Moving to Shared Equity: Locational Outcomes for Households in Shared Equity Homeownership Programs

Introduction

Alternative homeownership models such as shared-equity homeownership (SEH) provide numerous benefits for lower-income households, supplying a wealth-building opportunity for households that might otherwise be excluded from homeownership altogether (Acolin et al., 2021). However, the impact of shared-equity models on the locational outcomes of participating households has remained largely unexplored. While poverty deconcentration and the expansion of locational opportunity for low-income households are seen as key objectives for affirmatively furthering fair housing in programs such as Housing Choice Vouchers (HCV), SEH programs focus instead on place-based investments, often in disinvested neighborhoods. Despite this emphasis, it is important to understand how such programs might affect the locational outcomes of participating households. However, there is currently little evidence concerning how participation in SEH programs affects a household’s residential trajectory: whether entering or exiting SEH programs causes participating households to experience substantial changes in neighborhood demographics and access to opportunity. In this paper, we examine whether households participating in SEH programs experience differential outcomes in neighborhood characteristics along multiple dimensions, including neighborhood poverty, school quality, transportation and employment access, and demographic composition. The purpose of this study is to explore whether households participating in SEH programs experience any tradeoffs between the wealth-building opportunity afforded by SEH models and the neighborhood context of SEH units.

This study considers the following research questions. First, how do the neighborhood characteristics of SEH units compare with the prior and subsequent locations of participating households? Do participating households experience changes in neighborhood opportunity measures or demographic characteristics because of entering or exiting SEH programs? Second, are there meaningful differences between the residential trajectories of SEH households and either 1) households that entered traditional homeownership, or 2) households that remained renters? Does participation in SEH programs lead households to experience substantially different neighborhood outcomes than households in the private housing market? Although SEH programs are not designed primarily to address fair housing goals, locational outcomes remain an important consideration for households participating in SEH programs; these questions thus provide an opportunity to examine a possible underexplored role of alternative homeownership models in furthering broader housing policy goals.

Neighborhood Effects and Locational Outcomes

Decades of urban social research have documented the complex forces that contribute to residential stratification (Krysan and Crowder, 2017) and the effects of neighborhood context on individual life outcomes. While there are several hypothesized mechanisms through which these ‘neighborhood effects’ may be enacted (Ellen & Turner, 1997; Galster, 2012), we focus primarily on two mechanisms: 1) proximity to economic opportunities, and 2) quality of local services and social exposures. The most immediate potential impact of neighborhood context on socioeconomic status is the ability to access employment opportunities, which depends in part upon the spatial relationship between residential and employment locations as well as the availability and accessibility of transportation between those locations. Thus, the ‘spatial mismatch’ between residential and employment locations has been identified as one potential driver of racial and economic inequalities in urban areas (Kain, 1992; Ihlanfeldt & Sjoquist, 1998), particularly in inner-city neighborhoods where the dispersal of employment opportunities exacerbated the effects of concentrated poverty (Wilson, 1987). Enabling households to move into areas with greater economic opportunity may increase their employment outcomes and improve their economic well-being by decreasing the transportation costs associated with a particular location (Acevedo-Garcia et al., 2016). Greater spatial accessibility to employment may not translate into better employment outcomes, however, due to structural mismatches between available employment and worker qualifications that particularly impact women and people of color (Blumenberg, 2004; Hellerstein et al., 2008; Lens et al., 2019). Nonetheless, continued interest in spatial accessibility and job access motivates our inclusion of these indicators in our analysis. The effects of neighborhood context are also acute for children and adolescents, for whom exposure to concentrated poverty can contribute to increases in depression, behavioral issues, and educational challenges (Brooks-Gunn et al., 1993; Jencks & Mayer, 1989; Sampson et al., 2002). Longitudinal studies of subsequent life outcomes for children in the Moving To Opportunity (MTO) program – which provided a limited number of low-income households in five cities with vouchers targeted to low-poverty neighborhoods – have found that moving to lower-poverty neighborhoods had a positive effect on their future health, earnings, and other outcomes (Chetty et al., 2016; Kling et al. 2007; Ludwig et al., 2013). Neighborhood poverty may also improve employment outcomes for adults, based on limited evidence from the Gautreaux public housing desegregation initiative (Mendenhall et al., 2006; Rosenbaum, 1995). Poverty deconcentration and household access to opportunity have consequently emerged as important objectives of low-income housing policy (Khadduri, 2001). These goals have, however, been the subject of extensive theoretical and empirical contestation. Contemporary debates center the tension between people-based fair housing strategies and place-based community development strategies, which alternatively propose enabling low-income households to move to “opportunity” or investing in the neighborhoods where those households reside. Both approaches have been subject to critique: while community development strategies are critiqued for keeping low-income households in disadvantaged neighborhoods, fair housing strategies are critiqued for disrupting household stability and undermining place-based community ties (Johnson et al., 2002; Rosenbaum & Zuberi, 2010). While some call for integrating these goals (Dawkins, 2018; Julian, 2008; Turner, 2017), the focus on affirmatively furthering fair housing by moving low-income households may come at the expense of investing in affordable housing in low-income communities (Goetz, 2018; Imbroscio, 2012). Fair housing programs focused on poverty deconcentration via residential mobility have nevertheless dominated U.S. housing policy, despite mixed success in achieving the stated goal of improving household locational outcomes.

Outside of experimental programs such as Gautreaux and MTO, low-income household mobility has primarily been enacted through two policy strategies: the development of mixed-income communities, and the transition of low-income housing subsidies into vouchers for private rental housing (Popkin et al., 2000). Efforts to create mixed-income communities through HOPE VI, the Low-Income Housing Tax Credit (LIHTC) program, and the Rental Assistance Demonstration (RAD) program have had limited effects on improving neighborhood conditions for low-income households. With HOPE VI, only a fraction of original residents returned to the sites of redeveloped public housing projects, and displaced households generally ended up in neighborhoods with similar socioeconomic conditions as their original community (Goetz, 2010; Nguyen et al., 2016). RAD has resulted in most tenants remaining at or returning to the property after conversion, and although RAD conversions improve housing quality, neighborhood outcomes often remain the same (Hayes et al., 2021). Furthermore, it is only within the last decade that the scoring criteria for the LIHTC program has started placing emphasis on access to neighborhoods with better socioeconomic conditions (Walter et al., 2018).

Similar trends are observed in the context of the HCV program, which provides eligible low-income households with housing cost subsidies to pay for rental housing in the private market. Obstacles to locational mobility for voucher recipients include discrimination by landlords, a lack of resources for the housing search process, and subjection to involuntary forms of mobility such as eviction (DeLuca et al., 2013). Voucher holders tend to remain in neighborhoods with similar socioeconomic characteristics and do not fare significantly differently in terms of locational outcomes than similar renters without vouchers (Basolo & Nguyen, 2005; Devine et al., 2003; McClure, 2008; Pendall, 2000; Walter et al., 2015). Vouchers alone are thus not sufficient to improve the locational outcomes of low-income households, even with programmatic changes that have increased rent limits in more desirable neighborhoods without providing additional services to support the housing search process (Bergman et al., 2019; Reina et al., 2019). Although significant attention has been placed on the locational effects of mobility-oriented housing programs on low-income renters, only a limited strand of literature has examined neighborhood outcomes for low- and moderate-income homeowners. Homeownership can provide a valuable wealth-building opportunity for low-income households (Herbert et al., 2013). For low- and moderate-income households, however, accessing affordable homeownership may impose costs, including worse locational outcomes than if they continued to rent (Gabriel & Painter, 2008; Reid, 2007). Discrimination in mortgage origination (Galster, 2010), market segmentation and residential segregation (Flippen, 2010), and income and down payment constraints (Bostic & Lee, 2008) all contribute to the greater likelihood of low-income homeowners of color residing in distressed neighborhoods. Distressed segregated neighborhoods with lower-priced homes deflate housing appreciation and may in turn negate the wealth building potential of homeownership (Flippen, 2004; Reid, 2005; Shlay, 2006).

Locational choices of low-income and non-white homeowners have been examined extensively in terms of metropolitan location (Belsky & Duda, 2002; Gabriel & Painter, 2003; Gyourko et al., 1999; Immergluck, 1999), but fewer studies have examined if low-income households improve their neighborhood context when transitioning from rentership into homeownership. While low-income households of color may experience some improvements in locational characteristics when transitioning to homeownership, low-income white households largely do not improve their neighborhood context after buying a home (Cummings et al., 2002; Reid, 2005; Reid, 2007). This does not indicate that non-white households are located in more advantaged neighborhoods than white homeowners, but rather that such households tend to reside in more disadvantaged neighborhoods prior to homeownership relative to white households. Furthermore, several studies suggest that low-income homebuyers – including non-white homebuyers – are more likely to either end up in neighborhoods with higher poverty rates and higher levels of racial concentration (Immergluck, 1998; Van Zandt, 2007) or experience no change in socioeconomic conditions upon entering homeownership (Turnham et al., 2003; Turnham et al., 2004). Even when low-income renters were randomly assigned subsidized Individual Development Accounts (IDAs), no improvement was found in neighborhood amenities for households that used those saving accounts to purchase a home (Engelhardt et al., 2010). These patterns may derive in part from constraints on access to different urban spaces, with much higher levels of homeownership among Black households in central city neighborhoods relative to suburban geographies but also from borrowing constraints restricting affordable units to less desirable locations (Gabriel & Painter, 2008; Gyourko et al., 1999). Homeownership is not necessarily sufficient to generate better locational outcomes given housing market structures and the constraints faced by low- and moderate-income households in purchasing a home.

Shared-Equity Homeownership

SEH models provide opportunities for stable low-income homeownership while maintaining long-term housing affordability. SEH models generally share certain commonalities, including owner-occupation, income restrictions, restrictions on resale price, and the apportionment of equity to organizations or individuals beyond the owner occupant, all of which increase the short-term stability of residents as well as the long-term affordability of housing units (Ehlenz & Taylor, 2019; Temkin et al., 2013). SEH models include Community Land Trusts (CLTs), Limited Equity Housing Cooperatives (LECs), and deed-restricted housing units. These programs represent a relatively small but important intervention into low-income housing policy: as of 2013, there were estimated to be approximately 12,000 units in CLT programs throughout the United States, 167,000 units in LEC programs, and 31,000 deed-restricted units (Thaden, 2018; Wang & Balachandran, 2021). These programs have specific geographic and temporal scopes: the majority of LECs in the United States were formed in New York City from the late 1960s to the early 1980s due to landlord abandonment (Saegert & Benítez, 2005), whereas CLT programs have expanded slowly over time primarily in the Northeastern and Western regions of the United States (Sungu-Eryilmaz & Greenstein, 2007). CLT programs are generally facilitated by steward organizations that hold land in trust while allowing the housing on that land to be bought and sold by occupants, as long as those sales are subject to resale price restrictions that maintain the long-term affordability of the housing. At the same time, SEH programs provide a package of benefits that support upward mobility. Occupants of SEH programs cite stability as a key program advantage (Martin et al., 2020), which has been borne out by analyses finding similar or greater levels of stability among SEH households than conventional homeowners (Schneider et al., 2022; Wang et al., 2019), and less exposure to destabilizing events such as foreclosures (Thaden, 2011). Furthermore, recent evidence shows that participants in SEH programs build significantly more wealth than comparable renter households and only slightly less wealth than similar homeowners over the course of their tenures (Acolin et al., 2021). Collectively, these findings suggest that SEH programs provide a similar array of benefits to conventional homeownership for individual participants.

SEH programs offer benefits not only to individual participants but also to entire communities: in addition to providing households with access to affordable homeownership, these programs explicitly target investments to particular places. SEH programs can provide a range of community benefits, including preserving access to homeownership for future low-income households within the community, stabilizing neighborhoods experiencing disinvestment or gentrification, and preserving community assets into the future (Davis, 2006; Ehlenz & Taylor, 2019; Temkin et al., 2013). Despite this focus on place, evidence on the spatial context of SEH programs is limited. Choi et al. (2018) find that CLT units are less likely to be located in neighborhoods experiencing demographic changes commonly associated with gentrification, such as increases in median incomes, median home values, owner-occupation rates, and the share of the population that was white and college-educated. Nelson et al. (2020) find that the clustering of CLT units contributed to smaller declines in the sales prices of surrounding properties during the foreclosure crisis and increases in nearby sales prices after the recession. CLTs may include both owner-occupied and renter-occupied units, which limits the conclusiveness of these findings with respect to shared-equity homeownership. Examining a broader array of SEH programs, Wang et al. (2019) find that relative to similar owner-occupied units, SEH units were located in neighborhoods with higher average levels of labor market engagement and lower transportation costs, but also higher average poverty rates.

These patterns can be attributed to a combination of programmatic objectives, historical circumstances, and land acquisition costs. CLTs are often established in response to contexts of neighborhood disinvestment or fears of gentrification and displacement.ⁱ Hence, the reasons behind the formation of shared equity programs can have direct implications on the selection of their target areas. Additionally, some major funding sources that shared equity programs rely on (i.e. Community Development Block Grant, HOME grant, and Neighborhood Stabilization Program grant) are required to target certain disinvested and low-income neighborhoods. CLT programs grow primarily through the acquisition of land on the private market with public subsidies, donations, and collaborations with land banks (Davis & Jacobus, 2008). These factors limit the scope and locational context of CLT properties and impose fundamental constraints on locational choices for households entering SEH programs, particularly in more expensive urban housing markets where affordable land is relatively scarce. A compounding reality is that shared equity owners have historically been unable to access mortgage products insured by the Federal Housing Administration (FHA) (Stromberg & Stromberg, 2013), which are critical means for lower income homebuyers to live in a broader range of neighborhoods (Zhu & Ballesteros, 2021). We therefore hypothesize that households entering SEH units would experience decreases in certain opportunity metrics. However, we also hypothesize that other benefits associated with participation in SEH programs – particularly the opportunity to build housing wealth – will manifest in improved location outcomes for households exiting SEH units. The combination of place-based investment in specific communities and the household-level wealth-building benefits associated with shared-equity models might thus allow SEH programs to achieve multiple housing goals simultaneously: supporting access to opportunity for individual households while simultaneously bolstering community development.

Tracking Shared-Equity Moves

SEH household data is obtained from the Grounded Solutions Network HomeKeeper National Data Hub, which provides information about households and transaction entered by participating organizations (Wang et al., 2019). These data include address information for all SEH units and for certain programs also include the prior and subsequent addresses of participating households. We restrict the sample to households that purchased or resold a home between 1997 and 2019, and further restrict the sample to moves where the origin and destination could be identified at the level of the census tract.ⁱⁱ Addresses were geocoded using the US Census Bureau geocoding tool, with unmatched addresses subsequently geocoded through an iterative process combining Google’s geocoding API with manual address corrections. Addresses that could not be matched following this process were dropped from the analysis,ⁱⁱⁱ and geocoded addresses were spatially joined to 2010 census tracts.

Addresses for household locations before and after their time in SEH programs were inconsistently entered and formatted, requiring the multi-stage iterative geocoding process outlined above. Of 9,323 records of household characteristics prior to living in SEH programs, 6,892 contained some usable address information that could be geocoded. Additional restrictions were applied to these records to optimize the dataset for matching and analysis, such that the following records were excluded: sales taking place prior to 1997, households with the same address recorded in both time periods, households that were already homeowners when entering SEH, and households missing matching characteristics such as household income. These restrictions leave us with a dataset of 4,656 households entering SEH. Of 2,210 records of households exiting SEH, 843 contained any form of address and only 696 contained sufficient information to positively match the household with a census tract. After applying similar restrictions to these records, we obtain 491 households for which information is available both before and after their tenure in SEH. Given sample size limitations, we divide our analysis into two datasets: 1) all households for which location was observed prior to entering SEH, and 2) households for which location was observed both prior to entering and after exiting SEH. While the former dataset has a more robust sample size and provides more reliable estimates of changes between pre-SEH and SEH neighborhood characteristics, the latter dataset is crucial for understanding the neighborhood trajectories of households exiting SEH.

The 4,656 households in our dataset participated in 58 SEH programs across 28 states, consisting almost entirely of CLT units and deed-restricted units. The most significant source of observations is the Champlain Housing Trust based in Vermont, for which nearly 1,000 proram entrances are observed. Post-SEH observations are significantly more geographically constrained, with the 491 households participating in 27 SEH programs across 14 states. In both datasets, the majority of observations are associated with organizations located in Vermont, Minnesota, Oregon, Colorado, and Washington (Table 1). Most households lived nearby prior to entering SEH, with a median distance less than four miles, and less than 10% of households moving further than 13 miles to enter SEH units (Table 2). By contrast, the median household exiting SEH moved more than 10 miles and 10% moved more than 900 miles, with more than 40% moving outside of the county and more than 20% moving to a different state.

Table 1.

SEH organizations with ten highest observed entrance volumes retained in sample.

Organization	State	Entrances	Exits
Champlain Housing Trust	Vermont	779	273
One Roof Community Housing	Minnesota	421	80
City of Lakes CLT	Minnesota	320	17
Rocky Mountain CLT	Colorado	319	2
Proud Ground	Oregon	298	22
Colorado CLT	Colorado	271	4
Homestead CLT	Washington	253	18
Habitat for Humanity Greater San Francisco	California	226	9
First Homes Properties	Minnesota	197	18
Austin Habitat for Humanity	Texas	153	0

Table 2.

Distance characteristics of observed moves.

Type of Move	All Entrances	Entrances (Exit Also Observed)	Exits
Same Tract	529 (11.4%)	71 (14.5%)	53 (10.8%)
Same County, Different Tract	3,596 (77.2%)	355 (72.3%)	236 (48.1%)
Same State, Different County	431 (9.3%)	51 (10.4%)	92 (18.7%)
Outside State	100 (2.1%)	14 (2.9%)	110 (22.4%)
Total	4,656	491	941
10th Percentile Distance (Miles)	0.52	0.52	1
Median Distance (Miles)	3.48	3.87	9.84
90th Percentile Distance (Miles)	12.83	14.6	918.16

As the different geographic distributions of these datasets suggest, observations of households exiting SEH programs do not strictly resemble our observations of households entering the programs, nor of households in the overall HomeKeeper dataset. To assess the relative similarity of these three datasets – the full HomeKeeper dataset, households entering SEH, and households entering and exiting SEH – we examine several household-level variables that are available for all households in the HomeKeeper dataset. Households observed entering SEH largely mirror the full HomeKeeper dataset in terms of household size, presence of children and seniors in the household, and household income. However, the subset of households entering and exiting SEH tended to have fewer members, were less likely to include children or seniors, and had lower incomes than those only observed entering SEH. Although outcomes for households exiting SEH are crucial for our analysis, they are thus not perfectly representative of all households participating in SEH programs.

Neighborhood Outcomes for Shared-Equity Households

As a measure of neighborhood outcomes for SEH households, we use neighborhood indices developed by the US Department of Housing and Urban Development (HUD) for the Affirmatively Furthering Fair Housing (AFFH) rule. While this analysis is not intended as a program evaluation of SEH in terms of affirmatively furthering fair housing goals, the AFFH data provides several advantages, including uniform nationwide coverage, ample documentation, and comprehensive measures covering multiple dimensions of neighborhood opportunity. Defining neighborhoods based on 2010 census tracts, we include AFFH indices concerning poverty, school quality, transportation and employment access, labor market engagement, tenure composition, and racial composition (Table 3). These are drawn from AFFH data released in 2017 and made available by Urban Institute, with individual indicators based on data from 2008 through 2017. Job proximity and school quality, which are provided at the Census block group level, are aggregated to the tract level using weighting based on block group population. Although these indicators do not supply a complete picture of neighborhood ‘opportunity’, they provide an effective baseline understanding of many neighborhood characteristics that may affect a household’s long-term outcomes. Barriers to transportation and employment may have immediate impacts on household economic stability due to spatial mismatch, while concentrated poverty and low school quality may have long-term impacts on children. Racial and tenure composition provide a useful additional metric of neighborhood context.

Table 3.

AFFH Version 4 neighborhood indices (HUD Office of Policy Development & Research, 2017).

Neighborhood Measure	Data Source	Description
Low Poverty	American Community Survey (2009–2013)	Inverse percentile poverty rank; higher score indicates lower poverty
School Proficiency	Great Schools (2013–14); Common Core of Data (2013–14); Maponics (2016)	Percentile rank of school quality relative to state
Labor Market Engagement	American Community Survey (2006–2010)	Percentile rank based on employment level, labor force participation, and educational attainment
Jobs Proximity	Longitudinal Employer-Household Dynamics** (2014)	Percentile rank of spatial proximity to employment
Low Transportation Cost	Location Affordability Index (2008–2012)	Inverse percentile rank of transportation costs as a percentage of income for a 3-person single-parent renter household at 50% AMI; higher score indicates lower costs
Transit Trips	Location Affordability Index (2008–2012)	Percentile rank for public transit access for a 3-person single-parent renter household at 50% AMI
% Renters (2010)	Comprehensive Housing Affordability Strategy (2009–2013)	Percentage of households within tract identified as renters
% Black (2010)	Longitudinal Tract Database (2010)	Percentage of households within tract identified as Black
% Hispanic (2010)	Longitudinal Tract Database (2010)	Percentage of households within tract identified as Hispanic

We find that both entrances and exits from SEH units are associated with statistically significant changes in neighborhood characteristics. Entering SEH is associated with statistically significant increases in average poverty rates and transportation costs, and decreases in transit access, labor engagement, and school quality indices (Figure 1a). Differences were particularly large for school quality, poverty, and labor engagement indices – with a 5-point difference between pre-SEH and SEH neighborhoods. Entering SEH involves relocating to lower-opportunity neighborhoods on average, in terms of access to employment and resources, and in terms of neighborhood poverty. Such moves are also associated with decreases in neighborhood share of renters and increases in the share of the neighborhood that is Black or Hispanic, suggesting that these households are moving to neighborhoods that are more racially diverse but have higher rates of homeownership. These findings largely align with expectations given constraints on locations of units available for incorporation into SEH programs.

Figure 1.

Mean changes in neighborhood indicator values for (a) households observed entering SEH, and (b) households observed entering and exiting SEH.

Households with an observed post-SEH location experienced similar changes in neighborhoods characteristics when entering SEH programs. While statistical significance is limited to poverty, race, labor engagement, and school quality due to small sample size, all variables exhibit similar trajectories (Figure 1b). This validates the use of those households to evaluate the effect of exiting SEH programs on neighborhood outcomes. These households experienced significant decreases in poverty and increases in school quality upon exiting SEH, both of which point to moves into neighborhoods providing greater opportunity for children. By contrast, exits from SEH programs also led to further decreases in transit access and proximity to jobs and contributed to increases in neighborhood transportation costs. These findings apply when comparing post-SEH neighborhoods to both SEH neighborhoods and pre-SEH neighborhoods, indicating that households participating in SEH ultimately resided in neighborhoods with lower poverty and higher school quality, but also less transportation and employment access, relative to the neighborhoods they lived in prior to their participation in SEH programs. These findings may suggest that households leaving SEH moved to more affluent suburban areas, a supposition supported by a steady decrease in the average neighborhood population density for these households: from 3,948 people per square mile prior to entering SEH to 3,352 per square mile while in SEH, and ultimately to 2,747 per square mile after exiting SEH. Living in SEH thus appears tohave an impact on a household’s future neighborhood conditions, with those households ending up in neighborhoods that are more affluent and further from transportation and employment centers. Whether this can be considered a net benefit depends on whether questions of socioeconomic context or access are prioritized, as the benefits of lower neighborhood poverty rates may be offset by more limited and costly transportation and employment access.

Matched Pair Analysis

Methods

We use matched pair analysis to compare the residential trajectories of SEH participants to those of households with similar characteristics. A control group of households not participating in SEH programs is constructed from the Panel Survey of Income Dynamics (PSID) – a longitudinal household survey maintained by the Institute for Social Research at the University of Michigan – using restricted-access data that includes household data every other year and provides 2010 census tract identifiers associated with each household record. We retain households that moved between 1999 and 2017 for which census tract identifiers were available both before and after the move, further restricting this dataset to include only records with a household head 18 years of age or older, with less than $150,000 reported in annual income (similar to the highest reported income among retained HomeKeeper records), and non-zero longitudinal sampling weights. This analysis does not account for intermediate locations of households that moved multiple times between survey waves, but does allow us to examine their ultimate residential trajectories.

Moves recorded in the PSID are matched to Homekeeper records using a matching algorithm based on Mahalanobis distance via the matchIt R package (Ho et al., 2011). We match with replacement, such that each PSID household may be matched to multiple SEH households. Households are matched on household size, householder age, householder race, household income, whether they moved after 2008, and the median household income and median home value for both the region^iv and census tract of each household’s initial location. These variables are selected to attain similarity between PSID households and HomeKeeper households in terms of demographic and life-course characteristics and initial locational context, the latter of which is particularly important given that the PSID dataset may not be fully representative across urban form types (Makarewicz et al., 2020). We conduct four separate matches: 1) households entering SEH vs. PSID households transitioning from renting to homeownership, 2) households entering SEH vs. PSID households moving between rented units, 3) households entering and exiting SEH vs. PSID households transitioning from renting to homeownership during their first move, and 4) households entering and exiting SEH vs. PSID households moving between rented units during their first move. Given that the analysis of exits from SEH programs requires tracking two sequential moves, household exits are matched only with PSID households for which two sequential moves are available. However, records including household exits are still matched based only on initial tenure, given that our goal is to match households that are observationally similar prior to entering SEH. Our matching strategy is effective at balancing our variables of interest: the majority of matching variables have an absolute Standardized Mean Distance (SMD) less than 0.1, and all variables have an absolute SMD less than 0.17 (Figure 2).^v Using these matched datasets, we then analyze locational outcomes using our nine neighborhood-level indices. For each variable, we conduct weighted linear difference-in-differences models to account for baseline dissimilarities in neighborhood conditions between treatment and control groups and isolate the effect associated with moves into and out of SEH:

$$AFF{H}_i=SE{H}_i+Perio{d}_t+SE{H}_i\ast Perio{d}_t+Control{s}_i+e_{it}$$

Figure 2.

Match balance improvements for variables in each matching model, measured by absolute standardized mean difference.

For each of the four datasets, we run a separate model on each of the nine AFFH indicators, in which treatment condition SEH is a binary variable indicating whether a household is from the HomeKeeper dataset. For matched datasets containing households only observed prior to SEH, time period is defined as a binary variable indicating whether an observation is before or after the move into SEH. For matched datasets containing households observed entering and exiting SEH, time period is a three-category condition for which tenure in SEH is used as the reference category. Our primary coefficient of interest is the interaction between time period and treatment status, which indicates whether entering or exiting SEH is associated with a statistically significant difference in the change of neighborhood characteristics from one tenure location to another between datasets. To ensure optimal matching balance, we also include matching variables as controls and employ a weighting scheme which multiplies matching weights by PSID sampling weights for PSID observations. For exits from SEH, we also include tenure after the second move as an additional control to ensure that observed effects are not due to structural differences between those moving into homeownership and those moving into rented housing.

Results

Model results provide evidence that households participating in SEH experience different locational outcomes relative to comparable households. Entering SEH is associated with larger increases in poverty, renter share, and percent Black relative to similar households transitioning into homeownership (Table 4). These moves are also associated with smaller decreases in public transit access and jobs access and smaller increases in transportation costs, which indicates that while entering SEH entails slight decreases in transportation access, these declines are not as substantial as those experienced by similar households entering traditional homeownership. When comparing households entering SEH with PSID households that moved between rented units, we similarly find associations with increased poverty rates. However, households entering SEH are also associated with decreases in the neighborhood renter share, which suggests that SEH units are in neighborhoods with higher levels of homeownership (Table 5). Moves into SEH units are also associated with decreases in labor engagement and school quality indices and increases in percent Hispanic. That these metrics are significant for this model and not for the prior model is somewhat counterintuitive, given that one would expect a higher school quality and labor engagement for households transitioning to homeownership rather than remaining in rented housing. It is worth noting that the signs for these variables are consistent across both models; therefore, this discrepancy likely results from the smaller sample size of PSID households moving from renting to homeownership in the baseline model.

Table 4.

Households entering SEH vs. PSID households moving from renting to homeownership

Variable	Low Poverty	School Quality	Labor Engagement	Jobs Proximity	Transport Cost	Transit Trips	Percent Renters	Percent Black	Percent Hispanic
(Intercept)	38.791*** (10.112)	39.297*** (9.962)	41.599*** (9.741)	57.407*** (8.39)	60.487*** (7.971)	47.224*** (8.533)	42.931*** (8.325)	19.543*** (5.086)	22.302*** (6.182)
SEH	5.489*** (1.457)	3.427* (1.358)	18.487*** (1.512)	4.19*** (1.106)	15.975*** (1.211)	3.772*** (1.144)	3.151** (1.112)	−5.822*** (1.077)	−4.64*** (1.102)
Period	2.731 (1.958)	−1.1 (1.853)	−1.596 (2.069)	−4.23** (1.54)	−7.766*** (1.832)	−7.084*** (1.696)	−8.5*** (1.372)	−1.478 (1.387)	−0.348 (1.592)
SEH*Period	−7.056*** (2.024)	−2.671 (1.916)	−2.818 (2.127)	2.573 (1.599)	5.981** (1.862)	5.331** (1.744)	5.041*** (1.431)	3.847** (1.409)	2.439 (1.622)

Treatment Observations: 4656

Control Observations: 974

Table 5.

Households entering SEH vs. PSID households moving from renting to renting

Variable	Low Poverty	School Quality	Labor Engagement	Jobs Proximity	Transport Cost	Transit Trips	Percent Renters	Percent Black	Percent Hispanic
(Intercept)	54.847*** (9.432)	33.213*** (8.91)	54.966*** (9.252)	55.402*** (7.507)	71.702*** (6.873)	47.795*** (8.253)	41.434*** (7.301)	27.147*** (5.243)	13.451* (6.106)
SEH	6.363*** (0.828)	3.324*** (0.816)	19.549*** (0.829)	2.146** (0.654)	12.311*** (0.654)	1.26 (0.744)	0.764 (0.633)	−6.562*** (0.558)	−5.945*** (0.594)
Period	0.534 (1.074)	−0.962 (1.017)	−0.537 (1.072)	−0.523 (0.82)	−1.774 (0.928)	−1.874 (1.057)	−0.93 (0.821)	0.013 (0.77)	−0.153 (0.815)
SEH*Period	−4.859*** (1.119)	−2.809* (1.128)	−3.877** (1.182)	−1.134 (0.927)	−0.011 (0.987)	0.122 (1.133)	−2.53** (0.917)	2.357** (0.811)	2.244* (0.874)

Treatment Observations: 4656

Control Observations: 2179

For households exiting SEH, we find that exits are associated with significant decreases in both renter share and poverty, whether the comparison group from the PSID initially transitioned into homeownership or continued renting (Tables 6 and 7). Relative to households transitioning into traditional homeownership, SEH households also experienced increases in transportation costs upon exiting SEH units. The strong effects on decreasing poverty suggest that exits from SEH are associated with moves to neighborhoods with greater access to opportunity, and furthermore suggest that those exits provide improvement in neighborhood environments that are more significant than those experienced by comparable households not participating in SEH programs. Coefficients on SEH entrances are not significant in these models, but have signs and magnitudes matching the effects observed for the larger dataset of households entering SEH. This provides further indication that the subset of households observed after leaving SEH does not differ dramatically from the full set of observations in terms of neighborhood outcomes. These findings are constrained by the fact that these datasets are matched based only on the change in tenure during the first move, meaning that they do not reflect differences between households that ultimately ended up as homeowners after exiting SEH compared with those that ended up back in rented housing. However, this approach maintains consistency with the models focused on matching households based on their characteristics entering SEH, and results are unaffected by our inclusion of post-SEH tenure as an additional control.

Table 6.

Households entering and exiting SEH vs. PSID households moving from renting to homeownership

Variable	Low Poverty	School Quality	Labor Engagement	Jobs Proximity	Transport Cost	Transit Trips	Percent Renters	Percent Black	Percent Hispanic
(Intercept)	4.718 (43.72)	76.216 (44.122)	−18.057 (43.127)	82.31 (45.23)	−2.072 (39.642)	−49.788 (51.382)	−4.052 (28.877)	−4.89 (17.445)	0.8 (17.199)
SEH	−1.914 (2.364)	−2.562 (2.547)	16.726*** (2.853)	3.221 (2.553)	18.534*** (2.395)	2.142 (2.767)	9.72*** (1.606)	−0.657 (0.95)	−5.216*** (1.311)
Period (Pre)	−0.697 (3.16)	−1.314 (3.331)	3.304 (3.537)	1.074 (3.132)	9.684** (3.139)	3.549 (3.988)	7.107** (2.183)	0.541 (1.285)	−1.143 (1.514)
Period (Post)	−1.694 (3.334)	1.159 (3.299)	0.737 (3.679)	−2.167 (3.345)	2.84 (3.296)	0.497 (3.687)	1.835 (2.069)	2.045 (1.577)	−1.244 (1.538)
SEH*Period (Pre)	4.467 (3.521)	5.476 (3.69)	1.081 (3.799)	0.481 (3.446)	−9.676** (3.33)	−3.953 (4.236)	−4.647 (2.557)	−2.645 (1.441)	−0.287 (1.636)
SEH*Period (Post)	10.394** (3.675)	5.234 (3.678)	1.513 (3.953)	−2.018 (3.631)	−10.262** (3.541)	−6.554 (4.033)	−9.992*** (2.434)	−3.195 (1.757)	0.442 (1.674)

Treatment Observations: 491

Control Observations: 203

Table 7.

Households entering and exiting SEH vs. PSID households moving from renting to renting

Variable	Low Poverty	School Quality	Labor Engagement	Jobs Proximity	Transport Cost	Transit Trips	Percent Renters	Percent Black	Percent Hispanic
(Intercept)	69.063 (38.191)	−21.142 (33.224)	15.541 (37.817)	81.764** (29.988)	64.133* (32.302)	79.233* (38.249)	66.342* (32.174)	−10.055 (16.539)	24.845 (16.155)
SEH	−2.703 (2.481)	−1.003 (2.316)	12.641*** (2.24)	3.592 (2.089)	5.113* (1.989)	−8.73*** (2.317)	1.701 (2.149)	−2.202* (1.071)	−3.5*** (1.05)
Period (Pre)	0.923 (2.547)	5.388* (2.537)	−0.003 (2.427)	−1.154 (2.455)	−1.827 (2.042)	−4.255 (2.69)	−3.051 (2.357)	−1.456 (1.199)	−1.121 (1.163)
Period (Post)	−1.392 (2.805)	1.396 (2.509)	−2.748 (2.525)	−1.193 (2.367)	−3.969 (2.325)	−4.857 (2.539)	−2.159 (2.648)	1.15 (1.3)	−0.101 (1.278)
SEH*Period (Pre)	2.847 (3.01)	−1.226 (3.027)	4.388 (2.821)	2.709 (2.827)	1.835 (2.33)	3.852 (3.098)	5.511* (2.72)	−0.648 (1.378)	−0.309 (1.325)
SEH*Period (Post)	10.092** (3.243)	4.997 (3.037)	4.998 (2.934)	−2.992 (2.735)	−3.453 (2.671)	−1.2 (3.148)	−5.998* (2.946)	−2.3 (1.523)	−0.701 (1.447)

Treatment Observations: 491

Control Observations: 329

Supplementary Analyses

An important limitation of this analysis is the geographically and programatically biased distribution of the dataset, and in particular the large concentration of observations associated with the Champlain Housing Trust (CHT) in Vermont. Models excluding households associated with CHT housing units yield very similar results in terms of significance and magnitude, including increases in poverty, percent Black, and transit access, as well as lower transportation costs. For households exiting SEH, the model excluding CHT units finds effects associated with decreases in poverty; while other coefficients are non-significant, their signs and magnitudes are similar to those found in the primary models. This suggests that the locational effects associated with entrances and exits are robust and not overly determined by the particular organizational and spatial context of the country’s largest CLT. This result is encouraging given the outsize role of CHT in the HomeKeeper dataset, but more robust data collection is needed to fully explore program-level and regional differences in locational outcomes.

Another potential limitation is our use of the AFFH low poverty index as a primary measure of opportunity. While neighborhood poverty has been identified as an important factor in individual outcomes, particularly with respect to children, it serves as a somewhat indirect indicator for access to opportunity. Therefore, we run a parallel analysis on an intergenerational economic mobility indicator from the Opportunity Insights data tracker – the income at age 35 for children whose parents were at the 25th percentile of the national income distribution growing up in a particular neighborhood – which provides some indication of the extent to which living in a particular neighborhood affects children’s future life outcomes. We find patterns largely consistent with the results from the poverty index models: entering SEH is associated with moves to neighborhoods with statistically significantly lower income at age 35 and exiting SEH units is associated with moves to neighborhoods with higher income at age 35, though the latter are not statistically significant. This supports our conclusions regarding the effects of such moves on individual access to opportunity, though the results are not as statistically robust as for the low poverty index.

Finally, while the low poverty index is simply a normalized measure of tract-level poverty, it fails to account for the possibility of discontinuous effects associated with highly concentrated poverty (Jargowsky 1997). We therefore run additional models measuring the likelihood that a household will move into a neighborhood in the top decile of poverty rates according to AFFH data, corresponding with a non-student poverty rate of approximately 30 percent. We find that entering SEH is associated with statistically significant increases in the probability of moving into these high-poverty neighborhoods relative to PSID households, whether those households transitioned from renting to homeownership or remained renters. We further find that exiting SEH is associated with decreases in the probability of living in a high-poverty neighborhood, although these results are only statistically significant relative to households which transitioned into homeownership. This reaffirms that moves into SEH units are associated with greater exposure to concentrated poverty.

Discussion

Our analysis identifies some of the locational costs and benefits associated with participation in SEH programs. Based on our more robust dataset of households entering SEH, we identify a statistically significant decrease in opportunity indicators such as low poverty, school quality, and labor engagement associated with entering SEH units. Of particular interest, transitions into SEH neighborhoods are associated with increases in neighborhood poverty relative to comparable households that did not participate in SEH programs, whether those comparison households transitioned to homeownership or remained in rented housing. While entering SEH is also associated with decreases in transit access and increases in transportation costs, those households experience less significant increases in transportation barriers than comparable households entering homeownership.

There are several complementary explanations for these observed patterns, which are similar for other affordable housing programs such as LIHTC (Cummings & DiPasquale, 1999; Dawkins, 2013; Freeman, 2004; Walter et al., 2018). Shared-equity programs generally focus on place-based solutions rather than household-based mobility solutions, so it is not surprising that moves into shared-equity units might entail moving into areas with somewhat higher poverty. SEH programs are generally associated with non-profit organizations such as CLTs, many of which operate within particular communities for which affordable access to stable housing is a relevant concern. Funding sources for shared-equity programs may further require those programs to focus housing units in disinvested areas. Additionally, given the limited resources available to such organizations, it may only be possible to obtain land and housing units within lower-cost and higher-poverty neighborhoods. Meanwhile, although these neighborhoods may present less favorable transportation access relative to origin neighborhoods, they are also likely to be more centrally located and have better transit access than non-equity-restricted owner-occupied homes. The narrative concerning the effects of SEH programs on access to opportunity is therefore mixed: while such households are certainly exposed to greater concentration of poverty, they may also face fewer barriers to transportation than households entering traditional homeownership. The benefits of shared-equity homeownership – including tenure stability and the opportunity to build equity – likely outweigh any costs associated with locating in such neighborhoods. Indeed, Wang et al. (2019) find that households exiting shared-equity units are less likely than other households to list neighborhood conditions as a reason for their move.

Additionally, although households may experience some decreases in some neighborhood opportunity measures when entering SEH, we find that those households access higher-opportunity neighborhoods in their subsequent move. SEH households experience substantial decreases in poverty relative to comparable households in the PSID, such that they are in lower-poverty neighborhoods than they were before entering SEH. These households also move to neighborhoods with comparatively higher school quality and access to jobs, indicating that their tenure in SEH positioned them to access desirable neighborhoods and experience more favorable locational outcomes than households that remained in rented housing and even households that had transitioned to homeownership. These positive trajectories may be attributed to the combination stability and wealth-building benefits associated with SEH programs, which together provide a platform for upward mobility. In particular, recent evidence for the wealth-building potential of SEH programs using the same HomeKeeper data (Acolin et al., 2021) highlights the wealth gains associated with participation, which provide households with greater resources to access otherwise inaccessible housing market segments. SEH programs should therefore be seen as a net contributor to positive residential trajectories: the effects associated with entering SEH units are reflective of programmatic and resource constraints, but also can provide long-term benefits for participating households that are reflected in both their wealth and neighborhood characteristics. This provides evidence that SEH programs do not negatively impact household access to opportunity in the long term by virtue of their focus on disinvested neighborhoods, as tenure in SEH units may in fact broaden the potential residential trajectories of participating households.

There are several important limitations to this analysis that should be addressed in future research. The implications are limited by a small sample size, particularly for household observations after exiting SEH. This challenge is further exacerbated by the limited geographic distribution of observations, with many households and SEH units concentrated in a limited number of locations. Although models excluding CHT produced consistent results, the HomeKeeper dataset still reflects significant geographic and organizational biases. The differing levels of address data quality between organizations means that households from certain locations were far more likely to be excluded from this analysis. Our matching strategy also does not preclude the presence of unobserved differences between households occupying SEH units and households observed in the PSID; in particular, there may be some form of selection mechanism into SEH programs that cannot be accounted for via matching.

There are also substantial opportunities to extend this analysis, considering dimensions of locational outcomes beyond the relatively limited set of indicators considered here. One shortcoming of AFFH indicators is that they measure only one point in time and are not able to account for changes in neighborhood characteristics over time, nor do they measure other dimensions of neighborhood quality such as social networks and exposure to crime. Additionally, more robust future data collection would provide numerous alternatives to extend the present study. First, further research is needed on the differences between households that transitioned to ‘traditional’ homeownership after exiting SEH and households that transitioned back into rented housing, which could only be assessed to a limited extent due to small sample sizes. Second, more extensive locational histories of participating households before and after their SEH tenure would provide a better understanding of longer-term residential trajectories and would permit the introduction of additional match conditions based on propensity to move. Third, integrating the motivations of households entering or exiting shared-equity programs would provide further context for the observed residential trajectories, particularly given that neighborhood context can shape mobility motivations (Dantzler & Rivera, 2019; Jones & Dantzler, 2020). Finally, understanding the trade-offs between accessing homeownership and the locational constraints potentially imposed by SEH requires additional research, including qualitative studies to understand the experience of program participants and their level of satisfaction with their SEH location as well as their next location relative to the experience of similar households accessing traditional forms of homeownership.

Notwithstanding these limitations, this analysis provides valuable insight into locational outcomes and residential trajectories for households participating in SEH programs. The constraints faced by low- and moderate-income households when transitioning to homeownership can result in trade-offs between making that transition, accessing quality units, and locating in neighborhoods with desirable characteristics on a range of opportunity measures. Although SEH programs should be evaluated primarily on the degree to which they support community development and wealth-building goals, locational outcomes are a secondary outcome with potentially important implications for participating households. While SEH programs may site households in lower-opportunity neighborhoods, they do appear to improve long-term locational outcomes for participating households.. This illustrates the potential for homeownership models such as SEH to not only further goals of community development but also – in the long term – to further fair housing goals as well, breaking down traditional barriers between these housing approaches and emphasizing the interdependence of their goals (Turner, 2017). CLT programs already provide benefits not only for the communities in which they are located but also for individual residents, including stability, autonomy, confidence, and freedom (Martin et al., 2020). The present analysis suggests that the homeownership benefits associated with SEH models provide low- and moderate-income households with the opportunity to access a wider range of places among those who leave the shared-equity program. Further political and financial support for SEH models has the potential to support both community development and individual household outcomes simultaneously, as it supports household autonomy and opportunity while simultaneously serving valuable community development objectives.