Social Insurance and the Marriage Market

《Social Insurance and the Marriage Market》由会员分享，可在线阅读，更多相关《Social Insurance and the Marriage Market（86页珍藏版）》请在装配图网上搜索。

1、Social Insurance and the Marriage Market1 am grateful to Pierre-Andre Chiappori, Navin Kartik. and Wojciech Kopczuk for their generous support at all stages of this project I also thank S Anukriti, Clement de Chaisemartin, James Heckman, Lisa Jonsson, Per Krusell, Todd Kumler, Samuel Lee, Ben Marx,

2、Miguel Morin, David Munroe, Suresh Naidu, Torsten Persson, Giovanni Paci, Kiki Pop-Eleches, Malgorzata Poplawska, Maya Rossin-Slater, Bernard Salanie, Anna Sjogren. Reed Walker. Stephen Zeldes, Bjorn Ockert, and seminar participants at numerous universities and conferences for helpful discussions, s

3、uggestions and coininents Funding from the Center for Retirement Research at Boston College, Grant #5001537 from tho Social Security Administration, and the Hewlett Foundation/ITE Dissertation Fellowship in Population, Reproductive Health and Economic Development is gratefully acknowledgedSIEPR and

4、Department of Economics, Stanford University. Email: perssonpstanford.edu.Petra PerssonSeptember 13, 2013AbstractWhen social insurance eligibility depends on marital st at us, this is a government intervention into the marriage market I formally show that such intervention influences three behaviora

5、l margins in the marriage market, and test the theory exploiting a Swedish reform of survivors insurance an annuity paid to widows, but not divorcees, upon the husbands death First, I analyze bunching in the distribution of marriages and show that, by affecting the wedge between marriage and cohabit

6、ation, survivors insurance alters the composition of married couples up to 45 years before the annuitys expected payout Insunuice takeis larger in couples with higher ex post male mortality, holding constant the policys value at realization and all demographics that I observe, suggesting adverse sel

7、ection into government-provided insurance Second, I use a regression discontinuity design to show that removal of survivors insurance from existing marriage contracts caused divorces and, in surviving unions, a renegotiation of marital surplus. Third, because survivors insurance subsidized couples w

8、ith highly unequal earnings (capacities), its elimination raised the long-run assortativeness of matching I argue that such marriage market responses to social insurance design have important implications for when it is optimal to separate social insurance from marriage in modern societies.1 Introdu

9、ctionA major function of governments in many developed countries is to provide social insurance. The two largest social insurance programs in the United Stntcs, Social Security and Medicare, together account for more than 30 percent of federal spending. It is well recognized that the provision of so

10、cial insurance to protect against adverse income or health shocks distorts markets that offer private insurance against such shocksI instead focus on responses in the marriage market. I do this because social insurance often represents a twofold intervention, both into private insurance markets and

11、into the marriage market. This occurs whenever marital status influences eligibility for social insurance In the United States, for example, both Social Security and Medicare fit. into this category. I first ask how a link between social insurance and marriage affects the marriage markct，and then ex

12、amine the implications of my findings for when it is optimal to separate social insurance from marriage.Specifically, in the context of Sweden, I study a particular type of social insurance, survivors insurance Survivors insurance replaces part of the income that a household loses upon the death of

13、one household member Like many countries do today, Sweden used to provide survivors insurance through the marriage contract A widow was granted a lifetime annuity of survivors benefits upon her husbands death, but cohabitating partners or divorcees were not eligible. The value of this insurance was

14、significant; beneficiaries? average annual payout was $5000, for an average duration of eight years (Section 2 provides details) But to most couples, who entered marriage in their 20s or 30s. the insurance was not likely to payout until far in the future Marriage market responses to survivors insura

15、nce thus necessitate that couples have a long financial-planning horizon. I ask how this twofold interven- tion，into annuities markets and into the marriage market, affected the volume and nature of private contracting in the marriage market in Sweden. The volume of contracting is determined by entr

16、y into and exit from marriage The nature of private contracts has a range of dimensions, of which I study two: Who contracts with 一 th&t is, marries 一 whom? And, which contracting party gets what share of joint realized gains?In my empirical analysis，I exploit a 1989 reform that changed how survivor

17、s insurance was tied to the marriage cont.ract in Sweden The reform essentially eliminated survivors insurance, replacing the promise of a lifetime annuity with a promise of one (small) uadjustment transfer?5 Thus, the old marriage contTact, which came with a government-provided annuity that was exp

18、ected to pay out in old age，was replaced with a new marriage contract that came without it, but that otherwise was identical.Section 3 theoretically analyzes how this affects the marriage market when individuals are forward- looking My framework captures typical relationship dynamics Specifically, s

19、ingle individuals form couples; then, in each period, each couple gets new information about (shocks to) its match-specific marital surplus and updates its marital decision. Formally, I introduce matching into a collective household framework and model match quality as a stochastic process Ill this

20、framework, the inipact of a change in survivors insurance tied to marriage depends on whether an individual, at the time of the reform announcement is married cohabitating, or single (to be mat died in the future). For each group, the thcory delivers precise, distinct： tcstablc predictions about ini

21、tial responses and about future behavior conditional on these initial responses I test these using individual-level marital and tax records, described in Section 4. Some predictions vary depending on the properties of the stochastic match quality process. This also yields an implicit test of a large

22、ly open question about cohabitation, namely, whether cohabitation is a learning process in transition to marriage or an uend-state substitute for marriage. My results are consistent with a match quality process that is stochastic and exhibits positive autocorrelation; this supports the interpretatio

23、n of cohabitation as a learning processMy first set of results, presented in Section 5, concerns how tying social insurance to marriage affects entry into marriage Couples with a joint child who were not yet married at the reforms announcement in June 1988 were allowed to take up survivors insurance

24、 by marrying by December 31, 1989. By analyzing bunching in their distribution of new marriages, I can study how selection into marriage responded to a demand for survivors insurance. The distribution of new marriages displays a marriage boom in the last quarter of 1989 I estimate that a couple is o

25、n average 17 times more likely to marry in this quarter relative to the counterfketual scenario. Even couples below the age of 30 exhibit significant responses, implying financial-planning horizons as long as 45 years Consistent with the tlieoretic&l prediction that couples in which the male is more

26、 likely to die should exhibit stronger responses, I document larger increases in take-up of marriage 一 and thus, of survivors insurance 一 among couples for which the husbands ex post mortality is higher. This remains true when controlling for a range of observable factors that det ermine the governm

27、ents expec ted cos t of extending survivors insurance to a given couple, including the husband age at marriage. The positive correlation bet ween couples take-up of insurance through marriage and couples5 expected cost of coverage suggests the possible presence of asymmctrie information, which could

28、 pose a challenge to private markcts for annuities in Sweden.Further, I find that the reform impact on take-up of marriage is positive but ten times lower in a sample of couples in which one spouse reveals a same-sex preference years later, by entering a same-sex union after they were legalized in 1

29、995. If sexual preferences exhibit positive autocorrelation, this is consistent with the theoretical prediction that responses would be weaker in couples with weaker ex ante beliefs of staying togcther for life 一 the only circumstance in which the annuity would pay out.When match quality is stochast

30、ic, the theory suggests that the reform should affect who gets married. Specifically, the marriage boom should consist both of retimedmarriages that would have eventually taken place even in the absence of reform, and “extrn marriages that would never have occurred if the old marriage contract had r

31、emained available Intuitivcly, announcing a replacement of the existing marriage contract with a less desirable one attaches option value to marrying into the desirable contract. Couples then rush to marriage even though they, in the absence of reform, would wait to see if the relationship improves.

32、 To decompose the excess mass in the bunching region into responses along the “intensive55 (retiming) and “extensive57 (extra marriages) margins, I develop an empirical framework that exploits the shape of the density, but also a second dimension of information: the date of birth of a couples fiist

33、child. I couple this with specific assumptions about the irnpact of this second dimension of information on maritai behavior. Extm marriages accounted for more than a quarter of the response Inherent in the theoretical option value explanation for this is a prediction that marriages in the boom are

34、of a lower average match quality. Indeed： I show tliat such marriages are four percent&ge points more likely to encl within 15 years than other marriages with the same contract. Nevertheless, a sizable portion of extra marriages survives in the long lrun.11hen analyze the causal impact of (losing) s

35、urvivors insurance on exit from marriage. In Section 6, I study couples that were already married at the reform announcement, but for whom grandfathering rules induced variation in insurance tied to marriage. Specifically, couples that married before January 1, 1985 were allowed to keep the contract

36、 they married into; for most couples that married thereafter, this contract was revoked and replaced with the new one. This change was announced in June 1988 three and a half years after entry into marriage, rendering impossible any manipulation in response to a demand for survivors insurance. Using

37、 a regression discontinuity design that exploits both the eligibility cutoff and the timing of the reforms announcement, I show that the removal of survivors insurance from preexisting marriage contracts raised the long-run divorce T&te among these couples by six percentage points.My second set of r

38、esults concerns how tying social insurance to marriage afFects the nature of contracting in the marriage market; specifically, the division of joint marital gains, and matching I begin by studying couples division of surplus. Among couples tha.t married into the old contmct but then lost insuiance c

39、overage, the tlieory predicts a renegotiation of marital surplus in favor of the wife in surviving unions. Intuitively, the wifes expected utility from marriage is the sum of her utilities in the states of the world where her husband is alive and where he is dead The reform reduces her utility in th

40、e latter st ate; as compensation for this loss, her share of surplus in the former increases. I test this empirically by studying spousal labor supply, using a regression discontinuity design similar to the one described above. My results suggest that revocation of survivors insurance delayed elderl

41、y husbands7 retirement 一 even though it affected household wealth only in a state of the world in which the husband would not be alive. This suggests that while the reform induced a statutory loss solely on the wife, intra-household bargaining resulted in the economic incidence partly being borne by

42、 the husband, who gave up utility - that is, leisure - when he was aliveFinally, Section 7 analyzes long-term impacts on matching behavior. Because the annuity replaced household income that was lost due to the husband death, payments were higher in couples with more spousal specialization, in marke

43、t and non-market work Survivors insurance thus constituted a public subsidy on matchcs with highly unequal earnings (capacitics). I thcorctically show that under a standard supermodularity assumption on marital surplus, assortative matching arises in the absence of survivors insurance; however, when

44、 survivors benefits favor one-sided unassortative mat dies, assortativeness may break down. In a nutshell, removing survivors insurance should induce a larger share of skilled men to match with skilled worneiL To test this，I study the density of the share of highly skilled men that marry a woman of

45、lower skill. I show that the share of newly married couples that are assortatively matched on education increases by four percentage points following the introduction of the new marriage contract; fuither, this increase is permancnt. This suggests th&t the old survivors insurance program promoted un

46、assortative matching and spousal specialization in market and nonmarket work.These findings establish that tying social insurance to marriage has a significant impact on the volume and nature of co nt racting in the marriage marke t. In Sec tion 8, I formally examine the implications of these findin

47、gs for a social planner who wishes to alleviate old age poverty. In the presence of such marriage marke t responses, the social planner faces a t rade-off bet ween, on the one hand, protec ting women who do not participate in the labor market against poverty at the end of life, and distorting marria

48、ge market behavior, on the other. The gender wage gap is a key determinant for whethei it is optimal, in a given society, to separate social insurance from marriage.My study builds on and contributes to several strands of the litera.ture. First： my findings speak to the broader question of how benef

49、its and taxes should be designed for couples. Policies differ widely across developed countries: The United States, for example, connects a wide range of (tax) benefits to the marriage contract; in contTQst： in Sweden, taxation is individual and social insurance now is untied from marriage To unders

50、tand how to optimally design benefits for couples, one key step is to understand the nature of marriage market responses. I establish that tying social insurance to marriage has a significant impact on the volume of contracting in the marriage market. This finding contributes to previous literature

51、that has shown mixed evidence on the existence of marital responses to implicit penalties or subsidies inherent in tax codes.Whittington and Alm (2005), Bitler et al. (2006), and Light and Omori (2008) find that a smaller marriage penalty increases the Tate of marriage relative to divorce or cohabit

52、ation, whereas Bitler et al. (2004), Fitzgerald and Ribar (2005), and Knab et al. (2008) reach the opposite conclusion. In the context of Austria, Frimmel et al. (2012) show that the removal of cash transfers upon marriage raises the marriage rate. See Alni et al. (1999) and Moffitt (1998) for surve

53、ys of earlier contributions. To my knowledge, this paper is the first to document responses to benefits tht only pay out in the far future, and to &nalyze how tying benefits to the marriage contract affects matching My discussion of the optimal design of survivors benefits in the presence of marriag

54、e market responses relates to Klcven et al. (2009), who study optimal taxation of couples* labor income To focus on the spouses7 labor supply responses, they take the marriage decision as given and assume a unitary model of the houschol止 I begin to take a step in the oppositc but complementary direc

55、tion by focusing on the marriage margin and allowing for bargaining within the household.Casanova (2010) and Henriques (2012) also analyze the link bet ween marriage and social security, albei t w计 h d iff ere nt focuses. In the cont ext of the U.S., Casanova (2010) studies the incentives for joint

56、retirement and Henriques (2012) shows that husbands; claiming behaviors do not maximize total household benefits, but that they instead maximize their own benefi ts.Second, my analysis of endogenous selection into marriage suggests that take-up of survivors insur ance may be subject to adverse selec

57、tion. This argument builds on a literature on adverse selection in private insurance markets and the premise of government intervention in response.Akerlof (1970) and Rothschild and Stiglitz (1976) develop the theoretical argument. Chiappori and Salanie (2000) begin to examine 计 s key t hcore tical

58、prediction, a positive correlation bet ween (iernand and risk t ype. The empirical evidence, which I discuss in Section 5.2, is mixed. I focus not on a product offered in a private insurance market： but on a government-provided scheme that is supplied indirectly through the marriage contract.Third,

59、I contribute to the large literatui、e that analyzes how married couples, decisions depend on factors that influence each spouses relative bargaining position,1 by also analyzing impacts on earlier st ages of the mating process: matching and cohabi t&tionP Contrary to models in which matching is mode

60、led as the formation of married couples, here, the decision to form a couple is the decision to “start dating. This separation of couple formation from marriage is important in that it yields a marriage market equilibrium in which a strictly nonzero mass of couples chooses to cohabitate; a prerequis

61、ite for predicting for how social insurance impacts this empirically important group.Such factors include gender-targeted transfers; laws governing divorce, alimony, or access to contraception; and sex ratios in the marriage market. See, e.g., Angrist (2002), Chiappori et al. (2002), Dufio (2003), W

62、olfers (2006), Stevenson and Wolfers (2006); Stevenson (2008), Edlund and Kopczuk (2009), Voena (2011), and Attanasio and Lechene, (forthcoming). My results for cohabiting couples relate to Chiappori et al. (2012), who study a reform of alimony laws in Canada and find that it has distinct impacts on

63、 cohabitating and married couples. My discussion of matching relates to Becker (1973), who shows that spousal specialization in market and nonmarket work is optimal in the presence of comparative advantage and learning by doing. In Sweden in 2000, 30 percent of the parents of newborn children were m

64、arried, and 67 percent were cohabiting. See, e.g., Laibson (1997), Benartzi and Thaler (2004), Carroll et al. (2009), Beshears et al. (2010), Beshears et al. (2011), and Beshears et al. (2012).Fourth, a key distinction bet ween the extan t literature and my paper is the isolation of long-run incenti

65、ves. In contrast to a fiscal or legal change that affects the immediate payoffs from marriage, the reform that I study only affects transfers in the far future. The responses that I document thus rcflect forward-looking behavior, with implied financial-planning horizons of up to 40 years. My finding

66、s thus relate to the literature that analyzes how individuals prepare for financial security in retirement. Specifically, I show that when a publicly provided financial instrurnent is tied to marriage, financial- planning stra.tegies extend beyond asset accumulation to marital decisions. A twofold intervention into marriage and financial markets could therefore cre