Family Proximity, Childcare, and Women’s Labor Force Attachment

1. Family Proximity, Childcare, and Women’s Labor Force Attachment Janice Compton Robert A. Pollak June 2013 1 1

2. Introduction - 1 Part of a larger project on the geography of the family. There is a substantial literature focusing on coresidence and long-term care of the disabled elderly. In work with Janice Compton, Liliana Pezzin, and Barbara Schone, I’m looking at geographical proximity and long-term care of the disabled elderly. In this paper we focus on the implications of proximity of adult children to their parents for the labor supply of the adult children. 2 2

3. Introduction - 2 Our question: How does close geographical proximity to mothers and/or mothers-in-law affect the labor supply of married women with young children? We use data from two U.S. sources: 1. National Survey of Families and Households (NSFH) 2. US Census. . 3 3

4. Introduction - 3 We find that the predicted probability of employment and labor force participation is much higher – between 4 and 10 percentage points higher – for married women with young children who live close to their mothers and/or their mothers-in-law compared with those who do not. 4 4

5. Introduction - 4 We argue that there is clear and convincing evidence that the mechanism is the availability of childcare. We interpret the availability of childcare broadly enough to include not only regular scheduled childcare during work hours but also an insurance aspect of childcare – having a mother and/or mother-in-law in close proximity, who can meet irregular or unanticipated childcare needs. 5 5

6. Introduction - 5 Endogeneity issues: Issue 1: childcare decisions and labor supply decisions. To address this, we use IV estimation, using proximity as an instrument for childcare transfers from mother or mother-in-law. We argue that these estimates may underestimate the full effect of geographic proximity as they are based on observed transfers of care. Consider reduced form, estimating the effect of proximity on labor force attachment. 6 6

7. Introduction - 6 Endogeneity issues: Issue 2: potential endogeneity of proximity. This type of endogeneity is arguably less serious. We address this by comparing results from a sample of military wives in the census. Issue 3: potential endogeneity of fertility We have nothing to say about this. But think about work on housing prices and fertility that treats location as given. 7 7

8. Introduction - 7 To allay concerns that proximity affects labor force attachment through channels other than childcare, we show that proximity has no discernable effect on the labor force behavior of women without childcare needs – those without children and those with children over 12 years of age. 8

9. Introduction - 8 To be clear about our argument: We do not have a natural experiment. We do not have a structural model. We find that close proximity to mother and/or mother-in-law increases the labor supply of married women with young children. We argue that the concentration of these effects in this particular demographic group constitutes clear and convincing evidence that the mechanism underlying this association is the availability of childcare. 9 9

10. Introduction - 9 The style of argument (NSFH as example) 1. We find large effects of proximity on the labor force attachment of married women with young children. We do not find this for those with mothers in poor health. 2. We do not find effects of proximity on the labor force attachment of men, of married women with no children, or married women with only older children. 3. We do not find effects for unmarried women with young children. We argue that their labor force attachment is inelastic: they have to be in the labor force, so these women are not sensitive to the availability of childcare to meet irregular and unpredictable needs. 10 10

11. Introduction - 10 The style of argument (Census as example) Evidence from the Census (which does not provide information about distance). We use information about whether an individual is living in his or her birth state as a proxy for proximity to a parent. We look at married women with and w/o young children. We also construct a sample of “military wives” (civilian wives whose husbands are in the US military). We also look at recent migrants. The census lets us identify individuals who migrated from one state to another within the last 5 years. We compare those who have returned to their birth state with those who migrated elsewhere. More specifically,... 11

12. Introduction - 11 From the census we construct a subsample of recent migrants -- individuals who, five years prior to the census, were not living in either their birth state or their current state. We find that married women with young children who returned to their birth states or to their husbands' birth states have higher labor force participation than women who moved to a non-birth state.

13. The Literature - 1 There is a substantial literature on coresidence of adult children and elderly parents, focused on long-term care of the disabled elderly. Costa vs Ruggles on why there has been a decline in coresidence: Is it increasing affluence of the older generation? (Costa, 1999, JPubE) Or is it increasing economic independence of the younger generation? (Ruggles, 2007, ASR) 13

14. The Literature - 2 Proximity in economics; Not much published work Konrad et al. (2002 AER); Rainer and Siedler (2009 OEP) Loken, Lommerud, and Lundberg using Norwegian data (2013 Demography) Compton and Pollak using NSFH (MRRC WP) McGarry and Weimers using PSID 14 14

15. The Literature - 3 Some published papers using European data look at childcare using proximity to her mother as an instrument for childcare (they don’t have data on proximity to his mother): Dimova and Wolfe (2008, EuJPop) Zamarro (2009, RAND) These paper find small effects of childcare on the probability of work, but focus on regular hours of childcare only and don’t have date on proximity to mother-in-law. (This is important.) Using Canadian data Compton (2011) considers the direct effect of proximity on labor force attachment. The results are similar in magnitude to our findings for the U.S. 15 15

16. The Literature - 4 Proximity outside economics -- sociology; demography; geography. At the PAA in Washington April 2011 I discussed Seltzer, Bianchi, McGarry, and Yahirun, “Co-residence and Geographic Proximity of Mothers and Adult Children.” It has lots of references to the sociology literature. The proximity literature has been largely focused on long-term care of the disabled elderly. In this paper, we go in a different direction: child care and implications for labor force participation of women with young children. 16

17. Outline Evidence from NSFH (which provides information on distance between a married couple and his parents and her parents). NSFH is one of the few data sets containing this information. Having distance to both mother and mother-in-law turns out to be important. Evidence from the Census (which does not provide information about distance, but provides information about whether an individual is living in his or her birth state; we use this as a proxy for proximity to a parent). 17 17

18. NSFH - 1 National Survey of Families and Households (NSFH). We use the first two waves (1987-88) and (1992-94). Panel data. 13,007 households in first wave. We focus on those whose mothers and mothers-in-law are Alive and Living in the United States (ALUS). NSFH provides distance to mother and distance to mother-in-law. Few data sets provide this information. NSFH also provides information about the provision of “work related child care” and “non-work related child care” by his parents and by her parents. 18 18

19. NSFH - 2Some Facts to Keep in Mind 1. Child care depends on close proximity 2. Most adult Americans live surprising close to their parents. 3. There is a steep educational gradient. Some terminology (due to Costa and Kahn, QJE, 2000) Power couples: both spouses college graduates Low power couples: neither spouse a college graduate Part power couples: only one spouse a college graduate. 19 19

20. Adults (≥25) whose mothers are “ALUS” 20 20

21. Couples and Proximity - 1 Strong Education Gradient 21 21

22. Couples and Proximity - 2 1/4 of married women live within 5 miles of their mothers For three of the four education groups, 70% of the couples are either close to both mothers or close to neither mother. For the remaining group (part-power couples where she has the degree), 60% of the couples are either close to both mothers or close to neither. 22

23. NSFH - 3 Couples, Singles aged 25-60 whose mothers are ALUS 1837 unmarried women, 2954 married women Define “Close Proximity” as 25 miles or less Wave 2 (1992-1994) Distance (in miles) to mother and mother-in-law. Distance vs time as measure of proximity Distance to mother and mother-in- law is associated with labor force attachment 23 23

24. We need both distance to mother AND distance to mother-in-law.

25. Percent of Married women aged 25-60, with a child 12 or under, who are working full-time (NSFH) 25 25

26. Percent of Married women aged 25-60, with a child 12 or under, who are working full-time (NSFH) 26 26

27. NSFH - 4 The raw data are especially suggestive because the factors associated with living close to mother and mother-in-law are generally associated with lower labor force attachment. Women are younger and less educated They have less educated mothers and husbands They are more likely to be black or Hispanic They are less likely to live in an MSA and more likely to live in areas of higher unemployment. 27 27

28. NSFH - 5 Regression analysis: what happens when we introduce control variables (e.g., race, education, husbands’ characteristics, age and health of mother and/or mother-in-law). We consider two outcome variables: 1. Whether woman works or not (dichotomous) 2. Usual weekly hours (continuous) 28 28

29. NSFH - 6 We estimate the effect of predicted transfers of childcare (either work-related or non-work related) on the probability of work. Proximity to mother or mother-in-law is used as an instrument for the probability of receiving childcare. 29 29

30. NSFH - 7 No effect of childcare on the probability of work for unmarried women. For married women, no effect when we consider only childcare from her mother. But when we estimate the effect of childcare from either mother, we find a positive effect for both work-related and non-work related childcare. 30 30

31. NSFH - 8 Control Variables: X: Common to both regressions: age, age squared, whether husband works and his hours of work, husband’s income, whether self or husband currently has medical problems, race (black, Hispanic, white (omitted)), education categories (both spouses have college degrees, only she has a college degree, only he has a college degree, neither has a college degree (omitted)), age of youngest child, whether mother has a college degree. Z : instruments for work: region (Midwest, South, West, Northeast (omitted)), average commuting time in the county (to account for place-to-place differences in the amount of time it takes to travel), whether residing in an MSA, 1990 county level unemployment rate, and whether the respondent lived in a different city in the first wave of the data. :instruments for childcare: age categories of mother(s) (less than 60, 60-69, 70 and over (omitted)), whether mother(s) are in poor health, whether mother(s) are married and whether mother(s) live in close proximity. 31 31

32. NSFH - 9 Next investigate the reduced form relationship between proximity and labor force attachment. We find that proximity to mothers or mothers-in-law has a direct effect on labor force behavior (both employment and usual weekly hours). We find that proximity to mothers-in-law is an important part of the story. We distinguish among four groups: 1. Close to both mother and mother-in-law 2. Close to mother only 3. Close to mother-in-law only 4. Close to neither mother nor mother-in-law. 32 32

33. NSFH - 10 If we ignore mothers-in-law and just look at mothers, we are combining groups 1 and 2 and combining groups 3 and 4, so we have only two groups: 1 & 2. Close to mother and close to mother-in-law and close to mother only 3 & 4. Close to mother-in-law only and close to neither mother nor mother-in-law With this comparison, we find no significant effect of being in group 1&2 compared with being in group 3&4. But when we use all four groups, we find a strong and significant effect of being close to both mothers or close to mother-in-law only. 33 33

34. NSFH - 11 We partition the adult daughters into three groups mothers with young children (children 12 and under) mothers with only older children non-mothers For married mothers with young children, we find a large and statistically significant effect of close proximity to mother-in-law or to both mothers increases the predicted probability of employment by 10 percentage points. The coefficient on close proximity to her mother only is positive but insignificant. (In census data, we find significant effects for her mother only as well as for his mother only, with about the same effect size.) 34 34

35. NSFH - 12 When we restrict the sample to married women with young children whose mothers or mothers-in-law are in poor health, we find no effect on labor force attachment. This is consistent with our story about the the availability of child care as the mechanism through which proximity affects labor force attachment. (But it is also consistent with women taking care of their mothers or mothers-in-law.) 35 35

36. NSFH - 13 Could job-search networks be the mechanism through which proximity affects labor force attachment? No, because we see effects only for married women with young children -- not for men, not for married women with only older children, not for unmarried women. 36 36

37. NSFH - 14 Could husbands’ incomes or earnings be the mechanism through which proximity affects labor force attachment? (Are women who move away married to higher earning husbands?) No, it cannot be the mechanism because we control for husbands’ earnings.

38. NSFH - 15 Could child care by friends or other relatives be the mechanism through which proximity affects labor force attachment? Yes. The “child care insurance” story is consistent with other relatives or friends providing child care insurance. (But most actual childcare is provided by grandmothers.)

39. Census - 1 The census does not ask about distance to mother or mother-in-law. But the census does ask about birth state. As a proxy for a woman’s proximity to her mother, we use the woman’s residence in her birth state. As a proxy for a women’s proximity to her mother-in-law, we use her husband’s residence in his birth state. It’s not perfect, but it’s the best we’ve got. 39 39

40. Census - 2 Three separate analyses: 1. We use residence in birth state just the way we used “close proximity” in the NSFH to look at labor force attachment of various demographic subsamples (e.g., married women with young children). 2. We construct a subsample of “military wives” (civilian wives whose husbands are in the US military) and look at labor force attachment of those with young children. 40 40

41. Census - 3 3. From the census, we construct subsample of recent migrants -- individuals who, five years prior to the census, were not living in either their birth state or in their current state. We compare migrants who returned to their birth state with migrants with other destinations. We find that married women with young children who returned to their birth states or to their husbands' birth states have higher labor force participation than women who moved to a non-birth state.

42. Census - 4 We want to look at the effect on women’s employment and usual weekly hours of 1. Living in the birth state of both spouses 2. Living in only her birth state 3. Living in only his birth state 4. Living in the birth state of neither spouse. We also control for recent migration (i.e., living in a different state 5 years ago) and for the size of the state. We look at women age 25-45 (this makes it more likely that the mothers and mothers-in-law are ALUS). 42 42

43. Census - 5 We also construct a sample of “military wives” -- civilian women with husbands in the US military. Our sample size here is 14,833, of whom 10.2% live in only her birth state, 5.1% living in only his birth state, and 8.7% live in the birth state of both spouses. For obvious reasons, “neither” is very high compared with NSFH. The advantage of the military wives is that it controls, to some extent, for the endogeneity of location. (The husband’s location is arguable exogeneous; the wife’s, not so much. Notice the difference in the probably of living in her birth state vs his. This is presumably because the military assigns him; she may be local.) 43 43

44. Census - 6 Raw data - 1 For married women with young children, we find increasing labor force attachment as we move from birth state of neither spouse to birth state of both spouses. For other groups (e.g., married women with only older children), we find nothing. Effect is statistically significant for both his birth state only and her birth state only, and the effect sizes are about the same. (Recall that in NSFH, proximity to her mother only was not statistically significant and the effect size of proximity to his mother only was larger than proximity to her mother only.) 44 44

45. Census - 7 Raw data - 2 For military wives, we find a similar pattern for those with young children. For never married women with young children, we find a negative association between birth state residence and labor force attachment 45

46. Census - 8 Regression Control Variables: age, age squared, children (children 12 and under, only children over 12 in the household, no children in the household (omitted)), education (less than high school, high school diploma (omitted), more than high school, bachelor’s degree, more than bachelor’s degree), spouse education (groups same), disability, spouse disability, race (Black, Hispanic, white (omitted)), rented accommodations, whether in a metropolitan area, total income of spouse, whether in different state five years prior, size of current state (square miles), U.S. region. 46 46

47. Census - 9 Regression results: We find a positive effect of birth state residence for married women with young children, but not for those without young children. For the military wives, we find a positive effect of birth state residence for those with young children if they live in the birth state of both spouses. The effect sizes are not as large as those we found using NSFH, but are still substantial -- between 2.6 and 3.9 percentage points. 47 47

48. Census - 10 Finally, we construct subsample of 74,000 recent migrants -- individuals who, five years prior to the census, were not living in either their birth state or their current state For married women with young children, non-migrants have higher labor force attachment than migrants, as the tied mover hypothesis predicts. But when we compare married women with young children who migrate back into their birth state (return migrants) with those who move to another state (onward migrants), we find that the labor force attachment of return migrants is 4.5 percentage points higher than that of onward migrants. But no effects for other demographic groups. 48 48

49. Census - 11 Regression analysis for return migrants and onward migrants tells a similar story: the destination (birth state vs non-birth state) matters. 49 49

50. Census- Tied Mover - 1 Birth state and migration: disentangling proximity effects and the tied mover effect: Sjaastad (1962 JPE), Mincer (1978 JPE), Lichter (1983 Social F) The tied mover effect suggests that two-earner couples are less mobile than one-earner couples. More important for our purposes, the tied mover effect suggests that secondary earners (read: wives) who migrate will have lower labor force attachment, at least in the short run, than those who do not migrate. 50 50