We use household survey data from the UK to study how close middle-aged men and women in partnerships live to their parents and their partner’s parents. We find a slight tendency for couples to live closer to the woman’s parents than the man’s. This tendency is more pronounced among couples in which neither partner has a college degree and in which there is a child. In other respects, proximity to parents is gender-neutral, with the two partners having equal influence on intergenerational proximity. Better-educated couples live farther from their parents. And although certain family characteristics matter, intergenerational proximity is primarily driven by factors affecting mobility over long distances, which are mainly associated with the labor market, as opposed to gender or family circumstances.
An important determinant of the provision of in-kind help by adult children to parents, or vice versa, is the geographical proximity of the two generations. For people with a live-in partner, either married or cohabiting, intergenerational proximity reflects, in principle, the decisions of at least three sets of agents: the couple (who must negotiate between themselves) and two sets of parents.1 In practice, moving costs are often nonnegligible, and there are frictions in residential moves, making prior location choices important. In particular, local ties that are developed over time by living in the same place increase the costs of longer-distance mobility (McGinnis 1968).2 For this and other reasons, most residential moves are made by younger people in their 20s and 30s; and, in the UK at least, there is very little retirement mobility (see Chan and Ermisch forthcoming: figure 2). Given this, we will assume that parents are relatively fixed in their locations and that intergenerational proximity is driven primarily by children’s location decisions. The focus of this article is the negotiation between middle-aged married or cohabiting partners.
In general, partnership formation reduces mobility because movement needs to be negotiated between two people. However, the initial negotiation positions of the partners differ from case to case. For example, two people may both have moved, for independent reasons, to a new location where they meet and eventually form a partnership. In this case, the partners are on equal footing in terms of local ties. Alternatively, one partner might have moved into the local area of the other; in this case, the latter is likely to have stronger local ties and higher mobility costs than the former. We will elaborate this point in the upcoming section illustrating our hypotheses. Suffice it to say that because our data are cross-sectional, we cannot study the dynamics of intergenerational proximity directly.3 Instead, our focus is the eventual outcome of those dynamics.
This limitation of our data notwithstanding, an analysis framing residential location as a couple’s decision has at least two advantages over much of the existing literature, which is mainly individual-based (e.g., Hank 2007; Rainer and Siedler 2009, 2012; Shelton and Grundy 2000), save for a few notable papers (Blaauboer et al. 2011; Compton and Pollak 2013; Løken et al. 2013). First, the characteristics of both partners can be taken into account, potentially providing better estimates of how individuals’ circumstances affect intergenerational proximity. Second, we may be able to make inferences about the relative influence of the two partners.
Features of residential location other than proximity to parents are, of course, important in location decisions. These are in turn influenced by individual and household circumstances as well as the partners’ preferences and bargaining power. In particular, previous research has consistently found that better-educated individuals are geographically more mobile (see, e.g., Machin et al. 2012). For instance, in the data that we analyze herein, 6.6 % of couples in which both partners have a university degree moved in the following year, compared with 3.9 % of those in which neither partner has a degree. Given that residential moves tend to increase intergenerational distance (Rogerson et al. 1993), our expectation is that among educationally homogamous couples, those with more education live farther from their parents,4 perhaps because they operate in a geographically wider labor market, because they left home to study and partnered with someone far from their parental home, or for other reasons. As we will explain, educationally homogamous couples might serve as reference points in the analysis of the location decisions of heterogamous couples.
Each partner’s influence on residential location is the subject of a literature in economics and geography, much of which stemmed from Mincer’s (1978) seminal paper on family migration. One view, based on the assumption that the man’s career is more important for the couple’s resources, is that the man is dominant in location decisions that involve comparing different labor markets, and perhaps also in decisions about different locations in the same labor market. If both partners desire to live near their own parents, then male dominance would lead to locations that are, on average, closer to his parents. Such a tendency might be tempered by the fact that most residential moves are of a short distance, making it possible that women’s preferences are given more weight when considering different locations within a labor market. Because daughters usually have more contact with and give more help and care to parents than sons, women may have a stronger preference than men to live near parents. As Blaauboer et al. (2011) argued, the net effect of gender differences in power and in the strength of family ties could favor proximity to either the man’s or the woman’s parents.
An alternative view is that the bargaining power of the two partners is a function of the resources that they bring into the partnership. This implies that the weight given to the woman in location decisions (local as well as longer distance moves) would be increasing in her earning power relative to her partner’s.5 Løken et al. (2013) discussed bargaining power effects in the context of proximity to parents. Because it is relatively persistent differences in earning power that are important, we distinguish couples by their educational attainment. Provided that both partners prefer to live close to their own parents, we expect that the couple would, for given employment opportunities, live closer to the parents of the better-educated partner.
Our main research question is which partner’s education is more important in determining distance to parents, and in what direction. In addition to its intrinsic interest, an answer to this question sheds light on the relative bargaining power of men and women, and on the role of the labor market in location decisions. We also explore which individual and couple attributes other than gender and education affect intergenerational proximity.
We find a slight tendency for couples to live closer to the woman’s parents than to the man’s. However, a model postulating equal weights on the educational levels of the two partners fits the data very well. Equal weights strongly suggest that there is no bargaining effect that works through the partners’ educational levels. Better-educated couples tend to live farther from their parents. Certain family circumstances do predict intergenerational proximity. In particular, the presence of children favors locations that are closer to the woman’s parents. Overall, we conclude that intergenerational proximity is primarily driven by factors that affect mobility over longer distances (i.e., to places outside the local area), which are mainly associated with the labor market, as opposed to gender or family circumstances.
Previous Literature on Couples
Three studies have directly addressed the proximity of couples to the two sets of parents. Here, we focus on the nature of their data, the characterization of partners’ education, and the impacts of education on proximity, discussing other comparisons with our results later in this article. Blaauboer et al. (2011) used data from the Netherlands Kinship Panel Survey for persons aged 18–79. This survey collects information on the x – y coordinates of location, allowing the computation of the linear distance between the couple and parents. A disadvantage is that there is only one respondent who provides information on socioeconomic variables (e.g., education) and residential location for parents and parents-in-law as well as for himself/herself. The male partner’s education is found to have stronger effects on distance to both sets of parents than the female’s.
Compton and Pollak (2013: table 8) offered a relatively limited analysis of couples’ proximity to each partner’s mother (both alive living in the United States), using the National Survey of Families and Households, for persons aged 25 and older. They estimated a multinomial logit model distinguishing coresidence, near mother (within 30 miles), and farther (30 miles or more). Partners’ education is categorized as both having a college degree, one partner having a degree (two categories distinguishing which partner has it), and neither having a degree. Couples in which one or both partners have a degree live farther from their mothers, with the impact being weaker if only the woman has a degree.
Løken et al. (2013) used Norwegian registry data to study each partner’s distance from parents for a sample of women aged 34 who were married and whose parents did not live in the same postcode as their partner’s parents. They measured distance in the following categories relative to parents: same postal code, same municipality, same county, same region, or different region. The couples’ analysis was in terms of a multinomial logit for relative distance: same category, nearer to her parents, or nearer to his parents. The data distinguished between the college-educated and the rest, and the effect of husband’s education on proximity was several times larger than that of the wife.
Our study makes a number of contributions. First, it uses a large and nationally representative household survey for the United Kingdom and focuses on middle-aged (31- to 54-year-old) couples, who are more likely to be called upon for help from parents. Second, it uses self-reports from each partner, including their education and temporal distance from parents, as well as aspects of their history, including childbearing, number of siblings, and experience of divorce as a child. Third, it uses a model that provides a parsimonious parameterization of the impacts of each partner’s education, which allows a simple test of the relative influence of the two partners’ educational levels on proximity to their parents.
We analyze data on married or cohabiting couples interviewed in a new household panel survey from the UK: Understanding Society. In this nationally representative survey, all individuals aged 16 or older in the sampled households are interviewed annually. Individuals leaving their households are followed, and all adult members of the new households are also interviewed. The data collection of each wave, using computer assisted personal interviewing (CAPI), lasts 24 months; the first wave of data collection started in January 2009 and finished in January 2011. At present, four waves of data are available. In this article, we use data from the first wave, in which nearly 51,000 individuals were interviewed.6
This survey contains many questions that are salient for studying intergenerational proximity. Because all adults in the sampled households are interviewed separately, we have, for each member of the couple, self-reported information about proximity to his or her own parents. In this article, we focus on middle-aged heterosexual couples in which both partners have at least one living parent and the woman is between ages 31 and 54 (N = 3,816 couples). We focus on couples in this age range because their parents are of the ages (roughly mid-50s and older) when they may require in-kind help. Moreover, we wish to focus on those who have already completed their transition to adulthood. A significant proportion of younger people in their 20s have not yet left their parental home,7 and their location decision merits separate treatment in another study (cf. Løken et al. 2013).
We use data on household composition and information for noncoresident relatives to ascertain whether the respondent has a living mother or father. People with a mother (father) living outside the household are asked, “About how long would it take you to get to where your mother (father) lives? Think of the time it usually takes door to door.” Among all people aged 31–54, 17 % do not have a living parent, 49 % have both parents alive, 26 % have only a living mother, and 8 % have only a living father. And among those with two living parents, 87 % report that their parents are in the same proximity category, in large part because the parents live together; in 10 % of the cases, the mother lives closer to the child than does the father, whereas the opposite holds for the remaining 3 %. When the two parents are not living together and are in different proximity categories, we focus on the parent who lives closer to the child.
Table 1 illustrates the proximity data used in the analysis.8 The first two columns are for the full sample, and the last two columns are for a subsample in which both partners are white and UK-born, which excludes about one-third of the couples. In the full sample, a little more than one-half of the respondents live within 30 minutes of traveling time to their parents, but the proportion is higher among white, UK-born couples. About one-quarter of the full sample live more than two hours away from their parents, including those living abroad. The proportion is much smaller for the white, UK-born sample because it is much less common for their parents to be living abroad.
Table 1 also shows a slight tendency for the couples to live closer to the woman’s parents than the man’s. Indeed, in a cross-tabulation of distance to the two sets of parents (using the proximity categories of Table 1; not shown), we find that 33 % of the couples in the full sample live closer to the woman’s parents than to the man’s parents, but the opposite is true for 30 % of the couples.9 This pattern contrasts with findings from the Netherlands (Blaauboer et al. 2011) and Norway (Løken et al. 2013), but is qualitatively similar to American findings (Compton and Pollak 2013).10
In our main analyses, we do not work with the detailed proximity data shown in Table 1 because there is considerable interest in whether a person lives close enough to parents to see them frequently and to provide and receive help.11 Thus, we dichotomize the distance categories, contrasting coresident or within 15 minutes of traveling time (“near” parents, in short)12 against traveling time of 15 minutes or more (“far” from parents).
Better-educated people are likely to face a distribution of earning opportunities that has a larger variance, allowing them to be more choosy in the jobs that they accept and causing them to search longer and over a wider geographical area. Job opportunities requiring a higher level of education may also be more dispersed geographically. The higher income and greater wealth of the better-educated could also lead them to search for housing opportunities over a broader area. These tendencies lead us to expect that, all else being equal, better-educated people live farther from their parents. This is a clear prediction for educationally homogamous couples.
When the partners’ educational levels differ, one partner may have more say on the outcome. For instance, if the man works more hours and contributes more to household income, his education may hold more weight in location decisions. Another possibility is that the impact of each partner’s education is proportional to the earnings return to job search over a wide area corresponding to that educational level. In this case, the combined impact of the two partners’ education in proximity decisions may be an average of their educational levels, with possibly different weights for men and women. For example, with equal weights, couples in which the partners have a combination of a high and middle level of education will tend to live farther from both sets of parents than couples with a high and low level of education. Such an outcome would indicate the importance of labor market influences in proximity decisions.
Different educational levels may also affect bargaining power. Of course, we know little about people’s preferences for proximity to parents. These preferences may differ by gender (e.g., women may favor being near parents more than men). But if people generally prefer to live near their own parents, bargaining based on relative educational levels would produce countervailing effects to the labor market influences discussed in the previous two paragraphs. For example, among couples in which the woman is better educated than the man, a higher proportion of the couples would live closer to her parents than to his, compared with couples in which the same educational difference favors the man. In other words, there might be an asymmetry in relative proximity to the two sets of parents, depending on which partner is better educated.
A more specific bargaining-oriented indicator is the share of the woman’s income in the couple’s joint income, which averages 39 % in our data. For given partners’ educational levels, a woman earning a larger share of the joint income may be expected to live closer to her own parents.14
A different argument for educationally heterogamous couples arises from pre-partnership migration. Suppose, for exposition purposes, that single people move away from the area in which they grew up if and only if they obtain a degree.15 If this is the case, then partnerships formed between a university graduate and a nongraduate will be in the local area of the latter, who has higher moving costs because of local ties, including parents. As a consequence, such couples tend to live closer to the parents of the partner who is less well-educated. This line of reasoning also predicts an asymmetry in relative proximity to the two sets of parents but, contrary to the bargaining power argument, suggests that the couple will live closer to the parents of the less-educated partner.
Some circumstances other than the partners’ education and relative resources, such as the presence of children, may also influence relative proximity to parents. For example, daughters’ stronger family ties may encourage couples with children to live closer to her parents. We investigate such influences using the rich data on household characteristics and some aspects of each partner’s history, such as whether each has siblings and experienced parental separation as a child.
In the analyses presented here, we first explore proximity to woman’s parents and man’s parents separately (in each case, using the near–far dichotomy). Then we repeat the analysis using the joint distribution of distance to both sets of parents (i.e., X1, . . . , X4 in Fig. 1) as the dependent variable.
We distinguish three levels of education: high (bachelor’s or higher degree), intermediate (A-levels or further education, yet still less than a bachelor’s degree), and low (General Certificate of Secondary Education (GCSE) or lower).16 Table 3 shows that each of the three educational levels accounts for roughly one-third of men and women, with women being slightly better educated than men. Moreover, just over one-half (53 %) of the couples are educationally homogamous (i.e., found in cells on the main diagonal), and there are slightly more couples in which the woman is better educated than the man (25 %, found in cells above the main diagonal) than the other way around (22 %, below the main diagonal).
Table 4 reports, for each combination of the partners’ education, the proportion of couples living far from the woman’s parents, those living far from the man’s parents, and also the joint distribution of distance to both sets of parents. Of particular interest are the three pairs of situations in which the partners’ educational levels differ. In every case in which the woman is better educated than the man, the couple is more likely to live far from her parents (see the column labeled “Woman’s Parents”), compared with situations in which the educational difference is reversed. Similarly, if the man is better educated than the woman, the couple is more likely to live far from his parents (see the column labeled “Man’s Parents”). Turning to the joint distribution of distance to both sets of parents, when the woman is better educated than the man, the couple is less likely to live near her parents and far from his parents than if the educational difference is reversed (see the column labeled X3). Indeed, in two of the three cases, the couple is actually more likely to live far from her parents and near his parents (see the column labeled X4).17 These differences, albeit very small in most cases, cast doubt on the importance of bargaining power effects working through relative educational levels. Instead, they lend preliminary support to the pre-partnership migration argument.
Diagonal Reference Models
Our goal is to find a parsimonious model to summarize the data in Table 4. Before we turn to formal modeling, however, we note that in about one-third of the couples in our full sample, at least one partner is either foreign-born or nonwhite (see Table 1). Among this group, 44 % of the couples have at least one partner with a parent who is living abroad. For them, the choice of distance to parents is severely constrained. It is also possible that ethnic differences exist in the relative influence of partners’ education and in education group means, which may affect our estimates.18 Thus, in the formal modeling here, we restrict our analysis to a sample of 2,506 couples in which both partners are white and UK-born.19
It turns out that in the model predicting the log odds of living far from the woman’s parents (top panel), w = .41, (SE = .09) and w′ = .66, (SE = .10), which is consistent with the bargaining power argument. However, in the model predicting the log odds of living far from the man’s parents (middle panel), w = .32, (SE = .10) and w′ = .63, (SE = .11), which supports the pre-partnership migration argument. However, note that the CI of w and w′ still straddle .50 in all cases. Moreover, compared with Model 1, the deviance of Model 3 is reduced by 2.49 for distance to woman’s parents (top panel), and 2.95 for distance to man’s parents (middle panel). These are not statistically significant improvements for 1 degree of freedom (p = .11 and p = .09, respectively). Thus, there is no clear support for either the bargaining power argument or the pre-partnership migration argument. Overall, we prefer Model 1 to Model 3, and Model 2 to Model 1.
Our preferred Model 2 has strictly equal weights for men and women. Thus, contrary to the findings of Blaauboer et al. (2011) for the Netherlands, Løken et al. (2013) for Norway, or Compton and Pollak (2013) for the United States, we do not find that men’s education has a larger impact on relative proximity to parents. Instead, our result is consistent with the idea that the impact of each partner’s education is proportional to the earnings return to wide geographic job search for that level of education, and that the combined impact of the two partners’ education is a simple average of the impacts of the two educational levels. These results point to the importance of labor market considerations in location choice as well as the absence of bargaining or pre-partnership migration effects based on relative educational levels.
The top panel of Table 6 reports the estimates, under Model 2, of the log odds of educationally homogamous couples with high, intermediate, or low level of education living far from the parents. They imply that across the three education categories, the predicted probability of homogamous couples living far from the woman’s parents are .44 (low), .53 (intermediate), and .79 (high) respectively. The corresponding figures for being far from the man’s parents are .48, .58, and .80. Thus, the probability of being far from parents is slightly higher if the couple has intermediate rather than low level of education, but it is much higher if both partners have a degree.
The Influence of Other Factors on Proximity
In earlier individual-based analysis, Chan and Ermisch (forthcoming) found that the following variables are associated with proximity to parents among individuals aged 31–54: whether they have siblings, whether they experienced parental divorce by age 16, whether they have a child or children,23 housing tenure, and whether they moved in the past five years.24 Thus, in Model 4, we include these attributes in x. In addition, we include the woman’s age, the age difference between the partners, the age difference between the parent and the child for both partners, and the woman’s share of the couple’s joint income. We also include indicator variables for living in London, in the South East of England, and in a rural area. Descriptive statistics for these variables are shown in Table 9 in the appendix.
The parameter estimates of Model 4 are shown in the bottom panel of Table 6.25 The differences in the estimates of the diagonal reference terms across educational levels under Model 4 are very similar to those of Model 2. For example, in the left-hand column, the difference in the log odds for couples with high and intermediate levels of education are 1.25 under Model 2 and 1.28 under Model 4. In other words, the odds ratios implied by the diagonal reference parameters do not change much when covariates are added.
Regarding the covariates, middle-aged couples can benefit from help from grandparents in childcare if they live nearby, and the grandparents may have a strong interest in seeing their grandchildren often. Consistent with this argument, Table 6 shows that having a child reduces the probability of living far from the woman’s parents. However, there is no association between this variable and the distance to the man’s parents. This finding is broadly in line with that of Blaauboer et al. (2011) for the Netherlands, and might suggest that maternal grandparents are more involved in grandparenting.
When parents divorce or separate, at least one of them moves from the parental home, and repartnering may also strain relations with the children from the first marriage. Consistent with this view, we find that men who experienced parental separation or divorce as a child are more likely to live far from their parents. The corresponding parameter for women, although of similar magnitude, is not statistically significant from zero (p = .11), nor is it significantly different from that for men. The same holds for many of the covariates listed in the table. Thus, the gender differences reported in this section should be interpreted with caution.
With empirical support from many countries, Rainer and Siedler (2009, 2012) argued that children with sibling(s) are more likely to live farther from their parents than those who are an only child. Also, van der Pers and Mulder (2012) found that parents of an only child are more likely to live near that child than those of two or more children. In our analysis, men without siblings (i.e., only child) are less likely to live far from their parents, similar to Løken et al. (2013). Again, the corresponding parameter for women is of the expected sign but is not statistically significant.
These associations between family characteristics and distance to parents suggest that in certain circumstances, the location preferences of one of the partners are given more weight. For instance, that only-child status of the man shifts location toward his parents suggests that his preferences dominate in this situation, whereas the presence of children favors location nearer to the woman’s parents.
The impact of geographic mobility accumulates over time. Thus, we expect the distance between parent and child to increase with age (Rogerson et al. 1993). This is indeed what we see. Controlling for the partners’ age difference, older children live farther from their parents, as also found by Blaauboer et al. (2011) and Compton and Pollak (2013). Also, men with older parents live closer to their parents, but this is not true for women.
A couple’s underlying propensity to move is likely to be positively correlated with their recent history of residential change. Indeed, we find that couples who have moved in the past five years are more likely to live far from their parents, which is consistent with the view that mobility tends to move the generations farther apart.26
Compared with homeowners, private sector renters are more likely to live far from the woman’s parents. The same is true of tenants in social housing, although its parameter is marginally insignificant (p = .09). Possibly reflecting the fact that there are more interregional migrants in London and the South East of England, couples living in the urban parts of these areas tend to live farther from their parents. The same holds for couples living in rural areas compared with those living in cities. Exclusion of these region variables has little impact on the other parameter estimates.
Finally, the woman’s share of the couple’s joint income does not predict proximity to either set of parents. This is consistent with the finding of equal weights for the two partners’ education in the diagonal reference model. Together, they suggest that the location choice of couples is only mildly associated with gender, favoring women, and that there is little bargaining based on the partners’ relative income or educational levels.
Joint Distribution of Distance to Two Sets of Parents
In Model 4′, we add covariates to Model 2′. The results are reported in the bottom panel of Table 7. As in Table 6, the differences in the diagonal reference terms between the three educational levels are very similar for the top and bottom panels. Regarding the covariates, it is striking that all the statistically significant parameters pertain to the X1 versus X2 contrast.27 Couples with children are less likely to live far from both sets of parents. However, older couples, those who have moved in the past five years, and those living in rural areas, London, or the South East of England are more likely to live far from both sets of parents. These results are very similar to those reported in Table 6.
To illustrate the substantive magnitude of these associations, we report in Table 8 the predicted probabilities of X1, . . . , X4 given specific covariate values. Our baseline is a couple in which both partners are university graduates and the woman contributes 40 % of their joint income. She is age 40, he is age 43, and their parents are ages 70 and 73, respectively. They both have siblings, but they have no children of their own. Neither of them experienced parental divorce before the age of 16. Finally, they are homeowners in an urban area outside London and the South East of England, and they have not moved in the past five years.
Under this baseline scenario, the most likely outcome is for this couple to live far from both sets of parents (X1; probability = .60). Indeed, the probability of them living near both sets of parents, X2, is only 8 %. However, if this couple have children of their own, while other covariates are kept unchanged (scenario 2), then they would be a little more likely to live near both sets of parents (probability = .11 versus .08) or near the woman’s parents and far from the man’s parents (.18 versus .11) and would be somewhat less likely to live far from both sets of parents (.52 versus .60). However, X1 is still overwhelmingly the most likely outcome.
Furthermore, the difference that children make is small relative to that which is due to the couple’s own education. Thus, if the childless couple have GCSE or lower qualifications rather than university degrees (scenario 3), the probability of them living far from both sets of parents would drop from 60 % to 20 %, while that of living near both sets of parents would rise from 8 % to 35 %. A similarly large education contrast for couples with children can be seen between scenarios 2 and 4.
Finally, we computed the predicted probabilities of couples living in different areas or with different mobility history, and report them in scenarios 5–8. Compared with the baseline scenario 1, the differences in the predicted probabilities are in the expected direction and are of nontrivial magnitudes, but these differences are still smaller than those resulting from the couple’s educational levels.
To sum, we find a slight tendency for couples to live closer to the woman’s parents than to the man’s. This tendency is more pronounced among couples in which neither partner has a degree and in which there is a child. In other respects, proximity to parents is gender-neutral.
Educational attainment has a large influence on geographic mobility and thus on intergenerational proximity, with better-educated couples tending to live farther from their parents. We find that each partner’s educational level contributes equally to the proximity outcome. Our results are consistent with the idea that the impact of each partner’s education is proportional to the earnings return to wide geographic job search for that level of education, and that the combined impact of the two partners’ education is a simple average of the impacts of the two educational levels. There is no additional influence according to who has the higher qualification—that is, no bargaining power or pre-partnership migration effect associated with relative educational levels. Similarly, the partners’ income share does not affect proximity to parents.
Some circumstances related to family history and childbearing do, however, shift location closer to one set of parents or the other. In particular, the presence of children favors location closer to the woman’s parents. We conclude that proximity to parents is primarily driven by factors that affect mobility over long distances, which are mainly those associated with the labor market.
We are grateful to Heather Turner for helpful advice on the R gnm package. Early versions of this article were presented at seminars at Oxford and Barcelona, and at the 2014 annual conferences of the Population Association of America and the British Society for Population Studies. We thank the participants of these meetings and anonymous referees for helpful comments. Our research is supported by the Economic and Social Research Council’s Secondary Data Analysis Initiative, Phase 1, Award Number ES/K002902/1.
If either set of parents is divorced, the situation becomes even more complex.
Evidence for this is the negative impact of local friendship networks on longer-distance movement (Belot and Ermisch 2009).
To study the dynamics of intergenerational proximity, we would need residence mobility data for both partners for up to 36 years (from ages 18 to 54).
The associations between distance from parents and an individual’s education are consistently positive (see, e.g., Blaauboer et al. 2011; Chan and Ermisch forthcoming; Compton and Pollak 2013; Hank 2007; Løken et al. 2013; Shelton and Grundy 2000).
See Ermisch and Pronzato (2008) for such effects on men’s child support payments in Britain; on bargaining power effects, see Basu (2006), Chiappori et al. (2002), Couprie (2007), Lundberg and Pollak (1996), Lundberg et al. (1997), and Rangel (2006).
Further details of the survey are available online (http://www.understandingsociety.ac.uk).
Coresidence with parents is not uncommon among younger respondents in Understanding Society. For example, Chan and Ermisch (forthcoming) showed that about 18 % of people aged 25–29 coreside with their parents, compared with 7 % for people aged 30–34.
Ethnic minorities are oversampled in Understanding Society. All results presented are weighted to reflect the sampling design and nonresponse, using the weight variable a_indinus_xw.
For the white, UK-born sample, the corresponding figures are 35 % and 28 %, respectively.
Comparisons with the American data are inexact because of different age ranges (25 and older in the U.S. sample versus 31–54 for the UK sample) and different distant measures (miles in the U.S. sample versus traveling time in the UK sample). In the U.S. data, the median distance to the woman’s mother is 20 miles, compared with 25 miles for distance to the man’s mother (Compton and Pollak 2013: table 3).
Chan and Ermisch (forthcoming: figure 1) showed that adult children aged 31–54 are much more likely to see their parent daily if they live within 15 minutes of each other (25 % versus 8 % for the 15–30 minute distance category). There is a sharp decline in daily and weekly contact as proximity decreases beyond 30 minutes traveling distance. The third (2011–2012) wave of Understanding Society indicates that adult children aged 31–54 are much more likely to give some form of regular or frequent in-kind help to their parents if they live within 15 minutes of each other than if they live 15–30 minutes apart: 63 % versus 50 %. They are also more likely to receive in-kind help from parents if they live within 15 minutes: 54 % versus 44 %.
We acknowledge that coresidence is qualitatively different from living near but in a separate household (Compton and Pollak 2013); indeed, we demonstrated this to be the case in a previous study using individual data from the same source as used here (Chan and Ermisch forthcoming). But coresidence is too rare (less than 1 % of the couples) to be treated as a separate category in our analyses. Compton and Pollak (2013) also found small numbers coresiding in their couples’ sample.
Not all choices are available to all couples. For example, if the two sets of parents live very close to each other, then X3 and X4 would not be possible.
A woman’s higher share could also be interpreted as an indicator of higher labor market aspirations, which would encourage geographic mobility and thus locations farther from her parents. We recognize that woman’s share of household income may be an endogenous variable because it partly depends on location. The results are not affected by its exclusion from the models reported in this article.
Using data from the 18 annual waves of the British Household Panel Survey, we find that among movers under the age of 30, the mean distance moved is 65 km for persons with a university degree, 44 km for persons with an intermediate level of education, and 21 km for those with lower-level qualifications. If these young movers are also single, the distances for the three education groups are 70 km, 52 km, and 22 km, respectively. Details are available from the authors on request. See Ermisch (2009), who showed that young people from richer homes move farther from their parents when they leave, particularly when they are single.
In England, Wales, and Northern Ireland, GCSE refers to the “school-leaving” qualifications, typically gained by pupils at age 16; achieving A-levels, typically at age 18, is the qualification for university matriculation. Scotland has its own qualifications system, which has been converted to its equivalents for the rest of the UK in this analysis.
In contrast to what we find here, Compton and Pollak (2013: table 5), using data from the National Survey of Families and Households, found that when the woman has a degree and the man does not, the percentage of couples who live near her mother is 5.8 percentage points higher compared with couples in which the educational difference is reversed (in their data, we define near as the mother living within 30 miles). In the UK data, the corresponding difference is –5.6 percentage points for the 15-minute near–far threshold and 0.3 percentage points for a 30-minute near–far threshold.
Chan and Ermisch (forthcoming) showed very large differences between ethnic groups in intergenerational proximity, even after controlling for education and other covariates.
If UK-born nonwhites are included in the analysis, N increases from 2,506 to 2,803. The main results of this article are not affected by whether nonwhites are included in the sample. Details are available from the authors on request.
Generically, the model can be expressed in terms of a latent continuous variable. Let yjrc represent the latent travelling distance to parents for a partner in couple j. We assume that yjrc = wδrr + (1 − w)δcc + ej. An assumption about the distribution of ej is needed, such as a logistic or standard normal distribution. With the logistic assumption and the near–far dichotomy, πrc is the probability that yjrc > 0, πrc = exp(wδrr + (1 − w)δcc) / [1 + exp(wδrr + (1 − w)δcc)], which implies that log(πrc / (1 − πrc)) = wδrr + (1 − w)δcc. Thus, the parameters δkk = log(πkk / (1 − πkk)), k = 1, 2, 3; that is, they are the logit coefficients along the diagonal. The latent variable formulation extends easily to ordered logit or probit models when there are more distance categories.
We fit diagonal reference models with the R package gnm (Turner and Firth 2011). We have also used Stata to fit the same set of models using probit rather than the logit link function. The results we obtained are very similar to those reported here.
The unconstrained model contains nine parameters, one for each log(πrc / (1 − πrc)), compared with four parameters in the diagonal reference model.
We use the woman’s report of their parental status.
This variable refers to whether either of the partners moved in the past five years.
We cannot reject the parameter restrictions in Model 4 relative to Model 1 plus covariates, or Model 3 plus covariates. Details are available from the authors on request. In a bivariate probit model that allows the residual error terms in the two partners’ distance-to-parent equation to be correlated, the results are similar. The correlation between the errors is estimated to be .25. The conclusions are also similar when we estimate ordered probit and logit models using all seven categories of Table 1.
Regarding distance to woman’s parents, the parameter of “move in the past five years” is marginally not significant with p = .06.
For the X3 versus X2 contrast, the following parameters are significant at the 10 % level: woman experienced parental divorce by age 16 (p = .10), woman’s age (p = .07), and having moved in the past five years (p = .05). The same applies to the following parameters for the X4 versus X2 contrast: having a child (p = .06), man experienced parental divorce by age 16 (p = .06), and man’s age difference with his parents (p = .09).