Introduction

In the United States, parent–child coresidence has been increasing for several decades. In the 1960s, for example, fewer than one-third of young adults (ages 18–29) were residing with their parents. By 2000, nearly 40 percent of young adults were living in the parental home, and rates have accelerated in recent years due to the combined effects of the Great Recession and coronavirus pandemic. According to recent estimates, more than half of young adults are now living with their parents (Fry et al., 2020). However, these are not just young adults that have always lived in the parental home, but rather large shares of those now coresiding are doing so after first experiencing a spell of residential independence.

The combination of increased rates of coresidence and the knowledge that many children coreside after first living in an independent household has resulted in more scholarly attention towards the correlates of the residential transitions of home-leaving (i.e., residential independence) and home-returning (i.e., boomeranging) (Gillespie, 2020; Gillespie & Lei, 2021; Houle & Warner, 2017; Lei & South, 2016; Sandberg-Thoma et al., 2015; South & Lei, 2015). For the most part, this research has focused on correlates of residential transitions out of and back into the parental home including contemporaneous life course transitions, individual well-being, and characteristics of the family context. Yet, there has been less work on the larger geographic context where home-leaving and home-returning transitions unfold (but see Bayrakdar & Coulter, 2018; Hughes, 2003).

Given that scholars have long argued for the importance of the larger geographic context for residential mobility and neighborhood (or locational) attainment (e.g., South & Crowder, 1997), it follows that these same features may also impact the timing or likelihood of home-leaving and home-returning transitions. After all, to leave home and live independently requires somewhere to go. As such, the primary contribution of the current study is to explore the associations between characteristics of the larger geographic context (i.e., housing and labor markets) and the likelihood of leaving and (among those who leave) returning to the parental home. Importantly, we also consider whether these place-level features matter differently for young adults who make these residential transitions closer to home (i.e., within the same county) or further away (i.e., across county lines).

Existing Research on Residential Transitions

Recent trends in coresidence with parents has been met with a growing body of research exploring correlates of both home-leaving and home-returning. In broad terms, this research has examined these residential transitions in the context of individual and family characteristics, including contemporaneous events tied to the life course, personal well-being, and family dynamics.

Individual and Family Context

For many adolescents and young adults starting their own residential trajectories, the first exit from the parental home is often tied to other life events that unfold during the transition to adulthood. For instance, completing high school can be a trigger for exiting the parental home, and is often coupled with transitions to college and entrance into the labor market (Goldscheider et al., 1999; South & Lei, 2015). Conversely, labor market struggles or dropping out of college can be triggers to return to the parental home (Houle & Warner, 2017; Sandberg-Thoma et al., 2015). Relationship transitions and trajectories are also tied to residential transitions, with entry into cohabitation or marriage fostering home-leaving and relationship dissolution fostering home-returning (Goldscheider et al., 2014; Houle & Warner, 2017; South & Lei, 2015).

Additionally, individual mental and physical well-being are associated with residential transitions, especially among those already living on their own. Residentially independent young adults who report higher levels of emotional distress or more frequent drug and alcohol use are more likely than their peers to return to the parental household (Sandberg-Thoma et al., 2015). Having a parent in poor health is also associated with a return home, as is the experience of physical and sexual victimization (South & Lei, 2015).

Finally, research demonstrates that the larger family context can impact young adult residential transitions (Gillespie et al., 2022). Earlier work, for instance, found that youth from single-parent families were more likely than their peers from two-parent families to exit the parental home (Buck & Scott, 1993; Goldscheider & Goldscheider, 1998). Beyond family structure, family resources have also been linked to the likelihood of residential independence. Youth from families with more resources may find that they can lean on their parents during the early period of residential independence to help maintain an independent household (South & Lei, 2015). Finally, the nature and quality of the parent–child relationship is also relevant, although research here has been somewhat inconsistent. Gillespie (2020), for instance, found that a strong bond between parents and children increased the likelihood of home-leaving, while South and Lei (2015) found that maternal closeness decreases the likelihood of home-leaving.

Geographic Context: Housing and Labor Market Characteristics

Extending previous research that focuses primarily on individual and family contexts, we investigate the role of place-level characteristics on young adult home-leaving and returning. The little extant research in this area suggests that features of housing and labor markets may influence the mobility decisions underlying this critical life course transition (e.g., Bayrakdar & Coulter, 2018; Hughes, 2003). Bayrakdar and Coulter (2018), for example, found that young adults in Britain were less likely to leave the parental home when living in unfavorable housing and labor markets. In a study of household living arrangements, Hughes (2003) likewise showed that pricier housing markets diminished the chances of becoming residentially independent. We therefore view these cost-of-living features as key geographic determinants of young adults’ decisions to leave or return to the parental home and build on these studies in two ways. First, we focus on local area characteristics among a nationally representative contemporary cohort of young adults in the United States. And second, given that many residentially young adults subsequently return to the parental home, we separately examine the residential transitions of home-leaving and home-returning.

Theoretically, we call on the housing availability model, which considers how geographic areas are able (or unable) to provide adequate and desirable housing for current and prospective residents. In other words, residents of areas with an abundance of available (and affordable) housing will be more mobile than their peers in areas with fewer housing options (South et al., 2011). Young adults exiting and potentially returning to the parental home comprise a highly mobile group, as the accumulation of life course transitions in young adulthood often results in shifting housing needs (Warner & Sharp, 2016).

Thus, a number of place-level features should be associated with both home-leaving and home-returning among young adults. First, assuming that few contemporary young adults are going to leave the parental home and enter directly into homeowner status, the sheer volume of available (i.e., vacant) housing in the county of residence should impact the decision to exit the parental home. More specifically, counties with higher proportions of vacant housing units are expected to foster pathways to sustained residential independence. Housing costs, specifically an affordable rental market, should also contribute to sustained residential independence, as young adults will accrue opportunities and be able to financially sustain living in the housing market beyond their first independent household. Additionally, prior research on the salience of housing availability for residential mobility decisions has also stressed the importance of new housing construction in potential destinations (South & Crowder, 1997). This feature may be particularly relevant for home-leaving, as recently built homes may be more attractive for young adults entering the labor force and searching for their first independent household.

Hypothesis 1

Living in counties with more available and affordable housing options will be associated with a higher probability of leaving the parental home and a lower probability of returning to it.

In addition to housing market characteristics, labor market dynamics have been shown to influence mobility decisions, including boomerang migration (Bayrakdar & Coulter, 2018; Chan et al., 2021; Choi, 2023). There is evidence that external shocks, such as the Great Recession and coronavirus pandemic, are associated with a lower propensity of leaving the parental home (Lee & Painter, 2013; Matsudaira, 2016), and an elevated risk of boomeranging (Chan et al., 2021; Choi, 2023). Labor market conditions, particularly the local unemployment rate, should be associated with home-leaving and home-returning, such that a weaker labor market will provide fewer opportunities to leave home and be associated with co-residing with parents (Bayrakdar & Coulter, 2018).

Hypothesis 2

Living in counties with fewer labor market opportunities will be associated with a lower probability of leaving the parental home and a higher probability of returning to it.

Finally, it is also important to consider that the influence of contextual features on residential transitions may be conditioned by the distance from the parental home. In particular, a long history of research has considered variation in the correlates of short-distance moves (i.e., moving within the same county) and long-distance moves (i.e., moving across county boundaries) (e.g., Rossi, 1955). It has been argued that human capital and labor market characteristics may be more relevant for migration decisions, with households willing to move longer distances for more suitable employment (Clark & Davies Withers, 1999, 2007; Kulu & Milewski, 2008). Conversely, shorter-distance mobility decisions are frequently triggered by changes in housing needs or to bring current housing in line with housing aspirations (Clark & Davies Withers, 1999, 2007; Rossi, 1955). Recent research also indicates that short-distance moves are often motivated by living in closer proximity to family members (Gillespie & Mulder, 2020; Spring et al., 2017). Labor market conditions could therefore be more impactful for leaving the parental home and moving to a different county, whereas housing market features could be more salient for leaving the home but remaining in the same county.

Hypothesis 3

Housing market characteristics will be more important for residential transitions that begin and end in the same county, while labor market characteristics will be more important for residential transitions that begin and end in a different county.

Existing theory and research thus suggest that some contextual features may work to facilitate home-leaving and sustained residential independence, while others may make it challenging to remain in an independent household leading to a move back to the parental home. We explore these potential associations using a combination of individual- and county-level data from a contemporary cohort of young adults in the United States.

Data and Methods

Exploring the associations between residential transitions (i.e., home-leaving and home-returning) and the geographic context requires longitudinal data on both individuals and their geographic areas. Individual data for this study are drawn from the National Longitudinal Survey of Youth—1997 cohort (NLSY97). Several studies have utilized the NLSY97 data to explore individual and family correlates of home-leaving and home-returning (Gillespie, 2020; Houle & Warner, 2017; Sandberg-Thoma et al., 2015; Warner & Houle, 2018; Warner & Remster, 2021). Data collection for the NLSY97 began in 1997 with 8,984 respondents ages 12–18. Respondents have been interviewed a maximum of 18 times, with the last round of data collection completed in 2019 (residential transition data available through the 2017 interview).

Access to restricted data through the Bureau of Labor Statistics provides county and state FIPS codes for each respondent at each interview wave. These codes are used to merge the public-use individual-level data with geographic data. To appropriately capture the housing and labor market dynamics that we feel may be relevant for residential transitions, we utilize geographic data at the county-level. These data are drawn from a combination of the US Census (1990, 2000, 2010, 2020) and supplemented with the 2016–2020 American Community Survey (ACS) 5-year estimates.

We restrict the original sample in multiple ways to create working samples for the separate residential transitions of home-leaving and home-returning. First, 430 respondents are dropped because they do not provide information on residential transitions, either because they did answer the question series, or they dropped out of the survey before the series was administered. An additional 143 respondents are dropped because they report leaving the parent home either before 1997 or before they turned 16 (when time at risk begins). For home-leaving, this results in a working sample of 8,411 respondents. The working sample for home-returning is first limited to those who report ever living independently and were part of the home-leaving working sample (n = 7761). An additional 154 respondents are dropped because they did not provide any information on subsequent returns to the parental home or dropped out of the survey, resulting in a working sample for home-returning of 7607 respondents. To account for missing data, we use multiple imputation in Stata using the ICE command (Royston, 2005). This procedure iteratively replaces missing values on all variables with predictions derived from random draws of the observed data, creating multiple complete data sets (Allison, 2001). We average results across ten imputation samples and account for random variation across samples to calculate standard errors (Royston, 2005). We also note that the multiply imputed results presented here are consistent with results based on listwise deleted data.

Residential Transitions

Residential transitions in the NLSY97 data are captured based on a series of questions where respondents are asked to reflect on when they were first part of an independent household and (for many) when they returned back to the parental home (Dey & Pierret, 2014). Respondents are first prompted with a definition of a “permanent household” to ensure that (1) respondents are working off a shared conceptualization of residential independence and (2) residences that many consider to be temporary (e.g., college dorms) are excluded. With this conceptualization in mind, respondents are then asked about “living on their own.” Starting retrospectively in 2003 and then updated at each subsequent interview, respondents are asked if they have been the head of a household, or shared that role equally with others, for a period of at least 3 months. Those who answer in the affirmative are asked a follow-up question about returning to the parental home. This question prompts respondents to indicate if they subsequently moved back in with their parents or into someone else’s household, again for a period of at least 3 months.Footnote 1

Two additional features are key for the present study. First, respondents who report either transition of home-leaving or home-returning are asked to report the year the transition occurred. This means that the data can capture the year of first residential independence and the year of a subsequent return to the parental home, which allows us to match those self-reported transitions to time-varying and year-specific characteristics at the individual- and county-levels. Second, access to the restricted geographic data allows us to examine the geographic scope of the residential transitions. Prior research has shown that short- and long-distance residential moves can be motivated by different factors (Pelikh & Kulu, 2018; Pendakur & Young, 2013), and residential transitions are generally instances of residential mobility. Because we have geographic identifiers at each wave, we further distinguish residential transitions as same- or different-county. A same-county residential transition occurs if a respondent reports either home-leaving or home-returning, and their county in the year of the transition is the same as their county at the preceding interview (t − 1). A different-county residential transition occurs if a respondent reports home-leaving or home-returning, and their county in the year of the transition is different than the county at the preceding interview.Footnote 2

Housing and Labor Market Characteristics

Based on existing theory and research, we focus on four county-level characteristics that capture a county’s cost-of-living, housing, and labor markets. To create these measures, we use data from a combination of the US Census and the ACS. To align with the longitudinal nature of the NLSY97 data, we used linear interpolation to estimate county measures in non-Census years. The local housing market is measured by the percent of housing units that are vacant (i.e., the vacancy rate) and the percent of all housing units built within the last 10 years (i.e., new housing construction). Cost-of-living related to the housing market is captured with a measure of gross median rent in the county of residence (in hundreds of 2010 dollars). Finally, the local labor market is captured by the county unemployment rate.

Control Measures

While not a focal measure of the current study, we include two county-level control variables emphasized by the housing availability model (South et al., 2011). First, we control for the county’s population size (logged). Second, we control for the percent of the county population that is non-White.

Following prior research, we also control for a range of time-stable and time-varying individual characteristics shown to impact residential transitions. We account for gender (female = 1) and race (white = referent) as prior research demonstrates gender and racial/ethnic variation in the timing and likelihood of residential transitions (Lei & South, 2016; Stone et al., 2013). Prior research also demonstrates some broad regional trends in residential transitions (South & Lei, 2015), so we control for a time-varying indicator of Census region (Northeast = referent). Given the importance of the family context in residential transitions (Gillespie, 2020), we include three measures of parents and families. The first is parental attachment, a composite measure of respondent’s responses to three questions: respondent thinks highly of mother/father, respondent wants to be like mother/father, and respondent enjoys time with mother/father. Because this measure is only available at five time points (1997, 1998, 1999, 2001, and 2003), we take a respondent’s average parental attachment across available waves. We also control for parental education (high school degree or less = referent) and background family structure (two-parent = referent).

Key determinants of residential transitions are contemporaneous life course transitions and statuses (Sandberg-Thoma et al., 2015; South & Lei, 2015). As such, we include time-varying measures of age, earnings (thousands of dollars), employment status, parental status, marital status (unmarried = referent), and educational attainment (less than high school = referent). Prior research also shows that individual well-being is associated with residential transitions (Sandberg-Thoma et al., 2015), and so we control for emotional distress, a summary measure of the amount of time during the last month that respondents reported (1) being nervous, (2) feeling calm and peaceful (reverse coded), (3) feeling downhearted/blue, (4) being a happy being (reverse coded), and (5) feeling so down that nothing could cheer them up. Responses for each ranged from 1 (all of the time) to 4 (none of the time), and higher values indicate greater emotional distress. Finally, as an additional indicator of parent–child dynamics, we control for a time-varying measure of instrumental support from parents, coded 1 if a respondent reported receiving financial support from parents in the past year.

Analytic Strategy

We follow prior research by separately predicting home-leaving and home-returning using discrete time proportional hazard models (Gillespie, 2020; Warner & Houle, 2018). Respondents are at-risk for home-leaving in the survey wave corresponding to their 16th birthday, and remain in the risk set until the year they report first leaving the parental home or the time series ends with the 2017 interview. Respondents in our working sample over the age of 16 at the first wave (n = 519) are included in the study if they had not yet left the parental home (with time at-risk starting in 1997 at their first interview), and are omitted if they reported already having lived independently prior to 1997 (n = 18). By isolating the year when a respondent turns 16, we can then measure time-varying individual- and county-level control variables using the year that corresponds to the time at-risk. For those respondents who establish independent households, the time series for home-returning begins in the year they indicate first leaving the parental home. Respondents remain at risk until they return home, or the time series ends in 2017. Again, time-varying controls are taken in the year corresponding to each year that a respondent is at-risk for returning home. In models predicting same-county and different-county residential transitions, we use a multinominal regression approach.

Across models, we conducted exploratory analyses to determine the proper functional form of the baseline hazard. Among the various options, including linear and non-linear measures for time at-risk, a fully nonparametric model (with dummy variables for time at-risk) was the best fit for the data. All models are estimated in Stata and standard errors are clustered by county to adjust for nonindependence of observations among respondents of the same county. As noted above, all results are based on multiply imputed data.Footnote 3

Results

The presentation of results begins with descriptive information on home-leaving, home-returning, and county-level covariates of interest (Table 1). Multivariate discrete time hazard models then explore the associations between the local housing and labor markets and the likelihood of experiencing the residential transitions of home-leaving (Table 2) and home-returning (Table 3). Each of these tables also contains results from multinomial models predicting same- or different-county residential transitions. To provide more context on types of transitions, and the role of county-level characteristics, we also plot average marginal effects of interest in Figs. 1 (home-leaving) and 2 (home-returning).

Table 1 Descriptive statistics for home-leaving and home-returning samples
Table 2 Discrete time hazard models predicting home-leaving (source: NLSY97)
Table 3 Discrete time hazard models predicting home-returning (source: NLSY97)
Fig. 1
figure 1

Predicted probability of home-leaving to the same-or a different-county based on levels of country characteristics. Derived from versions of Table 2 with focal county- characteristics recoded to levels based on percentile distribution, time at-risk set to observation 6 (peak observation for home-leaving across sample), and all other individual and county control variables set to mean values

Fig. 2
figure 2

Predicted probability of home-returning to the same-or a different- country based on levels of country characteristics. Derived from versions of Table 3 with focal county-characteristics recoded to levels based on percentile distribution, time at-risk set to observation 2 (peak observation for home-leaving across sample), and all other individual and county control variables set to mean values

Descriptive Results

Based on their own reports, a clear majority of NLSY97 respondents leave the parental home and establish an independent household at least once. A total of 7693 respondents (or 91.5% of the home-leaving working sample of 8411 respondents) report leaving the home and living independently for a period of at least three months. And of this group, many subsequently return to the parental home. Among those who achieve residential independence, 59.7% report returning to the parental home for a period of at least three months (4539 of the 7607 respondents who are in the home-returning working sample). Additional descriptive statistics, based on imputed data and structured by the samples at-risk for home-leaving and home-returning, are presented in Table 1.

All descriptive statistics in Table 1 are in the person-year format, and thus represent averages across all waves of available at-risk data. As such, the results indicate that about 15% of all respondents report leaving the parental home in any given year they are at-risk for home-leaving. Respondents at-risk for leaving home live, on average, in counties where 8.96% of all housing units are vacant and 16.17% of all housing units were built in the last 10 years. While at-risk for leaving home, respondents lived in counties with a median gross rent of approximately $775 (in 2010 dollars) and an unemployment rate of 6.45%.

Table 1 also shows that, following their exits from the parental home, approximately 9% of residentially independent young adults return to the parental home in any given year they are at-risk for home-returning. Across all home-returning at-risk observations, respondents live in counties where 9.93% of all housing units are vacant, and 15.76% of all housing units were built in the prior 10 years. The median gross rent is consistent across the home-leaving and home-returning samples, while the unemployment rate is higher in the years respondents are at-risk for home-returning compared to the years they are at-risk for home-leaving. While living independently, respondents’ counties, on average, have a median gross rent of approximately $774 and an unemployment rate of 6.88%. While not shown in the table, we also note that, descriptively, most transitions unfold in the same county. A total of 80.4% of home-leaving transitions and 74.5% of home-returning transitions do not cross county lines.

Multivariate Results

To determine if county-level measures are associated with residential transitions, and especially net of established individual-level correlates of these transitions, we turn to discrete time event history models and multinomial discrete time models predicting home-leaving (Table 2) and home-returning (Table 3). In each table, Model 1 predicts all residential transitions, based on the focal county-level covariates, the hazard, and all time-stable and time-varying control measures. Models 2 and 3 predict same- and different-county residential transitions, compared to the base outcome of no residential transition. To conserve space, we present only coefficients for our variables of interest, and have included coefficients for control measures in the electronic supplementary materials.Footnote 4

Model 1 of Table 2 shows that, when considered alongside established correlates of residential transitions, only rental costs are significantly associated with the likelihood of home-leaving. While the coefficients for all other county measures are in the anticipated directions, they fail to reach conventional levels of statistical significance. Thus, when considering all instances of home-leaving, respondents are less likely to leave the parental home as the housing costs of the local area increase. Respondents residing in counties with higher median rents are less likely than their peers to exit the parental home.

When we separate home-leaving transitions by their geographic scope (i.e., same- or different-county), however, there is more evidence of the role of county-level characteristics. This is especially the case for same-county home-leaving transitions, which account for the majority of all home exits. Summarized in Model 2, the results show that, as anticipated, residing in areas with more vacant housing units is associated with a higher likelihood of leaving home but staying in the same county. The negative association between rental costs and all home-leaving transitions is concentrated in these same-county residential transitions, and the county unemployment rate is also tied to these same-county exits. More specifically, respondents residing in counties with higher rental costs and greater unemployment rates are less likely than their peers to leave the parental home and remain in the same county.

Among the smaller group of home-leaving transitions that cross county lines (Model 3), only new housing units is a significant predictor. Increases in new housing units are associated with an increased likelihood that a respondent exits the parental home to a different county. Finally, because the reference category for the coefficients in Models 2 and 3 is no home-leaving, we compared the associations between the county-level measures and the same- and different-county home exits. In the table, bolded coefficients and standard errors are significantly different across the types of residential transitions based on a test of equality of coefficients (Paternoster et al., 1998). The associations between vacant housing units and types of home-leaving are not significantly different, but all other county-level associations are significantly different across same- and different-county home-leaving. New housing units are especially relevant for different-county home-leaving, while rental prices and unemployment rates are especially relevant for same-county home-leaving.

To provide additional context to these findings, we plot predicted probabilities of home-leaving to the same or a different county at specified levels of our structural variables of interest in Fig. 1. The figure has two panels, one for same-county home-leaving (Panel A) and one for different-county home-leaving (Panel B). In each panel, predicted probabilities are based on the percentile distributions of the county measures, and we focus on the low (< 25th percentile) and high (> 75th percentile) points of the distributions. All model covariates from Table 2 are set to their mean values with the exception of the plotted county measure and the time at risk (which is set to time 6 when the likelihood of home-leaving peaks). For instance, Panel A of Fig. 1 contains eight distinct predicted probabilities, with each solid bar representing the probability of home-leaving to the same county based on the low-level of the county characteristics and each striped bar representing the probability of home-leaving to the same county based on the high-level of the county characteristics. An asterisk indicates if the difference between the low- and high-levels is statistically significant (based on a comparison of the 95% confidence intervals).

While vacant housing units, rental costs, and the unemployment rate were all significant predictors of same-county home-leaving, the results in Panel A show that the most meaningful differences between low- and high-levels of these characteristics are for vacant housing units and median rent. The predicted probability of home-leaving, at the time when it is most common, is 18.9% higher for respondents living in areas with high-levels of vacant housing units compared to respondents living in areas with low-levels of rental units. The probability of home-leaving for those residing in areas with low-levels of median rent is 21.1% higher than the probability for those residing in areas with high-levels of median rent.

As reflected in Table 2, the probabilities in Panel B of Fig. 1 show that only new housing units appear to make a significant difference in leaving home to a different county. Here, respondents in areas with high-levels of new housing units are significantly more likely than their peers in areas with low-levels of new housing units to leave the parental home to a new county. The two panels also demonstrate that same-county home-leaving is much more common than different-county home-leaving. Overall, then, the results for home-leaving provide some support for the expectation that structural features play a role in the pathways that young adults take out of the parental home. To determine if these structural features are also connected to the likelihood of returning to the parental home, we turn to hazard models predicting home-returning in Table 3.

Model 1 of Table 3 indicates that, in a fully adjusted model, again only rental costs are linked to the likelihood of home-returning. Much like with the results for home-leaving, all other structural features are in the anticipated direction, but only the coefficient for median rent is statistically significant. When all home-returning residential transitions are considered together, respondents in areas of high rent are more likely to return home (net of their income and instrumental support from parents, among other controls) than their peers in areas of lower rent.

As with home-leaving, however, separately exploring same- and different-county home-returning provides a more complete picture, and these results are summarized in Models 2 and 3. New housing stock protects against a transition back home among those living in the same county in the preceding year, while fostering home-returning among those who were living in a different county. The link between median rent and home-returning observed for all transitions appears to be most concentrated among returns for those living in a different county. Based on tests of equality of coefficients, the associations between new housing units and median rent are significantly different based on the type of home-returning transition. For new housing units, the association is negative for youth returning home from the same county and positive for youth returning home from a different county. For rent, the association is positive for both types, but significantly larger for different-county home returns.

The relevance of median rent for these longer distance returns becomes more apparent when we consider the likelihood of home-returning at low- and high-levels of our county indicators in Fig. 2. As with home-leaving, we examined home-returning by the 25th and 75th percentiles of each county indicator, and the results are shown in two panels to capture same-county (Panel A) and different-county (Panel B) returns. The differences between low- and high-levels of new housing units and local unemployment near, but do not reach, statistical significance for same-county returns (Panel A). While residents of areas with low-levels of new housing units are 14.9% more likely than their peers in areas with high-levels of new housing units to return home—the 95% confidence intervals slightly overlap for these estimates. Panel B, however, shows a larger and significant difference in returning home from a different county based on rental costs. Respondents in areas of high median rent are 63.8% more likely than their peers in areas of low rent to move back home from a different county.

Discussion

Coresidence with parents has become the most common living outcome among American young adults aged 18–29 (Fry, 2016). This is the result of a long-term shift in residential transitions that, for many, includes a return to the parental home after an initial spell of residential independence. To understand the individual and family dynamics that drive these changing patterns in home-leaving and home-returning, a growing body of research has taken up the issue of contemporary residential independence and boomeranging back to the parental home (Gillespie, 2020; Houle & Warner, 2017; Sandberg-Thoma et al., 2015; South & Lei, 2015; Stone et al., 2013). With a few exceptions (Bayrakdar & Coulter, 2018; Hughes, 2003), this research has largely neglected the fundamental dynamic that home-leaving and home-returning are largely processes of mobility and migration. And as residential moves, home-leaving and home-returning take place in a context that varies from place to place with respect to features such as the local housing and labor markets. The present study thus builds on existing research with a more thorough examination of the county-level correlates of home-leaving and home-returning among a contemporary cohort of American young adults.

Across our models predicting home-leaving and home-returning, the results provide mixed support for our expectations. We hypothesized that more available and affordable housing options would increase home-leaving and decrease home-returning. After accounting for time-stable and time-varying individual correlates of home-leaving, along with potential confounders at the county-level, the results indicate that young adults are more likely to leave home if they live in a county with lower median rental costs. Also in support of Hypothesis 1, higher average rents are associated with an increased likelihood of returning back to the parental home. That rental costs play a role in residential transitions is consistent with findings using British data (Bayrakdar & Coulter, 2018), and has implications for the residential trajectories that are core to the transition to adulthood. Few American young adults, especially contemporary young adults, will be in a position to launch from the parental home into a home they own independently or with others. Indeed, nearly two-thirds of those 35 and under in the US are renters (Desilver, 2021). That said, current young adults in the US are entering housing markets where rental costs are increasing faster than wages, leading to high rates of rent burden for younger households (where more than 30% of income goes towards rental costs) (Pew, 2018). Thus, there appear to be few American housing markets that can foster successful home-leaving for their young adults.

Our second hypothesis was that tighter labor markets (as measured by the unemployment rate), would decrease home-leaving and increase home-returning. While the relationship between the unemployment rate and all home-leaving and home-returning transitions was in the anticipated direction, it appears that the unemployment rate is only relevant for those home-exits that are local, or that start and end in the same county. Contrary to Hypothesis 3, a higher unemployment rate decreases the likelihood of same-county home-leaving, but is largely unrelated to different-county residential transitions. We anticipated that characteristics of the labor market would impact longer distance residential transitions because longer distance migration has been tied to these characteristics (Clark & Davies Withers, 1999). We suspect that this might reflect the life stage of these moves, especially leaving the parental home for the first time. The local (i.e., same-county) unemployment rate may play a larger role for same-county home-leaving because youth leaving the home for the first time often stay in the same county and need employment to meet their new housing costs.

Our final expectation, driven by research on residential mobility and internal migration, was that housing market characteristics would be more relevant for short distance residential transitions and that labor market characteristics would be more relevant for long distance residential transitions. Here, the findings were more mixed. As anticipated, higher levels of vacant housing units and lower rental costs are associated with a greater likelihood of leaving the parental home and remaining in the same county, while new housing units are associated with a decreased likelihood of returning to the same county. On the other hand, new housing units are associated with an increased likelihood of both home-leaving and home-returning to a different county, and higher rents also increase the likelihood of returning from a different county.

These mixed findings related to Hypothesis 3 show that housing and labor market features may be weighed differently for young adults depending on the nature and distance of their residential transition. Newly constructed housing, especially, is linked to different county home-leaving and home-returning, which could suggest that young adults may view this type of housing as more desirable when considering longer-distance exits from the parental home. In addition, we speculate that new homes may foster cross-county home-returning to the extent that boomeranging children view the parental home as a short-term housing arrangement. That is, new housing units in the area of the parental home may be an enticement for those previously living farther from home. Additional research should explore how residential trajectories unfold for those who do and do not return home, and consider if there is variation based on how far the child moved away for their first residential independence.

These findings should be considered alongside some limitations with the data and approach. First, the data are drawn from a single cohort in the United States, and as a result, may not be representative of other cohorts or contexts. Second, the wording of the questions to capture residential transitions include the phrase “someone else’s household” when querying respondents on returning to the parental home. It is therefore possible that respondents coded as having boomeranged may have exited residential independence, but may not have moved back in with parents. That said, these questions have been utilized by the Bureau of Labor Statistics (the organization charged with overseeing the data) and published research to capture the residential transitions of first residential independence and subsequent boomeranging back to the parental home (Dey & Pierret, 2014; Gillespie, 2020). Finally, it is worth mentioning that our use of county of residence across interviews to capture the geographic scope of residential transitions may include moves that are relatively short in distance. This is especially the case for respondents of large urban areas where counties are smaller compared to less populated area. Additional research should take a closer look at the scope of home-leaving and home-returning moves, and especially variation in the correlates of these moves at both the individual- and structural-levels.

Conclusion

Even with these limitations, results from this study extend prior research on contemporary residential transition out of and back into the parental home in important ways. Much research in this area has focused on correlates of home-leaving and home-returning rooted in life course transitions, individual well-being, and the background family context. Our primary contribution is that housing costs (reflected in average rental costs) are an important correlate of residential transitions, and that housing availability plays a role for residential transitions that are distinguished by their geographic scope. Given the importance of the rental market in the housing careers of young adults, our primary contribution that higher rental costs impede residential independence has important implications for the contemporary transition to adulthood.