties (Moore et al. 2008; Morland et al. 2002), dis pa rately concentrates value. Discrimination in the hous ing search pro cess (Fischer and Lowe 2014; Korver­Glenn 2021; Pager and Shepherd 2008; Rugh and Massey 2010), as well as long­stand ing pat terns of res i den tial seg­ re ga tion (Massey and Denton 1993; Reardon et al. 2018), restricts non­Whites’, espe cially low­income non­Whites’, access to neigh bor hoods with strong wealth advan tages. This research documenting the breadth of rac ism in hous ing sug gests that the seg re ga tion tax (Rusk 2001) results from spe cifc dis ad van tages in neigh bor hoods

This content downloaded from �����������63.236.101.18 on Sun, 21 Jan 2024 14:41:17 +00:00������������

All use subject to https://about.jstor.org/terms

297Wealth, Race, and Place: Neighborhood (Dis)advan tage and Inequality

into which tra di tion ally dis ad van taged pop u la tions, espe cially Blacks (Newman and Holupka 2016), are seg re gated.

Because wealth deter mi nants are mul ti ple and vary by race, it is impor tant to con sider effect het ero ge ne ity (e.g., Levy 2019; Wodtke et al. 2016). Housing is an out sized com po nent of non­Whites’ wealth (Kuebler 2013), so neigh bor hood effects through hous ing may be par tic u larly impor tant for racial and eth nic minor i ties. Alter­ natively, with the increas ing con cen tra tion of wealth (Saez and Zucman 2016) and larger pay off to homeownership among Whites (Krivo and Kaufman 2004), wealth ben e fts may be con cen trated in the most-advan taged neigh bor hoods, among Whites, or among homeowners.

Data

I use the restricted­use National Longitudinal Survey of Youth 1979 cohort (NLSY79) to ana lyze neigh bor hood effects on wealth accu mu lated at roughly age 50. The NLSY79 is a nation ally rep re sen ta tive panel sur vey of nearly 10,000 indi vid u als aged 14–21 in 1979. The NLSY79 sur veyed respon dents annu ally from 1979 to 1994 and bien ni ally after ward. During the ini tial wave(s), the NLSY79 also col lected infor­ ma tion from par tic i pants’ par ents. The NLSY79 has sev eral use ful fea tures for this anal y sis. First, it includes a rep re sen ta tive sam ple of His pan ics in the ini tial sam pling frame, per mit ting anal y sis of an impor tant but understudied group in the neigh bor­ hood effects and wealth lit er a tures. Second, each wave col lects data on res i den tial neigh bor hoods and a range of var i ables pre dic tive of ND. Third, wealth inequalities and racial gaps sta bi lize when a cohort reaches age 50 (Urban Institute 2015), so the NLSY79 rep re sents a recent cohort at this age. For this anal y sis, I use the lon gi tu di nal sam ple of roughly 7,300 indi vid u als who com pleted a sur vey in 2012. This num ber rep re sents a 79% response rate for those alive from the main lon gi tu di nal sam ple, a low level of attri tion for a study span ning 33 years.1

To merge data on par tic i pants’ neigh bor hoods, the restricted­use NLSY79 pro vi des wave-spe cifc res i den tial cen sus tract iden ti f ers using 2010 bound aries.2 I use neigh­ bor hood socio eco nomic data from the decen nial censuses and the fve-year Amer i can Community Survey (ACS) cen tered on 2010, which are pro vided by the Longitudinal Tract Database (LTDB) (Logan et al. 2014) and Social Explorer.3 I impute inter cen sal

1 Selecting 2012 respon dents as the ana lytic sam ple reduces miss ing data and allows the use of NLSY­ constructed sam pling weights that adjust for observed var i a tion in attri tion and account for oversampling in the ini tial frame. The ongo ing panel com prises 9,964 indi vid u als frst sur veyed in 1979. Since the ini tial sur vey, 689 respon dents were recorded as deceased. Of all other non re spon dents in 2012, 903 refused to par tic i pate, 466 could not be located, 125 were deemed too dif f cult to inter view, and 481 did not respond for other rea sons. Yearly attri tion for the NLSY79 is low compared with similar surveys, and evi dence sug gests that lon gi tu di nal panel stud ies can rea son ably esti mate cur rent pop u la tion sta tis tics decades after their incep tion (Schoeni et al. 2013). 2 Neighborhood clus ter ing is very low. For 84% of all per son-years in the sam ple, the respon dent is the only indi vid ual liv ing in their tract in that year. Ninety­nine per cent of all per son­years have four or fewer respon dents in a tract in a year. 3 Data from Social Explorer are avail able at https: / /www .socialexplorer .com /.

This content downloaded from �����������63.236.101.18 on Sun, 21 Jan 2024 14:41:17 +00:00������������

All use subject to https://about.jstor.org/terms

298 B. L. Levy

data using lin ear inter po la tion, har mo nize data to 2010 bound aries using the LTDB, and merge this infor ma tion to respon dents.

The depen dent var i able is fam ily wealth in 2012, when respon dents were roughly age 50 (mean age = 51.5; range = 47–56).4 The NLSY79 cal cu lates wealth as total assets minus debts. Assets include homes, auto mo biles, businesses, estates, stocks, bonds, and cash. Debts include prop erty, mort gage, and other debts. The NLSY79 top codes assets for the top 2% of respon dents as their group mean wealth, which is a poten tial lim i ta tion for ana ly ses of the wealth i est but is unlikely to bias aver age neigh bor hood effects. This wealth mea sure was assessed after the Great Recession, which had out sized impacts on non­Whites’ wealth (Pfeffer et al. 2013). Although 2012 may rep re sent a high­water mark for racial inequal ity, it is impor tant ana lyt i­ cally. Non­White house holds recov ered hous ing wealth at slower rates (Bricker et al. 2014; Thomas et al. 2018), and their height ened vul ner a bil ity to eco nomic down turns rep re sents an impor tant aspect of inequal ity. Given its right skew, I trans form wealth using the inverse hyper bolic sine (IHS), which is akin to the nat u ral log with the excep tion that the IHS is defned for zero and neg a tive num bers. The IHS trans for- ma tion also guards against the undue influ ence of out li ers.

The pri mary inde pen dent var i able (“treat ment”) is neigh bor hood (dis)advan tage. I cal cu late ND using fac tor anal y sis of seven neigh bor hood char ac ter is tics: pov erty, unem ploy ment, female­headed house holds, wel fare receipt, adults with out a high school diploma, adults with a col lege degree (neg a tive load ing), and work ers hold ing man a ge rial or pro fes sional jobs (neg a tive load ing). I mea sure these at the tract level and use the frst com po nent’s score for ND, which aligns with recent neigh bor hood effects research (e.g., Wodtke et al. 2011). In addi tion to ND, I mea sure expo sure to met ro pol i tan or micropolitan area5 racial res i den tial seg re ga tion using an entropy index (Theil’s H) based on tract­level shares of the pop u la tion that are White, Black, Asian, His panic, and mul ti ra cial/other (see Reardon and Firebaugh 2002). If seg re ga­ tion causes wealth disparities per se, then this should explain any asso ci a tion between ND and wealth.

Another focal inde pen dent var i able is race/eth nic ity. I use the racial/eth nic ori- gin with which the respon dent most closely iden ti fed in 1979 to code indi vid u als as His panic, non­His panic Black, non­His panic White, or non­His panic other race. I use screener-reported race/eth nic ity from 1978 to com plete miss ing or ambig u ous data. Screener­reported race is highly valid according to respon dent­pro vided race. Among nonmissing respon dents, 97.4% of indi vid u als have the same racial cat e gory for both mea sures.

This anal y sis includes con trol var i ables reflecting char ac ter is tics with a major effect on neigh bor hood attain ment (Harding 2003; Quillian 2003; Sampson and Sharkey 2008; Wodtke et al. 2011). Time-invari ant con trols include race/eth nic ity, sex

4 Wealth data are also avail able for 2016. Among those sur veyed in 2012 and 2016, wealth val ues and per­ cen tiles cor re late strongly between waves. Yet, attri tion due to death increased sub stan tially between 2012 and 2016; roughly 2.3% of 2012 respon dents were deceased by the 2016 wave. One quar ter of all attri tion due to death by 2016 occurred between the 2012 and 2016 waves. Given that wealth inequal ity sta bi lizes around age 50 and death is endog e nous to wealth, I use 2012 wealth as the out come. 5 I assign tracts to core­based met ro pol i tan or micropolitan areas using a Missouri Census Data Center cross walk: https: / /mcdc .missouri .edu /applications /geocorr .html.

This content downloaded from �����������63.236.101.18 on Sun, 21 Jan 2024 14:41:17 +00:00������������

All use subject to https://about.jstor.org/terms

299Wealth, Race, and Place: Neighborhood (Dis)advan tage and Inequality

(male/female), nativ ity, for eign lan guage spo ken dur ing child hood (yes/no), paren tal edu ca tion, paren tal employ ment skill level, child hood fam ily struc ture, and base line mea sures of wealth and ND. Nativity is a dummy var i able distinguishing frst- and sec ond­gen er a tion ado les cents from third­plus­gen er a tion ado les cents. Parental edu­ ca tion mea sures the highest edu ca tional attain ment of res i dent par ent(s): less than a high school diploma, high school diploma, some col lege, or bach e lor’s degree or higher. Parental job skill is the highest job skill level of res i dent par ent(s): unskilled, clerk/sales, skilled man ual, or man ager/pro fes sional. Childhood fam ily struc ture is one of the fol low ing categories: always lived with two bio log i cal par ents, always lived with one and never the other bio log i cal par ent, or some other liv ing arrange­ ment. Family wealth was frst mea sured in 1985, so wealth and ND in 1985 are base- line con trols. I adjust all dol lar val ues to 2012 con stant dol lars using the con sumer price index.

Time­vary ing con trols are char ac ter is tics of the respon dent, their fam ily, and the head of their house hold, which can be the respon dent. Respondent con trols are edu­ ca tional attain ment (same categories as noted ear lier), mar i tal sta tus (never mar ried, mar ried, or other), and age. Family con trols are wealth, fam ily size, income­to­needs ratio, inher i tance value, home value, home debt, home equity, and dummy var i ables for inher i tance receipt, homeownership sta tus, mov ing since the prior sur vey wave, and pub lic assis tance receipt. Wealth is fam ily wealth at the prior wave. Income­to­ needs is the ratio of fam ily income to the fed eral pov erty thresh old. Inheritance value is the IHS of the total value of estates, trusts, and inher i tances that the respon dent or spouse received in the last year. Home val ues are respon dent­reported mar ket val ues of pri mary homes at the prior wave; for rent ers, home val ues are zero.6 Home debt is the total value of mort gages, back taxes, and other debts owed on the res i den tial home by the respon dent and their spouse at the prior wave. Home equity is the dif fer ence between home val ues and home debts. I trans form home value, debt, and equity using the IHS. Head of house hold con trols are the num ber of jobs worked, the per cent age of weeks worked, and the num ber of hours worked per week in the last year.

Whereas I use the 2012 wave of the NLSY79 to mea sure the depen dent var i able, I use most post-base line waves (1986–2010, exclud ing 1992, 2004, and 2008)7 for the ND treat ment. Cumulative ND is the aver age ND score across these years. I address miss ing data using mul ti ple impu ta tion with chained equa tions and 10 imputed data sets. I retain obser va tions with an imputed depen dent var i able (see Wodtke et al. 2016). Section A of the online appen dix sum ma rizes data missingness. Data on neigh­ bor hood demo graph ics are more likely to be miss ing ear lier in the sam ple because many parts of the United States, espe cially rural loca tions, were not divided into cen sus tracts until 1990. Still, prob lems from missingness are not likely to be pro­ nounced, as I dis cuss later. I weight all ana ly ses and sta tis tics using 2012 sam pling weights. Section B of the online appen dix pres ents weighted sum mary sta tis tics for the frst imputed data set.

6 Self­reported val ues are gen er ally quite accu rate, and research has found that errors in esti mated val ues are uncor re lated with indi vid ual demo graph ics and neigh bor hood char ac ter is tics (Bucks and Pence 2006; Kiel and Zabel 1999). 7 I exclude 1992, 2004, and 2008 because they lack data on lagged fam ily wealth, home value, home debt, and home equity.

This content downloaded from �����������63.236.101.18 on Sun, 21 Jan 2024 14:41:17 +00:00������������

All use subject to https://about.jstor.org/terms

300 B. L. Levy

Figure 1 pres ents a weighted boxplot of raw 2012 wealth val ues by race/eth nic ity at roughly age 50. Wealth has strong pos i tive skew across all groups, confrming the need for the IHS trans for ma tion. Median wealth for Whites is sub stan tially greater than that for Blacks, His pan ics, or other races; sim i lar disparities exist at the 25th and 75th per cen tiles of the race-spe cifc dis tri bu tions. In terms of aver ages by group (not shown), Whites, Blacks, His pan ics, and other­race indi vid u als have mean wealth val ues of roughly $420,000, $90,000, $200,000, and $270,000, respec tively. Mean Black and His panic wealth com bined is 29% of mean White wealth, which aligns with other esti ma tes.8

Figure 2 pres ents weighted boxplots of cumu la tive ND by race/eth nic ity, sim i larly reveal ing large disparities. There is lit tle to no over lap in the interquartile ranges of ND between Whites and Blacks or His pan ics. Whereas few Whites live in highly dis ad van taged neigh bor hoods—those with an ND score 1 stan dard devi a tion or more above the national mean—the major ity of Blacks and His pan ics live in highly dis ad­ van taged neigh bor hoods. Moreover, roughly one quar ter of Whites but few Blacks or His pan ics live in highly advan taged neigh bor hoods.

8 The 2013 Survey of Consumer Finances (SCF), which does not top code wealth, reports mean wealth of non­Whites as 26.1% that of Whites. The dif fer ence with my cal cu la tions results at least par tially from the age com po si tion of Whites and non­Whites. The NLSY79 com pares sim i larly aged indi vid u als. The SCF is a cross­sec tional sur vey of fam i lies; because Whites are older than non­Whites (the median age dif fer ence between Whites and non­Whites is roughly 10 years), the SCF reports a larger over all racial wealth gap than would likely occur within an age cohort.

Fig. 1 Boxplots of raw wealth, by race. Statistics are weighted by 2012 respondent NLSY79 sampling weights. Boxes identify the group-specifc 25th percentile, median, and 75th percentile. Whiskers identify the 5th and 95th percentiles.

This content downloaded from �����������63.236.101.18 on Sun, 21 Jan 2024 14:41:17 +00:00������������

All use subject to https://about.jstor.org/terms

301Wealth, Race, and Place: Neighborhood (Dis)advan tage and Inequality

Overall, there is a mod est, neg a tive unweighted cor re la tion (R = –.280) between cumu la tive ND and the IHS of 2012 fam ily wealth. This cor re la tion varies sub stan­ tially by race. The cor re la tion is –.256 among Whites, com pared with only –.130 among Blacks, –.175 among His pan ics, and –.071 among other­race indi vid u als. Figure 3 pres ents weighted medi ans and interquartile ranges of IHS wealth for dif­ fer ent ranges of cumu la tive ND by race/eth nic ity. At each range of ND, Whites have gen er ally higher lev els of wealth than Blacks or His pan ics. Further, the wealth ben e ft of a reduc tion in ND appears to be mod estly larger at lower lev els of ND, par tic u larly for Whites and His pan ics.

Fig. 2 Boxplots of cumulative neighborhood disadvantage, by race. Statistics are weighted by 2012 respondent NLSY79 sampling weights. Boxes identify the group-specifc 25th percentile, median, and 75th percentile. Whiskers identify the 5th and 95th percentiles.

Fig. 3 Median and interquartile range of wealth, by cumulative neighborhood disadvantage and race/ ethnicity. Statistics are weighted by 2012 respondent NLSY79 sampling weights. Boxes identify the group-specifc 25th percentile, median, and 75th percentile. Values are omitted where the sample size within the range of cumulative neighborhood disadvantage is less than 50. Values for other­race individu­ als are omitted due to the overall small sample size. IHS = inverse hyperbolic sine.

This content downloaded from �����������63.236.101.18 on Sun, 21 Jan 2024 14:41:17 +00:00������������

All use subject to https://about.jstor.org/terms

302 B. L. Levy

Methods

To ana lyze fur ther the rela tion ship between ND and wealth, I adopt a coun ter fac tual approach designed to yield causal con clu sions. Traditional regres sion­con trol approaches for time-vary ing covariates can sig nif cantly bias esti ma tes of focal rela tion ships in an unknown direc tion and mag ni tude through overcontrolling and col lider strat i f ca tion (Wodtke et al. 2011). Inverse prob a bil ity of treat ment (IPT)–weighted mar ginal struc tural mod els (MSMs) solve both prob lems and per­ mit adjust ment for time­vary ing covariates (Robins 1998; Robins et al. 2000). IPT­weighted MSMs use two­stage pro pen sity score tech niques, which regress treat ment sta tus at each wave on the o ret i cally informed con trols in the frst-stage regres sion. By weighting obser va tions in the treat ment effect (sec ond­stage) regres­ sion by the inverse prob a bil ity that they received the observed treat ment sequence, IPT­weighted MSMs adjust for time­vary ing covariates with out requir ing their inclu sion in the sec ond­stage model.

I use a recent approach for IPT­weighted MSMs with a con tin u ous treat ment. In the frst stage, I divide treat ment into dec iles to esti mate IPT weights (IPTWs). Monte Carlo sim u la tions show the eff ciency of the IPTWs for use with a con tin u ous treat- ment in sec ond­stage mod els (Naimi et al. 2014). I pre dict treat ment dec ile using an ordered logit model. Results from the ordered logit pro vide the predicted prob a bil ity of being in each dec ile of treat ment at each wave con di tional on observed covariates, and I assign each per son­year the predicted prob a bil ity of expo sure cor re spond ing to the observed dec ile. Following Wodtke and col leagues (2011), I cal cu late sta bi lized IPTWs as fol lows:

SW =

pr(NDDt = nddt | NDDt−1,ND0 ,W0 ,Z0 ) pr(NDDt = nddt | NDDt−1,ND0 ,W0 ,Z0 ,Xt ,Xt−1)

t=1 T∏ .

(1)

The denom i na tor is the full pro pen sity score model esti mat ing treat ment dec ile (NDD) at time t based on lagged treat ment dec ile (NDDt−1), base line ND (ND0), base­ line IHS wealth (W0), time­invari ant covariates (Z0), time­vary ing covariates (Xt), and a one­wave lag of time­vary ing covariates (Xt−1).9 The numer a tor of Eq. (1), which excludes time­vary ing covariates, sta bi lizes the IPTWs. Recognizing racial inequalities in neigh bor hood attain ment (Massey and Denton 1993), I inter act all con trols in the numer a tor and denom i na tor with dummy var i ables iden ti fy ing Black and His panic indi vid u als to allow coef f cients to vary for each group. The cumu la- tive IPTW for an indi vid ual is the prod uct of each wave’s sta bi lized IPTW. I cen sor cumu la tive IPTWs at the 10th and 90th per cen tiles by recoding weights below and above those thresh olds to the rel e vant thresh old. This deci sion bal ances con cerns

9 I con trol for cur rent and one­wave lags of time­vary ing con trols to bal ance con cerns of endogeneity with those of data loss. Each addi tional year of lagged con trols elim i na tes one addi tional year of treat ment data owing to the lack of data on lagged con trols. For the wealth, home value, home debt, and home equity con­ trols, I use only lagged mea sures because con tem po ra ne ous val ues may be endog e nous to ND. I exclude lagged age given col lin ear ity with cur rent age.

This content downloaded from �����������63.236.101.18 on Sun, 21 Jan 2024 14:41:17 +00:00������������

All use subject to https://about.jstor.org/terms

303Wealth, Race, and Place: Neighborhood (Dis)advan tage and Inequality

about selec tion bias with those of extremely large IPTW var i ance (Cole and Hernán 2008).10

To ana lyze the rela tion ship between wealth and cumu la tive ND, I use the cumu la­ tive IPTWs esti mated with Eq. (1) as a prob a bil ity weight (mul ti plied with sam pling weights) to esti mate the fol low ing:

Wt = β0+β1

NDj t −1j=1

J = t−1∑ ⎛ ⎝⎜

⎞ ⎠⎟ +β2ND0 +β3W0 +βzZ0 + ε.

(2)

This model con trols for base line fam ily wealth, base line ND, and time­invari ant covariates—allowing those asso ci a tions to vary across three racial/eth nic groups (Blacks, His pan ics, and Whites/oth ers). IPTWs adjust for time-vary ing con trols. I also test qua dratic and cubic treat ment terms to allow the rela tion ship between cumu­ la tive ND and IHS wealth to be non lin ear. In addi tion, I explore racial var i a tion in the rela tion ship by interacting the treat ment with dummy var i ables for race/eth nic ity.11

The validity of causal con clu sions hinges on a key assump tion: no unmea sured confounding. All var i ables affect ing treat ment sta tus must be con trolled or highly cor re lated with con trols. Traditional regres sion mod els also require this assump tion, so IPTW MSMs are not unique in this respect. Regardless, it is a strong assump­ tion. To guard against selec tion bias, I include all key neigh bor hood selec tion var i­ ables iden ti fed by prior research: race/eth nic ity, income, edu ca tion, and liv ing in a female­headed house hold (Quillian 2003; Sampson and Sharkey 2008). I also con­ trol for lagged wealth to guard against reverse cau sa tion and con trol for base line ND to help adjust for unob served, time­invari ant con found ers con sis tently related to ND. Although no obser va tional study can dis count selec tion effects com pletely, the exten sive con trols included here help to min i mize endogeneity issues. Moreover, the frst-stage pro pen sity score model pre dicts ND rea son ably well (see sec tion C of the online appen dix).

After esti mat ing over all and race-spe cifc neigh bor hood effects, I inves ti gate effect mod er a tion by cumu la tive homeownership or expo sure to met ro pol i tan/micropolitan res i den tial seg re ga tion. The poten tial mod er a tors are time­vary ing, so I esti mate effect mod er a tion using IPT­weighted regres sion with resid u als (Wodtke and Almirall 2017). The IPTW cal cu la tion pro ceeds as described in Eq. (1) with the rel e vant mod­ er a tor (Mt, Mt−1) included in the numer a tor and denom i na tor. Then, I residualize the mod er a tor to purge its asso ci a tion with prior treat ment, as described in Eq. (3), which is weighted by the cumu la tive sta bi lized IPTW at time t:

δ̂Mt = Mt − Ê Mt | NDt−1,Wt−1,ND0 ,W0 ,Z0( ). (3)

10 Results are sub stan tively sim i lar if IPTWs are cen sored at the 5th/95th per cen tiles. IPTWs cen sored at the 1st/99th per cen tiles, how ever, have extreme var i ance, which pro duces impre cise, unsta ble esti ma tes. Supplemental ana ly ses not shown here indi cate that this results from the many waves of treat ment (14 for each indi vid ual) included in the anal y sis, poten tially in com bi na tion with non lin e ar ity or non ad di tiv ity in the pro pen sity score model for neigh bor hood dis ad van tage (Lee et al. 2011). 11 For neigh bor hood effects with racial/eth nic het ero ge ne ity, I inter pret predicted change in wealth for each group only for its group-spe cifc range (5th to 95th per cen tile) of ND.

This content downloaded from �����������63.236.101.18 on Sun, 21 Jan 2024 14:41:17 +00:00������������

All use subject to https://about.jstor.org/terms

304 B. L. Levy

After cre at ing residualized val ues of the mod er a tor, I esti mate mod er ated treat ment effects using the IPT-weighted regres sions with resid u als in Eq. (4):

Wt = β0 +β1 NDj t −1j=1

J = t−1∑ ⎛ ⎝⎜

⎞ ⎠⎟ +β2

NDj t −1j=1

J = t−1∑ ⎛ ⎝⎜

⎞ ⎠⎟

Mj

t −1j=1 J = t−1∑

⎛ ⎝⎜

⎞ ⎠⎟ +β3

δ̂Mj

t −1j=1 J = t−1∑

⎛

⎝ ⎜

⎞

⎠ ⎟

+β4ND0 +β5W0 +βzZ0 + ε. (4)

If treat ment effects are non lin ear or vary by race, I include the nec es sary terms and inter ac tions with the mod er a tor.

Finally, I explore the role of ND in the racial wealth gap. I esti mate a series of lin ear mod els that (1) describe over all, unad justed racial wealth inequal ity in 2012, (2) assess how wealth inequal ity changes when I adjust for time­invari ant and time­vary ing con trols, and (3) assess the extent to which wealth inequal ity can be explained by cumu la tive ND. These last mod els are not causal. Rather, they illus trate how ND and other var i ables may be salient for racial wealth inequal ity.

Results

Table 1 pres ents a sequence of mod els of IHS wealth at roughly age 50. Model 1 esti ma tes the uncon di tional, lin ear asso ci a tion between cumu la tive ND and wealth. Given the IHS trans for ma tion of wealth, param e ter esti ma tes are inter pret able as semi­elas tic i ties (Bellemare and Wichman 2020).12 A one­unit decrease in cumu la tive ND is asso ci ated with a 152% increase in wealth (−1 × –1.52 = 1.52 = 152%). Model 2 reveals mod est non lin e ar ity in the rela tion ship, which is stron gest at low lev els of ND.13 Table 2 pro vi des aver age adjusted predicted IHS wealth, based on the regres­ sion mod els in Table 1, at three illus tra tive val ues of cumu la tive ND: the 5th per cen­ tile, the median, and the 95th per cen tile (roughly −2.1, −0.1, and 2.9, respec tively, for the full sam ple). On the basis of model 2, an indi vid ual at the median of cumu la tive ND has approx i ma tely 420% more wealth than a sim i lar indi vid ual at the 95th per- cen tile of ND. Someone at the 5th per cen tile of ND has roughly 370% more wealth than a sim i lar indi vid ual at the median of ND.

Models 3 and 4 add time-invari ant con trols and their inter ac tions with race/eth nic ity. The asso ci a tion between cumu la tive ND and IHS wealth declines mod er ately but remains strong. Models 5 and 6 include time­invari ant con trols and add IPTWs to adjust for time­vary ing confounding. Time­vary ing con trols explain a larger share of the focal rela tion ship than do time­invari ant con trols. Nevertheless, the IPT­weighted MSMs fnd a neg a tive rela tion ship between cumu la tive ND and wealth that per sists at a sub stan tively impor tant mag ni tude. Model 6 indi cates that the over all rela tion ship is sig nif cantly non lin ear. Based on model 6, a reduc tion in cumu la tive ND from the median to the 5th per cen tile is asso ci ated with a 135% increase in wealth, whereas a

12 For raw wealth val ues above approx i ma tely $10, the IHS is essen tially equiv a lent to the nat u ral log plus a con stant (roughly 0.693). The IHS and nat u ral log trans for ma tions also are per fectly cor re lated (R = 1.0000). Roughly 8% of indi vid u als reporting wealth in 2012 reported val ues between –$10 and $10; of these, all but fve indi vid u als reported zero wealth. 13 A cubic term does not sub stan tively improve the model.

This content downloaded from �����������63.236.101.18 on Sun, 21 Jan 2024 14:41:17 +00:00������������

All use subject to https://about.jstor.org/terms

305Wealth, Race, and Place: Neighborhood (Dis)advan tage and Inequality

Table 1 Linear mod els of inverse hyper bolic sine (IHS) wealth by cumu la tive neigh bor hood dis ad van tage (ND) at approx i ma tely age 50

Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 Model 7 Model 8

ND −1.52*** −1.65*** −1.12*** −1.23*** −0.42*** −0.51*** −0.68*** −0.69*** (0.06) (0.07) (0.10) (0.11) (0.12) (0.14) (0.19) (0.19)

ND2 0.09*** 0.09*** 0.07* −0.02 (0.02) (0.02) (0.03) (0.08)

ND × Black 0.75** 0.60 (0.28) (0.43)

ND × His panic 0.28 −0.37 (0.31) (0.40)

ND × Other Race 0.95† 1.26* (0.56) (0.62)

ND2 × Black 0.06 (0.10)

ND2 × His panic 0.22* (0.11)

ND2 × Other Race −0.31 (0.22)

Constant 9.45*** 9.25*** 8.84*** 8.74*** 9.20*** 9.11*** 9.25*** 9.24*** (0.10) (0.12) (0.77) (0.78) (0.85) (0.86) (0.86) (0.88)

Time­Invariant Controls

x x x x x x

IPTWs x x x x

Notes: Standard errors are shown in paren the ses. All mod els are weighted using 2012 par tic i pant sam pling weights. N = approx i ma tely 7,300 for all mod els. IPTW = inverse prob a bil ity of treat ment weight. †p < .10; *p < .05; **p < .01; ***p < .001

reduc tion in ND from the 95th per cen tile to the median is asso ci ated with only a 92% increase in wealth (see Table 2). Figure 4 plots aver age adjusted predicted wealth from the 5th per cen tile to the 95th per cen tile of cumu la tive ND based on esti ma tes in model 6. ND is most strongly asso ci ated with wealth at low (neg a tive) lev els of ND. Changes in ND val ues between 0.1 and 2.9 are asso ci ated with sta tis ti cally indis tin­ guish able changes in wealth.

Model 7 fnds racial/eth nic het ero ge ne ity in esti mated neigh bor hood effects. Among Whites, a one­unit decrease in cumu la tive ND is asso ci ated with a 68% increase in wealth. This fg ure implies that for a White adult at the sam ple median of White wealth ($175,000), a one­unit reduc tion in ND increases wealth at age 50 by $119,000. Among His pan ics, the rela tion ship between ND and wealth is mod estly neg a tive, but Table 2 reveals that changes in ND from the group 5th per cen tile to the 95th per cen tile do not yield sta tis ti cally sig nif cant changes in wealth. Among Black and other-race adults, the asso ci a tion between ND and IHS wealth is essen tially null—or even mod­ estly, non sig nif cantly pos i tive (see Tables 1 and 2). Unlike the over all model, model 8 indi cates that there is not sig nif cant non lin e ar ity in the race-spe cifc esti mated neigh- bor hood effects. In fact, racial/eth nic het ero ge ne ity in the focal rela tion ship appears to explain the non lin e ar ity in over all esti ma tes. As Figure 2 shows, Whites are dis pro por­ tion ately con cen trated in advan taged neigh bor hoods. Because changes in ND sig nif - cantly impact only Whites’ wealth, ignor ing racial/eth nic het ero ge ne ity in the over all model yields an appar ent non lin ear rela tion ship that is stron gest at low lev els of ND.

This content downloaded from �����������63.236.101.18 on Sun, 21 Jan 2024 14:41:17 +00:00������������

All use subject to https://about.jstor.org/terms

306 B. L. Levy

Table 2 Predicted inverse hyper bolic sine (IHS) wealth by cumu la tive neigh bor hood dis ad van tage (ND) at approx i ma tely age 50 based on esti ma tes in Table 1

Model 1

Model 2

Model 3

Model 4

Model 5

Model 6

Model 7

Model 8

ND, Overall 5th per cen tile (−2.1) 12.64 13.10 11.78 12.21 10.26 10.62 (0.16) (0.20) (0.22) (0.27) (0.29) (0.38) Median (−0.1) 9.60 9.41 9.55 9.35 9.42 9.27 (0.10) (0.12) (0.10) (0.11) (0.13) (0.15) 95th per cen tile (2.9) 5.05 5.22 6.20 6.43 8.16 8.35

(0.20) (0.19) (0.33) (0.32) (0.37) (0.37) ND, Whites 5th per cen tile (−2.3) 11.18 11.16 (0.56) (0.60) Median (−0.5) 9.95 10.03 (0.41) (0.43) 95th per cen tile (1.3) 8.72 8.75

(0.50) (0.51) ND, Blacks 5th per cen tile (−0.7) 6.68 6.78 (1.82) (1.83) Median (1.9) 6.85 6.64 (1.85) (1.90) 95th per cen tile (4.9) 7.06 7.12

(2.04) (2.01) ND, His pan ics 5th per cen tile (−1.4) 10.64 11.54 (1.94) (1.91) Median (1.0) 9.67 8.80 (1.98) (2.02) 95th per cen tile (3.8) 8.55 8.54

(2.25) (2.24) ND, Other Race 5th per cen tile (−1.5) 8.22 7.46 (1.21) (1.40) Median (0.1) 8.66 9.11 (0.71) (0.73) 95th per cen tile (1.7) 9.09 9.07

(1.02) (0.96)

Notes: Parenthetical per cen tile val ues are approx i mate. Standard errors are shown in paren the ses. All esti­ ma tes are weighted using 2012 sam pling weights.

A related inter pre ta tion of these results is that neigh bor hood effects may operate as a spline func tion but that only Whites reside in advan taged neigh bor hoods that sig nif - cantly alter wealth accu mu la tion. It is not pos si ble to adju di cate defn i tively between these inter pre ta tions with the pres ent data.

Table 2 (model 7) shows that had a White adult at the group-spe cifc 95th per- cen tile of cumu la tive ND instead expe ri enced the group-spe cifc 5th per cen tile, they would have an esti mated 246% increase in wealth. The dol lar value of this increase depends on the indi vid ual’s observed wealth. Figure 5 plots the esti mated impact

This content downloaded from �����������63.236.101.18 on Sun, 21 Jan 2024 14:41:17 +00:00������������

All use subject to https://about.jstor.org/terms

307Wealth, Race, and Place: Neighborhood (Dis)advan tage and Inequality

of a 95th­to­5th per cen tile shift in ND for two wealth val ues: the group 25th per­ cen tile (panel a) and median (panel b). A White adult with $31,000 in wealth (panel a) who expe ri enced the group 95th per cen tile of cumu la tive ND would instead be expected to have $107,000 in wealth had they expe ri enced the 5th per cen tile of ND. A White adult with median wealth ($175,000), how ever, would instead be expected to have $603,000 in wealth had they expe ri enced the same change. Figure 5 reveals that Whites’ out sized wealth returns to reduc tions in ND result from two phe nom ena: higher over all wealth lev els and greater raw—and rel a tive—returns to reduc tions in ND. The results are robust to alter na tive approaches for miss ing data (see sec tion D of the online appen dix). A sen si tiv ity anal y sis also pro vi des evi dence that the neg­ a tive asso ci a tion between cumu la tive ND and wealth is likely to be causal among Whites (see sec tion E of the online appen dix).

Having found a sig nif cant rela tion ship between ND and wealth, I now exam ine poten tial effect mod er a tion. Table 3 pres ents IPT­weighted regres sions with resid u als ana lyz ing mod er a tion by cumu la tive homeownership (model 9) and cumu la tive met­ ro pol i tan/micropolitan seg re ga tion (model 10). Given the pre ced ing results, I allow the mod er ated rela tion ships to fur ther vary by race/eth nic ity. Although it misses the thresh old for sta tis ti cal sig nif cance, model 9 reveals sub stan tively impor tant mod er a tion of the rela tion ship between cumu la tive ND and wealth by cumu la tive homeownership. The asso ci a tion between ND and per cent age change in wealth is gen er ally stron ger among those never or rarely owning their home from 1985 to 2010.

Among Whites never owning, a one­unit decrease in ND is asso ci ated with a 142% increase in wealth by age 50. By con trast, among Whites always owning their

Fig. 4 Average adjusted predictions of overall wealth, by cumulative neighborhood disadvantage, based on estimates from model 6. Estimates span the 5th percentile to the 95th percentile of the distribution of cumulative neighborhood disadvantage. Estimates are combined across 10 multiple imputation data sets using Rubin’s rules. Error bars indicate 95% confdence intervals. IHS = inverse hyperbolic sine.

This content downloaded from �����������63.236.101.18 on Sun, 21 Jan 2024 14:41:17 +00:00������������

All use subject to https://about.jstor.org/terms

308 B. L. Levy

Fi g.

5 A

ve ra

ge a

dj us

te d

pr ed

ic te

d w

ea lth

g iv

en c

ha ng

e in

c um

ul at

iv e n

ei gh

bo rh

oo d

di sa

dv an

ta ge

fr om

th e

gr ou

p 95

th p

er ce

nt ile

to th

e gr

ou p

5t h

pe rc

en til

e ( ba

se d

on m

od el

7 ).

Th e

ra ce

-s pe

ci fc

2 5t

h pe

rc en

til e

an d

m ed

ia n

w ea

lth , r

es pe

ct iv

el y,

a re

a pp

ro xi

m at

el y

$3 1,

00 0

an d

$1 75

,0 00

fo r W

hi te

s, $0

a nd

$ 12

,0 00

fo r B

la ck

s, $5

00 a

nd $

48 ,0

00 fo

r H is

pa ni

cs , a

nd

$7 ,0

00 a

nd $

99 ,0

00 fo

r o th

er -r

ac e

in di

vi du

al s.

Th e

ra ce

-s pe

ci fc

5 th

a nd

9 5t

h pe

rc en

til es

o f c

um ul

at iv

e ne

ig hb

or ho

od d

is ad

va nt

ag e,

re sp

ec tiv

el y,

a re

a pp

ro xi

m at

el y

−2 .3

a nd

1 .3

fo

r W hi

te s,

−0 .7

a nd

4 .9

fo r B

la ck

s, −1

.4 a

nd 3

.8 fo

r H is

pa ni

cs , a

nd −

1. 5

an d

1. 7

fo r o

th er

­r ac

e in

di vi

du al

s. Es

tim at

es a

re c

om bi

ne d

ac ro

ss 1

0 m

ul tip

le im

pu ta

tio n

da ta

se ts

u si

ng

R ub

in ’s

ru le

s. Er

ro r b

ar s i

nd ic

at e

95 %

c on

fd en

ce in

te rv

al s f

or p

re di

ct ed

w ea

lth g

iv en

a c

ha ng

e in

n ei

gh bo

rh oo

d di

sa dv

an ta

ge .

This content downloaded from �����������63.236.101.18 on Sun, 21 Jan 2024 14:41:17 +00:00������������

All use subject to https://about.jstor.org/terms

309Wealth, Race, and Place: Neighborhood (Dis)advan tage and Inequality

Table 3 Inverse prob a bil ity of treat ment–weighted regres sion with resid u als esti mat ing neigh bor hood effect mod er a tion

Model 9 Model 10

ND −1.42*** −1.26** (0.36) (0.43) ND × Black 0.69† 0.92 (0.40) (0.78) ND × His panic 0.38 1.30†

(0.47) (0.73) ND × Other Race 1.24 2.67†

(1.09) (1.59) Effect Moderation ND × homeownership 0.72 (0.46) ND × homeownership × Black 1.29* (0.62) ND × homeownership × His panic 0.14 (0.68) ND × homeownership × other race −0.82 (1.77) ND × seg re ga tion 1.49 (1.53) ND × seg re ga tion × Black −0.83 (2.41) ND × seg re ga tion × His panic −3.72 (2.60) ND × seg re ga tion × other race −7.17 (6.01) Residualized Moderators δ̂ homeownership 5.09*** (0.57) δ̂ homeownership × Black −3.15* (1.42) δ̂ homeownership × His panic −1.32 (1.31) δ̂ homeownership × other race 1.31 (2.14) δ̂ seg re ga tion −0.60 (1.74) δ̂ seg re ga tion × Black −1.05 (5.32) δ̂ seg re ga tion × His panic −4.00 (5.19) δ̂ seg re ga tion × other race −2.29

(6.88) Constant 9.13*** 8.96*** (0.92) (0.91)

This content downloaded from �����������63.236.101.18 on Sun, 21 Jan 2024 14:41:17 +00:00������������

All use subject to https://about.jstor.org/terms

310 B. L. Levy

Model 9 Model 10

Time­Invariant Controls x x Main IPTWs x x Regression With Residuals x x

Notes: Both treat ment (ND) and mod er a tors (homeownership and seg re ga tion) are cumu la tive over the anal y sis win dow. Standard errors are shown in paren the ses. All mod els are weighted using 2012 par tic i pant sam pling weights. N = approx i ma tely 7,300 for all mod els. IPTW = inverse prob a bil ity of treat ment weight. †p < .10; *p < .05; **p < .01; ***p < .001

Table 3 (continued)

home, a one-unit decrease in ND is asso ci ated with a sta tis ti cally sig nif cant, although smaller, 70% increase in wealth (−1 × [–1.42 + 0.72] = 0.70). These asso ci a tions must, how ever, be interpreted in the con text of the siz able disparities in raw wealth by hous­ ing ten ure. Whites always owning their home have 25th per cen tile and median wealth of roughly $158,000 and $343,000, respec tively. Whites never owning their home have respec tive val ues of $0 and $3,200. Mean wealth totals reveal siz able inequal ity as well: $607,000 among Whites always owning ver sus $59,000 among Whites never owning their home. Thus, although the per cent age increase in wealth asso ci ated with reduc tions in ND appears sub stan tively larger among White non­homeowners, the esti mated impact on raw wealth is sub stan tially greater among White long­term homeowners. A White long­term homeowner with group median wealth could expect to gain $240,000 more in wealth with a one-unit reduc tion in ND. A White adult never owning their home and hold ing group median wealth could expect to gain only $4,500 more in wealth with a one-unit reduc tion in ND.

Among His pan ics never owning a home, ND neg a tively relates to wealth. A one-unit decrease in cumu la tive ND is expected to increase wealth by 104% (−1 × [–1.42 + 0.38] =1.04). The median wealth of His panic adults never owning their home is $1,500; so, although dou bling this value is not insig nif cant, it is not trans for ma tive. For His pan ics always owning their home, a one­unit reduc tion in ND is asso ci ated with a non sig nif cant 18% increase in wealth (−1 × [–1.42 + 0.38 + 0.72 + 0.14] = 0.18). Among other-race indi vid u als, ND is not sig nif cantly related to wealth regard less of homeownership sta tus.

For Black adults, the rela tion ship between cumu la tive ND and wealth varies sig­ nif cantly by homeownership sta tus. Among Black non-homeowners, sam ple median and mean wealth are $0 and $18,000, respec tively, so even the mar gin ally sig nif - cant 73% increase in wealth asso ci ated with a one­unit reduc tion in ND is sub stan­ tively quite small for many indi vid u als. Among Black long­term homeowners—a small sub set of the sam ple—a one­unit decrease in ND is asso ci ated with a 1.27­unit reduc tion in IHS wealth (−1 × [–1.42 + 0.69 + 0.72 + 1.29] = −1.27). This result slightly misses tra di tional sig nif cance thresh olds (it is sig nif cant at p < .2). Still, a 127% increase in wealth per one­unit increase in ND for Black adults always owning their home would be sub stan tively impor tant given that their median wealth is $64,000. I return to this poten tially sur pris ing result in the next sec tion.

Model 10 explores mod er a tion by cumu la tive expo sure to racial res i den tial seg­ re ga tion. For con text, the sam ple 5th and 95th per cen tiles of cumu la tive seg re ga tion

This content downloaded from �����������63.236.101.18 on Sun, 21 Jan 2024 14:41:17 +00:00������������

All use subject to https://about.jstor.org/terms

311Wealth, Race, and Place: Neighborhood (Dis)advan tage and Inequality

Table 4 Linear mod els of inverse hyper bolic sine (IHS) wealth at approx i ma tely age 50 explor ing the racial wealth gap

Model 11 Model 12 Model 13 Model 14 Model 15

Black −4.57*** −2.72*** −1.05*** −1.41*** −2.37*** (0.23) (0.32) (0.31) (0.33) (0.34)

His panic −2.75*** −0.93* 0.04 −0.64* −0.76* (0.29) (0.47) (0.42) (0.32) (0.33)

Other Race −1.40** −0.66 −0.08 −0.69 −0.58 (0.48) (0.48) (0.41) (0.48) (0.48)

ND −1.33*** −1.67*** (0.08) (0.12)

ND × Black 0.91*** (0.17)

ND × His panic 0.58** (0.19)

ND × Other Race 0.75 (0.49)

Constant 10.32*** 8.89*** 1.01 9.73*** 9.58*** (0.13) (0.58) (4.62) (0.14) (0.15)

Time­Invariant Controls x x Time­Varying Controls x

Notes: ND = cumu la tive neigh bor hood dis ad van tage. Standard errors are shown in paren the ses. All mod els are weighted using 2012 par tic i pant sam pling weights. N = approx i ma tely 7,300 for all mod els.

*p < .05; **p < .01; ***p < .001

are 0.08 and 0.46, respec tively. Unlike the mod er a tion by homeownership, I fnd no evi dence of sig nif cant mod er a tion by seg re ga tion expo sure. Estimates sug gest that the neigh bor hood–wealth rela tion ship could be stron ger for Whites in places with low seg re ga tion. By con trast, this asso ci a tion might be stron ger for His pan ics in highly seg re gated areas, although there is siz able uncer tainty in this esti mate. This model also con trols for residualized seg re ga tion expo sure, which is not sig nif cantly asso ci ated with wealth. Its nonsignifcance sug gests that seg re ga tion does not cause wealth disparities per se or explain the impact of ND.

Finally, the ordi nary least squares mod els in Table 4 explore the role of ND in the racial wealth gap. Note that IHS wealth remains the depen dent var i able, reduc ing the influ ence of out lier wealth val ues dis pro por tion ately held by Whites. Model 11 dem on strates that Blacks, His pan ics, and other-race adults have sig nif cantly lower uncon di tional IHS wealth than Whites. Model 12 adds all time­invari ant con trol var i ables, includ ing base line wealth and ND; the groups’ wealth gaps with Whites decline by one half to two thirds. Model 13 adds time­vary ing con trols from the 2008 and 2010 waves to model 12. The His panic–White and other–White gaps essen tially dis ap pear, but the Black–White wealth gap remains sig nif cant. Time-invari ant and time­vary ing con trols explain only 77% of the Black–White gap.

Model 14 adds only the mea sure of cumu la tive ND to model 11. This model explains 69% of the Black–White wealth gap, as well as 77% and 51% of the His panic–White and other–White wealth gaps, respec tively. Thus, ND explains at least as much of the

This content downloaded from �����������63.236.101.18 on Sun, 21 Jan 2024 14:41:17 +00:00������������

All use subject to https://about.jstor.org/terms

312 B. L. Levy

White–non­White wealth gaps as all time­invari ant con trols together. Moreover, ND explains nearly as much of the Black–White wealth gap as all con trols. These results do not imply that ND causes this siz able por tion of the gaps; other covariates cer­ tainly play a role. Still, that such a large por tion of the gaps can be explained by one var i able is remark able. Model 15 allows the rela tion ship between ND and wealth to vary by race/eth nic ity. The His panic–White and other–White wealth gaps observed in model 14 are largely unchanged. Yet, the Black–White wealth gap grows by more than a point, reit er at ing Blacks’ unequal returns to neigh bor hood advan tage.

Exploratory Analysis of Home Values for Black Homeowners

Model 9 indi cates a poten tial coun ter in tu i tive, pos i tive asso ci a tion between cumu la­ tive ND and wealth among Black long­term homeowners. The small sam ple size of Black long-term homeowners (roughly 50) might explain this fnd ing’s nonsignif­ cance at tra di tional thresh olds. Alternatively, the asso ci a tion may be an arti fact of the spe cifc sam ple. Given the sub stan tive mag ni tude of the asso ci a tion, I inves ti gate it fur ther here—spe cif cally, the pos si bil ity for dif fer en tial appre ci a tion of home val ues.

I explore how the poten tial for appre ci a tion relates to ND for Black ver sus White homeowners using the 1990 decen nial cen sus and the fve-year ACS cen tered on 2012. I esti mate tract­level lin ear regres sions for the bivar i ate rela tion ships between 1990 ND and two out comes: (1) the raw increase in median home val ues from 1990 to 2012 (real 2012 dol lars, adjusted using the con sumer price index) and (2) the per cent age change in median home val ues from 1990 to 2012 (real dol lars).14 I spec ify a cubic poly no mial for ND to allow for non lin e ar ity and esti mate each regres sion twice using dif fer ent weights: (1) the tract’s num ber of homes owned by a non­Hispanic Black house holder in 1990 and (2) the tract’s num ber of homes owned by a non­Hispanic White house holder in 1990. With weighting, these esti­ mate the aver age rela tion ship between ND and raw/rel a tive appre ci a tion for (1) neigh bor hoods in which Black homeowners reside and (2) those in which White homeowners reside, thus reflecting poten tial wealth gain if homeowners remain in their neigh bor hoods.15 These mod els are explor atory and should be interpreted cau tiously given the poten tial for moves over a 20­year period and the exis tence of racial disparities in home val ues within a neigh bor hood.

Figure 6 plots aver age predicted raw appre ci a tion (panel a) and rel a tive appre ci a­ tion (panel b) based on these mod els (see sec tion F of the online appen dix for regres­ sion coef f cients). Plots range from the weighted 5th to 95th per cen tiles of 1990

14 Raw appre ci a tion is not heavily skewed. To min i mize out lier bias, I recode raw appre ci a tion val ues below or above the 1st and 99th per cen tiles, respec tively, to those val ues. I also recode rel a tive per cent age appre ci a tion val ues above 10 (1,000%) to 10. Findings for raw appre ci a tion are not sen si tive to this deci­ sion, and fnd ings for rel a tive appre ci a tion are sub stan tively sim i lar using alter na tive recode thresh olds (2.5, 5, or 20). 15 Although this anal y sis occurs at a dif fer ent level (neigh bor hood) than the main mod els, these data pro­ vide far greater sta tis ti cal power to detect how poten tial appre ci a tion for Black homeowners might vary by ND. Long­term homeowners in the NLSY79 would be homeowners in 1990.

This content downloaded from �����������63.236.101.18 on Sun, 21 Jan 2024 14:41:17 +00:00������������

All use subject to https://about.jstor.org/terms

313Wealth, Race, and Place: Neighborhood (Dis)advan tage and Inequality

Fig. 6 Average predicted tract­level future median home appreciation from 1990 to 2012, by 1990 neigh­ borhood disadvantage for Black homeowners and White homeowners. Predictions occur at race-specifc homeowners’ (weighted) 5th to 95th percentiles of 1990 neighborhood disadvantage at increments of 5 percentiles (i.e., 5th, 10th, 15th . . . 90th, 95th percentiles). Estimates are based on regression models shown in section F of the online appendix.

This content downloaded from �����������63.236.101.18 on Sun, 21 Jan 2024 14:41:17 +00:00������������

All use subject to https://about.jstor.org/terms

314 B. L. Levy

ND for Black and White homeowners. For both groups, poten tial raw appre ci a tion from 1990 to 2012 is neg a tively related to ND, but this rela tion ship is much stron­ ger among White homeowners’ neigh bor hoods. Among Black homeowners in 1990, those liv ing in neigh bor hoods at the race- and homeownership-spe cifc 30th per cen- tile of ND would have seen neigh bor hood median appre ci a tion by 2012 of just a few thou sand dol lars more than those liv ing in neigh bor hoods at the 95th per cen tile of dis ad van tage. This fnd ing explains why rel a tive appre ci a tion varies so strongly by ND for Black homeowners’ 1990 neigh bor hoods. Home val ues tend to be lower as ND increases; thus, whereas predicted rel a tive appre ci a tion is roughly 25% in neigh­ bor hoods at the 30th per cen tile of dis ad van tage among 1990 Black homeowners, it is over 42% in neigh bor hoods at the 95th per cen tile. Among White homeowners’ 1990 neigh bor hoods, rel a tive appre ci a tion increases only mod estly with ND. Panel b also reveals that where 1990 ND scores over lap for Black and White homeowners, rel a­ tive appre ci a tion is con sis tently higher in White homeowners’ neigh bor hoods. Only in highly dis ad van taged neigh bor hoods—where 1990 White homeowners rarely reside—do Black homeowners see rel a tive appre ci a tion outpace that among White homeowners’ neigh bor hoods. This pat tern of results is con sis tent with dif fer en tial appre ci a tion playing a role in the observed siz able, pos i tive (though non sig nif cant) asso ci a tion between cumu la tive ND and wealth for Black long­term homeowners.

Conclusion

This research adds to a robust lit er a ture exam in ing neigh bor hood effects on life chances. Disadvantaged neigh bor hoods have well­documented neg a tive effects on edu ca tional, eco nomic, and behav ioral out comes early in the life course (Chetty et al. 2016; Sharkey and Faber 2014; Wodtke 2013; Wodtke et al. 2011). The gen eral focus on how neigh bor hoods affect chil dren and youth leaves open ques tions regard ing how neigh bor hoods affect adults, on whom there are com par a tively few stud ies (e.g., Osterman 1991; Vartanian and Buck 2005). The pres ent research com ple ments and extends prior work by pro vid ing, to my knowl edge, the frst rig or ous quan ti ta tive anal y sis of neigh bor hood effects on wealth.

I ana lyze the impact of neigh bor hood dis ad van tage from emerg ing adult hood to mid dle adult hood on wealth at age 50, argu ing that neigh bor hoods play a crit i cal role in wealth accu mu la tion after account ing for reverse cau sa tion and selec tion effects. Lower ND is asso ci ated with sig nif cantly greater wealth accu mu la tion, but the ben- e fts are con cen trated among Whites, espe cially White homeowners. Among White adults con tin u ously owning their home in this study, the median indi vid ual holds $343,000 in wealth at age 50; had this indi vid ual expe ri enced a one-unit reduc tion in cumu la tive ND, which is equiv a lent to going from median ND to 20th per cen tile ND, they would instead have an esti mated $583,000 in wealth—a 70% increase. A sen si tiv ity anal y sis sug gests that neigh bor hood effects on Whites’ wealth are poten­ tially causal. Black and His panic adults who rarely or never own homes do see wealth ben e fts from reduc tions in ND, but these ben e fts are mod est given the groups’ over all low lev els of wealth. Perhaps sur pris ingly, Black long­term homeowners have a poten­ tially strong, pos i tive asso ci a tion between cumu la tive ND and wealth, although the asso ci a tion misses tra di tional sig nif cance thresh olds and is based on a small sam ple.

This content downloaded from �����������63.236.101.18 on Sun, 21 Jan 2024 14:41:17 +00:00������������

All use subject to https://about.jstor.org/terms

315Wealth, Race, and Place: Neighborhood (Dis)advan tage and Inequality

An explor atory anal y sis sug gests that stron ger rel a tive home appre ci a tion for Black homeowners in dis ad van taged neigh bor hoods might play a role in such a rela tion ship.

This study’s fnd ings are unlikely to be unique to the NLSY79 cohort. Residential seg re ga tion remains pro nounced (Lee et al. 2014; Reardon et al. 2018), and homes are a key source of wealth (Kuebler 2013; Shapiro 2006). Thus, neigh bor hoods will con tinue to be an impor tant aspect of U.S. wealth inequal ity. One advan tage of using the NLSY79 is its rep re sen ta tive sam ple of His pan ics. Future research could extend this anal y sis to other cohorts using alter na tive sur veys. Given the var i a tion in expla­ na tions for racial wealth inequal ity across the wealth dis tri bu tion (Maroto 2016), research ana lyz ing how neigh bor hood effects vary across the wealth dis tri bu tion would also be worth while. Future work might also explore mech a nisms for neigh bor­ hood effects on wealth, the dynamic rela tion ship between neigh bor hood con di tions and wealth across the life course, and how ND spe cif cally affects Black long-term homeowners.

These ques tions not with stand ing, the pres ent results impli cate meso­level neigh bor hood con di tions as a key aspect of U.S. wealth inequal ity and the racial wealth gap. Research on wealth inequal ity, which is char ac ter ized by macro­ and micro­level expla na tions (Keister 2005; Keister and Moller 2000; Semyonov and Lewin­Epstein 2013), would be enriched by inte grat ing such meso­level fea tures. Analyses of the racial wealth gap iden tify dif fer ences in homeownership, edu ca tional attain ment, income, and inter gen er a tional resources as key driv ers (Campbell and Kaufman 2006; Herring and Henderson 2016; Keister 2000; Killewald and Bryan 2018; Oli ver and Shapiro 1995). This study quantifes the impor tance of ND. Elim­ inating disparities in ND and equal iz ing returns to neigh bor hood con di tions would reduce the racial wealth gap mark edly. Blacks are dou bly dis ad van taged through expo sure to ND and returns to neigh bor hood advan tage. By con trast, Whites are the pri mary ben e f cia ries of these neigh bor hood-based inequalities.

The rac ist his tory of U.S. hous ing has linked wealth, race, and place over many decades (Conley 2010; Oli ver and Shapiro 1995). Exclusionary prac tices through the mid­twen ti eth cen tury, such as redlining and urban renewal, restricted Black Amer i cans’ abil ity to accu mu late wealth while advan tag ing Whites (Faber 2020; Lipsitz and Oli ver 2010). Since the mid­twen ti eth cen tury, a “pred a tory inclu sion” of Black Amer i cans in the hous ing mar ket has focused on the extrac tion of Black wealth via higher inter est rates, subprime mort gages and other exploit ative loans or con tracts, and fore clo sure disparities (Taylor 2019; see also Hall et al. 2015a; Korver­Glenn 2021; Rugh et al. 2015). This his tory aligns with the inter gen er a tional her i ta bil ity of wealth that Pfeffer and Killewald (2018) observed, as well as the dis pa rate returns to neigh bor hood advan tage for Whites found in this study.

Given these facts, a com monly touted strat egy for alle vi at ing the racial wealth gap— increas ing minor ity homeownership (Asante­Muhammad et al. 2016; Shapiro 2006; Shapiro et al. 2013; Sullivan et al. 2015)—may be constrained in its poten tial effec­ tive ness. Integrating low­income and non­White pop u la tions into more­advan taged neigh bor hoods can be ben e f cial for achieved sta tus (Chetty et al. 2020; Chetty et al. 2016), but inte gra tion efforts can also val o rize Whiteness and the mid dle class as ideal while neglecting the needs of low­income and non­White pop u la tions (Pattillo 2009). When inte gra tion occurs through gen tri f ca tion of low-income, non- White neigh bor hoods, it often leads to the cul tural and polit i cal disempowerment

This content downloaded from �����������63.236.101.18 on Sun, 21 Jan 2024 14:41:17 +00:00������������

All use subject to https://about.jstor.org/terms

316 B. L. Levy

of remaining long­term res i dents (e.g., Hyra 2017). Recognizing Whites’ per sis tent pref er ence for racial homophily (Krysan et al. 2009), any deseg re ga tion efforts must ensure that Blacks, His pan ics, and other non­White groups have pri macy in trans lat­ ing neigh bor hoods and homes into wealth.

Independent of deseg re ga tion, invest ment in low­wealth com mu ni ties is crit i cal, espe cially in pre dom i nantly non­White neigh bor hoods (Sharkey 2013). Obama­era place­based mod els, such as Choice Neighborhoods and Promise Neighborhoods, could be use ful, although they would require sig nif cantly greater funding and scale. Targeted mort gage assis tance and loan for give ness could fur ther reme di­ ate the last ing effects of rac ism in the hous ing mar ket. More gen er ally, how ever, ade quately addressing the per va sive ness of racial inequal ity resulting from ongo­ ing struc tural rac ism and the his tory of slav ery, racial ter ror, and Jim Crow likely requires rep a ra tions. Even with sig nif cant pol icy inter ven tion, these prob lems will not be solved soon. Residential seg re ga tion and disparities in neigh bor hood con­ di tions remain dra matic, and they will affect gen er a tions of Amer i cans for years to come. ■

Acknowledgments This research received sup port from a Dissertation Completion Fellowship awarded by the Graduate School at the University of North Carolina at Chapel Hill and a post doc toral fel low ship supported by the National Science Foundation (grant SES-1637136), as well as the Population Research Training grant (T32 HD007168) and the Population Research Infrastructure Program grant (R24 HD050924) awarded to the Carolina Population Center at the University of North Carolina at Chapel Hill by the Eunice Kennedy Shriver National Institute of Child Health and Human Development. The author thanks Kathleen Mullan Harris, Robert J. Sampson, Kyle Crowder, Guang Guo, Douglas Lee Lauen, Erica Meade, Ted Mouw, the Harvard Sociology Urban Data Lab, and sev eral anon y mous review ers for their help ful com ments on early ver sions of this man u script.

References

Anacker, K. B. (2010). Still pay ing the race tax? Analyzing prop erty val ues in homog e nous and mixed­race sub urbs. Journal of Urban Affairs, 32, 55–77.

Asante­Muhammad, D., Collins, C., Hoxie, J., & Nieves, E. (2016). The ever-grow ing gap: Without change, Afri can-Amer i can and Latino fam i lies won’t match White wealth for cen tu ries (Report). Washington, DC: Institute for Policy Studies, CFED.

Bellemare, M. F., & Wichman, C. J. (2020). Elasticities and the inverse hyper bolic sine trans for ma tion. Oxford Bulletin of Economics and Statistics, 82, 50–61.

Bricker, J., Dettling, L. J., Henriques, A., Hsu, J. W., Moore, K. B., Sabelhaus, J., . . . Windle, R. A. (2014). Changes in U.S. fam ily finances from 2010 to 2013: Evidence from the Survey of Consumer Finances (Federal Reserve Bulletin, Vol. 100 No. 4). Retrieved from https: / /www .federalreserve .gov /pubs / bulletin /2014 /pdf /scf14 .pdf

Bucks, B., & Pence, K. (2006). Do homeowners know their house val ues and mort gage terms? (FEDS Working Paper No. 2006­03). Washington, DC: Board of Governors of the Federal Reserve System.

Campbell, L. A., & Kaufman, R. L. (2006). Racial dif fer ences in house hold wealth: Beyond Black and White. Research in Social Stratification and Mobility, 24, 131–152.

Chetty, R., Hendren, N., Jones, M. R., & Porter, S. R. (2020). Race and eco nomic oppor tu nity in the United States: An inter gen er a tional per spec tive. Quarterly Journal of Economics, 135, 711–783.

Chetty, R., Hendren, N., & Katz, L. F. (2016). The effects of expo sure to bet ter neigh bor hoods on chil­ dren: New evi dence from the Moving to Opportunity exper i ment. Amer i can Economic Review, 106, 855–902.

This content downloaded from �����������63.236.101.18 on Sun, 21 Jan 2024 14:41:17 +00:00������������

All use subject to https://about.jstor.org/terms

317Wealth, Race, and Place: Neighborhood (Dis)advan tage and Inequality

Coate, D., & Schwester, R. W. (2011). Black­White appre ci a tion of owner­occu pied homes in upper income sub ur ban inte grated com mu ni ties: The cases of Maplewood and Montclair, New Jersey. Journal of Housing Research, 20, 127–139.

Cole, S. R., & Hernán, M. A. (2008). Constructing inverse prob a bil ity weights for mar ginal struc tural mod­ els. Amer i can Journal of Epidemiology, 168, 656–664.

Collins, C., & Hoxie, J. (2015). Billionaire bonanza report: The Forbes 400 and the rest of us (Report). Washington, DC: Institute for Policy Studies.

Conley, D. (1999). Getting into the Black: Race, wealth, and pub lic pol icy. Political Science Quarterly, 114, 595–612.

Conley, D. (2001). Capital for col lege: Parental assets and post sec ond ary school ing. Sociology of Education, 74, 59–72.

Conley, D. (2010). Being Black, liv ing in the red: Race, wealth, and social pol icy in America. Berkeley: University of California Press.

Crowder, K., & South, S. J. (2008). Spatial dynam ics of White flight: The effects of local and extralocal racial con di tions on neigh bor hood out­migra tion. Amer i can Sociological Review, 73, 792–812.

Denton, N. A. (2001). The role of res i den tial seg re ga tion in pro mot ing and maintaining inequal ity in wealth and prop erty. Indiana Law Review, 34, 1199–1211.

Dettling, L. J., Hsu, J. W., Jacobs, L., Moore, K. B., & Thompson, J. P. (2017). Recent trends in wealth- hold ing by race and eth nic ity: Evidence from the Survey of Consumer Finances (FEDS Notes report). Washington, DC: Board of Governors of the Federal Reserve System. https: / /doi .org /10 .17016 /2380 ­7172 .2083

Di, Z. X., Belsky, E., & Liu, X. (2007). Do homeowners achieve more house hold wealth in the long run? Journal of Housing Economics, 16, 274–290.

Du Bois, W. E. B. (1899). The Philadelphia Negro. Philadelphia: University of Pennsylvania Press. Emerson, M. O., Yancey, G., & Chai, K. J. (2001). Does race mat ter in res i den tial seg re ga tion? Exploring

the pref er ences of White Amer i cans. Amer i can Sociological Review, 66, 922–935. Faber, J. W. (2020). We built this: Consequences of New Deal era inter ven tion in America’s racial geog ra­

phy. Amer i can Sociological Review, 85, 739–775. Faber, J. W., & Ellen, I. G. (2016). Race and the hous ing cycle: Differences in home equity trends among

long­term homeowners. Housing Policy Debate, 26, 456–473. Fischer, M. J., & Lowe, T. S. (2014). Homebuyer neigh bor hood attain ment in Black and White: Housing

out comes dur ing the hous ing boom and bust. Social Forces, 93, 1481–1512. Flippen, C. (2004). Unequal returns to hous ing invest ments? A study of real hous ing appre ci a tion among

Black, White, and His panic house holds. Social Forces, 82, 1523–1551. Flippen, C. (2010). The spa tial dynam ics of strat i f ca tion: Metropolitan con text, pop u la tion redis tri bu tion,

and Black and His panic homeownership. Demography, 47, 845–868. Galster, G. C., Cutsinger, J. M., & Malega, R. (2008). The costs of con cen trated pov erty: Neighborhood

prop erty mar kets and the dynam ics of decline. In N. P. Retsinas & E. S. Belsky (Eds.), Revisiting rental hous ing: Policies, pro grams, and pri or i ties (pp. 93–113). Cambridge, MA: Joint Center for Housing Studies, Harvard University; Washington, DC: Brookings Institution Press.

Galster, G. C., Tatian, P., & Smith, R. (1999). The impact of neigh bors who use Section 8 cer tif cates on prop erty val ues. Housing Policy Debate, 10, 879–917.

Hall, M., Crowder, K., & Spring, A. (2015a). Neighborhood fore clo sures, racial/eth nic tran si tions, and res i den tial seg re ga tion. Amer i can Sociological Review, 80, 526–549.

Hall, M., Crowder, K., & Spring, A. (2015b). Variations in hous ing fore clo sures by race and place, 2005– 2012. Annals of the Amer i can Academy of Political and Social Science, 660, 217–237.

Harding, D. J. (2003). Counterfactual mod els of neigh bor hood effects: The effect of neigh bor hood pov erty on drop ping out and teen age preg nancy. Amer i can Journal of Sociology, 109, 676–719.

Harris, D. R. (1999). “Property val ues drop when Blacks move in, because . . .”: Racial and socio eco nomic deter mi nants of neigh bor hood desir abil ity. Amer i can Sociological Review, 64, 461–479.

Herring, C., & Henderson, L. (2016). Wealth inequal ity in Black and White: Cultural and struc tural sources of the racial wealth gap. Race and Social Problems, 8, 4–17.

Hipp, J. R., Petersilia, J., & Turner, S. (2010). Parolee recid i vism in California: The effect of neigh bor hood con text and social ser vice agency char ac ter is tics. Criminology, 48, 947–979.

Howell, J., & Korver­Glenn, E. (2021). The increas ing effect of neigh bor hood racial com po si tion on hous­ ing val ues, 1980–2015. Social Problems, 68, 1051–1071.

This content downloaded from �����������63.236.101.18 on Sun, 21 Jan 2024 14:41:17 +00:00������������

All use subject to https://about.jstor.org/terms

318 B. L. Levy

Hyra, D. S. (2017). Race, class, and pol i tics in the Cappuccino City. Chicago, IL: University of Chicago Press. Jargowsky, P. A. (1996). Take the money and run: Economic seg re ga tion in U.S. met ro pol i tan areas.

Amer i can Sociological Review, 61, 984–998. Keister, L. A. (2000). Race and wealth inequal ity: The impact of racial dif fer ences in asset own er ship on

the dis tri bu tion of house hold wealth. Social Science Research, 29, 477–502. Keister, L. A. (2005). Getting rich: America’s new rich and how they got that way. New York, NY:

Cambridge University Press. Keister, L. A., & Moller, S. (2000). Wealth inequal ity in the United States. Annual Review of Sociology,

26, 63–81. Kiel, K. A., & Zabel, J. E. (1999). The accu racy of owner­pro vided house val ues: The 1978–1991 Amer i can

Housing Survey. Real Estate Economics, 27, 263–298. Killewald, A., & Bryan, B. (2016). Does your home make you wealthy? Russell Sage Foundation Journal

of the Social Sciences, 2(6), 110–128. Killewald, A., & Bryan, B. (2018). Falling behind: The role of inter­ and intragenerational pro cesses in wid­

en ing racial and eth nic wealth gaps through early and mid dle adult hood. Social Forces, 97, 705–740. Killewald, A., Pfeffer, F. T., & Schachner, J. N. (2017). Wealth inequal ity and accu mu la tion. Annual

Review of Sociology, 43, 379–404. Korver­Glenn, E. (2021). Race bro kers: Housing mar kets and seg re ga tion in 21st cen tury urban America.

New York, NY: Oxford University Press. Krivo, L. J., & Kaufman, R. L. (2004). Housing and wealth inequal ity: Racial-eth nic dif fer ences in home

equity in the United States. Demography, 41, 585–605. Krysan, M., Couper, M. P., Farley, R., & Forman, T. A. (2009). Does race mat ter in neigh bor hood pref er­

ences? Results from a video exper i ment. Amer i can Journal of Sociology, 115, 527–559. Kuebler, M. (2013). Closing the wealth gap: A review of racial and eth nic inequalities in homeownership.

Sociology Compass, 7, 670–685. Lee, B. A., Iceland, J., & Farrell, C. R. (2014). Is ethnoracial inte gra tion on the rise? Evidence from met ro-

pol i tan and micropolitan America since 1980. In J. R. Logan (Ed.), Diversity and disparities: America enters a new cen tury (pp. 415–456). New York, NY: Russell Sage Foundation.

Lee, B. A., Mat thews, S. A., Iceland, J., & Firebaugh, G. (2015). Residential inequal ity: Orientation and over view. Annals of the Amer i can Academy of Political and Social Science, 660, 8–16.

Lee, B. K., Lessler, J., & Stu art, E. A. (2011). Weight trim ming and pro pen sity score weighting. PLoS One, 6, e18174. https: / /doi .org /10 .1371 /journal .pone .0018174

Levy, B. L. (2019). Heterogeneous impacts of con cen trated pov erty dur ing ado les cence on col lege out­ comes. Social Forces, 98, 147–182.

Lipsitz, G., & Oli ver, M. L. (2010). Integration, seg re ga tion, and the racial wealth gap. In C. Hartman & G. D. Squires (Eds.), The inte gra tion debate: Competing futures for America cit ies (pp. 153–167). New York, NY: Routledge.

Logan, J. R., Xu, Z., & Stults, B. J. (2014). Interpolating U.S. decen nial cen sus tract data from as early as 1970 to 2010: A lon gi tu di nal tract data base. Professional Geographer, 66, 412–420.

Logan, J. R., Zhang, W., Turner, R., & Shertzer, A. (2015). Creating the Black ghetto: Black res i den tial pat terns before and dur ing the Great Migration. Annals of the Amer i can Academy of Political and Social Science, 660, 18–35.

Ludwig, J., Duncan, G. J., Gennetian, L. A., Katz, L. F., Kessler, R. C., Kling, J. R., & Sanbonmatsu, L. (2012). Neighborhood effects on the long­term well­being of low­income adults. Science, 337, 1505–1510.

Maroto, M. (2016). Growing far ther apart: Racial and eth nic inequal ity in house hold wealth across the dis tri bu tion. Sociological Science, 3, 801–824.

Massey, D. S., & Denton, N. A. (1993). Amer i can apart heid: Segregation and the mak ing of the under- class. Cambridge, MA: Harvard University Press.

Moore, L. V., Diez Roux, A. V., Evenson, K. R., McGinn, A. P., & Brines, S. J. (2008). Availability of rec­ re a tional resources in minor ity and low socio eco nomic sta tus areas. Amer i can Journal of Preventive Medicine, 34, 16–22.

Morland, K., Wing, S., Diez Roux, A., & Poole, C. (2002). Neighborhood char ac ter is tics asso ci ated with the loca tion of food stores and food ser vice places. Amer i can Journal of Preventive Medicine, 22, 23–29.

This content downloaded from �����������63.236.101.18 on Sun, 21 Jan 2024 14:41:17 +00:00������������

All use subject to https://about.jstor.org/terms

319Wealth, Race, and Place: Neighborhood (Dis)advan tage and Inequality

Naimi, A. I., Moodie, E. E. M., Auger, N., & Kaufman, J. S. (2014). Constructing inverse prob a bil ity weights for con tin u ous expo sures: A com par i son of meth ods. Epidemiology, 25, 292–299.

Newman, S. J., & Holupka, C. S. (2016). Is tim ing every thing? Race, homeownership and net worth in the tumul tu ous 2000s. Real Estate Economics, 44, 307–354.

Oli ver, M. L., & Shapiro, T. M. (1995). Black wealth/White wealth: A new per spec tive on racial inequal ity. New York, NY: Routledge.

Oli ver, M. L., & Shapiro, T. M. (2006). Black wealth/White wealth: A new per spec tive on racial inequality (Tenth­anniversary ed.). New York, NY: Routledge.

Osterman, P. (1991). Welfare par tic i pa tion in a full employ ment econ omy: The impact of neigh bor hood. Social Problems, 38, 475–491.

Pager, D., & Shepherd, H. (2008). The soci ol ogy of dis crim i na tion: Racial dis crim i na tion in employ ment, hous ing, credit, and con sumer mar kets. Annual Review of Sociology, 34, 181–209.

Pattillo, M. (2009). Investing in poor Black neigh bor hoods “as is.” In M. A. Turner, S. J. Popkin, & L. Rawlings (Eds.), Public hous ing and the leg acy of seg re ga tion (pp. 31–46). Washington, DC: Urban Institute.

Peng, L., & Thibodeau, T. (2013). Risk seg men ta tion of Amer i can homes: Evidence from Denver. Real Estate Economics, 41, 569–599.

Peterson, R. D., & Krivo, L. J. (2010). Divergent social worlds: Neighborhood crime and the racial-spa tial divide. New York, NY: Russell Sage Foundation.

Pfeffer, F. T., Danziger, S., & Schoeni, R. F. (2013). Wealth disparities before and after the Great Reces­ sion. Annals of the Amer i can Academy of Political and Social Science, 650, 98–123.

Pfeffer, F. T., & Killewald, A. (2018). Generations of advan tage. Multigenerational cor re la tions in fam ily wealth. Social Forces, 96, 1411–1442.

Pfeffer, F. T., & Schoeni, R. F. (2016). How wealth inequal ity shapes our future. Russell Sage Foundation Journal of the Social Sciences, 2(6), 2–22.

Piketty, T. (2014). Capital in the twenty-first cen tury (A. Goldhammer, Trans.). Cambridge, MA: Belknap Press of Harvard University Press. (Original work published 2013)

Quillian, L. (2003). How long are expo sures to poor neigh bor hoods? The long­term dynam ics of entry and exit from poor neigh bor hoods. Population Research and Policy Review, 22, 221–249.

Reardon, S. F., Bischoff, K., Owens, A., & Townsend, J. B. (2018). Has income seg re ga tion really increased? Bias and bias cor rec tion in sam ple­based seg re ga tion esti ma tes. Demography, 55, 2129–2160.

Reardon, S. F., & Firebaugh, G. (2002). Measures of mul ti group seg re ga tion. Sociological Methodology, 32, 33–67.

Robins, J. M. (1998). Marginal struc tural mod els. In Amer i can Statistical Association (Ed.), 1997 Proceedings of the Section on Bayes ian sta tis ti cal sci ence (pp. 1–10). Alexandria, VA: Amer i can Sta­ tistical Association.

Robins, J. M., Hernan, M. A., & Brumback, B. (2000). Marginal struc tural mod els and causal infer ence in epi de mi ol ogy. Epidemiology, 11, 550–560.

Ross, C. E., & Mirowsky, J. (2001). Neighborhood dis ad van tage, dis or der, and health. Journal of Health and Social Behavior, 42, 258–276.

Rugh, J. S., Albright, L., & Massey, D. S. (2015). Race, space, and cumu la tive dis ad van tage: A case study of the subprime lend ing col lapse. Social Problems, 62, 186–218.

Rugh, J. S., & Massey, D. S. (2010). Racial seg re ga tion and the Amer i can fore clo sure cri sis. Amer i can Sociological Review, 75, 629–651.

Rusk, D. (2001). The “seg re ga tion tax”: The cost of racial seg re ga tion to Black homeowners (Report). Washington, DC: Brookings Institution, Center on Urban and Metropolitan Policy.

Saez, E., & Zucman, G. (2016). Wealth inequal ity in the United States since 1913: Evidence from cap i tal­ ized income tax data. Quarterly Journal of Economics, 131, 519–578.

Sampson, R. J., Morenoff, J. D., & Gannon­Rowley, T. (2002). Assessing “neigh bor hood effects”: Social pro cesses and new direc tions in research. Annual Review of Sociology, 28, 443–478.

Sampson, R. J., & Sharkey, P. (2008). Neighborhood selec tion and the social repro duc tion of con cen trated racial inequal ity. Demography, 45, 1–29.

Schoeni, R. F., Stafford, F., McGonagle, K. A., & Andreski, P. (2013). Response rates in national panel sur veys. Annals of the Amer i can Academy of Political and Social Science, 645, 60–87.

Semyonov, M., & Lewin­Epstein, N. (2013). Ways to rich ness: Determination of house hold wealth in 16 countries. Euro pean Sociological Review, 29, 1134–1148.

This content downloaded from �����������63.236.101.18 on Sun, 21 Jan 2024 14:41:17 +00:00������������

All use subject to https://about.jstor.org/terms

320 B. L. Levy

Shapiro, T. M. (2004). The hid den cost of being Afri can Amer i can: How wealth per pet u ates inequal ity. New York, NY: Oxford University Press.

Shapiro, T. M. (2006). Race, homeownership, and wealth. Washington University Journal of Law & Policy, 20, 53–74.

Shapiro, T., Meschede, T., & Osoro, S. (2013). The roots of the wid en ing racial wealth gap: Explaining the Black–White eco nomic divide (Research and Policy Brief). Waltham, MA: Institute on Assets and Social Policy.

Sharkey, P. (2013). Stuck in place: Urban neigh bor hoods and the end of prog ress toward racial equal ity. Chicago, IL: University of Chicago Press.

Sharkey, P., & Faber, J. W. (2014). Where, when, why, and for whom do res i den tial con texts mat ter? Mov­ ing away from the dichot o mous under stand ing of neigh bor hood effects. Annual Review of Sociology, 40, 559–579.

Spilerman, S. (2000). Wealth and strat i f ca tion pro cesses. Annual Review of Sociology, 26, 497–524. Stu art, G. (2003). Discriminating risk: The U.S. mort gage lend ing indus try in the twen ti eth cen tury. Ithaca,

NY: Cornell University Press. Sullivan, L., Meschede, T., Dietrich, L., & Shapiro, T. (2015). The racial wealth gap: Why pol icy mat ters

(Report). New York, NY: Demos. Taylor, K.­Y. (2019). Race for profit: How banks and the real estate indus try undermined Black

homeownership. Chapel Hill: University of North Carolina Press. Thomas, M. E., Moye, R., Henderson, L., & Horton, H. D. (2018). Separate and unequal: The impact of

socio eco nomic sta tus, seg re ga tion, and the Great Recession on racial disparities in hous ing val ues. Sociology of Race and Ethnicity, 4, 229–244.

Turner, T. M., & Leua, H. (2009). Homeownership, wealth accu mu la tion and income sta tus. Journal of Housing Economics, 18, 104–114.

Urban Institute. (2015). Nine charts about wealth inequal ity in America (Report). Retrieved from http: / / apps .urban .org /features /wealth -inequality -charts /

Vartanian, T. P., & Buck, P. W. (2005). Childhood and ado les cent neigh bor hood effects on adult income: Using sib lings to exam ine dif fer ences in OLS and fxed effect mod els. Social Service Review, 79, 60–94.

Wodtke, G. T. (2013). Duration and tim ing of expo sure to neigh bor hood pov erty and the risk of ado les cent par ent hood. Demography, 50, 1765–1788.

Wodtke, G. T., & Almirall, D. (2017). Estimating mod er ated causal effects with time­vary ing treat­ ments and time­vary ing mod er a tors: Structural nested mean mod els using regres sion with resid u als. Sociological Methodology, 47, 212–245.

Wodtke, G. T., Elwert, F., & Harding, D. J. (2016). Neighborhood effect het ero ge ne ity by fam ily income and devel op men tal period: Evidence from a coun ter fac tual model of high school grad u a tion. Amer i can Journal of Sociology, 121, 1168–1222.

Wodtke, G. T., Harding, D. J., & Elwert, F. (2011). Neighborhood effects in tem po ral per spec tive: The impact of long­term expo sure to con cen trated dis ad van tage on high school grad u a tion. Amer i can Sociological Review, 76, 713–736.

Brian L. Levy blevy4@gmu .edu

Department of Sociology and Anthropology, George Mason University, Fairfax, VA, USA; https: / /orcid .org /0000 -0002 -5784 -2204

This content downloaded from �����������63.236.101.18 on Sun, 21 Jan 2024 14:41:17 +00:00������������

All use subject to https://about.jstor.org/terms