University of Southern California Dissertations and Theses

Essays in labor economics: demographic determinants of labor supply

(USC Thesis Other)

Essays in labor economics: demographic determinants of labor supply

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content Essays in Labor Economics Demographic Determinants of Labor Supply by Ali Abboud A dissertation presented to the F aculty of the USC Graduate School University of Southern California in partial fulllment of the requirements for the degree of Doctor of Philosophy in the sub ject of Economics. May 2019 Cop yrigh t 2019 Ali Abb oud Acknowledgments I am grateful to Matthew Kahn, Je Nugen t and Vittorio Bassi for their guid- ance, supp ort and encouragemen t in completing this w ork. I ha v e also b eneted immensely from discussions with F ann y Camara, P aulina Oliv a, Sandra Rozo, and other facult y mem b ers at USC. I ha v e b een luc ky to share the past six y ears with w onderful friends. I appreciate the commen ts and suggestions of m y fello w studen ts - esp ecially Andreas, Mahrad, Mic hele, Rac hel and Vladimir - their mark is presen t all o v er this man uscript. My life in Los Angeles w as blessed b y the presence of friends who lo ok, sp eak and think lik e me. Ali, A yda, Jenane, Joanna, Mohamad, Na jw a, No or Al-Huda and Zeena y our presence in m y life made home closer. I thank Lena for her lo v e and patience. Y our presence in the latest c hapter of m y life brough t jo y and condence in a turbulen t phase. I’m lo oking forw ard for the remaining c hapters with y our presence b y m y side. Finally , I am grateful to m y lo ving family . Nabil and Rami ha v e b een the b est brothers I could ha v e ev er wished for, their supp ort and lo y alt y are b ey ond limits. My paren ts, Hassan and Haifa w ere the greatest supp orters of m y academic v en ture from the da y they taugh t me ho w to read and write till this da y . i Abstract This dissertation comprises three essa ys in lab or economics. The rst pap er studies the consequences of unplanned births on w omen’s careers. T o answ er this question, I exploit v ariation in access to legal ab ortion that could p oten tially af- fect the rates of unplanned births. I then use this random v ariations in fertilit y realizations to ev aluate the eect of fertilit y sho c ks on w omen’s earnings. The early legalization of ab ortion in v e US states led to v ariation in access to ab or- tion across states and birth cohorts, whic h allo ws the estimation of the eect of accessing ab ortion at a certain age on w omen’s fertilit y . The evidence suggests that early access to ab ortion led to a signican t dela y in the age of start of moth- erho o d. I also do cumen t an increase in completed fertilit y among blac k w omen who receiv ed access to ab ortion early in their fertilit y cycle. I then nd that lab or earnings increase b y 13% as a result of the dela y of an unplanned start of mother- ho o d. Results from the eect of age of start of motherho o d on lab or supply and o ccupation status suggest that most earnings gains are due to b etter o ccupations rather than increase in hours w ork ed. The second pap er explores ho w lab or hours exibilit y aect the willingness of married w omen to participate in the lab or mark et. I exploit a la w c hange in F rance reducing full time w orkw eek hours from 39 to 35 hours. I sho w that the lab or force participation among married w omen increases b y 3 p ercen tage p oin ts as a result of the la w c hange. I then set up a discrete c hoice mo del of lab or force participation. P articipation costs v ary with the total n um b er of hours sp en t in the w orkplace. Moreo v er, I allo w for exible correlation b et w een w ages and hours sp en t in the w orkplace. Estimates of the structural mo del sho w a p ositiv e correlation b et w een w ages and hours sp en t in the w orkplace and that longer hours are more taxing for w omen than men. The last pap er prop oses a theoretical framew ork that links longevit y and qual- it y of life to individual c hoices of h uman capital in v estmen t and retiremen t. The mo del predicts an increase in sc ho oling y ears and a dela y in retiremen t in resp onse to an increase in life exp ectancy and health qualit y . These ndings rationalize recen t trends in emplo ymen t, where emplo ymen t has b een decreasing a y ounger ages and increasing at older ages. ii Contents 1 Ev olution of W omen’s Lifetime Earnings in Resp onse to Early F ertilit y Sho c ks 1 1.1 In tro duction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 1.2 Ab ortion Legalization at State and F ederal Lev els . . . . . . . . . . 6 1.3 Iden tication Strategy . . . . . . . . . . . . . . . . . . . . . . . . . 7 1.3.1 Sources of V ariation . . . . . . . . . . . . . . . . . . . . . . 8 1.3.2 Estimation of the Eect of A ccess to Ab ortion on F ertilit y Outcomes . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10 1.3.3 Threats to Iden tication and V alidit y of Researc h Design . . 12 1.4 Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13 1.5 Eect of Ab ortion A ccess on F ertilit y . . . . . . . . . . . . . . . . . 16 1.5.1 Completed F ertilit y . . . . . . . . . . . . . . . . . . . . . . . 16 1.5.2 Birth Timing . . . . . . . . . . . . . . . . . . . . . . . . . . 17 1.5.3 Blac k W omen Marriage Outcomes . . . . . . . . . . . . . . . 20 1.5.4 Robustness Chec ks . . . . . . . . . . . . . . . . . . . . . . . 21 1.6 Consequences of Unplanned Start of Motherho o d on W omen’s Career 23 1.6.1 Data on Lab or Mark et Outcomes . . . . . . . . . . . . . . . 24 1.6.2 Estimation Results . . . . . . . . . . . . . . . . . . . . . . . 26 1.7 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29 2 W orking More when W orking Less: The Eect of Reduced W ork- w eek Hours on P articipation of Married W omen 49 2.1 In tro duction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49 2.2 Reduced W orkw eek Hours: The Aubry La w . . . . . . . . . . . . . 51 2.3 Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53 2.4 Eect of Lab or Hour Reduction on P articipation . . . . . . . . . . . 54 2.4.1 Descriptiv e Evidence and St ylized F acts . . . . . . . . . . . 54 2.4.2 Reduced F orm Evidence . . . . . . . . . . . . . . . . . . . . 57 2.5 A Mo del of P articipation Choice . . . . . . . . . . . . . . . . . . . . 59 2.6 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61 iii 3 The Eect of Extended Longevit y and Impro v emen t in Health Qualit y on Career Span 70 3.1 In tro duction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70 3.2 A Mo del of Health, Sc ho oling and Retiremen t . . . . . . . . . . . . 72 3.2.1 Mo del Setup . . . . . . . . . . . . . . . . . . . . . . . . . . . 72 3.2.2 Optimal Sc ho oling and Retiremen t Choices . . . . . . . . . . 74 3.2.3 Changes in Health Qualit y and Mortalit y . . . . . . . . . . . 77 3.3 Sim ulation Results . . . . . . . . . . . . . . . . . . . . . . . . . . . 79 3.4 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79 A App endix to Chapter 1 84 A.1 Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84 A.1.1 F ertilit y Outcomes . . . . . . . . . . . . . . . . . . . . . . . 84 A.1.2 Lab or Mark et Outcomes . . . . . . . . . . . . . . . . . . . . 85 A.2 T ables and Figures . . . . . . . . . . . . . . . . . . . . . . . . . . . 87 A.2.1 Estimation Results b y Y ear of Age . . . . . . . . . . . . . . 87 A.2.2 Placeb o T est: Ruling Out Changes in F ertilit y T rends . . . . 94 A.3 Robustness to P oten tial Migration . . . . . . . . . . . . . . . . . . . 95 A.3.1 Estimation Using P oten tial Non-Mo v ers . . . . . . . . . . . 95 A.3.2 Evidence on Selectiv e migration . . . . . . . . . . . . . . . . 100 B App endix to Chapter 2 102 B.1 Lik eliho o d F unction . . . . . . . . . . . . . . . . . . . . . . . . . . . 102 C App endix to Chapter 3 103 C.1 Solution to the dynamic problem . . . . . . . . . . . . . . . . . . . 103 iv List of T ables 1.1 Summary Statistics of F ertilit y and Lab or Mark et Outcomes 31 1.2 Illustration of the Iden tication . . . . . . . . . . . . . . . . . 32 1.3 Cross States Dierences in Completed F ertilit y (Age Group) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33 1.4 Dierence-in-Dierences Estimates of the Eect of Ab or- tion A ccess on Completed F ertilit y . . . . . . . . . . . . . . . 34 1.5 Cross States Dierences in Births Timing (Age Group) . . 35 1.6 Dierence-in-Dierences Estimates of the Eect of Ab or- tion A ccess on Births Timing . . . . . . . . . . . . . . . . . . . 36 1.7 Cross States Dierence in Blac k W omen’s Marriage Out- comes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37 1.8 Dierence-in-Dierences Estimates of the Eect of Ab or- tion A ccess on Blac k W omen’s Marriage . . . . . . . . . . . . 38 1.9 Eect of Start of Motherho o d on Earnings and Lab or Supply 39 1.10 Eect of Start of Motherho o d on Lab or Outcomes of Em- plo y ed W omen . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40 2.1 Summary Statistics . . . . . . . . . . . . . . . . . . . . . . . . . 62 2.2 Changes in Lab or Hours W ork ed . . . . . . . . . . . . . . . . . 63 2.3 W eekly Hours W ork ed . . . . . . . . . . . . . . . . . . . . . . . 64 2.4 Eect of W orkw eek Hours Reduction on Married W omen’s P articipation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65 2.5 Structural Estimates . . . . . . . . . . . . . . . . . . . . . . . . 66 3.1 Sim ulation P arameterization . . . . . . . . . . . . . . . . . . . 81 A.1 Cross States Dierences in Completed F ertilit y (Age Y ears) 87 A.2 Dierence-in-Dierences Estimates of the Eect of Ab or- tion A ccess on Completed F ertilit y . . . . . . . . . . . . . . . 88 A.3 Cross States Dierences in Births Timing (Age Y ears) . . . 89 A.4 Dierence-in-Dierences Estimates of the Eect of Ab or- tion A ccess on Births Timings . . . . . . . . . . . . . . . . . . 90 v A.5 Cross States Dierences in Completed F ertilit y (P oten tial Non-Mo v ers) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96 A.6 Dierence-in-Dierences Estimates of the Eect of Ab or- tion A ccess on Completed F ertilit y (P oten tial Non-Mo v ers) 97 A.7 Cross States Dierences in Births Timing (P oten tial Non- Mo v ers) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98 A.8 Dierence-in-Dierences Estimates of the Eect of Ab or- tion A ccess on Births Timing (P oten tial Non-Mo v ers) . . . 99 A.9 Evidence on Selectiv e Migration . . . . . . . . . . . . . . . . . 101 vi List of Figures 1.1 F ertilit y Outcome T rends . . . . . . . . . . . . . . . . . . . . . . . 41 1.2 Eect of A ccess to Ab ortion on Completed F ertilit y . . . . . . . . . 42 1.3 Eect of A ccess to Ab ortion on Age at Start of Motherho o d . . . . 43 1.4 Eect of A ccess to Ab ortion on Birth Spacing . . . . . . . . . . . . 44 1.5 Eect of A ccess to Ab ortion on Completed F ertilit y b y Race . . . . 45 1.6 Eect of A ccess to Ab ortion on Age at Start of Motherho o d b y Race 46 1.7 Eect of A ccess to Ab ortion on Birth Spacing b y Race . . . . . . . 47 1.8 Eect of A ccess to Ab ortion on Marriage Outcomes for Blac k W omen 48 2.1 Changes in Distributions of W eekly Hours W ork ed b y Men and W omen 67 2.2 Y early Lab or F orce P articipation of Men and W omen b y Marrital Status . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68 2.3 Lab or F orce P articipation of Men and W omen b y Age and Marital Status in 1990 and 2002 . . . . . . . . . . . . . . . . . . . . . . . . 69 3.1 Sim ulation Results for Changes in Life Exp ectancy . . . . . . . . . 82 3.2 Sim ulation Results for Changes in Health Qualit y . . . . . . . . . . 83 A.1 Eect of A ccess to Ab ortion on Completed F ertilit y . . . . 91 A.2 Eect of A ccess to Ab ortion on Age at Start of Motherho o d 92 A.3 Eect of A ccess to Ab ortion on Birth Spacing . . . . . . . . 93 A.4 Eect of Ab ortion A ccess on Age of Motherho o d . . . . . . 94 vii Chapter 1 Evolution of W omen’s Lifetime Earnings in Response to Early F ertility Shocks 1.1 Introduction T w o imp ortan t features mark ed the liv es of w omen in the United States in the past 60 y ears. The a v erage n um b er of c hildren to whic h an American w oman ga v e birth dropp ed b y half from 3.65 births p er w oman in 1960 to 1.80 in 2017. During the same p erio d, American w omen’s participation in the lab or force increased from 38% to 57%. The relationship b et w een fertilit y and w omen’s career c hoices has dra wn the atten tion of economists and demographers, leading to the emergence of a large b o dy of empirical w ork in v estigating the eect of family size and birth ev en ts on w omen’s lab or mark et outcomes. Ho w ev er, studies that quan tify the eect of a fertilit y sho c k on w omen’s careers remain scarce. There are sev eral reasons wh y fertilit y sho c ks and their consequences for w omen merit consideration. First, unplanned pregnancies accoun t for a large share of preg- nancies. F or instance, in 2006 half the pregnancies in the U.S. w ere unin tended 1 , 1 The U.S. has one of the highest rates of unin tended pregnancies among ric h coun tries. The rate of unin tended pregnancy among w omen of fertile age in 1987 w as 54 pregnancies p er 1,000 w omen. It decreased slo wly to reac h 45 p er 1,000 in 1994 (Hensha w, 1998). The trend in unin tended pregnancy got in v erted in the follo wing decade to regain it’s previous lev el of 54 unin tended pregnancy p er 1,000 w omen in 2008, to drop again to 45 p er 1,000 w omen in 2011 (Finer and Zolna, 2016). 1 with the rate of unin tended pregnancies v arying extensiv ely b y age group (Mosher et al. , 2012). Second, the consequences of unplanned births migh t dier from the eects of planned births. A planned pregnancy is an endogenous c hoice that re- sults from a cost b enet analysis and in tra-household bargaining (Bec k er, 1993). W omen who self-select in to pregnancy are therefore lik ely to ha v e a lo w er birth cost 2 . Birth cost dep ends on the opp ortunit y cost of the time dedicated to par- en ting whic h is most lik ely larger for high earning w omen (Wilde et al. , 2010). F ertilit y sho c ks nonetheless can o ccur at dieren t rates among w omen with dif- feren t p oten tial earning and birth p enalties. Moreo v er, the birth cost migh t v ary during a w oman’s lifetime: if pregnancy timing w as completely deterministic, then it w ould o ccur in p erio ds where the cost of birth is the lo w est. W ard and Butz (1980) nd that couples time births to a v oid p erio ds when female w age rates are exp ected to b e high. Ho w ev er, a fertilit y sho c k can o ccur in phases that migh t b e crucial in determining the shap e of the earnings prole. F or example, an unplanned birth during teenage y ears can disrupt sc ho oling, decreasing future earnings as a result of lo w er h uman capital. Another crucial phase is the early lab or mark et exp erience (Blundell et al. , 2016), where an unplanned birth of a c hild can driv e a w oman outside the lab or mark et at a y oung age making it harder for her to rejoin the lab or mark et later. The main c hallenge in iden tifying the eect of fertilit y sho c ks on w omen’s earn- ings arise from omitted v ariables. Unobserv ed factors, suc h as p erception of gender role and fertilit y and career preferences, are lik ely determinan t of w omen’s career tra jectories and ma y b e correlated with the probabilit y of unplanned pregnancies. Moreo v er, a w oman’s career p oten tial can aect her optimal con traceptiv e eort c hoice, leading to self selection in to dieren t lev els of unplanned pregnancy risks. Giv en the impracticalit y of an exp erimen tal design that w ould randomly assign fertilit y sho c ks, researc hers ha v e to rely on quasi-exp erimen tal v ariation to iden- tify the desired eect. There are t w o kinds of approac hes that ha v e b een dev elop ed in the literature. The rst emplo ys an instrumen tal v ariable approac h exploiting v ariations resulting from random biological sho c ks suc h as birth of a t win (Rosen- 2 Kuziemk o et al. (2018) nds that w omen, esp ecially those with higher education, underes- timate the consequences of pregnancy on their lab or supply . Ho w ev er, they do not refute that w omen an ticipate a birth p enalt y when making their fertilit y decisions 2 zw eig and W olpin, 1980; Bronars and Grogger, 1994; Angrist and Ev ans, 1998) or miscarriages (Hotz et al. , 2005; Miller, 2011). The t wins instrumen t constitutes a sho c k on the in tensiv e margin of fertilit y . If the w age p enalt y is suered b y the birth ev en t exclusiv ely then a t win births will not capture the marginal p enalt y due to the sho c k if the pregnancy w as planned. The miscarriage instrumen t is most lik ely correlated with baseline health of pregnan t w omen, whic h is an imp ortan t determinan t of future earnings. Moreo v er, b oth instrumen ts rely on ev en ts that ha v e a v ery lo w frequency of o ccurrence. The second approac h, that I pursue in this pap er, exploits p olicy c hanges that could lead to v ariation in fertilit y sho c ks realizations. One suc h p olicy is ab ortion legalization (Angrist and Ev ans, 2000). In this pap er, I estimate the eect of an early fertilit y sho c k unplanned start of motherho o d on w omen’s earnings. I exploit b et w een state v ariation in timing of ab ortion legalization to iden tify the eect of ab ortion accessibilit y on v arious fertilit y outcomes. The cross state and birth cohort v ariations in exp osure allo w me to iden tify the eect of access starting at a giv en age. The estimates sho w a signican t dela y of half y ear in start of motherho o d for w omen who obtained access to ab ortion b efore the age of 21. These eects are large giv en that the in sample age at whic h motherho o d starts is 23 on a v erage. Estimates of completed fertilit y sho w a precisely estimated zero eect of ab ortion accessibilit y at all ages on completed fertilit y of white w omen. In con trast, I nd a substan tial increase in completed fertilit y for blac k w omen that obtained early access to ab ortion. While the latter result seems coun ter in tuitiv e at rst glance, in v estigation of marriage outcomes sho w that, while the probabilit y of marriage is not aected b y access to ab ortion, h usbands of blac k w omen who had early access to ab ortion had signican tly higher earnings and w ere more lik ely to hold college degrees. The com bined ndings on fertilit y outcomes suggest that the eect of ab ortion accessibilit y is restricted to the early y ears of the fertilit y cycle, where it causes a substan tial dela y in the start of motherho o d b y a v oiding an unplanned start of motherho o d. I then pro ceed to estimating the eect of age at the start of motherho o d on earnings. An instrumen tal v ariable approac h is adopted to estimate the causal eect, using state ab ortion exp osure as an instrumen t. I nd that p ostp oning motherho o d for one y ear increases y early lab or earnings b y an a v erage of $2; 194, whic h corresp onds to a 13% increase from the mean. Moreo v er, the ndings sho w 3 that impro v emen t in o ccupation status accoun t for most of the observ ed gains in earnings. This is attested to b y the signican t increase in the rep orted o ccupation index caused b y dela y ed start of motherho o d, whereas the eects found on lab or supply c hoices are mostly insignican t. This study con tributes broadly to the literature on the eect of motherho o d on w omen’s careers. This includes a large b o dy of literature that ev aluates the eect of motherho o d on lab or supply and earnings (Rosenzw eig and Sc h ultz, 1985; Hotz and Miller, 1988; Lundb erg and Rose, 2000; Sigle-Rush ton and W aldfogel, 2007; Lundb org et al. , 2017; Ch ung et al. , 2017; Aaronson et al. , 2018; Klev en et al. , F orthcoming). Sp ecically , m y pap er con tributes to the literature ev aluat- ing the consequences of fertilit y sho c ks on w omen’s careers. Previous pap ers in this literature nd mixed eects of fertilit y sho c ks on w omen’s earnings and lab or supply . F or instance, using miscarriage as an instrumen t for a v oided teenage preg- nancies Hotz et al. (2005) 3 nd an increase in lab or earnings and ann ual hours w ork ed at older age for w omen who b ecame mothers while they w ere teenagers. In con trast Klepinger et al. (1999) nd that adolescen t motherho o d leads to a sig- nican t decrease in sc ho oling and early lab or mark et exp erience resulting in lo w er lab or earnings. Similarly , using exp osure to ab ortion, Angrist and Ev ans (2000) do cumen t an increase in high sc ho ol completion and lab or supply of blac k w omen who a v oided teenage pregnancy . The fo cus of these studies is restricted to fertil- it y sho c ks among teenage w omen, in v estigating sc ho oling as a main mec hanism. Ho w ev er, I in v estigate p oten tial fertilit y sho c ks throughout the whole fertilit y cy- cle. My results sho w that access to ab ortion has an eect b ey ond teenage y ears on dela ying unplanned start of motherho o d. The imp ortance of in v estigating the eect of rst birth b ey ond teenage y ears is highligh ted in Herr (2016), who nds that a rst birth disrupting a w oman’s early career migh t b e more consequen tial than a rst birth b efore en tering the lab or mark et. This pap er is closest to Miller (2011) who studies the eect of motherho o d timing on career path. My ndings on the eect of dela ying start of motherho o d on y early earnings concur with her ndings. In Miller’s study , ho w ev er, the increase in earnings is mostly attributed to an increase in lab or supply , while increase in w age rates accoun t for one third of the increase in total earnings. Whereas I nd 3 see also Hotz et al. (1997) 4 that most of the eect streams from the increase in w age rate. This pap er also con tributes to a second line of literature ev aluating the con- sequences of ab ortion legalization. Previous studies of ab ortion p olicy c hanges in the U.S. and elsewhere assess the eect of ab ortion legalization on c hildren selec- tion (Grub er et al. , 1999; Donoh ue and Levitt, 2001; Ananat et al. , 2009), c hildren outcomes (P op-Elec hes, 2006), mother’s health (Clark e and Muhlrad, 2018) and fertilit y (Kane and Staiger, 1996; Levine et al. , 1999; Angrist and Ev ans, 2000; Levine and Staiger, 2004). T w o particular features distinguish m y pap er. First, I study the eect of ab ortion access throughout the fertilit y cycle in con trast to pre- vious literature fo cusing on w omen’s fertilit y during their teenage y ears. Second, I in v estigate the eect of the access on b oth fertilit y quan tit y and timing. Most previous pap ers in v estigate y early rate at state or coun t y lev els. This could b e misleading as a drop in y early fertilit y rate can b e due to ab ortion dela ying timing of births rather than decreasing completed fertilit y . Studying the consequence of ab ortion accessibilit y on w omen fertilit y outcomes is of particular p olicy relev ance. This is esp ecially true in ligh t of the recen t reg- ulatory trends put in place b y m ultiple states to limit access to ab ortion. Since the ab ortion rate attained its historic p eak of 29.3 ab ortions p er 1,000 w omen aged 15-44 in 1981, usage has b een in steady decline for the past three decades. It is widely b eliev ed that the decline in ab ortion use can b e fully or partially attributed to state ab ortion restrictions. The rst t yp e of regulations target ab or- tion pro viders b y limiting public funds to facilities pro viding ab ortion pro cedures as w ell as other regulations that increase the cost of op eration. Another set of regulations target ab ortion patien ts directly , including h usband consen t, paren tal consen t for minors, mandatory counseling and w aiting p erio d b et w een counseling and ab ortion date. The n um b er and restrictiv eness of state regulations ha v e b een asso ciated with a sharp decrease in ab ortion pro viders (Jones and K o oistra, 2011; Jones and Jerman, 2014, 2017), along with an increase in a v erage distance to near- est pro vider (Bearak et al. , 2017) and the a v erage out-of-p o c k et cost of ab ortion pro cedures (Jones et al. , 2018) as w ell as an increase in an tiab ortion harassmen t (Jones and K o oistra, 2011). The literature ev aluating the eect of state regula- tions remains scarce. One particular exp erience, T exas House Bill 2 of 2013, has b een studied in a series of recen t pap ers (Quast et al. , 2017; Grossman et al. , 5 2017; Cunningham et al. , 2018; Fisc her et al. , 2018). Their ndings suggest that the increase in distance to the nearest ab ortion pro vider, caused b y the bill, de- creased the ab ortion rate in the state, while the evidence on the eect of the la w on fertilit y is mixed. While this pap er do es not sp eak directly to the eect of state regulations, its ndings can shed ligh t on their p oten tial consequences. The remainder of this pap er is organized as follo ws. A brief history of ab ortion related legislativ e c hanges in the United States is pro vided in section 2. Section 3 discusses the details of the strategy emplo y ed to iden tify the eect of access to ab ortion on fertilit y . Section 4 presen ts the data used in estimation. Estimation results of the eect of ab ortion access on fertilit y are presen ted in section 5. In section 6, I presen t the estimation approac h and results for the eect of early fertilit y sho c ks on earnings. The last section concludes. 1.2 Abortion Legalization at State and F ederal Lev- els In 1973 the United States Supreme Court in the case of Ro e v W ade ruled that State and F ederal restrictions on ab ortions in the United States violate the 14th amendmen t and henceforth gran ted w omen legal access to ab ortion in all States. Prior to that landmark decision ab ortion accessibilit y largely dep ended on a w oman state of residency . Bet w een 1967 and 1970 a partial lib eralization of ab ortion la ws w ere adopted in 15 states. Most of these c hanges p ermitted ab ortion in cases of rap e, incest and pregnancy complications that could b e life threatening to the mother. Sklar and Berk o v (1974) conclude that legal c hanges in what they call reform states led to a signican t reduction in o v erall fertilit y , esp ecially among unmarried w omen. Angrist and Ev ans (2000) use fertilit y v ariation resulting from state reforms to ev aluate the eect of teenage fertilit y on sc ho oling and lab or outcomes. They do cumen t a signican t p ositiv e eect of ab ortion accessibilit y on sc ho oling and emplo ymen t of blac k w omen. The most imp ortan t legal c hanges prior to national legalization to ok place in 1970, when v e States (Alask a, California, Ha w aii, New Y ork and W ashington), 6 rep ealed their an ti ab ortion la ws 4 . This pap er uses legislativ e c hanges in the legal status of ab ortion in the early 70s to measure the eect of ab ortion accessibilit y on fertilit y and lab or mark et outcomes of w omen. There are t w o main reasons for considering Rep eal states rather than the Re- form states as the treatmen t group for our analysis. First, as rep orted in Sklar and Berk o v (1974) Rep eal states had the highest rates of legal ab ortions p er thousand w omen in 1971 with New Y ork (27.1), Ha w aii (23.6), California (23.5), W ashington (19.7) and Alask a (17.4). Among the Reform states only Oregon (15.7), Dela w are (13.7) and Maryland (11.4) had rates that w ere close to those of Rep eal states while the rest of them had a rate of ab ortion lo w er than 10 p er thousand w omen 5 . Second, in Rep eal states ab ortion b ecame legal on request while in Reform states ab ortion accessibilit y w as conditional on sp ecic circumstances. Giv en the in terest in understanding the eect of ab ortion accessibilit y on lab or outcomes, optional rather than necessary ab ortion is more relev an t for our purp ose. 1.3 Identication Strategy A t ypical w oman’s repro ductiv e cycle extends for around thirt y y ears. Ovu- lation starts at around the age of 14 and con tin ues through her forties, with the lik eliho o d of conceiving a c hild dropping drastically after the age of 42. F ertilit y decisions are made b y w omen throughout this phase of their life. These decisions could b e summed up as c ho osing the total n um b er of c hildren they wish to ha v e and the timing at whic h they c ho ose to ha v e them. W omen often mak e these c hoices sim ultaneously with their career c hoices, as b oth are in terrelated. Career aspirations migh t aect their fertilit y c hoices, but also fertilit y realizations migh t aect their career paths. Ho w ev er, fertilit y realizations are not fully deterministic, as there is alw a ys the c hance of unplanned pregnancies. Sev eral biological and b eha vioral factors cause the probabilit y of un w an ted pregnancy to v ary during the repro ductiv e cycle. Moreo v er, the consequences of an unplanned pregnancy on a w oman’s career, if an y , migh t dep end on the phase of her life at whic h it o ccurs. 4 In California it w as the opinion of the State Supreme Court in P eople v Belous that state la ws banning ab ortion are unconstitutional. 5 See Sklar and Berk o v (1974), table 3 7 Hence, the eect of legal access to ab ortion on realized fertilit y , and consequen tly on lab or mark et c hoices, can v ary widely b y the age at whic h w omen obtain access to it. The remainder of this section describ es ho w the eect of access to ab ortion at ev ery y ear of age can b e iden tied. 1.3.1 Sources of V ariation T o iden tify the eect of legal access to ab ortion at ev ery y ear of age, exogenous v ariations to ab ortion accessibilit y throughout w omen’s fertile life are needed. This is exactly the t yp e of v ariation that w omen, who w ere fertile in the 1970s and the 1980s, ha v e exp erienced as a result of the legal c hanges describ ed ab o v e. The three y ear lag in nation wide rep eal generated v ariation across states in the age at whic h w omen of similar birth cohorts could obtain access to ab ortion. F urthermore, dierences in age at legalization within state led to v ariation in access across birth cohorts. Consider for instance a w oman b orn in 1955 and living in one of the rep eal states. She w as 15 y ears old when she obtained access to legal ab ortion for the rst time. A w oman b orn one y ear earlier and residing in the same state, obtained access to legal ab ortion in the same y ear as her state p eer, ho w ev er that happ ened when she w as 16 y ears old. Mean while their birth cohort p eers living in a non-rep eal state did not ha v e access to ab ortion un til national legalization three y ears later, whic h also happ ened with one y ear of age dierence. Comparing outcomes of these w omen across states and birth cohorts allo ws iden tication of the eect of access to ab ortion at age 15. This approac h of using v ariation in in tensit y of exp osure to legal ab ortion across birth cohorts and states is similar to the approac h emplo y ed in Duo (2001). W omen residing in dieren t states migh t ha v e dieren t preferences for fertilit y and career paths. W e could also exp ect suc h heterogeneit y to exist across birth co- horts. The v alidit y of the dierence-in-dierences strategy rely on the assumption that in the absence of treatmen t, there are no dierences b et w een rep eal and non- rep eal states in the rate of c hange in taste for fertilit y and lab or mark et c hoices b et w een subsequen t cohorts. This assumption w ould not hold if the treatmen t assignmen t (states c ho osing to rep eal ab ortion ban) w as conditional on dierences in the trend of b et w een cohorts c hange in preferences. F or instance, if states’ re- 8 p eal in 1970 w as due to stronger shifts in preferences to w ards smaller family size among y ounger w omen in these states compared to w omen of similar age in the other states, then the estimates of the eect of ab ortion access w ould b e biased, sho wing larger (in absolute v alue) eects for y ounger w omen compared to older w omen. It should also b e noted, that fertilit y outcomes are age censored and lab or mark et outcomes correlate with age. Hence, when comparing across birth cohorts, it is crucial that fertilit y realizations are ac hiev ed and lab or mark et outcomes are compared at similar ages. F ailure to do so will confound the treatmen t eect with the age at observ ation. An illustration of the iden tication strategy is presen ted in T able 1.2. The a v erages of t w o fertilit y outcomes are compared across three birth cohort groups in rep eal and non-rep eal states. The y oungest group are those b orn b et w een 1950 and 1955, who w ere b et w een the ages of 15 and 20 when rep eals at the state lev el to ok place. The second group of w omen are those b orn b et w een 1940 and 1945, who w ere b et w een the ages of 25 and 30 when state rep eals to ok place. The eldest group consists of w omen b orn b et w een 1930 and 1935, who w ere b et w een the age of 35 and 40 when state rep eals w ere put in place. The data are from the 1970, 1980 and 1990 census (5 p ercen t Public Use Micro data Samples). I use the 1970 observ ations for the eldest group, the 1980 for the middle group and the 1990 observ ation for the y oungest group. This guaran tees that the comparison happ ens for observ ations at similar ages, when the w omen in our sample are b et w een the ages of 35 and 40. T w o fertilit y outcomes are considered; a w oman’s age at the time she ga v e birth to her rst c hild and the total n um b er of c hildren she ga v e birth to. The y oungest group obtained access to legal ab ortion early in their fertilit y cycle, the second group obtained access in the middle of their fertilit y cycle, while for the eldest group, ab ortion w as legalized when they w ere in the nal phases of their fertilit y cycle. If the impact of access to ab ortion diminishes with age at exp osure, fertilit y of the y ounger group should b e signican tly altered compared to the older cohorts, and there should b e no signican t dierence for the t w o eldest groups. F or eac h outcome, the rst and second ro ws in T able 1.2 presen t the a v erages for b oth groups of w omen in rep eal and non-rep eal states resp ectiv ely . The third ro w presen ts the dierences b et w een rep eal and non-rep eal state for eac h group. 9 W omen in rep eal states in all age groups dela y the age at whic h they giv e birth to their rst c hild compared to w omen in non-rep eal states. The dela y for the y oungest group is 13 mon ths on a v erage, and 7 mon ths for the other t w o groups. The dierence-in-dierences estimates in Mo dule A sho w that there is a signican t dierence of 6 mon ths b et w een w omen that obtained access at the b eginning of their fertilit y cycle and those that obtained access in the middle of their fertilit y cycle, while the dierence b et w een the latter and the w omen that did not obtain access to ab ortion is insignican t. Similarly in Mo dule B of the table, w e observ e smaller family size for w omen in rep eal states. T otal fertilit y is around 10 p ercen t lo w er in rep eal states compared to non-rep eal states across all age groups. The dierence-in-dierences estimates b et w een the t w o eldest groups is insignican t, whereas the dierence b et w een the y oungest and the second group is statistically signican t but the magnitude of the eect is small. This simple exercise pro vides evidence that earlier access to legal ab ortion has a signican t eect on fertilit y , most notably on increasing the age at whic h mothers giv e birth to their rst c hild. In what follo ws a more formal discussion of the iden tication strategy is presen ted. 1.3.2 Estimation of the Eect of Access to Abortion on F er- tility Outcomes Let y A i(sb) b e a fertilit y outcome observ ed at age A of w omen i, b orn in y ear b in state s. T o iden tify the direct eect of access to ab ortion at dieren t y ears of age on fertilit y , I estimate the follo wing equation: y A i(sb) = a X I=a I Repeal i(s) I 70 i(b) + a X J=a J J 73 i(b) + i(s) + i(sb) (1.1) whereRepeal i(s) is a dumm y v ariable indicating residency in a state s that rep ealed it’s ab ortion ban in 1970. a and a are the ages at whic h fertilit y starts and ends resp ectiv ely . I 70 (b) are a series of indicator v ariables that tak e v alue one if the individual of birth cohort b is of age I in 1970 (date at whic h state rep eal is assumed to b e in eect). Similarly J 73 (b) are indicator v ariables that tak e v alue one if an individual of birth cohort b is of age J in 1973 (date at whic h ab ortion is legalized at the federal lev el). s are states xed eects. 10 In this regression framew ork, dierencing outcomes of w omen across rep eal and non rep eal states and across birth cohorts iden ties the eect of access to ab ortion at earlier age. The c hart b elo w illustrates the iden tication strategy using t w o birth cohort groups of w omen b orn k y ears apart, with the y oungest b eing of age i2 (a; a) in 1970. The dierence i i+k 6 is the eect of obtaining access to ab ortion at age i compared to obtaining access at age i +k . Rep eal Non Rep eal Dierence y ounger cohort i + i+3 i+3 i older cohort i+k + i+k+3 i+k+3 i+k Dierence i i+k The dierence-in-dierences estimates compare the eect of access to ab ortion at t w o dieren t ages during the fertilit y cycle. Giv en that w omen are fertile for almost 30 y ears during their lifetime, this pro vides a large n um b er of estimates for ev ery y ear of age. One particular estimate of in terest is i = i a , whic h iden ties the eect of access to ab ortion when a w oman is of age i compared to obtaining access to ab ortion when she is at the end of her fertilit y cycle. W e could as w ell in v estigate the age trend in the co ecien ts; a decline in the magnitude of co ecien ts withi indicates that early access to ab ortion has a stronger eect on the outcomes of in terest. Alternativ ely , an increasing (in absolute v alue) trend implies that access to ab ortion has a stronger eect at a later age. F or the remainder of this pap er, it is assumed that w omen are fertile b et w een the ages of 15 and 40. This assumption is not restrictiv e ev en if it is kno wn for a fact that w omen can still conceiv e c hildren b ey ond this p erio d. Nev ertheless, most pregnancies are conceiv ed within this age in terv al, and more imp ortan tly , this assumption is more ab out restricting our in terest to the eect of exp osure to ab ortion during this age group rather than actually imp osing a fertilit y restriction. 6 This "dierences-in-dierences" estimation strategy is v alid under the follo wing iden tifying assumption E[y A s=R;b+k y A s=R;b jT = 0] = E[y A s= R;b+k y A s= R;b jT = 0], where T is an indicator of treatmen t whic h in this framew ork is early exp osure to ab ortion due to legislativ e c hanges at the state lev el. This assumption w ould not hold if legislativ e c hanges at State lev els w ere endogenously enacted due to c hange in preferences for fertilit y and lab or mark et c hoices b et w een subsequen t birth cohorts. 11 F or the main analysis I use an alternativ e sp ecication of equation (1.1), where instead of lo oking at the eect of exp osure b y age y ear I in v estigate the eect of exp osure at coarser age group. Age y ears b et w een 15 and 40 are put in groups of 3 y ears of age eac h. Similar to the individual age y ear exp osure ab o v e, v ariation in exp osure b y age group and across states allo w iden tication of the eect of exp osure at a certain age group. The follo wing equation is estimated: y sg = 9 X g=1 g Repeal S AG 70 g + 9 X g=1 g AG 73 g + 1 age + 2 age 2 + s + sg (1.2) where AG g (AG 1 = [15 17];AG 2 = [18 20];:::AG 9 = [39 40]) is a indicator for age group. Using age group instead of individual y ears of age allo w us to con trol for age, I do so b y including age and age 2 . Similar to the p er age y ear case ab o v e the eects w e are in terested in iden tifying are the eect of exp osure to ab ortion starting age phase g captured b y g = g 9 . The standard errors sg are clustered at the state lev el. 1.3.3 Threats to Identication and V alidity of Research De- sign As stated earlier, the dierence-in-dierences strategy describ ed ab o v e pro vides consisten t estimates of the eect of ab ortion accessibilit y on fertilit y outcomes, under the assumption that in the absence of state rep eal the trends in fertilit y preferences are similar across states. This assumption w ould not hold if the rep eal of the ab ortion ban at the state lev el w as conditional on some dieren tial trend in fertilit y outcomes. F or example, if the rep eal of ab ortion bans in the dieren t states w as in resp onse to a more rapid increase in teenage pregnancy among y ounger birth cohorts in rep eal states com- pared to non-rep eal states, then the estimate w ould suer from a serious selection bias problem that understates the eect of ab ortion access. If the rep eal w as due to stronger preference for smaller family size among the y ounger birth cohorts in rep eal states, then the estimate confound the eect of ab ortion access and fertilit y preferences whic h w ould most lik ely cause an up w ard bias in the estimation of the 12 eect. T o v erify the v alidit y of the researc h design, I in v estigate the trends in fertilit y outcomes for cohorts that had dieren t lev els of access to ab ortion. A v erages of the t w o main fertilit y outcomes of in terest - completed fertilit y and age at start of motherho o d - are computed separately for rep eal and non-rep eal states for dieren t birth cohorts. The trends in completed fertilit y are sho wn in Figure 1.1a, and Figure 1.1b sho ws the trends in age of start of motherho o d. The t w o v ertical lines represen t the cutos for the birth cohorts with dieren t lev els of exp osure across rep eal and non-rep eal states. The birth cohorts to the left of the rst v ertical line ha v e no access to ab ortion during their fertilit y cycle in b oth groups of states, while the birth cohorts to the righ t of the second v ertical line ha v e similar access to ab ortion in b oth groups of states throughout their fertilit y cycle, as they are b oth exp osed to the federal accessibilit y p ost Ro e v. W ade. The birth cohorts in b et w een are the ones with the v arying lev els of exp osure to ab ortion. The plots clearly sho w that for b oth fertilit y outcomes, the trends in fertilit y are similar across b oth state groups for the birth cohorts that had no access or late access to ab ortion. As for the birth cohorts that had early access to ab ortion (1950-1959), w e can see that the trend in age at start of motherho o d b ecome steep er in the rep eal states compared to non-rep eal states, suggesting that ab ortion access led to a dela y in the start of motherho o d. As for completed fertilit y , Figure 1.1a sho ws con v ergence in completed fertilit y for the birth cohort that had dierence in ab ortion access early in their life, suggesting that ab ortion access early in fertilit y cycle led to increase in completed fertilit y . 1.4 Data T o estimate equation (1.1) I w ould ideally lik e to ha v e access to a panel dataset of w omen who w ere at v arious phases of their fertilit y cycle during the ab ortion reforms of the early 70’s. In addition to p ermitting prop er estimation of the equation of in terest, suc h data w ould allo w me to study the dynamic eect of ab ortion access. Ho w ev er, no suc h data set is a v ailable. Instead, I rely on cross sectional samples to estimate the eect in question. The follo wing section discusses the main data set used in this analysis, and ho w the estimation is adapted for the 13 cross sectional nature of the data. The analysis is conducted using data from four observ ation y ears (1970, 1980, 1990 and 2000) of the census 5 p ercen t Public Use Micro data Samples (PUMS Bureau of the Census). Giv en the dates of state and federal p olicy c hanges and the assumption that w omen’s fertilit y cycles span b et w een the ages of 15 and 40, I restrict the fo cus of the study to w omen b orn in the United States b et w een the y ears 1930 and 1955. This insures that at the time of the state lev el rep eal, the y oungest birth cohort is 15 y ears of age and the oldest cohort is 40. Three y ears later, when the federal court rep eal tak es place the age span is 18 to 43. There are t w o main adv an tages of using the census data. First, they pro vide a large sample size that allo ws b etter estimation precision. This is imp ortan t since the ob jectiv e is to estimate the eect of exp osure to ab ortion at eac h y ear during the fertilit y cycle, ev en if ab ortion has a considerable eect o v er the lifetime of a w oman, the eect migh t b e mo dest for some y ears of age. Second, census data rep orts a wide range of information on fertilit y and lab or mark et c hoices for w omen of v arious birth cohorts. Despite these adv an tages, the use of the census for the purp ose of this study p oses t w o main c hallenges . First, the census rep orts birth state and the state of residency at observ ation, whereas I w ould lik e to observ e the state of residency during the fertilit y phase, since it’s the la ws of this state that determines if a w oman has access to legal ab ortion of not. F or the main analysis I assume that the state of residency during the fertile p erio d of life is the same as the state of birth. This w ould p oten tially bias the estimates if in ter state migration decisions w ere based on a systematic relationship b et w een a w oman’s fertilit y preference and a v ailabilit y of legal ab ortion in the migration destination. The state of residency at observ ation is then used to construct an indicator for p oten tial non-mo v ers 7 and do the estimation using this subsample as a robustness c hec k. The second c hallenge arise out of the cross-sectional nature of the data used. As noted in the equation sp ecication, iden tication of the causal eect requires observ ation of fertilit y measures at an age where the fertilit y outcome has b een realized. Birth cohorts used in the estimation span o v er 25 y ears, meaning that their fertilit y outcomes w ere realized at dieren t p oin ts in time. Hence, observ ation of w omen 7 62% of w omen in the sample reside at observ ation in the same state as their birth state. 14 of dieren t birth cohorts at dieren t sample y ears is required in order to obtain a prop er measure of fertilit y outcomes. I consider three fertilit y outcomes. The rst is completed fertilit y whic h is dened as the total n um b er of c hildren a w oman giv es birth to during her life- time. The other t w o fertilit y outcomes are the age of start of motherho o d and spacing b et w een the rst t w o c hildren. V ariables construction and further sample restrictions for estimating the eect of ab ortion access on fertilit y are discussed in detail in app endix A.1.1. Summary statistics of fertilit y outcomes are rep orted in Mo dule A of T able 2.1. On a v erage w omen in rep eal states ga v e birth to a smaller n um b er of c hildren compared to w omen in non-rep eal states. The a v erage n um b er of c hildren p er w oman are 2:46 in non-rep eal states and 2:21 in rep eal states. Age of start of motherho o d sho ws as w ell a signican t dierence b et w een rep eal and non-rep eal states. W omen in rep eal states are on a v erage one y ear older than w omen in non- rep eal states when they giv e birth to their rst c hild. Lo oking at these outcomes b y race rev eals heterogeneit y b et w een blac k and white w omen. On a v erage, blac k w omen ga v e birth to more c hildren and started ha ving c hildren at a y ounger age than white w omen, a dierence that is p ersisten t across rep eal and non-rep eal states. Bet w een state dierences sho w that in terms of c hildren b orn the largest dierence is among blac k w omen, for whom non-rep eal states a v erage of c hildren b orn w as larger than the rep eal state a v erage b y 0:68. The corresp onding dierence b et w een states in the n um b er of c hildren for white w omen is m uc h smaller ( 0:2 c hildren). The a v erage age at the birth of the rst c hild sho ws a signican t dela y of 0:85 y ears (10 mon ths) for white w omen in rep eal states compared to white w omen in non-rep eal states, while for blac k w omen the dierence is smaller and is around 7 mon ths. The a v erages of birth spacing sho w that for w omen who ha v e t w o c hildren the a v erage spacing b et w een c hildren is around 0:24 y ears (3 mon ths) larger in non-rep eal states compared to rep eal states. The a v erage spacing for white w omen is around 3:5 y ears whereas for blac k w omen it is around 4:6 y ears. 15 1.5 Eect of Abortion Access on F ertility This section presen ts estimation results of the eect of access to ab ortion on fertilit y outcomes. Equation (1.2) is estimated for the full sample and separately for blac k and white w omen for all fertilit y outcomes of in terest. Estimates for completed fertilit y are rep orted in T able 1.3, while estimates for the birth timing outcomes are rep orted in T able 1.5. The dierence-in-dierences estimates for the eect of ab ortion access on completed fertilit y and timing of birth are rep orted in T ables 1.4 and 1.6 resp ectiv ely . All estimation results are rep orted graphically in Figures 1.2 to 1.7. Estimates of equation (1.1) supplemen t the main analysis and the results are rep orted in App endix A.2.1. Impact of ab ortion access on completed fertilit y is discussed rst, follo w ed b y the eect on birth timing outcomes. 1.5.1 Completed F ertility Estimated co ecien ts of equation (1.2), rep orted in T able 1.3 and Figure 1.2a, sho w that w omen in rep eal states had lo w er completed fertilit y rates compared to their birth cohort p eers in non-rep eal states. Ho w ev er, the dierence b et w een rep eal and non-rep eal states is smaller for the y ounger birth cohorts. As a matter of fact, the dierence in completed fertilit y b et w een states for the y oungest birth cohort is only signican t at the 10% lev el. Giv en that the y ounger cohorts had earlier access to ab ortion, the decreasing trend in these dierences suggests that cohorts that had earlier access to ab ortion had an increase in completed fertilit y . More formally , the dierence-in-dierences estimates of the eect of ab ortion access on completed fertilit y (T able 1.4 and Figure 1.2b) sho w that early ab ortion access increases completed fertilit y . W omen who obtained access to ab ortion b et w een the age of 15 and 23 had on a v erage 0:06 more c hildren than w omen who did not ha v e access to ab ortion during their fertilit y cycle. There is no signican t eect of ab ortion on completed fertilit y for w omen who obtained access at a later age. The small, but statistically signican t eect, masks imp ortan t heterogeneit y . When lo oking at the eect b y race, it can b e clearly seen that the result for the full sample is driv en b y the eect that ab ortion access had on completed fertilit y of blac k w omen. White w omen had an a v erage completed fertilit y of 2:35 p er w oman, 16 with no signican t dierence b et w een rep eal and non-rep eal states (Figure 1.5). The dierence in completed fertilit y of blac k w omen sho ws that among earlier birth cohorts, blac k w omen b orn in rep eal states had signican tly lo w er n um b er of c hildren during their lifetime, whereas for the later birth cohorts con v ergence in completed fertilit y in rep eal and non-rep eal states is do cumen ted (T able 1.3 and Figure 1.5). The results sho w that access to ab ortion b efore the age of 20 increases fertilit y of blac k w omen b y an a v erage of half a c hild (Figure 1.5c). The eect diminishes as access is dela y ed, but a signican tly p ositiv e eect of ab ortion access on completed fertilit y p ersists un til the age of 28 (see T able A.2). The n ull eect of ab ortion on completed fertilit y of white w omen is not surprising. It has b een h yp othesized that, ev en if ab ortion w ere una v ailable, w omen w ould still ha v e b een able to con trol the total n um b er of c hildren they giv e birth to during their lifetime. In resp onse to an unplanned pregnancy that leads to an unplanned birth, a w oman can alw a ys readjust her later fertilit y c hoices and still ac hiev e her in tended family size. Ho w ev er, the eect that ab ortion access has on completed fertilit y of blac k w omen is puzzling. P oten tial explanations for these results are explored later in the pap er. 1.5.2 Birth Timing Age at Birth of First Child As p oin ted out in the discussion ab o v e, w omen migh t b e able to con trol com- pleted fertilit y in the absence of ab ortion. Nev ertheless, since ab ortion is the only v olun tary metho d to a v oid an un timely birth once an un timely pregnancy has tak en place, ab ortion accessibilit y can ha v e a signican t eect on birth timing. The rst birth timing outcome studied is the age at start of motherho o d. The eect of ab ortion access on this fertilit y realization has dra wn particular atten- tion in the literature, the fo cus b eing on the probabilit y of teenage motherho o d. Estimation results for age at start of motherho o d are rep orted in T able 1.5 and Figure 1.3a. The a v erage w oman in the sample ga v e birth to her rst c hild at age 24. W omen in non-rep eal states attain motherho o d at a y ounger age than their birth cohort p eers in rep eal states. F or birth cohorts that w ere older than 21 at the time of state legal c hanges, the dierence in age at start of motherho o d is half a 17 y ear. Ho w ev er, for w omen y ounger than that the dierence across states increases to almost a full y ear. Dierence-in-dierences estimates (see Figure 1.3b) sho w that obtaining access to legal ab ortion b efore the age of 21 dela ys the start of motherho o d b y 6 mon ths, and the eect fades out when access is dela y ed b ey ond that age. The eect on white w omen is consisten t with the one found using the full sample (Figure 1.6c). Estimation results for blac k w omen (Figure 1.6b) sho w that the dela y in age at rst c hild is restricted to those who obtained access to ab ortion b y the age of 18. The results for blac k w omen also sho w a statistically signican t negativ e estimates for age groups 27 to 29 and 30 to 32. This result is surprising, as it suggests that access to ab ortion b et w een the ages of 27 and 32 led to an earlier start of motherho o d among blac k w omen, whereas I w ould ha v e an ticipated an insignican t eect. I b eliev e this result is due to a measuremen t error in the outcome v ariable. As explained in App endix A.1.1 age at start of motherho o d is constructed b y taking the dierence b et w een the age of the mother at observ ation and the age of the eldest c hild in the household. Sample restrictions w ere set to minimize the p ossibilit y of measuremen t errors in v ariable construction. If w omen in the sample ga v e birth early in their life, then their eldest c hild is more lik ely to ha v e left the household b y the time of observ ation. Giv en that blac k w omen w ere on a v erage y ounger when they ga v e birth to their rst c hild (T able 2.1), they are more susceptible to this t yp e of measuremen t error. In other w ords, the measuremen t of age at rst c hild for the oldest age group (33-35), whic h is used as the reference group, ma y b e o v erstated. This measuremen t error led to do wn w ard bias in the estimated eect of ab ortion access on age of start of motherho o d. Therefore, the rep orted estimates should b e though t of as lo w er b ound of the eect of ab ortion access on age of start of motherho o d for blac k w omen. Angrist and Ev ans (2000) do cumen t a 7 p ercen tage p oin t decrease in proba- bilit y of motherho o d b efore the age of 20 for blac k w omen who obtain access to ab ortion while they are teenagers. The results in this pap er reconcile with their ndings for blac k w omen conditional on teen ab ortion users dela ying the birth of the rst c hild b y 7 y ears. This means that blac k w omen who used ab ortion to prev en t an early start of motherho o d during their teenage y ears later started motherho o d at around the age of 24 (a v erage age at start of motherho o d in the 18 sample). While I nd a p ositiv e eect of ab ortion access on dela ying motherho o d for white w omen, they nd a small and insignican t eect on the probabilit y of ha ving c hildren b efore the age of 20. In m y opinion there are t w o reasons wh y the eect found in this pap er is larger. First, the dierence in treatmen t group denition. This pap er restricts the treatmen t group to states that rep ealed ab ortion bans and made ab ortion a v ailable to w omen at request while their treatmen t group includes all states that had reforms allo wing w omen partial access to ab ortion. As men tioned in section 2, ab ortion use diered considerably b et w een these t w o groups of states, with higher usage rates in states that had a complete rep eal of ab ortion bans. A second reason is that the a v erage eect found in this pap er migh t b e capturing some spillo v er eects that are not relev an t for the outcome considered in their study . A ccess to ab ortion migh t aect the age of start of motherho o d of some w omen ev en if they did not particularly use ab ortion to end an un w an ted pregnancy . P eer eects streaming from ab ortion users can aect the preference of other w omen ab out fertilit y c hoices (K ohler et al. , 2001). While this kind of p eer eect w ould dela y the a v erage age at whic h w omen in the p opulation giv e birth to their rst c hild, they w ould not necessarily aect the rate of teenage pregnancies. Birth Spacing The second birth timing outcome in v estigated is the birth spacing for w omen who had t w o c hildren. The age in the sample used for this estimation ranges b et w een 29 and 40, with the ma jorit y of the observ ation b eing in their mid 30s. Before analyzing the estimation results, it should b e noted that the ndings from this particular subsample do not pro vide a full understanding of birth spacing c hoices and ho w they are aected b y ab ortion access. On one hand, w omen in this sample could p oten tially still ha v e more c hildren and hence the v ariable considered do es not constitute a comprehensiv e measure of birth spacing. On the other hand, these are w omen who had only t w o c hildren b y the mid-p oin t of their fertilit y cycle. While this is the norm in curren t da ys, for the birth cohorts in question this sample migh t b e highly selectiv e in terms of family size and career preferences. The purp ose of estimating this outcome is to complemen t the ndings on start of motherho o d and to see whether dela ying start of motherho o d had a subsequen t 19 eect on birth of the next c hild. I mostly nd that ab ortion access starting at an y age has no eect on spacing b et w een the rst t w o c hildren (T able 1.6 and Figure 1.4). Consequen tly , regardless of the age at whic h a w oman giv es birth to her rst c hild, this w ould not aect the n um b er of y ears a w omen will w ait to giv e birth to her second c hild. This result is not surprising giv en that the literature on the eect of fertilit y on w omen’s earnings nds that the w age p enalt y is incurred as so on as the rst c hild is b orn (Lundb erg and Rose (2000); Klev en et al. (F orthcoming)). This implies that once the rst c hild is b orn and the w age p enalt y is suered, the opp ortunit y cost of a second c hild is no w smaller, whic h could mean that once motherho o d is realized there is no longer an y b enet to dela ying subsequen t births. It should b e noted that for blac k w omen the results sho w a jump in spacing for t w o age groups (Figure 1.7b); blac k w omen who receiv ed access to ab ortion b et w een the ages of 21 and 23 and b et w een the ages of 27 and 29 had a statistically signican t increase in spacing of 5 mon ths b et w een the rst and second c hild. One p ossible explanation for this observ ation is that while these birth cohorts did not ha v e access to ab ortion in order to prev en t an unplanned start of motherho o d, they still b eneted from it to dela y the birth of the second c hild in an eort to mitigate the adv erse consequences of the rst unplanned birth. 1.5.3 Black W omen Marriage Outcomes As sho wn in the previous section, early access to ab ortion prev en t an unplanned start of motherho o d at a y oung age for b oth blac k and white w omen. Man y teenage or y oung mothers end up b eing single mothers, whic h could p oten tially aect man y subsequen t outcomes, for instance the probabilit y and the qualit y of marriage. An unplanned teenage pregnancy leading to a birth can p oten tially decrease the c hance of remarriage of the mother. It can as w ell lead to marriage with a partner of lo w er sc ho oling and income. Hence, escaping teenage motherho o d for w omen can p oten tially increase the probabilit y of forming stable families in the future as w ell as higher household income, whic h could p oten tially increase the demand for c hildren as a result of the p ositiv e income eect. In this section, I pro vide suggestiv e evidence supp orting marriage as a p oten tial c hannel leading to the observ ed increase in completed fertilit y of blac k w omen. I 20 do so b y in v estigating the eect of ab ortion access on marriage outcomes of blac k w omen. Using the sample of blac k w omen used in the birth timing estimation, I estimate equation (1.2) for a v ariet y of marriage outcomes. The dierence-in-dierences estimates rep orted in T able 1.8 sho w no signi- can t eect of ab ortion access on probabilit y of marriage. T w o earnings outcomes are in v estigated, h usband y early earnings and o ccupation status. While ab ortion access do es not seem to ha v e a signican t eect on o ccupation status, h usbands of blac k w omen who receiv ed early access to ab ortion had signican tly larger lab or earnings. I also in v estigate the eect of ab ortion access on h usbands sc ho oling. College completion rates for h usbands of blac k w omen who obtained early access to ab ortion are signican tly higher. Blac k w omen who receiv ed access to ab ortion b efore the age of 26 are married to men who are 8 p ercen t more lik ely to ha v e a college degree. Giv en the sample mean of college degree completion in the sample this a v ery large eect. Results of blac k w omen’s completed fertilit y and marriage outcomes sho w that the cohorts that exp erienced an increase in completed fertil- it y are the cohorts that ha v e h usbands with higher college completion rates and earnings. These ndings giv e supp ort to the prop osition stated earlier ab out the increase in completed fertilit y b eing the result of an increase in household income brough t ab out b y a higher earning h usband. 1.5.4 Robustness Checks In all the estimations rep orted ab o v e, a w oman’s state of birth is used to determine her exp osure to ab ortion. The p ossibilit y of b et w een state migration implies that the treatmen t v ariable (in teraction of rep eal with age at rep eal) is measured with error. This measuremen t error could bias the estimates of the eect of ab ortion access on fertilit y outcomes. What could b e concerning in particular is the p ossibilit y that migration decisions are correlated with the state legislativ e c hanges. Suc h correlation could arise as a result of selectiv e migration of w omen due to state decisions to rep eal or uphold ab ortion bans. It migh t as w ell b e due to migration decisions due to other factors that could correlate with the legal status of ab ortion in the state. F or instance if rep eal state colleges w ere more app ealing for females, there w ould a systematic migration from rep eal to non- 21 rep eal states for college age w omen. Information on state of residency at v arious p oin ts in time a v ailable in the IPUMS is used to c hec k the sensitivit y of the results found ab o v e. The details of these robustness c hec ks are pro vided in app endix A.3 and discussed briey in what follo ws. Ho w ev er, it should rst b e noted that a systematic measuremen t error due to selectiv e migration from non-rep eal to rep eal states w ould most lik ely cause a do wn w ard bias. Giv en the magnitude of the estimates found, I b eliev e it is highly unlik ely that suc h measuremen t error exist. As discussed in the data section the IPUMS rep orts b oth the state of birth and the state of residency at observ ation. While these observ ation do not pro vide full information on the complete migration history of observ ed individuals, they inform us ab out whic h w omen did indeed migrate at some p oin t during their life. Then equation (1.2) is estimated using the subsample of p oten tial non-mo v ers, whic h exclude all w omen that are kno wn to ha v e migrated. Estimation results rep orted in app endix A.3.1 are consisten t with the results found in the main estimation. T o further c hec k that there w as no systematic migration among w omen of v arious age b et w een rep eal and non-rep eal c hanges at the time of the legal c hanges, I tak e adv an tage of state of residency information a v ailable in the 1970 sample of the census, whic h rep ort b oth the state of residency at observ ation and 5 y ears earlier (1965). App endix A.3.2 in v estigates w omen’s migration decisions using men as a comparison group. Migration decisions of men w ere not p oten tially aected b y the state ab ortion legal status or an y other factor that could b e app ealing to w omen. I nd that there is no dierence b et w een men and w omen of all birth cohorts in the prop ensit y to emigrate from rep eal to non-rep eal states. As for the migration from non-rep eal to rep eal, men who w ere b et w een the ages of 19 and 24 at the time of legal c hanges w ere signican tly more lik ely to migrate compared to w omen, while no signican t dierence in the prop ensit y to migrate for the other birth cohorts. This section establishes a signican t eect of early access to ab ortion on de- la ying age of start of motherho o d. The results suggest that ab ortion access had no signican t eect on other fertilit y outcomes, with the exception of completed fertilit y of blac k w omen. Impro v emen ts in marriage qualit y , h usbands with higher earnings and education as a result of early exp osure to ab ortion seems to b e the most lik ely explanation for the increased completed fertilit y of blac k w omen. This 22 phenomenon is most lik ely the b ypro duct of a v oided unplanned start of mother- ho o d. In other w ords, the ndings in this section establish that the direct eect of ab ortion access on w omen’s fertilit y is through the a v oidance of early fertilit y sho c k. In the next section, I use the v ariation in ab ortion exp osure as an instru- men t to estimate the eect of age of start of motherho o d on w omen’s earnings. One migh t w orry that the observ ed eect on age of start of motherho o d is due to a fertilit y trend c hange among y ounger birth cohorts that coincided with the legal c hanges. This p ossibilit y can b e ruled out b y observing the p ost-trend for age of start of motherho o d in Figure 1.1b. The 1960-1969 birth cohorts had similar access to legal ab ortion across states, since w omen in these birth cohorts w ere to o y oung at the time of state legalization and w ere therefore all exp osed to federal ab ortion legalization. While the trends are not fully parallel, the sharp dierence in the trend of age of motherho o d observ ed for the early exp osure group (1950- 1959) is no longer presen t for this group. Moreo v er, I extend the sp ecication of equation (1.2) to include w omen b orn b et w een 1956 and 1958. The added cohorts w ere b et w een the ages of 12 and 14 at the time of the state legal c hanges and hence are unlik ely to b e aected b y the treatmen t. These birth cohorts exp erienced equal access to ab ortion across states, as they all obtained access to ab ortion at the start of their fertilit y cycle through federal legalization. Estimation results are rep orted in App endix A.2.2 for the full sample and b y race (Figure A.4). The results sho w that access to ab ortion b et w een the ages of 12 and 14 had no eect on age of start of motherho o d. This result pro vides further evidence that the eect found ab o v e is indeed due to ab ortion access and not due to a trend c hange in fertilit y preferences. 1.6 Consequences of Unplanned Start of Mother- hood on W omen’s Career The baseline reduced form relationship b et w een fertilit y realization y F i(s) and a lab or mark et outcome y L i(s) of a w oman i living in state s can b e written as follo ws y L i(s) = 0 + 1 y F i(s) +X + s + i(s) (1.3) 23 WhereX is a v ector of observ able and s is a state of residency xed eect. Omit- ted v ariable bias as w ell as sim ultaneit y of fertilit y and lab or supply c hoices imply that OLS estimates are unlik ely to reco v er a consisten t estimates of 1 . Since random assignmen t of fertilit y ev en ts is unfeasible, the b est course of action to iden tify the eect of fertilit y realizations on lab or mark et outcome is to exploit a quasi natural exp erimen t that leads to random v ariation in fertilit y realizations. The discussion ab o v e sho ws that the early rep eal of ab ortion is a plausible natural exp erimen t to estimate the desired eect. Angrist and Ev ans (2000) use state v ari- ation in access to ab ortion to estimate the eect of teenage pregnancy on sc ho oling and lab or mark et outcomes. Similarly , I used v ariation in access to legal ab ortion as an instrumen t to measure the eect of age of start of motherho o d on a v ariet y of lab or mark et outcomes. Estimation of equation (1.2) for age of motherho o d is used as a rst stage of the 2SLS estimation strategy . The tted v alues for the fertilit y outcome are then used to estimate the second stage equation. V alidit y of the instrumen tal v ariable approac h is conditional on the instrumen t satisfying the exclusion restriction, meaning that the eect of ab ortion access on lab or mark et outcomes should b e exclusiv ely through the eect of access on the realizations of the early fertilit y sho c ks. This approac h pro vides a consisten t estimate of the a v erage eect of fertilit y real- izationy F on lab or mark et outcomes y L . More imp ortan tly for the purp ose of this pap er, as noted b y Angrist et al. (1996), these a v erage eects are for the subsample of w omen whose fertilit y realization w as altered b y ab ortion legalization (LA TE). In other w ords, the estimate 2SLS estimate ^ 1 for age of start of motherho o d can b e in terpreted as the a v erage eect of dela ying an unplanned start of motherho o d for w omen who c hose to use ab ortion. 1.6.1 Data on Labor Market Outcomes The main v ariable of in terest is y early lab or earnings whic h is readily a v ailable in ev ery sample of the IPUMS. The cross sectional nature of the data has b een one of the main c hallenges in this pap er. Prop er estimation of the eect in question relies on appropriate c hoice of sample. The sample used in this estimation is the same as the one used in estimating the eect of ab ortion access on birth 24 timing outcomes. Sev eral reasons justify this sample selection. First, the fo cus is on estimating the eect of age at start of motherho o d on lab or earnings using exp osure to ab ortion as an instrumen tal v ariable, and as argued earlier this is the prop er sample to estimate the rst stage. Second, w ages v ary with age of individuals, and to prop erly iden tify the eect of an earlier fertilit y sho c k on a w oman’s earnings, w e should compare earnings outcomes for w omen of similar age. The age range of w omen in the selected sample v aries b et w een 29 and 41, with most of them concen trated around the age of 35. The earning data rep orted are nominal earnings measured in three samples that are 10 y ears apart, I hence deate the earnings and express them in 2012 dollars to mak e them comparable. The main limitation of estimating the eect of fertilit y sho c ks on earnings using cross sectional data is the inabilit y to capture the dynamics. What is observ ed in the sample is a one time snapshot of a w oman’s lab or earnings when she is around the age of 35. If fertilit y sho c ks ha v e an y eect on earnings, w e w ould w an t to understand the mec hanisms that lead to this eect. One particular mec hanism this pap er aims to explore is the lab or mark et exp erience. While the data set at hand do es not allo w in v estigation of this mec hanism directly , I pro vide suggestiv e evidence on p oten tial mec hanisms using an o ccupation index measure rep orted in the IPUMS and observ ations of lab or force participation and w eekly hours w ork ed. Details on sample denition and lab or outcome v ariables construction are re- p orted in app endix A.1.2. Summary statistics of lab or mark et outcomes are re- p orted in Mo dule B of T able 2.1. Blac k w omen b orn in non-rep eal states ha v e higher lab or participation than blac k w omen b orn in rep eal states. White w omen’s lab or force participation is the same in b oth groups of states and is lo w er than the lab or force participation of blac k w omen. A v erages for hours w ork ed are com- puted for the full sample and for the subsample of w orking w omen. In the full sample, w omen b orn in rep eal states w ork 22:51 hours a w eek while w omen b orn in non-rep eal states w ork 23:17. Blac k w omen ha v e the largest b et w een state dif- ference, with blac k w omen in non-rep eal states w orking on a v erage 1.35 more hours p er w eek than blac k w omen in rep eal states. In the w orking w omen subsample, w omen b orn in non-rep eal states w ork around one more hour p er w eek compared to w omen in rep eal states. There is no dierence b et w een blac k w omen who w ork on a v erage 38:2 hours a w eek. White w omen in rep eal states w ork the least n um- 25 b er of hours p er w eek at an a v erage of 35:57, compared to an a v erage of 36:38 for white w omen b orn in non-rep eal states. Put together, these statistics sho w that there is no large dierence in lab or supply decisions b et w een rep eal and non-rep eal states dierence, with the exception of the higher lab or force participation of blac k w omen in non-rep eal states. Lab or earnings are signican tly higher in rep eal states. On a v erage, a w oman b orn in a rep eal state earns $18; 502 compared to $16; 010 a v erage y early earnings for a w oman b orn in a non-rep eal state. Lo oking at the dierence p er earnings b y race, I note that the dierence in earnings is larger among blac k w omen. The dif- ference in earnings among rep eal and non-rep eal states is $3; 270 for blac k w omen and $2; 480 for white w omen. Hourly w age rates are similarly higher in rep eal states, not surprisingly giv en that there are no signican t dierences in lab or hours w ork ed of w orking w omen and they are constructed using y early earnings and a v- erage w eekly hours w ork ed of w orking w omen. The o ccupation index v ariable is a t w o digit v ariable taking v alues b et w een 0 and 96, with higher v alues reecting o ccupations with higher w ages and b etter education. The rep orted a v erages sho w that w omen in rep eal states ha v e signican tly b etter o ccupations than w omen in non-rep eal states. The dierence in o ccupation is largest among blac k w omen. 1.6.2 Estimation Results In this section, I presen t and discuss estimation results of equation (1.3). OLS and 2SLS estimates of 1 for lab or earnings and lab or supply c hoices are rep orted in T ables 1.9 and 1.10. Y early Lab or Earnings Unsurprisingly , estimation results sho w a p ositiv e signican t asso ciation b e- t w een age of start of motherho o d and y early lab or earnings. P ostp oning the birth of the rst c hild b y one y ear is asso ciated with $282 more in y early earnings for a w oman in her mid 30s. F or the subsample of blac k w omen, the asso ciation b et w een earnings and age of motherho o d is t wice the size found for white w omen. The OLS results are not v ery informativ e for the purp ose of answ ering the question p osed in this pap er. In man y cases, w omen c ho ose the age at whic h they en ter motherho o d 26 based on the p oten tial eect of this ev en t on their career. It is lik ely that w omen with higher p oten tial earning abilit y self select in to later start of motherho o d. It is also lik ely that b oth fertilit y c hoices and lab or mark et earnings are aected b y un-observ able c haracteristics, suc h as views on gender roles. More imp ortan tly , in this pap er I am in terested in determining the eect of fertilit y sho c k due to the unplanned start of motherho o d on earnings rather than the eect of the age of start of motherho o d itself. The estimates from the instrumen tal v ariable approac h are therefore more relev an t for this purp ose, as these estimates will iden tify the eect of an a v erted unplanned start of motherho o d on a w oman’s earnings. The results sho w substan tial and statistically signican t gains in earnings from dela ying an unplanned start of motherho o d. The 2SLS estimate for the full sample sho ws that a one y ear dela y in unplanned start of motherho o d increases y early earnings b y $2; 194, whic h is a 13 p ercen t increase from the mean. The eect is signican t and of the same order of magnitude for b oth blac k and white w omen. The larger magnitude of the 2SLS estimates in comparison to the OLS estimates is not surprising, giv en that the v ariation in age of start of motherho o d in the IV estimation is arising from w omen who had access to ab ortion and c hose to use it. Among w omen with equal access, ab ortion usage is not random. W omen who c hose to ab ort an unplanned rst c hild are most lik ely w omen who p oten tially w ould ha v e suered a high w age p enalt y from suc h an early start of motherho o d. The gains in earnings could b e the result of higher lab or supply or higher pa ying o ccupation or a com bination of b oth. As a rst step in in v estigating the source of these gains, I lo ok at the eect of age of start of motherho o d on hourly w age rates. The rst ro w in table 1.10 rep orts regression estimates for the eect of age at start of motherho o d on hourly w age rates. A one y ear dela y in the start of motherho o d increase the w age rate b y a v erage of $2. There is a sligh t heterogeneit y in the eect b y race, with the gains for blac k w omen b eing larger than those for white w omen. White w omen’s w age rates increase b y 13% from the mean, while for blac k w omen the increase in w age rates is 16%. These n um b ers sho w that the increase in w age rate accoun ts for all the gains in lab or earnings for white w omen. F or blac k w omen the p ercen tage increase in w age rate is higher than the p ercen tage increase in y early lab or earning, whic h migh t b e due to income eect on lab or supply decision. 27 The w age rate v ariable is constructed using y early lab or earnings v ariable and rep orted w eekly lab or hours. P oten tial measuremen t errors in rep orted lab or hours could mean that the rep orted estimates suer from division bias. Therefore, I directly in v estigate the eect of age at start of motherho o d on lab or supply and o ccupation status. Occupation v ersus Lab or Supply Estimation results rep orted in the second ro w of T able 1.10 sho w a signican t increase in o ccupation status as a result of dela ying the unplanned start of moth- erho o d. Similar to the results found on earnings, the IV estimates are larger than the OLS estimates, whic h as stated earlier is most lik ely due to w omen c ho osing to dela y birth b y means of ab ortion are those with b etter career prosp ects. The eects are particularly large for blac k w omen, for whom dela ying the age of start of motherho o d b y one y ear put them at equal o ccupation lev el as the a v erage w orking white w oman. Equation (1.3) is estimated for three lab or supply outcomes. Estimation results of the eect of age at start of motherho o d on lab or force participation and w eekly lab or hours are rep orted in T able 1.9. While the estimate of the eect on w eekly lab or hours conditional on emplo ymen t are rep orted in the third ro w of T able 1.10. The results suggest that the eect of early fertilit y sho c ks on later lifetime lab or supply are minimal if an y . The most signican t eect is on lab or force participation of white w omen, where I nd a signican t 2:8% increase as a result of dela ying start of motherho o d b y 1 y ear. Dela ying fertilit y has an opp osite eect on blac k w omen lab or force participation, where the p oin t estimate sho w a 3:1% decrease in lab or force participation as a result of a one y ear dela y in motherho o d, ho w ev er the estimate is only signican t at 10 p ercen t signicance lev el. Similarly , the eect on lab or hours w ork ed seems to b e small. F or the sample of w orking w omen dela ying age of start of motherho o d increase hours supplied of emplo y ed blac k w omen b y sligh tly more than half an hour p er w eek, while white w omen decrease w eekly hours w ork ed b y sligh tly less than an hour p er w eek. F or the full sample of w omen, including those not w orking, dela ying age of start of motherho o d has no signican t eect on w eekly hours w ork ed. The mild eect found on lab or supply at b oth extensiv e and in tensiv e margins, 28 in addition to the p ositiv e and signican t causal eect of age of start of mother- ho o d on o ccupation status, lead me to b eliev e that the do cumen ted earning gains are mostly due to impro v emen t in o ccupation status. While the cross sectional nature of the IPUMS do es not allo w observ ation of the lifetime earning prole, the ndings ab o v e suggest strong dynamic eects of an early fertilit y sho c k on lifetime earnings. By prev en ting an unplanned start of motherho o d, w omen attain signif- ican tly higher earnings when they reac h their mid thirties. The fact that these gains are due to obtaining b etter o ccupations that pa y higher w ages indicate that if they w eren’t a v oided fertilit y sho c ks w ould ha v e lead to p ermanen t decline in lifetime earnings. 1.7 Conclusion In this pap er I asses the eect of fertilit y sho c k on earning b y exploiting a dierence in timing of p olicy c hange at the state and federal lev el. The early rep eal of ab ortion ban in 5 states led to v ariation in ab ortion access at dieren t y ears of age. This v ariation resulted in dierence in fertilit y sho c ks realizations. Sp ecically , w omen who receiv ed early access to ab ortion w ere less lik ely to start their motherho o d early . I then exploit this random v ariation in fertilit y sho c ks to iden tify the eect of an unplanned start of motherho o d on earnings. I nd statistically and economically signican t p ositiv e eect of dela ying motherho o d on w omen earnings. The evidence in this pap er suggest that most of the increase in earnings is due to higher w age rate resulting from a b etter o ccupation. The ndings in this pap er pro vide a b etter understanding of the p oten tial con- sequences of ab ortion legalization in the United States on fertilit y and careers of w omen. While eect of ab ortion access on total fertilit y seems to b e limited if an y , ab ortion has a considerable eect on fertilit y timing, more sp ecically on the age at whic h w omen en ter motherho o d. These eect are most signican tly early in the fertilit y cycle of w omen. Although this pap er do es not study the eect of state restrictions targeting ab ortion users and pro viders, it can sp eak to some of their consequences on w omen fertilit y . While to da y no ban is imp osed on access to ab ortion, man y p olicies are enacted to limit it’s accessibilit y . Moreo v er, these p olicies can p enalize w omen unev enly b y making the cost of access higher for some 29 of them compared to other. Y oung w omen, who as sho wn in this pap er are the largest b eneciaries in terms of con trolling their fertilit y are also the most vulner- able to state restrictions as they tend to ha v e less resources to aord more costly ab ortions. As is also sho wn in this pap er, ab ortion access is not just ab out con trol o v er fertilit y cycle. V ariations in fertilit y realization due to ab ortion utilization has signican t eects on w omen career. There are sustainable gains in earnings do cu- men ted as a result of dela ying motherho o d, esp ecially if en trance to motherho o d w as due to an unplanned pregnancy . 30 T ables T able 1.1: Summary Statistics of F ertilit y and Lab or Mark et Outcomes This table presen ts descriptiv e statistics for the main outcome v ariables studied. Data are from the 5% Public Use Micro data Samples (IPUMS), construction of v ariables and sample restrictions are describ ed in details in app endix A.1. The a v- erages are rep orted separately for rep eal and non-rep eal states, for the full sample and b y race. F ull Sample White W omen Blac k W omen Mo dule A: F ertilit y Outcomes Complete d F ertility Rep eal 2.21 2.18 2.23 Non Rep eal 2.46 2.38 2.91 A ge at Birth of First Child Rep eal 24.80 24.90 22.83 Non Rep eal 23.83 24.05 22.21 Birth Sp acing Rep eal 3.51 3.45 4.53 Non Rep eal 3.75 3.66 4.65 Mo dule B: Lab or Mark et Outcomes L ab or F or c e Particip ation Rep eal 0.676 0.671 0.699 Non Rep eal 0.680 0.671 0.739 W e ekly L ab or Hours Rep eal 22.51 22.29 23.66 Non Rep eal 23.17 22.89 25.01 W e ekly L ab or Hours if W orking Rep eal 35.82 35.57 38.29 Non Rep eal 36.65 36.38 38.26 Y e arly W age Earnings Rep eal 18,502 18,215 20,575 Non Rep eal 16,010 15,835 17,305 Hourly W age R ate Rep eal 15.45 15.34 16.54 Non Rep eal 13.12 13.10 13.35 Oc cup ation Index Rep eal 43.04 43.23 40.67 Non Rep eal 39.74 40.47 35.27 31 T able 1.2: Illustration of the Iden tication This table presen ts a comparison of fertilit y outcomes for w omen with dieren t lev els of exp osure to ab ortion. Data is extracted from the IPUMS (1970, 1980 and 1990). The no exp osure group consists of w omen b orn b et w een 1930 and 1935, the mid fertilit y cycle exp osure group consists of w omen b orn b et w een 1940 and 1945 and the early exp osure group consists of w omen b orn b et w een 1950 and 1955. Birth cohorts 1950-1955 observ ed in 1990 Birth cohorts 1940-1945 observ ed in 1980 Birth cohorts 1930-1935 observ ed in 1970 Mo dule A: Age at First Child Rep eal State (1) 25.72 24.41 24.46 (4.85) (3.94) (3.75) Non-Rep eal State (2) 24.60 23.75 23.69 (4.69) (3.85) (3.66) Dierence (1)-(2) 1.12 0.66 0.77 [0.00] [0.00] [0.00] Dierence-in-Dierences 0.46 -0.11 [0.00] [0.38] Mo dule B: T otal F ertilit y Rep eal State (3) 1.75 2.18 2.81 (1.36) (1.50) (1.90) Non-Rep eal State (4) 1.91 2.41 3.08 (1.36) (1.61) (2.08) Dierence (3)-(4) -0.16 -0.23 -0.26 [0.00] [0.00] [0.00] Dierence-in-Dierences 0.07 0.03 [0.00] [0.78] Note: Standard deviations are rep orted in paren thesis under the sample means and p-v alue for the test of signicance of the dierences are rep orted in brac k ets under. 32 T able 1.3: Cross States Dierences in Completed F ertilit y (Age Group) This table rep orts estimation results of equation (1.2) for the completed fertilit y outcome v ariable, for the full sample and b y race. Completed fertilit y is dened as the total n um b er of c hildren a w oman giv e birth to during her lifetime. It is measured as the total n um b er of c hildren w omen rep orted to ha v e giv en births to in the 1990 sample. Mean 1 2 3 4 5 6 7 8 9 [15-17] [18-20] [21-23] [24-26] [27-29] [30-32] [33-35] [36-38] [39-40] F ull Sample Complete d F ertility 2.42 -0.05 -0.05 -0.06 -0.07 -0.07 -0.09 -0.08 -0.12 -0.11 (0.03) (0.02) (0.02) (0.02) (0.02) (0.02) (0.02) (0.03) (0.03) N=1,761,209 Blac k W omen Complete d F ertility 2.86 -0.10 -0.05 -0.17 -0.18 -0.31 -0.43 -0.65 -0.64 -0.57 (0.03) (0.03) (0.06) (0.02) (0.10) (0.09) (0.04) (0.12) (0.07) N=180,627 White W omen Complete d F ertility 2.35 0.02 0.01 0.01 -0.01 0.01 0.01 0.03 -0.02 -0.03 (0.02) (0.01) (0.02) (0.01) (0.02) (0.02) (0.01) (0.02) (0.02) N=1,535,533 Note: All regressions ha v e b een w eigh ted b y p opulation w eigh ts. Clustered robust standard errors are rep orted in paren thesis. * p< 0:1; ** p< 0:05; *** p< 0:01. 33 T able 1.4: Dierence-in-Dierences Estimates of the Eect of Ab ortion A ccess on Completed F ertilit y This table rep orts the dierence-in-dierences estimates for the eect of ab ortion access on completed fertilit y . The columns rep ort the eect of ab ortion access starting a certain age phase compared to receiving access to ab ortion at the end of the fertilit y cycle. These co ecien ts are computed using estimates from T able 1.3. Mean 1 2 3 4 5 6 7 8 [15-17] [18-20] [21-23] [24-26] [27-29] [30-32] [33-35] [36-38] F ull Sample Complete d F ertility 2.42 0.06 0.06 0.05 0.05 0.04 0.03 0.03 -0.001 (0.03) (0.03) (0.04) (0.03) (0.03) (0.03) (0.02) (0.02) Blac k W omen Complete d F ertility 2.86 0.47 0.52 0.40 0.39 0.26 0.14 -0.08 -0.06 (0.09) (0.09) (0.11) (0.07) (0.08) (0.08) (0.06) (0.09) White W omen Complete d F ertility 2.35 0.04 0.04 0.03 0.01 0.03 0.03 0.06 0.01 (0.04) (0.03) (0.03) (0.03) (0.03) (0.02) (0.01) (0.01) Note:* p< 0:1; ** p< 0:05; *** p< 0:01. 34 T able 1.5: Cross States Dierences in Births Timing (Age Group) This table rep orts estimation results of equation (1.2) for age at birth of rst c hild and births spacing, for the full sample and b y race. F or information on v ariables construction and sample denition refer to App endix A.1.1. Mean 1 2 3 4 5 6 7 [15-17] [18-20] [21-23] [24-26] [27-29] [30-32] [33-35] F ull Sample A ge at First Child 24.00 0.99 0.92 0.60 0.57 0.43 0.37 0.46 (0.08) (0.13) (0.08) (0.04) (0.02) (0.06) (0.04) N=800,730 Birth Sp acing 3.71 -0.59 -0.61 -0.56 -0.57 -0.56 -0.54 -0.59 (0.03) (0.05) (0.02) (0.03) (0.03) (0.02) (0.06) N=354,611 Blac k W omen A ge at First Child 22.26 0.83 0.27 0.01 0.06 0.08 -0.08 0.28 (0.07) (0.06) (0.14) (0.11) (0.10) (0.16) (0.12) N=80,000 Birth Sp acing 4.64 -0.60 -0.72 -0.23 -0.72 -0.28 -0.64 -0.63 (0.16) (0.10) (0.08) (0.09) (0.12) (0.09) (0.13) N=27,156 White W omen A ge at First Child 24.20 1.11 1.06 0.74 0.74 0.60 0.54 0.60 (0.07) (0.14) (0.06) (0.03) (0.02) (0.06) (0.05) N=706,113 Birth Sp acing 3.62 -0.62 -0.63 -0.62 -0.61 -0.62 -0.58 -0.64 (0.02) (0.04) (0.02) (0.03) (0.03) (0.02) (0.06) N=321,787 Note: All regressions ha v e b een w eigh ted b y p opulation w eigh ts. Clustered robust standard errors are rep orted in paren thesis. * p< 0:1; ** p< 0:05; *** p< 0:01. 35 T able 1.6: Dierence-in-Dierences Estimates of the Eect of Ab ortion A ccess on Births Timing This table rep orts the dierence-in-dierences estimates for the eect of ab ortion access on birth timing outcomes. The columns rep ort the eect of ab ortion access starting a certain age phase compared to receiving access to ab ortion at the end of the fertilit y cycle. These co ecien ts are computed using estimates from T able 1.5. Mean 1 2 3 4 5 6 [15-17] [18-20] [21-23] [24-26] [27-29] [30-32] F ull Sample A ge at First Child 24.00 0.53 0.46 0.14 0.11 -0.03 -0.09 (0.07) (0.16) (0.11) (0.06) (0.04) (0.04) Sp acing 3.71 -0.00 -0.02 0.03 0.02 0.03 0.05 (0.08) (0.07) (0.07) (0.08) (0.08) (0.06) Blac k W omen A ge at First Child 22.26 0.54 -0.01 -0.27 -0.23 -0.20 -0.37 (0.15) (0.16) (0.25) (0.09) (0.07) (0.10) Sp acing 4.46 0.03 -0.09 0.41 -0.08 0.35 -0.01 (0.26) (0.16) (0.15) (0.15) (0.16) (0.17) White W omen A ge at First Child 24.20 0.51 0.46 0.14 0.13 -0.00 -0.06 (0.06) (0.18) (0.11) (0.07) (0.06) (0.04) Sp acing 3.62 0.02 0.01 0.02 0.03 0.02 0.05 (0.07) (0.07) (0.07) (0.09) (0.09) (0.06) Note:* p< 0:1; ** p< 0:05; *** p< 0:01. 36 T able 1.7: Cross States Dierence in Blac k W omen’s Marriage Outcomes This table rep orts estimation results of equation (1.2) for marriage outcome of blac k w omen. Marriage is a dumm y v ariable that tak es v alue 1 if a w oman is married at the time of observ ation and 0 otherwise. The three other outcome v ariables are dened for married blac k w omen only . Husband College Completion is a dumm y v ariable that tak es v alue 1 if the h usband has a college degree and 0 otherwise. Husbands Y early Earnings is a con tin uous measure of y early lab or earnings deated and expressed in 2012 dollars. Husband Occupation Index is a measure of o ccupation status that tak es v alue b et w een 0 and 100, with larger v alues indicating o ccupations with higher median w age rates. Mean 1 2 3 4 5 6 7 [15-17] [18-20] [21-23] [24-26] [27-29] [30-32] [33-35] Marrie d 0.48 -0.09 -0.09 -0.09 -0.11 -0.08 -0.13 -0.06 (0.01) (0.00) (0.00) (0.01) (0.01) (0.00) (0.02) N=146,494 Husb ands Col le ge Completion 0.07 0.03 0.04 0.04 0.04 -0.03 -0.02 -0.04 (0.02) (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) N=69,682 Husb ands Y e arly Earnings 16,891 4,672 4,891 2,427 189 -688 240 -1,495 (465) (651) (476) (535) (534) (1,112) (513) N=69,679 Husb ands Oc cup ation Index 65.97 -6.68 -1.79 -4.12 -2.60 -2.65 -5.49 -6.89 (1.32) (1.20) (0.53) (0.45) (1.27) (2.74) (2.73) N=50,962 Note: All regressions ha v e b een w eigh ted b y p opulation w eigh ts. Clustered robust standard errors are rep orted in paren thesis. * p< 0:1; ** p< 0:05; *** p< 0:01. 37 T able 1.8: Dierence-in-Dierences Estimates of the Eect of Ab ortion A ccess on Blac k W omen’s Marriage This table rep orts the dierence-in-dierences estimates for the eect of ab ortion access on blac k w omen marriage outcomes. The columns rep ort the eect of ab ortion access starting a certain age phase compared to receiving access to ab ortion at the end of the fertilit y cycle. These co ecien ts are computed using estimates from T able 1.7. Mean 1 2 3 4 5 6 [15-17] [18-20] [21-23] [24-26] [27-29] [30-32] Marriage 0.48 0.02 -0.02 -0.03 -0.05 -0.02 -0.06 (0.02) (0.02) (0.02) (0.01) (0.03) (0.02) Husb and Col le ge Completion 0.07 0.06 0.07 0.08 0.08 0.00 0.01 (0.03) (0.01) (0.01) (0.02) (0.01) (0.01) Husb and Y e arly Earnings 16,891 6,167 6,386 3,922 1,684 807 1,736 (562) (1,030) (906) (689) (616) (1,484) Husb and Oc cup ation Index 65.97 0.21 5.10 2.77 4.29 0.24 1.40 (3.81) (3.76) (3.13) (2.56) (1.64) (0.60) Note: * p< 0:1; ** p< 0:05; *** p< 0:01. 38 T able 1.9: Eect of Start of Motherho o d on Earnings and Lab or Supply This table rep orts OLS and 2SLS estimation results of equation (1.3) for the eect of age of start of motherho o d on y early lab or earnings, lab or force participation and w eekly hours w ork ed. The sample includes all w omen b oth w orking and not w orking, with lab or hours and earnings assigned the v alue of 0 if a w oman is not w orking. The estimation results for the full sample are rep orted in columns (1) and (2), for the subsample of blac k w omen are rep orted in columns (3) and (4), and for the subsample of white w omen are rep orted in columns (5) and (6). F ull Sample Blac k W omen White W omen OLS 2SLS OLS 2SLS OLS 2SLS (1) (2) (3) (4) (5) (6) Y e arly L ab or Earnings 282 2,194 516 1,989 252 1,980 (10) (263) (20) (514) (12) (243) Aver age Y e arly Earnings 16,433 17,559 16,253 L ab or F or c e Particip ation -0.008 0.027 0.002 -0.031 -0.009 0.028 (0.000) (0.010) (0.000) (0.016) (0.000) (0.010) Aver age L ab or F or c e Particip ation 0.679 0.736 0.671 W e ekly Hours W orke d -0.34 0.50 0.13 -0.64 -0.41 0.43 (0.02) (0.43) (0.02) (0.66) (0.02) (0.42) Aver age W e ekly Hours W orke d 23.06 24.90 22.79 Num b er of Observ ations 800,730 80,000 706,113 Note: All regressions ha v e b een w eigh ted b y p opulation w eigh ts. Clustered robust standard errors are rep orted in paren thesis. * p< 0:1; ** p< 0:05; *** p< 0:01. 39 T able 1.10: Eect of Start of Motherho o d on Lab or Outcomes of Emplo y ed W omen This table rep orts OLS and 2SLS estimation results of equation (1.3) for the eect of age of start of motherho o d on hourly w age rate, Occupation Index and W eekly hours w ork ed. The sample is restricted to w orking w omen only . The estimation results for the full sample are rep orted in columns (1) and (2), for the subsample of blac k w omen are rep orted in columns (3) and (4), and for the subsample of white w omen are rep orted in columns (5) and (6). F ull Sample Blac k W omen White W omen OLS 2SLS OLS 2SLS OLS 2SLS (1) (2) (3) (4) (5) (6) Hourly W age R ate 0.47 1.96 0.36 2.18 0.49 1.79 (0.02) (0.11) (0.01) (0.48) (0.03) (0.10) Aver age Hourly W age R ate 13.5 13.6 13.5 Oc cup ation Index 1.15 3.03 0.96 3.42 1.19 2.79 (0.02) (0.13) (0.03) (0.52) (0.03) (0.12) Oc cup ation Index Aver age 40.30 35.67 40.95 W e ekly Hours W orke d if Employe d -0.21 -0.64 0.00 0.68 -0.26 -0.78 (0.02) (0.22) (0.02) (0.22) (0.02) (0.21) Aver age W e ekly Hours W orke d if Employe d 37.33 38.19 37.23 Num b er of Observ ations 474,609 53,434 411,933 Note: All regressions ha v e b een w eigh ted b y p opulation w eigh ts. Clustered robust standard errors are rep orted in paren thesis. * p< 0:1; ** p< 0:05; *** p< 0:01. 40 Figures Figure 1.1: F ertilit y Outcome T rends This gure sho ws fertilit y outcomes trends for birth cohorts that receiv ed v arious lev els of access to legal ab ortion. The t w o v ertical lines demarcate the birth cohorts that had v arying lev el of exp osure to legal ab ortion across states. Observ ations of total n um b er of c hildren b orn, age of mother and age of eldest c hild in the household from the 1960, 1970, 1980, 1990 and 2000 IPUMS samples are used to constructed the measures of completed fertilit y and age at birth of rst c hild for the 1920-1929, 1930-1939, 1940-1949, 1950-1959 and 1960-1969 birth cohorts resp ectiv ely . 0 1 2 3 4 Completed Fertility 1920−1929 1930−1939 1940−1949 1950−1959 1960−1969 Birth Cohorts Non−Repeal States Repeal States (a) Completed F ertilit y 23 23.5 24 24.5 25 25.5 Age at Birth of First Child 1920−1929 1930−1939 1940−1949 1950−1959 1960−1969 Birth Cohorts Non−Repeal States Repeal States (b) Age at Birth of First Child Note: Completed fertilit y for the 1960-1969 birth cohorts is not rep orted due to the v ariable not b eing a v ailable in the 2000 IPUMS sample. 41 Figure 1.2: Eect of A ccess to Ab ortion on Completed F ertilit y −.2 −.1 0 .1 .2 Difference in Completed Fertility 15−17 18−20 21−23 24−26 27−29 30−32 33−35 36−38 39−40 Age at Access to Abortion (a) Completed F ertilit y T rend b y Age Group 0.06 0.06 0.05 0.04 0.04 0.03 0.03 −0.01 −0.15 −0.10 −0.05 0.00 0.05 0.10 0.15 Diff−in−Diff Estimates 15−17 18−20 21−23 24−26 27−29 30−32 33−35 36−38 Age at Access to Abortion (b) Estimates of the eect of ab ortion access on Completed F ertilit y - Age Group 42 Figure 1.3: Eect of A ccess to Ab ortion on Age at Start of Motherho o d −1.5 −1 −.5 0 .5 1 1.5 Age at First Child 15−17 18−20 21−23 24−26 27−29 30−32 33−35 Age at Access to Abortion (a) Age at Birth of First Child T rend b y Age Group 0.53 0.46 0.14 0.11 −0.03 −0.09 −1.00 −0.50 0.00 0.50 1.00 Diff−in−Diff Estimates 15−17 18−20 21−23 24−26 27−29 30−32 Age at Access to Abortion (b) Estimates of the eect of ab ortion access on Age at Birth of First Child - Age Group 43 Figure 1.4: Eect of A ccess to Ab ortion on Birth Spacing −1 −.5 0 .5 1 Difference in Birth Spacing 15−17 18−20 21−23 24−26 27−29 30−32 33−35 Age at Access to Abortion (a) Birth Spacing T rend b y Age Group −0.00 −0.02 0.03 0.02 0.03 0.05 −0.20 −0.10 0.00 0.10 0.20 Diff−in−Diff Estimates 15−17 18−20 21−23 24−26 27−29 30−32 Age at Access to Abortion (b) Estimates of the eect of ab ortion access on Births Spacing - Age Group 44 Figure 1.5: Eect of A ccess to Ab ortion on Completed F ertilit y b y Race −1 −.5 0 .5 1 Completed Fertility 15−17 18−20 21−23 24−26 27−29 30−32 33−35 36−38 39−40 Age at Access to Abortion Black Women White Women (a) Completed F ertilit y T rends for Blac k and White W omen 0.47 0.52 0.40 0.39 0.26 0.14 −0.08 −0.06 −0.80 −0.40 0.00 0.40 0.80 Diff−in−Diff Estimates 15−17 18−20 21−23 24−26 27−29 30−32 33−35 36−38 Age at Access to Abortion (b) Dierence-in-Dierences Estimates for Blac k W omen 0.04 0.03 0.03 0.01 0.03 0.03 0.06 0.01 −0.80 −0.40 0.00 0.40 0.80 Diff−in−Diff Estimates 15−17 18−20 21−23 24−26 27−29 30−32 33−35 36−38 Age at Access to Abortion (c) Dierence-in-Dierences Estimates for White W omen 45 Figure 1.6: Eect of A ccess to Ab ortion on Age at Start of Motherho o d b y Race −1.5 −1 −.5 0 .5 1 1.5 Age at First Child 15−17 18−20 21−23 24−26 27−29 30−32 33−35 Age at Access to Abortion Black Women White Women (a) Age at Birth of First Child T rends for Blac k and White W omen 0.54 −0.01 −0.27 −0.23 −0.20 −0.37 −1.00 −0.50 0.00 0.50 1.00 Diff−in−Diff Estimates 15−17 18−20 21−23 24−26 27−29 30−32 Age at Access to Abortion (b) Dierence-in-Dierences Estimates for Blac k W omen 0.51 0.46 0.14 0.13 −0.00 −0.06 −1.00 −0.50 0.00 0.50 1.00 Diff−in−Diff Estimates 15−17 18−20 21−23 24−26 27−29 30−32 Age at Access to Abortion (c) Dierence-in-Dierences Estimates for White W omen 46 Figure 1.7: Eect of A ccess to Ab ortion on Birth Spacing b y Race −1.5 −1 −.5 0 .5 1 1.5 Difference in Birth Spacing 15−17 18−20 21−23 24−26 27−29 30−32 33−35 Age at Access to Abortion Black Women White Women (a) Birth Spacing T rends for Blac k and White W omen 0.03 −0.09 0.41 −0.08 0.35 −0.01 −1.00 −0.50 0.00 0.50 1.00 Diff−in−Diff Estimates 15−17 18−20 21−23 24−26 27−29 30−32 Age at Access to Abortion (b) Dierence-in-Dierences Estimates for Blac k W omen 0.02 0.01 0.02 0.03 0.02 0.05 −1.00 −0.50 0.00 0.50 1.00 Diff−in−Diff Estimates 15−17 18−20 21−23 24−26 27−29 30−32 Age at Access to Abortion (c) Dierence-in-Dierences Estimates for White W omen 47 Figure 1.8: Eect of A ccess to Ab ortion on Marriage Outcomes for Blac k W omen −.2 −.1 0 .1 .2 Difference in Probability of Marriage 15−17 18−20 21−23 24−26 27−29 30−32 33−35 Age at Access to Abortion (a) Dierence in Rate of Marriage −0.02 −0.02 −0.03 −0.05 −0.02 −0.06 −0.10 −0.05 0.00 0.05 0.10 Diff−in−Diff Estimates 15−17 18−20 21−23 24−26 27−29 30−32 Age at Access to Abortion (b) Estimates of the eect of ab ortion access on Rate of Marriage −.2 −.1 0 .1 .2 Difference in Husband College Completion 15−17 18−20 21−23 24−26 27−29 30−32 33−35 Age at Access to Abortion (c) Dierence in Husbands College Completion 0.06 0.07 0.08 0.08 0.00 0.01 −0.20 −0.10 0.00 0.10 0.20 Diff−in−Diff Estimates 15−17 18−20 21−23 24−26 27−29 30−32 Age at Access to Abortion (d) Estimates of the eect of ab ortion access on Husband College Completion −10000 −5000 0 5000 10000 Difference in Husband Labor Earnings 15−17 18−20 21−23 24−26 27−29 30−32 33−35 Age at Access to Abortion (e) Dierence in Husbands Earnings 6167 6386 3922 1684 807 1736 −10000 −5000 0 5000 10000 Diff−in−Diff Estimates 15−17 18−20 21−23 24−26 27−29 30−32 Age at Access to Abortion (f ) Estimates of the eect of ab ortion access on Husband Earnings −15 −10 −5 0 5 10 15 Difference in Husband Occupation Status 15−17 18−20 21−23 24−26 27−29 30−32 33−35 Age at Access to Abortion (g) Dierence in Husbands Occupation 0.21 5.10 2.77 4.29 0.24 1.40 −15.00 −10.00 −5.00 0.00 5.00 10.00 15.00 Diff−in−Diff Estimates 15−17 18−20 21−23 24−26 27−29 30−32 Age at Access to Abortion (h) Estimates of the eect of ab ortion access on Husband Occupation 48 Chapter 2 W orking More when W orking Less: The Eect of Reduced W orkweek Hours on Participation of Married W omen 1 2.1 Introduction Gender dierences in w ages, lab or force participation and hours w ork ed ha v e alw a ys b een k ey questions in lab or economics (Killingsw orth and Hec kman, 1986). The past few decades ha v e witnessed signican t impro v emen ts in w omen’s righ ts and education, and c hanges in the o v erall p erceptions in gender roles. Y et the gender w age gap still p ersists. Moreo v er, w omen are still less lik ely to b e em- plo y ed and, when emplo y ed, they supply signican tly few er hours than their male co w ork ers. Goldin (2014) argues that the remaining gender w age gap results from the dis- prop ortionate comp ensation for more hours w ork ed. In other w ords, men earn higher w ages b ecause they are willing to sp end longer hours at the w orkplace than w omen. Cortes and P an (2017) test this theory and nd plausible causal rela- tionship b et w een return to o v erw orking and the gender w age gap. This raises the question of the dierence b et w een men and w omen in their willingness to w ork longer hours. Household sp ecialization resulting from comparativ e adv an tage in 1 Join t W ork with Serena Cannan and Pierre Mouganie 49 house pro duction pro vides a p ossible explanation (Bec k er, 1993). W omen, esp e- cially when they are married and ha v e c hildren, ha v e a high in-house pro ductivit y , whic h increases the opp ortunit y cost of participating in the lab or mark et. In other w ords, the cost of an hour sp en t at the w orkplace is higher for w omen than it is for men. This mec hanism as a p oten tial explanation for dierence in lab or supply and lab or force participation is what this pap er aims to in v estigate. A gro wing b o dy pf empirical literature in v estigate the issue of time cost on lab or supply of married w omen. One particular approac h has b een to exploit exogenous v ariation to household pro ductivit y of married w omen b y ev aluating the eect of a v ailable c heap er substitutes to in-house pro duction on lab or supply of w omen (Cortes and T essada, 2011; Cortes and P an, 2013). In this pap er, I follo w a more direct approac h to ev aluate the cost of w orking longer hours on married w omen’s decisions to participate in the lab or mark et. In this pap er, W e estimate the eect of a reduction in hours exp ected to sp end in the w orkplace on lab or force participation of married w omen. W e exploit a w orkw eek hours reduction la w that led to a considerable decrease in w eekly hours w ork ed in F rance b et w een the y ears 2000 and 2002. Reduced form estimates sho w an a v erage of 3 p ercen tage p oin t increase in participation of married w omen as a result of the p olicy c hange. W e then estimate a discrete c hoice mo del of lab or force participation. The decision to participate dep ends on p oten tial individual lab or earnings and a participation cost that v aries with the hours exp ected to sp end in the w orkplace. The cost function sp ecication allo ws for heterogeneous cost of participation b y gender and marital status. Moreo v er, the mo del admits a exible correlation b et w een w ages and hours sp en t in the w orkplace. Estimates of the structural mo del sho w a p ositiv e correlation b et w een w ages and hours sp en t in the w orkplace, and that participation is more taxing for w omen than men. A previous study has used the w orkw eek hours reduction in F rance to study individual lab or supply decisions. Goux et al. (2014) use the same p olicy c hange to study in ter-dep endencies in sp ousal lab or supply . While the main outcome w e study is dieren t from the on they study , their results are of relev ance for our purp ose. In particular, they nd that married w omen’s lab or supply do es not resp ond to c hanges in their h usband’s lab or supply . Consequen tly , when a man w orks few er hours, this has no eect on their wife’s use of time, whic h implies that 50 men do not substitute for w omen’s w ork in the household. The remainder of this pap er is organized as follo ws: section 2 briey discuss the c hanges in w orkw eek hours that to ok place in F rance in the late 90s. The data set used is presen ted in Section 3. In section 4, reduced form evidence on the eect of shorter w orkw eek hours on lab or force participation of married w omen are presen ted. Section 5 presen ts and discusses estimation results of a discrete c hoice mo del of lab or force participation. The last section concludes the pap er. 2.2 Reduced W orkweek Hours: The Aubry Law The early legislativ e elections of 1997 brough t a coalition of the left to p o w er in F rance. The new go v ernmen t, led b y Lionel Jospin, sough t to resolv e high unemplo ymen t. The 35-hours w orkw eek, whic h came to b e kno wn as Loi Aubry I (after Martin Aubry the Lab or Minister), w as enacted in 1998 and it aimed at reducing unemplo ymen t b y means of w ork sharing. The new lab or hours w ere implemen ted in t w o stages. Large rms (20+w ork ers) w ere required to implemen t the 35-hour w ork w eek starting Jan uary 2000, while small rms (less than 20 w ork ers) had to follo w in Jan uary 2002. A dditionally , nancial incen tiv es w ere giv en to rms (large and small) for immediate reduction in w orking hours. In 2000, a second la w (Aubry I I) w as enacted conrming the 35- hour w ork w eek and oered a series of nancial incen tiv es to complying rms in the form of tax cuts and subsidies. The go v ernmen t allo w ed bargaining among so cial partners (rms and w ork ers) in order to reac h satisfying outcomes for b oth parties. F or instance, ann ualisation of w orking hours w as p ermitted, in whic h a exible 1,600-hour w orking y ear limit w as allo w ed instead of a more rigid 35-hour w orking w eek limit. A detailed discussion of the p olitical atmosphere and the reform that accompanied the w orkw eek hours c hanges of that era is pro vided in Ask enazy (2013). It should b e noted that the Aubry La w w as nev er fully implemen ted, since the triumph of the righ t in the 2002 parliamen tary election brough t an end to the so cialist go v ernmen t programs. This pap er’s aim is to in v estigate ho w the length of hours exp ected to b e sp en t outside the household aects a w oman’s lab or mark et activit y . Three main features mak e the c hanges brough t b y the Aubry La w an ideal natural exp erimen t to answ er 51 this question. The rst feature is that the la w actually ac hiev ed its stated ob jectiv e of reducing hours w ork ed, with an estimated 2 hours reduction in total w orking time (F ull time + Ov ertime) in rms that complied with the la w (P asseron, 2000, 2002). The eect of the p olicy on w age earnings w ere small. P asseron (2000) do cumen ts a 1% w age increase among the rms that signed the Aubry agreemen t in its rst y ear. This second feature rules out p oten tial income mec hanisms that could lead to an increase in lab or force participation. The la w oered a subsidy for rms conditional on their commitmen t to creating jobs, leading to a 7% a v erage increase in jobs o v er t w o y ears (P asseron, 2000, 2002). This last feature guaran tees that individuals who migh t b e more inclined to participate in the lab or force as a result of reduced w orkw eek hours will not b e discouraged b y the lac k of emplo ymen t opp ortunities. In short, b et w een the y ears 2000 and 2002, jobs required less hours of attendance at the w orkplace with virtually no c hange in w ages and more job v acancies to hire p oten tial new en tran ts in to the lab or mark et. As stated ab o v e, the p olicy c hange under consideration w as binding for priv ate rms with more than 20 w ork ers in the p erio d b et w een 2000-2002. Nonetheless, most public sector institutions 2 whic h emplo y ed around 16% of the F renc h lab or force at that time implemen ted the 35-hour w orking w eek (Ask enazy, 2013). Al- though the la w ga v e small rms (less than 20 w ork ers) un til 2002 to adjust for the new w orkw eek hours, a large n um b er of these rms mo v ed to 35-hour w ork w eek early on. In the rst trimester of 2000, around 50% of rms with as few as 10 emplo y ees signed the Aubry agreemen t (P asseron, 2000). The fact that the p olicy w as implemen ted b y emplo y ers that w ere b ey ond those designated in the la w indi- cates the p olicy led to an adjustmen t of the a v erage exp ected hours to b e sp en t in the w ork place ev ery w eek for the whole lab or mark et. The o v erreac hing eect of the p olicy is imp ortan t for the empirical analysis emplo y ed b elo w, since otherwise iden tifying the treated group w ould b e a ma jor c hallenge. If, for instance, the ef- fect of the p olicy w as restricted to a small group of emplo y ers, determining whic h p oten tial lab or mark et en tran t has adjusted their w orking hours exp ectation w ould b e virtually imp ossible, as that requires observing whic h group of emplo y ers the individual is considering. This information is not a v ailable in most data sets. 2 Sc ho ol teac hers, one fth of the public sector jobs, w ere the only public sector emplo y ees not to b enet from reduced w orking hours. 52 2.3 Data The data used in the empirical analysis are from the F renc h Lab or F orce Surv ey (LFS), administered b y the National Institute of Statistics and Economic Studies (INSEE). The LFS is a national surv ey that includes data on a v ariet y of lab or mark et outcomes and c hoices of in terest. Most imp ortan tly , the surv ey rep orts lab or mark et activit y status, hours w ork ed and y early lab or earnings. The surv ey also includes a wide range of information on individual c haracteristics, suc h as education, region of residency and marital status. The LFS is an un balanced panel, where eac h individual is surv ey ed for three consecutiv e time units and then replaced. The sample used includes observ ations recorded ann ually in the surv ey b et w een 1990 and 2002. The sample size is around t w o million observ ations, whic h are almost equally spread o v er the 13 observ ation y ears 3 . Observ ations in the sample prior to Jan uary 2000 - the date Aubry I w as put in to eect - are around 80% of the total n um b er observ ations. Giv en that the main in terest in this pap er is to in v estigate lab or mark et activit y , w e restrict the sample to individuals who w ere b et w een the ages of 18 and 75 at time of observ ation. Three main lab or mark et measures are directly rep orted in the surv ey: lab or force participation status, w eekly hours w ork ed and mon thly salary . The LFS also rep orts age and marital status. w e construct a marriage dumm y v ariable that tak es v alue 1 if the individual is married and 0 otherwise. Single, wido w ed and div orced are group ed together since the fo cus of the analysis is on the eect of household sp ecialization and ho w lab or mark et participation of w omen c hange in the presence of a partner. The surv ey rep orts the highest degree earned. Using this information, w e construct dumm y v ariables for univ ersit y degree holders, high sc ho ol diploma holders and middle sc ho ol diploma holders. Summary statistics of lab or mark et outcomes and individual c haracteristics are rep orted in T able 2.1. On a v erage, w omen w ork less than men b efore and after the c hange in w orkw eek hours. W omen are less lik ely than men to b e activ e in the lab or mark et, and when they are emplo y ed they w ork few er hours. The a v erage n um b er of hours w ork ed b y 3 There are around 150,000 observ ations in ev ery y ear, but the n um b er of observ ations is sligh tly higher in the last four y ears. 53 b oth men and w omen in the p ost-Aubry era are lo w er than the a v erage n um b er of hours w ork ed b efore the w orkw eek hours c hange. The dierence in hours w ork ed b efore and after the c hange is 1:58 and 1:19 hours for men and w omen resp ectiv ely . Lab or force participation of men is 1 p ercen tage p oin t lo w er after the reduction in w orkw eek hours, whereas for w omen it increased 2 p ercen tage p oin ts. Mon thly earnings of men are higher than w omen b oth b efore and after the w orkw eek hour c hanges. The gender w age gap in b oth time p erio ds is p ersisten t, with the a v erage salary of men b eing 33% higher that the a v erage salary of w omen. Individual c haracteristics rep orted in mo dule B sho w similar a v erage c haracter- istics for the samples b efore and after the c hanges, with the exception of a lo w er prop ortion of married individuals and a higher prop ortion of univ ersit y degree holders in the later samples. These dierences are most lik ely due to the more recen t sample y ears b eing comp osed of later b orn birth cohorts for whic h norms of marriage and college attendance are dieren t. 2.4 Eect of Labor Hour Reduction on Participa- tion This section presen ts evidence on the eect of reduction in w orkw eek hours on lab or supply of married w omen. First, w e presen t descriptiv e evidence of c hanges in the distribution of lab or activit y and hours w ork ed that accompanied the c hanges in w orkw eek regulations in F rance. W e then pro ceed to sho w that the reduction in w orkw eek hours led to a signican t increase in married w omen’s lab or force participation. 2.4.1 Descriptive Evidence and Stylized F acts Decrease in W eekly Hours W ork ed In Jan uary 2000, full time w orkw eek hours dropp ed from 39 to 35 hours. While w orking more than 35 hours p er w eek w as still a p ossibilit y , the la w rendered it more costly for emplo y ers, as these o v ertime hours w ere to b e rem unerated at higher rates and sub ject to higher taxation. T o explore the c hange in hours w ork ed 54 asso ciated with the reduced w orkw eek hours the follo wing equation is estimated h it = 0 + 1 Aubry t + it (2.1) where h it is w eekly hours w ork ed b y individual i at y ear t. Aubry t is a dumm y v ariable that tak es v alue 0 b efore the y ear 2000 and 1 thereafter. Equation (2.1) is estimated for the full sample and separately for men and w omen. OLS estimates of 1 are rep orted in T able 2.2. Regression results sho w that among the emplo y ed, w eekly hours w ork ed after w orkw eek reduction w as put in place is on a v erage one and one-half hours less than what it used to b e b efore. The Aubry la w is asso ciated with a large reduction in a v erage w eekly hours w ork ed for w omen. The a v erage hours w ork ed for w omen w ere 1:58 hours less during the Aubry era, from a mean of 35:08 w eekly hours of w ork. In comparison, men w ork ed 1:19 less hours ev ery w eek from a mean of 42:37. The ab o v e discussion suggests that, as predicted, the Aubry la w is asso ciated with emplo y ees sp ending less time in the w orkplace. Moreo v er, the reduction in w eekly hours w ork ed is not uniform across genders, as suggested b y the distribu- tions of w eekly hours w ork ed. Plots of the distribution of hours w ork ed b y men and w omen b efore and after the implemen tation of the reduced w orkw eek hours are presen ted in Figure 2.1. Stark dierences b et w een men and w omen are observ- able. Prior to the w orkw eek reduction, men’s hours w ork ed p eek ed at 39 hours a w eek, with smaller densit y spik es around 45, 50 and 60 hours a w eek. In the p ost-Aubry era, the distribution of hours w ork ed b y men is t w o p eek ed at 35 and 39 hours p er w eek. While few er men w ork ed more than 40 hours a w eek, p eeks at 45, 50 and 60 hours still p ersist. Similar to men, w eekly hours w ork ed b y w omen prior to 2000 p eek ed at 39 hours a w eek. Ho w ev er, v ery few w omen w ork ed more than 40 hours, with an almost uniform mass of w omen w orking b et w een 30 and 38 hours and a spik e in densit y around the 20 hours a w eek. After 2000, the densit y of hours w ork ed b y w omen is also t w o p eek ed at 35 and 39 hours. There are few er w omen w orking more than 40 hours a w eek, and the densit y of w omen w orking b et w een 20 and 35 hours is more spread out. Changes in the distribution of hours w ork ed can also b e seen in T able 2.3, 55 where dieren t quan tiles of the distribution of w eekly hours w ork ed are rep orted. As men tioned b efore, there is a reduction in a v erage hours w ork ed b y b oth men and w omen. Nev ertheless, the c hange in the distribution is quite dieren t. After the in tro duction of shorter w orkw eek hours, most men still w ork ed more than 35 hours a w eek and more than half of them w ork ed more than 39 hours a w eek. The only visible c hange for men is the 25 th p ercen tile dropping from 39 hours a w eek to 35. This w as noted in the distribution plot, with the men’s distribution going from single p eek ed at 39 hours to double h ump ed at 35 and 39 hours. As for the w omen’s distribution of w orking hours, the 25 th p ercen tile dropp ed from 30 to 28 and the median dropp ed from 39 to 35. F or b oth men and w omen, there w ere no ma jor c hanges in the upp er p ercen tiles. Changes in the Lab or F orce P articipation This pap er in v estigates the eect of reduced w orkw eek hours on female lab or force participation. This section lo oks at the c hanging trends in lab or force par- ticipation b y gender, marital status and age. The discussion in this section will help guide the formal empirical analysis of the causal relationship presen ted in the next section. Lab or force participation b y gender and marital status from y ears 1990 to 2002 are plotted in Figure 2.2. The plots sho w no ma jor c hanges in lab or force partici- pation after the reduction in w orkw eek hours for neither men or w omen regardless of their marital status. Moreo v er, there are no dierences in lab or force partic- ipation b et w een single and married individuals with the same gender. The only distinctiv e feature that could b e noted is the consisten tly higher lab or force par- ticipation of men compared to the lab or force participation of w omen, with 70% of the men b eing part of the lab or force compared to 50% of the w omen. This rst observ ation of the data migh t b e misleading and lead us to the conclusion that the reduction in w orkw eek hours w as not asso ciated with an y c hange in activit y among the p oten tial w orking age p opulation. The decision to participate in the mark et dep ends on the net return to lab or mark et activit y , whic h is determined b y p oten tial lab or mark et earnings and the returns to non-mark et activities. In the case of married w omen, ma jor non-mark et activities include fertilit y , c hildcare, and household pro duction of go o ds. A direct measure of the non-lab or activit y is 56 dicult to construct and most data sets w e are a w are of do not include reliable measures. Moreo v er, the data set used do es not include information on determi- nan ts of non-mark et activities that migh t b e of in terest for the purp ose of this study 4 . Ho w ev er, these determinan ts are highly correlated with a w oman’s age. Figure 2.3 presen ts plots of lab or force participation for men and w omen b y marital status and age for the y ears 1990 and 2002. These plots pro vide evidence of the p oten tial c hange in married w omen’s lab or force participation that w as asso ciated with the reduced w orkw eek hours, while the rates of lab or force par- ticipation b y age had remained virtually unc hanged b et w een the y ears 1990 and 2002 for single w omen as w ell as single and married men. The pattern has b een dieren t for married w omen as it is illustrated in Figure 2.3b. The c hange is most notably visible for married w omen in the age group of 35 to 55. In 1990, around 70% of married w omen aged 35 w ere in the lab or force. The rate of lab or force participation of married w omen starts decreasing at the age of 42 to reac h 50% b y age 55. The rates of lab or force participation of married w omen displa y a dieren t age trend in 2002. The rate remains p ersisten tly close to 80% un til age 52 and b y age 55 it remains higher than 60%. The do cumen ted increase in participation is concen trated among married w omen b et w een the ages of 35 and 55. This is the p erio d of life where most w omen giv e birth and allo cate a large p ortion of their time for c hildcare. 2.4.2 Reduced F orm Evidence Iden tication Strategy The discussion in the previous section made clear that the la w actually w as eectiv e in decreasing hours w ork ed for b oth men and w omen. While the a v erages of hours w ork ed for b oth genders decreased after the la w w as put in to eect, the c hanges asso ciated with the la w in the distribution of hours w ork ed w as dier- en t. Moreo v er, it sho ws evidence that Aubry la w w as asso ciated in c hanges in lab or force participation of married w omen with no clear c hanges in activit y of other demographic groups (including single w omen and married and single men). 4 suc h v ariables include n um b er and age of c hildren, cost of lo cal c hildcare, presence of nearb y relativ es that could substitute for c hildcare, cost of consumable go o ds and services that could substitute for in house pro duction. 57 The purp ose of this section is to pro vide plausible estimates of the a v erage eect of the reduction in w orkw eek hours on the lab or force participation of married w omen. The main c hallenge in iden tifying this eect is the prop er denition of the comparison group. The descriptiv e evidence ab o v e suggests that the eect of w orkw eek hours reduction is heterogeneous along t w o dimensions: gender and marital status. This nding guides the approac h in this section, in whic h w e adopt a dierence-in- dierence-in-dierences (DDD) approac h to estimate the eect of shorter hours w ork ed on lab or force participation of married w omen. T o do so, w e estimate the follo wing equation: a it = 0 + 1 Female i + 2 Married i + 3 Aubry t + 4 Married i Female i + 5 Female i Aubry t + 6 Married i Aubry t + 7 Married i Female i Aubry t +X + it (2.2) Wherea it indicates lab or activit y status, taking v alue 1 if individual i is in the lab or force at time t and 0 otherwise. Female i is a gender indicator v ariable, Married it is a dumm y v ariable indicating if individual i is married at time t and Aubry t is a dumm y v ariable indicating y ears in whic h w orkw eek hours reduction is in eect. X is a v ector of observ able con trols including education lev els, a p olynomial function of age as w ell as y ear and region xed eects. The a v erage eect of w orkw eek reduction on married w omen lab or force participation is iden tied b y 7 . Estimation Results Estimates of the eect of reduced w orkw eek hours on married w omen’s lab or force participation estimated in equation (2.2) are rep orted in columns 1 to 3 in T able 2.4. Estimation results sho w that in comparison to men and single w omen, lab or force participation of married w omen increased b y 3:6 p ercen tage p oin ts as a result of the reduction in w orkw eek hours. The estimate in column 1 is sligh tly smaller, sho wing an a v erage eect of 2:6 p ercen tage p oin t. T o further consolidate the results obtained from the main estimation, t w o al- ternativ e approac hes using alternativ e subsamples for comparison, are adopted. 58 The rst approac h consists of restricting the sample to married individuals only and estimating the follo wing equation a it = 0 + 1 Female i + 2 Aubry t + 3 Female i Aubry t +X + it (2.3) The a v erage eect of w orkw eek hours reduction on married w omen in this sp ec- ication is iden tied b y 3 . Estimation results of equation (2.3) are rep orted in columns 4 to 6 in T able 2.4. The results are consisten t with those found in the main estimation sho wing that married w omen’s lab or force participation increased b y 3:3 p ercen tage p oin ts as a result of the few er w orkw eek hours. The second approac h estimates the follo wing equation restricting the sample of female observ ations only a it = 0 + 1 Married i + 2 Aubry t + 3 Married i Aubry t +X + it (2.4) The eect of in terest is iden tied b y 3 . Estimation results are rep orted in columns 7 to 9 in T able 2.4. The estimated eect using single w omen as comparison group sho w a sligh tly smaller but still signican tly p ositiv e eect of reduced w orkw eek hours on participation of married w omen. The estimated a v erage treatmen t eect is a 2:1 p ercen tage p oin t increase in lab or force participation. 2.5 A Model of Participation Choice Section 4 sho ws that the reduction in house w ork ed created b y the Aubry la w led to an increase in lab or force participation of married w omen compared to single w omen and men. The nding that married w omen are more lik ely to b e activ e in the lab or mark et when they are exp ected to sp end few er hours at the w orkplace suggests that married w omen’s cost of participation increases when the hours of w ork increase at a higher rate than other demographics. In what follo ws, w e dev elop and estimate a discrete c hoice mo del of lab or force participation that pro vides evidence of this phenomenon. An individual i lab or mark et opp ortunit y at time t is c haracterized b y lab or earnings w it and exp ected hours of w ork h e it . The tuple (w it ;h e it ) is dra wn from a 59 join t random distribution f(w;h) and is observ ed b y the agen t b efore a partici- pation decision is made. Lab or mark et participation decision is based on a laten t v ariable y it =w it c it represen ting the net b enet of participating in the mark et whic h is equal to the dierence of lab or earnings and the cost of lab or mark et activit y . Lab or force participation decision is therefore a it = 8 < : 1 if y it 0 0 if y it < 0 Wherea it denote lab or mark et activit y , and tak es v alue 1 if activ e and 0 otherwise. The cost of participation in the lab or mark et dep ends on individual c haracteristics X i , w eekly lab or hours exp ected to w ork h e it and a random term it . c it =c(X i ;h e it ) + it where it N(0; 2 ). P articipation decision a it is observ ed in the data. Ho w ev er, lab or earnings and eectiv e hours w ork ed are observ ed only in cases of emplo ymen t. Solution to problems of this sort ha v e b een prop osed b y Hec kman (1974). The maxim um lik eliho o d problem consists of maximizing the function b elo w. Deriv ation of the maximization problem can b e found in app endix B. N X i=1 a i log(w i c(X i ;h e i ;)jw i ;h i )+ N X i=1 (1a i )log Z 1 0 Z 1 0 [1(w i c(X i ;h e i ;)jw;h)]f(w;h)dwdh Where (:) is the cum ulativ e distribution function of a standard normal distribu- tion and are the parameters of the cost function of in terest. The cost of participation function is assumed to ha v e the follo wing functional form c(X i ;h e it ;) = + 1 h e it + 2 female i h e it + 3 female i married i h e it The marginal cost of participation of exp ected hours of w ork for men is measured b y 1 , as for married w omen it is 1 + 2 + 3 . 60 Moreo v er, assume that (w i ;h i ) N(; ) where = ( w ; h ) is the v ector of means of the distribution of exp ected hours w ork ed and is the co v ariance matrix. Estimation results are rep orted in T able 2.5. The results conrm that ev ery additional hour of w ork increases the participation w ork. Moreo v er hours of w ork are more taxing for w omen ( 2 > 0), esp ecially if they are married ( 3 > 0). 2.6 Conclusion In this pap er, w e in v estigate the eect of reducing time sp en t in the w orkplace on the lab or mark et activit y of married w omen. The Aubry la w c hanged the ex- p ectation on hours to b e sp en t in the lab or mark et. The a v erage n um b er of w eekly hours w ork ed in F rance b y b oth men and w omen ha v e signican tly decreased since the la w to ok eect in 2000. This v ariation in exp ected hours of w ork resulted in an increase in lab or force participation among married w omen. w e then estimated a structural mo del of lab or force participation to iden tify the marginal cost of an additional w orking hour on lik eliho o d of participation. The evidence in this pap er sho ws that when the hours exp ected to supply increase participation b ecome more taxing for w omen than men. 61 T ables T able 2.1: Summary Statistics This table presen ts descriptiv e statistics for lab or mark et outcomes and individual c haracteristics. Data are from the F renc h Lab or F orce Surv ey (LFS), construction of v ariables and sample restrictions are describ ed in the data section in the main text. The a v erages are rep orted separately b y gender for the y ears b efore and after the reduction in w orkw eek hours. Pre-Aubry P ost-Aubry (1990-1999) (2000-2002) Mo dule A: Lab or Mark et Outcomes L ab or F or c e Particip ation W omen 0.534 0.552 Men 0.693 0.684 W e ekly Hours W orke d W omen 35.08 33.89 Men 42.37 40.79 Monthly Salary (in F r ench F r anc) W omen 6,501 7,381 Men 8,939 9,875 Mo dule B: Individual Characteristics A ge W omen 44.02 44.76 Men 43.11 43.98 Marrie d W omen 0.59 0.55 Men 0.61 0.57 University De gr e e W omen 0.15 0.20 Men 0.15 0.19 High Scho ol De gr e e W omen 0.13 0.14 Men 0.11 0.13 Midd le Scho ol De gr e e W omen 0.30 0.31 Men 0.36 0.37 Numb er of Observation 1,278,321 379,210 62 T able 2.2: Changes in Lab or Hours W ork ed This table rep orts estimation results of equation (2.1). The dep endan t v araible is w eekly hours w ork ed, whic h are self-rep orted in the surv ey b y emplo y ed individu- als. Estimation results are rep orted for the full sample and b y gender. F ull Sample W omen Men (1) (2) (3) Post-Aubry -1.49 -1.58 -1.19 (0.07) (0.09) (0.06) Me an Hours W orke d 39.19 35.08 42.37 Num b er of Observ ations 804,881 450,991 353,890 Note: Clustered robust standard errors are rep orted in paren thesis. * p< 0:1; ** p< 0:05; *** p< 0:01. 63 T able 2.3: W eekly Hours W ork ed This table rep orts quan tiles of the distribution of hours w ork ed conditional on emplo ymen t. The distributions are rep orted separately for men and w omen b efore and after the reduction in w orkw eek hours. Quan tile Pre-Aubry P ost-Aubry (1990-1999) (2000-2002) Mo dule A: W omen q 10 20 19 q 25 30 28 q 50 39 35 q 75 39 39 q 90 45 45 Mean 35.08 33.89 Mo dule B: Men q 10 35 35 q 25 39 35 q 50 39 39 q 75 45 45 q 90 60 55 Mean 42.37 40.79 Note: q p denote the pth p ercen tile. 64 T able 2.4: Eect of W orkw eek Hours Reduction on Married W omen’s P articipation This table rep orts estimates of the a v erage eect of w orkw eek hours reduction on married w omen’s lab or force participation. Columns (1) to (3) rep ort A TE estimated from equation (2.2). Columns (4) to (6) rep ort A TE estimated from equation (2.3). Estimates of A TE from equation (2.4) are rep orted in columns (7) to (9). F ull Sample Married Subsample F emale Subsample (1) (2) (3) (4) (5) (6) (7) (8) (9) A TE 0.026 0.036 0.036 0.036 0.033 0.033 -0.017 0.021 0.021 (0.005) (0.004) (0.004) (0.003) (0.003) (0.003) (0.005) (0.004) (0.003) Individual Con trols N Y Y N Y Y N Y Y Y ear and Region F.E. N N Y N N Y N N Y Num b er of Observ ations 1,654,216 975,828 857,336 Note: Clustered robust standard errors are rep orted in paren thesis. * p< 0:1; ** p< 0:05; *** p< 0:01. 65 T able 2.5: Structural Estimates This table rep orts estimates of the structural parameters of the discrete c hoice mo del presen ted in section 5 P arameter V alue 1 Marginal cost of participation of exp ected hours of w ork for men 0:067 2 Incremen tal cost of hours of w ork for w omen 0:017 3 Marriage p enalt y of w omen 0:004 2 w Lab or earnings V ariance 6; 800 2 h W eekly Hours of W ork V ariance 1:13 2 wh Correlation of W eekly Hours W ork ed and Lab or Earnings 1,285 66 Figures Figure 2.1: Changes in Distributions of W eekly Hours W ork ed b y Men and W omen 0 .1 .2 .3 .4 .5 Density 0 10 20 30 40 50 60 70 80 Weekly Hours Worked Pre−Aubry Post−Aubry (a) Distribution of Hours W ork ed b y W omen 0 .1 .2 .3 .4 .5 Density 0 10 20 30 40 50 60 70 80 Weekly Hours Worked Pre−Aubry Post−Aubry (b) Distribution of Hours W ork ed b y Men 67 Figure 2.2: Y early Lab or F orce P articipation of Men and W omen b y Marrital Status 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 Labor Force Participation 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 Year (a) Lab or F orce P articipation of Single W omen 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 Labor Force Participation 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 Year (b) Lab or F orce P articipation of Married W omen 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 Labor Force Participation 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 Year (c) Lab or F orce P articipation of Single Men 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 Labor Force Participation 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 Year (d) Lab or F orce P articipation of Married Men 68 Figure 2.3: Lab or F orce P articipation of Men and W omen b y Age and Marital Status in 1990 and 2002 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 Labor Force Participation 20 25 30 35 40 45 50 55 60 65 70 75 Age 1990 2002 (a) Lab or F orce P articipation of Single W omen 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 Labor Force Participation 20 25 30 35 40 45 50 55 60 65 70 75 Age 1990 2002 (b) Lab or F orce P articipation of Married W omen 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 Labor Force Participation 20 25 30 35 40 45 50 55 60 65 70 75 Age 1990 2002 (c) Lab or F orce P articipation of Single Men 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 Labor Force Participation 20 25 30 35 40 45 50 55 60 65 70 75 Age 1990 2002 (d) Lab or F orce P articipation of Married Men 69 Chapter 3 The Eect of Extended Longevity and Improvement in Health Quality on Career Span 3.1 Introduction The last 200 y ears w ere mark ed b y a signican t increase in life exp ectancy . F rom the time the rst lifetable w as constructed to to da y , life exp ectancy has dou- bled in w estern Europ e. In the United States, life exp ectancy at birth has increased from 39.4 in 1880 to 79.2 in 2015. Japan, who has the highest con temp oraneous life exp ectancy of 83.6 y ears, has kno wn con tin uous increase in life exp ectancy throughout the 20th cen tury , with the exception of a sharp dip during the W orld W ar I I. Gains in life longevit y w ere not restricted to industrialized coun tries; most coun tries in Asia, Africa and South America witnessed increases in life exp ectancy at dieren t rates, with some of these coun tries attaining lev els close to those seen in the US and Europ e. Impro v emen t in public health and biological kno wledge coupled with medical inno v ation led to remark able impro v emen ts in h uman health in the past t w o cen turies. Health impro v emen t w as mark ed b y a b etter qualit y of life and b y a decrease in mortalit y rates at ev ery age 1 . The past cen tury ha v e also b een mark ed b y rapid c hanges in lab or supply b y men and w omen. Banerjee and Blau (2016) do cumen t the v arying trends in lab or supply decisions in the last 50 y ears of the 20th cen tury . One in teresting feature 1 Figures and trends rep orted in this paragraph are tak en from Roser (2019). 70 they note is the c hanging trend of lab or supply b y age in the mid 1980s. They nd that lab or supply decreased for men and w omen at y ounger age and increased at older age. This suggests that, starting in the mid 80s, y ounger birth cohorts ha v e b een dela ying their en try in to the lab or mark et and later birth cohorts ha v e b een p ostp oning their transition to retiremen t. This pap er aims to explore whether these c hanges in lab or supply decisions can b e explained b y the c hanges in life exp ectancy and health qualit y noted ab o v e. In this pap er, I prop ose a theoretical mo del of lifetime utilit y maximization that incorp orates sc ho oling and retiremen t decisions. The agen t maximizes life- time utilit y taking in to accoun t their health status. I deriv e optimalit y conditions and sho w that optimal sc ho oling and retiremen t decisions exist under standard regularit y conditions for the earnings function. I then discuss the p oten tial ef- fects of c hanges in longevit y and life qualit y on optimal agen t c hoices. Sim ulation results sho w that the suggested mo del can rationalize the do cumen ted c hange in lab or supply trends. A large b o dy of literature ev aluates the relationship b et w een health status and dynamic lab or supply . One main line of literature studies the eect of exoge- nous c hanges in life exp ectancy on h uman capital accum ulation (Ja y ac handran and Lleras-Muney, 2009; Oster et al. , 2013; Cerv ellati and Sunde, 2015), nding a p ositiv e eect of extended longevit y on sc ho oling decisions. A large n um b er of pap ers in the macro and structural lab or literature fo cuses on studying the relationship b et w een health and sc ho oling decisions (Kalemli-Oczan et al. , 2000; Kalemli-Oczan, 2003; Soares, 2005) and retiremen t decisions (Galama et al. , 2013; Kuhn et al. , 2015; Garcia-Gomez et al. , 2017). This pap er is closest to the pap ers b y Hazan (2009) and Cerv ellati and Sunde (2013) in it’s attempt to ha v e a global understanding of the eect of c hange in health on sc ho oling and dynamic lab or supply . This pap er is organized as follo ws: section 2 includes the main mo del, the main set up and the conditions to determine optimal y ears of sc ho oling. In sec- tion 3, results from sim ulations are presen ted and follo w ed b y discussion of the eect of c hanges in surviv al probabilit y and health qualit y prole on sc ho oling and retiremen t. The last section concludes. 71 3.2 A Model of Health, Schooling and Retirement This section dev elops a lifetime mo del of utilit y maximization. Agen ts mak e optimal c hoices of consumption, sc ho oling and retiremen t sub ject to v arying health proles. Conditions for existence and uniqueness of the optimal c hoices are pro- vided and discussed. I then discuss ho w y ears of sc ho oling and retiremen t age v aries with the c hanges in health prole. The mo del builds on the w ork in Murph y and T op el (2006), where to t w o dis- tinct t yp e of health measures are considered. The rst, S(t;), a surviv al function whic h measure the probabilit y of b eing aliv e at age t. The second health factor, H(t), measures the qualit y of life at ev ery y ear of age indep enden t of mortal- it y . Health proles of individuals dep end on the state of medical and scien tic tec hnology and en vironmen tal factors that they are exp osed to from the time of birth on w ards, and is not the result of individual decisions or in v estmen ts. Ex- ogenous c hanges in tec hnology and kno wledge o v er time con tribute to v ariation in health proles across dieren t birth cohorts. Medical progress can aect b oth or either one of the health dimensions w e consider. F or example, dev elopmen t of new HIV medication extended life exp ectancy of HIV patien ts and also had substan tial eect on life qualit y b y easing a large n um b er of the virus symptoms. Impro v e- men t in men tal health treatmen t and assistiv e tec hnologies for ph ysical disabilities impro v ed life qualit y of aected individuals without c hanging longevit y . Man y impro v emen ts in the medical eld led to increase in life exp ectancy without neces- sarily aecting health qualit y . One particular example that had substan tial eect on life exp ectancy of w omen is the reduction in maternal mortalit y that w as the result of a series of medical and surgical inno v ations, as w ell as increasing hospital accessibilit y . 3.2.1 Model Setup Consider a represen tativ e agen t for a birth cohort b. A t birth, the agen t learns the health H b (t) and surviv al S(t; b ) proles of their corresp onding birth cohorts. The agen t maximizes lifetime utilit y b y c ho osing consumption for ev ery y ear of agefC t g 1 t=0 , the age at whic h they lea v e sc ho ol t S , and nally the age at whic h 72 they retire from lab or mark et t R . The c hoice of sc ho ol exit age and retiremen t age divide the agen t life in to three distinct phases: sc ho oling, w orking life and retiremen t. Individual utilit y from consumption in ev ery p erio d u(C t ) is increasing and conca v e: u c (:)> 0, u cc (:) 0. An individual suers a disutilit y of lab or L during w orking life, and enjo ys a retiremen t leisure utilit y R. Exp ected lifetime utilit y for an individual b orn in y ear b U b := Z ts 0 u(C(t))H b (t)S(t; b )e t dt + Z t R t S [u(C(t))L]H b (t)S(t; b )e t dt + Z 1 t R [u(C(t)) +R]H b (t)S(t; b )e t dt (3.1) where is a time discoun t factor. The health qualit y factor in teracts with utilit y m ultiplicativ ely . I normalize H(t) to b e in the in terv al [0; 1], where H(t) = 1 for a p erfectly health y p erson, and H(t) = 0 indicates sev ere illness. Health qualit y function b ecomes a health qualit y discoun t factor. b is the hazard rate of mortalit y of birth cohort b. S(t; b ) =exp Z t 0 b (x)dx While in the lab or phase, an individual collects w age earnings w(H(t);t S ;t). W age earning at ev ery y ear of age t dep ends on h uman capital and kno wledge obtained through y ears of sc ho oling t S , and on individual health qualit y H(t). Giv en the empirical literature on returns to sc ho oling, I assume w t S (:) > 0 and w t S t S (:) 0. I further assume that the eect of health qualit y on earnings is increasing w H (:) and con v ex w HH (:) 0. During retiremen t, the agen t collects retiremen t b enets B(t). I assume a p erfect ann uit y mark et with in terest rate r . 73 Assets accum ulation is then describ ed b y the follo wing la w of motion _ A(t) = 8 > > > < > > > : C(t)S(t;) +rA(t); for tt S w(H(t);t S ;t)C(t) S(t;) +rA(t); for t S <tt R B(t)C(t) S(t;) +rA(t); for t R <t (3.2) Moreo v er, I imp ose the initial condition A(0) = 0 and a terminal condition (No- P onzi condition) lim t!1 A(t) = 0 Equation (3.2) with the initial and terminal conditions constitute the lifetime bud- get constrain t that the maximization problem should satisfy . The represen tativ e agen t problem is summarized in the follo wing optimization problem max fCtg 1 t=0 ;t S ;t R U b (C t ;t S ;t R ) s.t. (2) (P) 3.2.2 Optimal Schooling and Retirement Choices Solution of the dynamic con trol problem is pro vided in App endix C. The nec- essary conditions for optimal c hoices of fC (t);t S ;t R g are rep orted and discussed b elo w. Lifetime Consumption Prole The optimal consumption c hoice should satisfy the follo wing equation u c (C(t)) u c (C(t 0 )) = H(t 0 ) H(t) e (r)(tt 0 ) (3.3) This is an augmen ted Euler equation. It diers from the standard Euler equation b y accoun ting for the health qualit y discoun ting of utilit y , whic h en ters the equa- tion through the b et w een p erio ds ratio of health qualit y . The optimal condition is suc h that sp ending on consumption equates the cross p erio d presen t v alue of marginal utilit y of consumption discoun ted b y health qualit y . This condition im- plies that t w o factors determine relativ e consumption: the time discoun t factor and the relativ e health discoun ting factor. The pro duct of b oth these discoun t factors is 74 what determines relativ e consumption in equation (3.3), whic h p oses an empirical c hallenge to iden tifying their eects on consumption smo othing separately . In the basic case where =r , the Eurler equation b ecomes H(t)u c (C(t)) =H(t 0 )u c (C(t 0 )) As a result of dierences in health qualit y b et w een p erio ds, consumption is not p erfectly smo othed. Conca vit y of the utilit y function and the equation ab o v e implies that the optimal consumption path is suc h that higher consumption is allo cated for the y ears where health qualit y is higher. This result is not surprising giv en the earlier in terpretation of the health function as a discoun t factor, making consumption a complemen t to qualit y of health status. The follo wing prop osition giv es a more general result ab out the relationship b et w een health qualit y and consumption sp ending o v er the life cycle. Prop osition 1 If r and he alth deterior ates over lifetime _ H(t) 0, then optimal c onsumption pr ole is de cr e asing over time. Pro of. Let t>t 0 )H(t)H(t 0 ) u c (C(t)) u c (C(t 0 )) = H(t 0 ) H(t) e (r)(tt 0 ) 1 )C(t)C(t 0 ) As noted in Murph y and T op el (2006), this result concurs with studies of life cycle consumption, where consumption decline at older age ha v e b een do cumen ted. Life exp ectancy do es not aect relativ e consumption, as it is ob vious from equation (3.3). Ho w ev er, the follo wing sho ws that surviv al function is a main determinan t of y ears of sc ho oling and retiremen t age. As suc h, surviv al function aects the lifetime budget constrain t and consequen tly has an eect on lifetime consumption lev els. 75 Optimal Sc ho oling and Retiremen t Choices Optimal sc ho oling y ears and retiremen t age ft S ;t R g are determined b y the follo wing system of equations Z t R t S (t) @w @t S S(t;)e (tt S ) dt +LH(t S )S(t S ;) = t S w(H(t S );t S ;t S )S(t S ;) (3.4) t R w(H(t R );t S ;t R )B(t R ) = (L +R)H(t R ) (3.5) The non-linearit y of the system mak es it dicult to deriv e closed form solutions for the optimal v alues of t S and t R , without imp osing restrictiv e functional forms on H(t), S(t;) and w(H(t);t S ;t). Section 3 presen ts sim ulation results to illustrate the eect of c hanges in health qualit y and mortalit y on sc ho oling and retiremen t implied b y the ab o v e system of equations. The remainder of this section pro vides a discussion of the optimalit y conditions and the p oten tial trade-os that arise from exogenous c hanges in the health qualit y prole and surviv al function. As noted in the app endix, equation (3.4) is obtained from equating the curren t v alue of the Hamiltonians of the sc ho oling phase and w orking life phase. While equation (3.5) is obtained b y equating the curren t v alue of the Hamiltonians of the w orking life phase and retiremen t phase. (t) is the adjoin t v ariable of the assets la w of motion, ts is the v alue of the adjoin t v ariable at the time the agen t transition from sc ho oling to w orking life phase and t R is its v alue at transition to retiremen t. The adjoin t v ariable (t) is the marginal v alue of assets and, as sho wn in the app endix, it is con tin uous across the dieren t life phases and dep ends on the dierence b et w een time preference and in terest rate. The system pro vides a clear in terpretation of the agen t’s optimal transition c hoices. The left hand side of equation (3.4) is the marginal b enet of dela ying transition from sc ho oling to w orking life. The rst part of the sum is the presen t v alue of returns to h uman capital, whic h represen t the gains in lifetime earnings from the additional time sp en t in sc ho ol. These returns to sc ho oling are earned o v er the whole w orking life. The second part of the summation is the a v oided disutilit y of lab or obtained b y dela ying transition in to w orking life. The righ t hand side of equation (3.4) is the marginal loss in earnings from dela ying transition 76 in to w orking life. The second equation equates the marginal b enet and loss from c ho osing retiremen t at that p erio d. The left hand side of equation (3.5) is the marginal b enet of lab or earnings net of retiremen t b enets, whic h is what the individual gains from sta ying in the lab or mark et (marginal cost of retiremen t). The righ t hand side of the equation is the marginal utilit y of leisure at retiremen t, whic h is the health qualit y discoun ted sum of leisure obtained from retiremen t and the a v oided disutilit y of lab or. Lemma 2 in the app endix sho ws that the system of equation admits a solution under some regularit y conditions on the earnings function. Conditions 2 and 3 require that p oten tial w age earnings are considerably higher than retiremen t b en- ets early in life and that they later deteriorate and con v erge to 0 to w ards the end of life. The in v erted U shap e age w age prole often assumed to hold in the macro lab or literature satises these conditions. Moreo v er, there is ample empirical evi- dence that w ages decline at older age (Ben-P orath, 1967; Johnson and Neumark, 1996) and that this con tin uous w age decline pla ys an imp ortan t role in transition to retiremen t (Casano v a, 2012; Garcia-Gomez et al. , 2017). Most imp ortan tly , the rst condition in the lemma implies that earnings should b e smo othly increasing in education. Phenomena lik e sheepskin eect that ha v e b een do cumen ted in the education literature (Jaeger and P age, 1996; Belman and Heyw o o d, 1997) could p oten tially violate this assumption. Suc h phenomenon that could lead to jumps in the p oten tial returns to sc ho oling leading to abrupt c hanges in the marginal v alue of sc ho oling, whic h could p oten tially cause discon tin uit y in the left hand side of equation (3.4). Assuming a jump in returns from obtaining higher degrees these could result in the agen t c ho osing to indenitely dela y transition from sc ho oling to lab or force. This is unlik ely to happ en giv en the assumption that assets ha v e to con v erge to 0 and hence cannot remain indenitely negativ e. Moreo v er, with realistic surviv al functions that ha v e a rapid con v ergence to 0 after a certain age an optimal date of sc ho oling should b e attained. 3.2.3 Changes in Health Quality and Mortality In this section, I pro vide an in tuitiv e discussion of the consequences of c hanges in the surviv al function and health qualit y proles on sc ho oling and retiremen t age 77 according to the system of equations ab o v e. The main in terest is to understand ho w increasing life exp ectancy and impro v emen t in health qualit y at ev ery y ear of age can aect the optimal decisions of the agen t. Equation (3.4) can b e rewritten as follo ws Z t R t S (t) @w @t S S(t;) S(t S ;) e (tt S ) dt +LH(t S ) = t S w(H(t S );t S ;t S ) An increase in life exp ectancy at ev ery y ear of age implies that S(t;) is higher at ev ery age t. The v alue of the in tegral on the left hand side of the equation ab o v e increases when the surviv al function increase at ev ery t. Consequen tly , the marginal v alue of h uman capital accum ulation through sc ho oling increases, whic h should p oten tially increase the willingness to in v est additional time in sc ho oling. A dditional y ears of sc ho oling will increase the w age rate at ev ery y ear of age, and equation (3.5) then will imply that the optimal retiremen t age will b e higher, whic h in turn increases the v alue of sc ho oling and leads to an increase in the optimal y ears of sc ho oling. The conca vit y of w age function in h uman capital ensures that the cycle will not go on forev er. The main in tuition of this result is that if the agen t exp ects to liv e longer, they ha v e additional y ears to enjo y the b enets of accum ulated h uman capital whic h mak es sp ending more y ears in the sc ho oling phase more attractiv e. The higher w age earnings obtained from sc ho ol increase the opp ortunit y cost of retiremen t at later age, whic h mak es it optimal to dela y retiremen t. Slo w er deterioration of health qualit y (higher H(t) at ev ery t) increases the option v alue of retiremen t as can b e seen in equation (3.5). Impro v ed health qualit y leads to higher marginal utilit y of leisure at later y ears of age, making an earlier retiremen t less attractiv e as higher leisure enjo ymen t can b e attained at ev ery y ear of age. Moreo v er, this increase in retiremen t age will increase the span of the w orking life, whic h increases the span of returns to sc ho oling, thereb y increasing the marginal b enet of sc ho oling. A dded to that is the complemen tarit y of health qualit y and sc ho oling in the w age function, whic h adds to the b enet of additional h uman capital when qualit y of life is higher. Both these c hannels mak e it more optimal for the agen t to increase their sc ho oling in v estmen t in resp onse to slo w er deterioration of health qualit y . 78 3.3 Simulation Results This section presen ts and discusses sim ulation results to illustrate the predic- tions made ab o v e. T o p erform the sim ulation exercise, I parameterize the functions in the mo del ab o v e. A log utilit y function and a Mincerian earnings equations are c hosen in addition to constan t retiremen t b enets after the age of 65. T able 3.1 summarizes the sim ulation parameterization. T o rst ev aluate the eect of longevit y on sc ho oling and retiremen t, I x a ran- dom dra w of deteriorating health prole H(t)2 [0:7; 1] suc h that H(t+1)H(t). Recall that for the exp onen tial surviv al function life exp ectancy is equal to 1 . Optimal sc ho oling and retiremen t decisions are then computed for dieren t v al- ues of . The results are rep orted in Figure 3.1. The results sho w that as life exp ectancy increases, n um b er of y ears sp en t in sc ho ol increase (Figure 3.1a) and optimal retiremen t age is dela y ed (Figure 3.1b). Ho w ev er, b oth y ears of sc ho ol- ing and retiremen t age increase at a decreasing rate, whic h is mostly due to the conca vit y of the utilit y and earnings functions. T o illustrate the eect of c hange in health deterioration, I replicate the sim ula- tion ab o v e using dieren t ranges of health deterioration. Similar to the sim ulation ab o v e, I x random dra ws of deteriorating health proles H(t)2 [a; 1] and rep eat the sim ulations for dieren t v alues of a =f0:70; 0:80; 0:90; 1g. The results are rep orted in Figure 3.2. It is clear from the results that as health qualit y impro v es, it is optimal for the agen t to sta y longer in sc ho ol and dela y their retiremen t age. 3.4 Conclusion This pap er prop oses a theoretical mo del of lifetime c hoice of consumption, sc ho oling and retiremen t decisions. Theoretical and sim ulation results deriv ed from the mo del sho w that as life exp ectancy at ev ery age increase, it is optimal for a utilit y maximizing agen t to increase the y ears sp en t in sc ho ol and dela y the age at whic h they retire. Similar v ariations in sc ho oling and retiremen t can b e predicted in resp onse to an impro v emen t in health qualit y . Moreo v er, the optimalit y conditions deriv ed from the mo del sho w the underlying mec hanism that con tributes to suc h resp onse. The ndings from this mo del rationalize the 79 observ ed c hanges in emplo ymen t trends observ ed in Banerjee and Blau (2016). The mo del assumes p erfect kno wledge of health and surviv al prole, whic h is highly unlik ely to hold in real life. While individuals are not completely oblivious to their health proles, they are not fully a w are of their exact health status at ev ery age. Better mo deling of health status learning can impro v e the predictions of the mo del. Nev ertheless, the conclusions deriv ed from this basic mo del sho w the imp ortance that health can ha v e on ma jor decisions of lifetime lab or supply . Understanding birth cohort resp onses to structural c hanges in health and life ex- p ectancy is imp ortan t for man y p olicy relev an t issues, including so cial securit y and retiremen t b enets designs. 80 T ables T able 3.1: Sim ulation P arameterization P arameterization In terest rate r = 0:02 Time discoun t factor = 0:05 Utilit y function u(C) =log(C) W age function w(H(t);t s ;t) = (1 +H(t))[30; 000 + 25t s + 900(tt s ) 20(tt s ) 2 ] Retiremen t b enets 25; 000 if t> 65 Health Qualit y Randomly decreasing in [a;b] where a;b2R Surviv al function Exp onen tial with parameter 81 Figures Figure 3.1: Sim ulation Results for Changes in Life Exp ectancy 40 50 60 70 80 90 12 14 16 18 20 22 Life Expectancy Y ears of schooling (a) Sc ho oling Y ears 40 50 60 70 80 90 62 64 66 68 70 72 74 Life Expectancy Retirement Age (b) Retiremen t Age 82 Figure 3.2: Sim ulation Results for Changes in Health Qualit y 40 50 60 70 80 90 12 14 16 18 20 22 Life Expectancy Y ears of schooling (a) Sc ho oling Y ears - a=0.70 40 50 60 70 80 90 62 64 66 68 70 72 74 Life Expectancy Retirement Age (b) Retiremen t Age - a=0.70 40 50 60 70 80 90 12 14 16 18 20 22 Life Expectancy Y ears of schooling (c) Sc ho oling Y ears - a=0.80 40 50 60 70 80 90 62 64 66 68 70 72 74 Life Expectancy Retirement Age (d) Retiremen t Age - a=0.80 40 50 60 70 80 90 12 14 16 18 20 22 Life Expectancy Y ears of schooling (e) Sc ho oling Y ears - a=0.90 40 50 60 70 80 90 62 64 66 68 70 72 74 Life Expectancy Retirement Age (f ) Retiremen t Age - a=0.90 40 50 60 70 80 90 12 14 16 18 20 22 Life Expectancy Y ears of schooling (g) Sc ho oling Y ears - a=1 40 50 60 70 80 90 62 64 66 68 70 72 74 Life Expectancy Retirement Age (h) Retiremen t Age - a=1 83 Appendix A Appendix to Chapter 1 A.1 Data The data a v ailable for the purp ose of this study are the 1960, 1970, 1980, 1990 and 2000 samples of 5% Public Use Micro data Samples (IPUMS Bureau of the Census). In this section I describ e in detail the construction of the main outcome v ariables as w ell as the dieren t sample restrictions imp osed to estimate the eect of access to ab ortion throughout the fertilit y cycle. I rst restrict the fo cus of study to w omen b orn in the US b et w een 1930 and 1955. This restriction guaran tees that at the time of early state lev el legalization of ab ortion in 1970 the age of w omen observ ed extend b et w een 15 and 40, whic h is the assumed range of fertilit y cycle in this study . F urther restrictions are imp osed to estimate the eect on v arious fertilit y and lab or mark et outcomes the details of whic h are pro vided in what follo ws. A.1.1 F ertility Outcomes Iden tication of the eect of ab ortion access on fertilit y outcomes require observ a- tion of these outcomes once they ha v e b een ac hiev ed. Observing outcomes earlier than ac hiev emen t can lead to estimation bias. This is particularly true in the case of cross sectional data suc h as the one used in this pap er. F or instance, if in a sample y ear w e observ e the total n um b er of c hildren b orn to w omen prior to ac hieving their fertilit y cycle, estimation of the eect of ab ortion will lik ely b e o v erstated for the y ounger birth cohorts, since at the time of observ ation total n um b er of c hildren b orn to older w omen is a b etter measuremen t of completed fertilit y as compared to y ounger w omen. The rst outcome of in terest, completed fertilit y , is dened as the total n um b er of c hildren a w oman ga v e birth to during her fertilit y cycle. Hence an accurate measuremen t of completed fertilit y requires the observ ation of total n um b er of c hildren b orn to a w oman after she has concluded her fertilit y phase, whic h is usually around the age of 45. The census rep orts total n um b er of c hildren w omen ga v e birth to b y the time they w ere in terview ed. Giv en that birth cohorts are restricted to w omen b orn b et w een 1930 84 and 1955, the ideal measuremen t of completed fertilit y is the observ ed total n um b er of c hildren in the 2000 sample. Unfortunately this v ariable is not rep orted in that sample y ear, as suc h completed fertilit y is set to equal the total n um b er of c hildren a w oman ga v e birth to observ ed in the 1990 sample. This p oten tially create a measuremen t error for the completed fertilit y observ ation of the y ounger birth cohorts. Kno wing that most w omen giv e birth to their c hildren b y the age of 42, this measuremen t error is most lik ely restricted to w omen in the sample b orn after 1948. The second fertilit y outcome considered is the age at whic h w omen b ecome mother. This v ariable is constructed b y taking the dierence of the age of the eldest c hild and the age of the mother. There are t w o p oten tial c hallenges constructing this v ariable. First, it requires observ ation of w omen after they had giv en birth to at least one c hild. Second, the surv ey rep orts the age of the eldest c hild still living in the household, consequen tly the observ ation should b e at a time where the rst b orn c hild still liv e in the household. T o satisfy these t w o restriction w omen in the sample should b e at least 30 y ears old and not older than 40. The sample therefore includes w omen b orn b et w een 1935 and 1941 and observ ed in 1970, w omen b orn b et w een 1942 and 1948 and observ ed in 1980 and w omen b orn b et w een 1949 and 1955 and observ ed in 1990. I additionally restrict the sample to households where total n um b er of c hildren living in the household is equal to total n um b er of c hildren the mother ga v e birth to. This w as due to some w omen b eing older than 40 at observ ation and hence their eldest c hild migh t ha v e left the household. This restriction p oten tially exclude w omen who had c hildren v ery early in their life. It should also b e noted that the eldest birth cohorts in this restriction (1935-1941) are observ ed prior to legal c hange. I am assuming that for these birth cohort the birth of rst c hild o ccurred prior to ab ortion legalization and hence the legal c hanges had no eect on this particular fertilit y realization. This assumption is reasonable giv en that the a v erage age at birth of a c hild in the sample is 24 1 . Finally a third fertilit y outcome is considered. Restricting the sample to w omen who had t w o c hildren only , I construct birth spacing v ariable b y taking the dierence of the age of the eldest and y oungest c hild. A.1.2 Labor Market Outcomes The sample used in estimation of lab or mark et outcomes is the same as the one used for estimating the eect of ab ortion access on birth timing. There are t w o sets of lab or outcomes I study in this pap er; Lab or supply and lab or earning Outcomes. I use m ultiple lab or supply v ariables in order to study the eect on lab or supply at in tensiv e and extensiv e margins. The rst of these v ariables is lab or force participation status whic h is readily a v ailable in the data. W eekly hours w ork ed are rep orted as a con tin uous v ariable in the data for the 1980 and 1990 sample. Ho w ev er, in the 1970 sample the hours w ork ed are rep orted in in terv al. Hence in constructing the con tin uous measure of hours w ork ed I used the rep orted v alues in the 1980 and 1990 sample and tak e the midp oin t of the rep orted in terv al in the 1970 sample. W omen who are not w orking are assigned a 1 see also Mathews and Hamilton (2002) 85 v alue of zero hours w ork ed. Analyzes of lab or mark et outcomes is conducted using b oth the full sample and the subsample of w orking w omen. The surv ey rep orts nominal v alues of y early lab or earnings. Since I’m stac king three y ears of surv ey that span o v er 20 y ears, nominal w ages in the later samples will b e automatically larger due to ination. T o mak e w age earnings comparable across surv ey y ears, I deate the earnings and express them in 2012 dollars. I construct a w age rate v ariable for w orking w omen b y taking the ratio of y early lab or earnings and hours w ork ed. This v ariable is lik ely to suer from division bias as a result of measuremen t error in the hours w ork ed v ariable, so analyzes of results on this v ariable should b e studied with care. The last lab or mark et outcome v ariable I study is an o ccupation index v ariable rep orted in the IPUMS. The o ccupation index tak es v alues b et w een 0 and 100, with larger v alues indicating o ccupation with higher median earned income. 86 A.2 T ables and Figures A.2.1 Estimation Results by Y ear of Age T able A.1: Cross States Dierences in Completed F ertilit y (Age Y ears) Complete d F ertility F ull Sample Blac k W omen White W omen 15 -0.06 -0.15 0.01 (0.03) (0.05) (0.02) 16 -0.05 -0.03 0.01 (0.02) (0.07) (0.02) 17 -0.04 -0.13 0.02 (0.03) (0.03) (0.02) 18 -0.04 -0.09 0.02 (0.02) (0.03) (0.01) 19 -0.06 0.03 -0.01 (0.02) (0.03) (0.01) 20 -0.05 -0.10 0.01 (0.02) (0.05) (0.02) 21 -0.06 -0.09 0.00 (0.02) (0.08) (0.02) 22 -0.07 -0.13 -0.00 (0.02) (0.03) (0.02) 23 -0.05 -0.29 0.02 (0.02) (0.06) (0.02) 24 -0.05 0.01 0.01 (0.02) (0.08) (0.02) 25 -0.09 -0.36 -0.04 (0.03) (0.04) (0.01) 26 -0.07 -0.22 -0.01 (0.02) (0.10) (0.01) 27 -0.05 -0.30 0.02 (0.02) (0.08) (0.01) 28 -0.06 -0.21 0.01 (0.03) (0.05) (0.03) 29 -0.08 -0.46 -0.02 (0.02) (0.20) (0.02) 30 -0.10 -0.32 -0.01 (0.03) (0.19) (0.02) 31 -0.08 -0.57 0.01 (0.03) (0.08) (0.03) 32 -0.08 -0.44 0.02 (0.03) (0.11) (0.02) 33 -0.06 -0.57 0.03 (0.03) (0.10) (0.02) 34 -0.06 -0.56 0.04 (0.02) (0.08) (0.02) 35 -0.11 -0.83 0.01 (0.03) (0.10) (0.02) 36 -0.13 -0.63 -0.02 (0.04) (0.07) (0.03) 37 -0.107 -0.64 -0.00 (0.04) (0.07) (0.03) 38 -0.13 -0.64 -0.03 (0.03) (0.29) (0.03) 39 -0.08 -0.42 -0.00 (0.03) (0.19) (0.02) 40 -0.14 -0.73 -0.06 (0.05) (0.11) (0.03) Note: All regressions ha v e b een w eigh ted b y p opulation w eigh ts. Clustered robust standard errors are rep orted in paren thesis. * p< 0:1; ** p< 0:05; *** p< 0:01. 87 T able A.2: Dierence-in-Dierences Estimates of the Eect of Ab ortion A ccess on Completed F ertilit y Complete d F ertility F ull Sample Blac k W omen White W omen 15 0.07 0.58 0.06 (0.03) (0.15) (0.04) 16 0.09 0.69 0.06 (0.04) (0.09) (0.03) 17 0.10 0.59 0.07 (0.04) (0.12) (0.03) 18 0.10 0.64 0.06 (0.04) (0.12) (0.04) 19 0.08 0.76 0.04 (0.05) (0.11) (0.04) 20 0.09 0.63 0.05 (0.04) (0.10) (0.04) 21 0.08 0.63 0.05 (0.06) (0.07) (0.05) 22 0.07 0.60 0.04 (0.06) (0.09) (0.05) 23 0.09 0.43 0.07 (0.05) (0.07) (0.05) 24 0.09 0.74 0.06 (0.06) (0.07) (0.05) 25 0.05 0.37 0.01 (0.04) (0.13) (0.03) 26 0.07 0.50 0.04 (0.04) (0.20) (0.04) 27 0.09 0.43 0.07 (0.04) (0.18) (0.04) 28 0.07 0.52 0.06 (0.07) (0.14) (0.05) 29 0.06 0.27 0.03 (0.04) (0.30) (0.04) 30 0.04 0.41 0.04 (0.03) (0.28) (0.03) 31 0.06 0.16 0.06 (0.06) (0.16) (0.05) 32 0.06 0.29 0.07 (0.04) (0.17) (0.03) 33 0.08 0.15 0.08 (0.06) (0.09) (0.04) 34 0.08 0.17 0.09 (0.04) (0.17) (0.03) 35 0.03 -0.10 0.06 (0.03) (0.19) (0.02) 36 0.01 0.09 0.03 (0.02) (0.16) (0.03) 37 0.03 0.09 0.04 (0.03) (0.16) (0.02) 38 0.01 0.08 0.02 (0.05) (0.38) (0.05) 39 0.06 0.31 0.05 (0.03) (0.28) (0.04) Note:* p< 0:1; ** p< 0:05; *** p< 0:01. 88 T able A.3: Cross States Dierences in Births Timing (Age Y ears) A ge at First Child Birth Sp acing F ull Sample Blac k W omen White W omen F ull Sample Blac k W omen White W omen 15 0.98 0.98 1.09 -0.69 -0.82 -0.69 (0.12) (0.15) (0.12) (0.05) (0.13) (0.04) 16 0.98 0.46 1.15 -0.53 -0.27 -0.58 (0.10) (0.12) (0.09) (0.03) (0.47) (0.03) 17 0.99 1.06 1.08 -0.56 -0.61 -0.60 (0.06) (0.21) (0.04) (0.04) (0.10) (0.04) 18 0.94 0.30 1.07 -0.61 -0.63 -0.63 (0.08) (0.08) (0.05) (0.05) (0.18) (0.04) 19 0.99 0.31 1.13 -0.63 -0.96 -0.63 (0.14) (0.09) (0.14) (0.06) (0.24) (0.05) 20 0.81 0.21 0.97 -0.59 -0.40 -0.62 (0.21) (0.14) (0.24) (0.04) (0.28) (0.04) 21 0.66 0.10 0.81 -0.63 0.22 -0.67 (0.22) (0.32) (0.21) (0.06) (0.16) (0.04) 22 0.51 0.08 0.64 -0.55 -0.31 -0.62 (0.05) (0.07) (0.03) (0.02) (0.10) (0.03) 23 0.63 -0.12 0.78 -0.53 -0.31 -0.58 (0.05) (0.29) (0.03) (0.04) (0.15) (0.05) 24 0.61 0.12 0.75 -0.55 -0.92 -0.57 (0.04) (0.15) (0.03) (0.04) (0.06) (0.05) 25 0.62 -0.14 0.79 -0.53 -0.26 -0.57 (0.04) (0.25) (0.03) (0.02) (0.19) (0.03) 26 0.47 0.20 0.65 -0.66 -0.84 -0.70 (0.05) (0.19) (0.06) (0.03) (0.06) (0.03) 27 0.54 0.02 0.73 -0.58 -0.43 -0.63 (0.07) (0.19) (0.07) (0.03) (0.14) (0.02) 28 0.35 -0.12 0.54 -0.69 -0.91 -0.73 (0.05) (0.07) (0.05) (0.02) (0.22) (0.02) 29 0.39 0.24 0.54 -0.48 -0.13 -0.54 (0.06) (0.11) (0.07) (0.06) (0.11) (0.08) 30 0.35 -0.19 0.53 -0.48 -0.65 -0.52 (0.06) (0.09) (0.05) (0.04) (0.34) (0.04) 31 0.42 0.07 0.60 -0.61 -0.29 -0.68 (0.08) (0.30) (0.09) (0.05) (0.31) (0.05) 32 0.34 -0.10 0.50 -0.53 -0.81 -0.55 (0.13) (0.13) (0.15) (0.07) (0.21) (0.09) 33 0.49 0.18 0.66 -0.73 -1.14 -0.77 (0.08) (0.19) (0.08) (0.09) (0.22) (0.08) 34 0.56 0.69 0.66 -0.56 -0.94 -0.58 (0.07) (0.15) (0.05) (0.05) (0.14) (0.04) 35 0.32 -0.02 0.48 -0.46 0.24 -0.54 (0.05) (0.15) (0.05) (0.11) (0.25) (0.10) Note: All regressions ha v e b een w eigh ted b y p opulation w eigh ts. Clustered robust standard errors are rep orted in paren thesis. * p< 0:1; ** p< 0:05; *** p< 0:01. 89 T able A.4: Dierence-in-Dierences Estimates of the Eect of Ab ortion A ccess on Births Timings A ge at First Child Birth Sp acing F ull Sample Blac k W omen White W omen F ull Sample Blac k W omen White W omen 15 0.65 1.01 0.61 -0.23 -1.07 -0.15 (0.13) (0.20) (0.10) (0.14) (0.30) (0.12) 16 0.66 0.49 0.67 -0.06 -0.52 -0.03 (0.10) (0.15) (0.08) (0.13) (0.55) (0.09) 17 0.66 1.09 0.60 -0.10 -0.86 -0.05 (0.07) (0.30) (0.07) (0.11) (0.29) (0.10) 18 0.62 0.32 0.58 -0.15 -0.88 -0.09 (0.08) (0.18) (0.08) (0.11) (0.31) (0.11) 19 0.67 0.33 0.65 -0.17 -1.21 -0.09 (0.15) (0.18) (0.18) (0.10) (0.35) (0.10) 20 0.49 0.23 0.49 -0.13 -0.65 -0.08 (0.21) (0.20) (0.27) (0.11) (0.39) (0.09) 21 0.34 0.12 0.33 -0.16 -0.03 -0.12 (0.22) (0.41) (0.24) (0.10) (0.35) (0.09) 22 0.19 0.10 0.16 -0.08 -0.55 -0.08 (0.07) (0.16) (0.08) (0.13) (0.26) (0.12) 23 0.31 -0.10 0.30 -0.07 -0.56 -0.04 (0.08) (0.35) (0.07) (0.13) (0.29) (0.12) 24 0.29 0.14 0.27 -0.08 -1.17 -0.03 (0.07) (0.24) (0.07) (0.14) (0.26) (0.13) 25 0.30 -0.12 0.31 -0.07 -0.51 -0.03 (0.07) (0.27) (0.07) (0.12) (0.31) (0.12) 26 0.14 0.22 0.17 -0.19 -1.09 -0.15 (0.09) (0.19) (0.10) (0.13) (0.26) (0.12) 27 0.21 0.04 0.25 -0.12 -0.68 -0.08 (0.10) (0.23) (0.12) (0.10) (0.28) (0.10) 28 0.03 -0.10 0.07 -0.23 -1.15 -0.19 (0.07) (0.16) (0.08) (0.12) (0.36) (0.11) 29 0.06 0.26 0.06 -0.01 -0.11 0.00 (0.08) 0.16) (0.06) (0.15) (0.28) (0.15) 30 0.02 -0.17 0.05 -0.01 -0.89 0.02 (0.09) (0.18) (0.09) (0.12) (0.37) (0.13) 31 0.09 0.09 0.12 -0.15 -0.54 -0.14 (0.10) (0.33) (0.08) (0.13) (0.44) (0.12) 32 0.02 -0.07 0.02 -0.07 -1.06 -0.01 (0.14) (0.19) (0.13) (0.08) (0.34) (0.07) 33 0.16 0.21 0.18 -0.27 -1.38 -0.22 (0.09) (0.23) (0.07) (0.06) (0.27) (0.05) 34 0.24 0.71 0.18 -0.10 -1.19 -0.05 (0.09) (0.21) (0.06) (0.13) (0.32) (0.10) Note: All regressions ha v e b een w eigh ted b y p opulation w eigh ts. Clustered robust standard errors are rep orted in paren thesis. * p< 0:1; ** p< 0:05; *** p< 0:01. 90 Figure A.1: Eect of A ccess to Ab ortion on Completed F ertilit y −.2 −.1 0 .1 .2 Difference in Completed Fertility 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 Age at Access to Abortion (a) Completed F ertilit y T rend b y Y ear of Age 0.07 0.09 0.10 0.10 0.08 0.09 0.08 0.07 0.090.09 0.05 0.07 0.09 0.07 0.06 0.04 0.060.06 0.080.08 0.03 0.01 0.03 0.01 0.06 −0.20 −0.10 0.00 0.10 0.20 Diff−in−Diff Estimates 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 Age at Access to Abortion (b) Estimates of the eect of ab ortion access on Completed F ertilit y - Y ear of Age 91 Figure A.2: Eect of A ccess to Ab ortion on Age at Start of Motherho o d −1.5 −1 −.5 0 .5 1 1.5 Age at First Child 0 5 10 15 20 Age at Access to Abortion (a) Age at Birth of First Child T rend b y Y ear of Age 0.65 0.66 0.66 0.62 0.67 0.49 0.34 0.19 0.31 0.29 0.30 0.14 0.21 0.03 0.06 0.02 0.09 0.02 0.16 0.24 −1.00 −0.50 0.00 0.50 1.00 Diff−in−Diff Estimates 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 Age at Access to Abortion (b) Estimates of the eect of ab ortion access on Age at Birth of First Child - Y ear of Age 92 Figure A.3: Eect of A ccess to Ab ortion on Birth Spacing −1 −.5 0 .5 1 Difference in Birth Spacing 0 5 10 15 20 Age at Access to Abortion (a) Birth Spacing T rend b y Y ear of Age −0.23 −0.06 −0.10 −0.15 −0.17 −0.13 −0.16 −0.08 −0.07 −0.08 −0.07 −0.19 −0.12 −0.23 −0.01−0.01 −0.15 −0.07 −0.27 −0.10 −1.00 −0.50 0.00 0.50 1.00 Diff−in−Diff Estimates 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 Age at Access to Abortion (b) Estimates of the eect of ab ortion access on Births Spacing - Y ear of Age 93 A.2.2 Placebo T est: Ruling Out Changes in F ertility T rends I extend the sample to include w omen b orn b et w een 1956 and 1958. F or these birth cohorts there is no dierence in exp osure to ab ortion as they w ere b et w een the ages of 12 and 14 at the time of state legalization, and they all receiv e equal access to ab ortion at the b eginning of their fertilit y cycle whic h happ ens p ost Ro e v. W ade. The plots b elo w rep ort the dierence-in-dierences estimates obtained b y estimating equation (1.2) for the extended sample. Figure A.4: Eect of Ab ortion A ccess on Age of Motherho o d 0.17 0.52 0.46 0.14 0.11 −0.04 −0.09 −1.00 −0.50 0.00 0.50 1.00 Diff−in−Diff Estimates 12−14 15−17 18−20 21−23 24−26 27−29 30−32 Age at Access to Abortion (a) F ull Sample 0.05 0.54 −0.02 −0.28 −0.24 −0.21 −0.37 −1.00 −0.50 0.00 0.50 1.00 Diff−in−Diff Estimates 12−14 15−17 18−20 21−23 24−26 27−29 30−32 Age at Access to Abortion (b) Blac k W omen 0.23 0.50 0.45 0.14 0.13 −0.00 −0.06 −1.00 −0.50 0.00 0.50 1.00 Diff−in−Diff Estimates 12−14 15−17 18−20 21−23 24−26 27−29 30−32 Age at Access to Abortion (c) White W omen 94 A.3 Robustness to Potential Migration A.3.1 Estimation Using Potential Non-Movers In addition to state of birth, the IPUMS include the state of residency of w omen at the time of observ ation. This additional information do es not pro vide full information ab out the state of residency of w omen during their fertilit y cycle. If a w oman at the time of observ ation is observ ed in the same state as her birth state then she is assumed to ha v e b een living in her birth state during her fertilit y cycle. I then estimate equation (1.2) using the subsample of p oten tial non-mo v ers. Estimation results for the three main fertilit y outcomes are rep orted b elo w. Comparing these results with the results found in the main estimation, I nd that the direction of the eects are preserv ed with sligh t c hange in the magnitude and precision of the estimates. 95 T able A.5: Cross States Dierences in Completed F ertilit y (P oten tial Non-Mo v ers) Mean 1 2 3 4 5 6 7 8 9 [15-17] [18-20] [21-23] [24-26] [27-29] [30-32] [33-35] [36-38] [39-40] F ull Sample Complete d F ertility 2.45 -0.04 -0.06 -0.07 -0.10 -0.08 -0.09 -0.09 -0.13 -0.10 (0.03) (0.02) (0.02) (0.02) (0.02) (0.03) (0.03) (0.04) (0.06) N=1,082,124 Blac k W omen Complete d F ertility 2.91 -0.52 -0.47 -0.60 -0.71 -0.87 -1.01 -1.23 -1.29 -1.19 (0.04) (0.03) (0.03) (0.03) (0.09) (0.11) (0.06) (0.10) (0.07) N=109,018 White W omen Complete d F ertility 2.38 0.05 0.03 0.02 -0.02 0.02 0.03 0.05 0.01 0.01 (0.02) (0.02) (0.02) (0.02) (0.02) (0.02) (0.02) (0.03) (0.04) N=941,926 96 T able A.6: Dierence-in-Dierences Estimates of the Eect of Ab ortion A ccess on Completed F ertilit y (P oten tial Non-Mo v ers) Mean 1 2 3 4 5 6 7 8 [15-17] [18-20] [21-23] [24-26] [27-29] [30-32] [33-35] [36-38] F ull Sample Complete d F ertility 2.45 0.05 0.04 0.02 -0.00 0.02 0.01 0.01 -0.03 (0.05) (0.06) (0.06) (0.07) (0.05) (0.05) (0.04) (0.02) Blac k W omen Complete d F ertility 2.91 0.68 0.73 0.60 0.49 0.32 0.18 -0.04 -0.09 (0.10) (0.09) (0.08) (0.07) (0.10) (0.11) (0.06) (0.09) White W omen Complete d F ertility 2.38 0.04 0.01 0.01 -0.03 0.01 0.02 0.04 0.00 (0.04) (0.05) (0.05) (0.06) (0.04) (0.04) (0.02) (0.01) 97 T able A.7: Cross States Dierences in Births Timing (P oten tial Non-Mo v ers) Mean 1 2 3 4 5 6 7 [15-17] [18-20] [21-23] [24-26] [27-29] [30-32] [33-35] F ull Sample A ge at First Child 23.72 1.20 1.10 0.72 0.69 0.53 0.46 0.63 (0.07) (0.21) (0.07) (0.03) (0.06) (0.12) (0.11) N=508,242 Birth Sp acing 3.76 -0.71 -0.72 -0.65 -0.65 -0.67 -0.66 -0.79 (0.04) (0.05) (0.02) (0.03) (0.04) (0.03) (0.07) N=223,340 Blac k W omen A ge at First Child 22.08 1.38 0.53 0.46 0.54 0.47 0.47 0.94 (0.08) (0.21) (0.14) (0.13) (0.16) (0.15) (0.14) N=49,188 Birth Sp acing 4.71 -0.62 -0.66 0.01 -0.65 -0.16 -0.55 -0.61 (0.14) (0.07) (0.11) (0.16) (0.11) (0.12) (0.15) N=16,516 White W omen A ge at First Child 23.90 1.26 1.20 0.78 0.77 0.61 0.53 0.67 (0.07) (0.23) (0.06) (0.04) (0.06) (0.14) (0.12) N=448,915 Birth Sp acing 3.67 -0.73 -0.74 -0.71 -0.68 -0.72 -0.69 -0.82 (0.03) (0.04) (0.03) (0.04) (0.04) (0.02) (0.08) N=202,959 98 T able A.8: Dierence-in-Dierences Estimates of the Eect of Ab ortion A ccess on Births Timing (P oten tial Non-Mo v ers) Mean 1 2 3 4 5 6 [15-17] [18-20] [21-23] [24-26] [27-29] [30-32] F ull Sample A ge at First Child 23.72 0.57 0.47 0.09 0.06 -0.10 -0.17 (0.15) (0.32) (0.18) (0.13) (0.06) (0.04) Birth Sp acing 3.76 0.08 0.07 0.14 0.13 0.11 0.13 (0.08) (0.07) (0.09) (0.09) (0.10) (0.08) Blac k W omen A ge at First Child 22.08 0.44 -0.41 -0.48 -0.40 -0.47 -0.47 (0.13) (0.33) (0.26) (0.12) (0.10) (0.09) Birth Sp acing 4.71 -0.01 -0.05 0.62 -0.04 0.45 0.06 (0.24) (0.18) (0.20) (0.20) (0.20) (0.20) White W omen A ge at First Child 23.90 0.59 0.53 0.11 0.10 -0.06 -0.14 (0.18) (0.35) (0.18) (0.16) (0.07) (0.04) Birth Sp acing 3.67 0.09 0.09 0.11 0.14 0.10 0.13 (0.08) (0.07) (0.10) (0.11) (0.12) (0.08) 99 A.3.2 Evidence on Selective migration F urthermore I test if there w as an y selectiv e migration among w omen of dieren t age group to and from rep eal states. The 1970 sample of the IPUMS rep ort the state of residency in 1970 as w ell as the state of residency in 1965. This allo w me to construct t w o migration dumm y v ariables. The rst v ariable M R i(b) tak es v alue 1 in case individual i of birth cohort b migrated from a non-rep eal to a rep eal state and 0 otherwise. The second migration v ariable M NR i(b) tak es v alue 1 if individual i of birth cohort b migrated from a rep eal to a non-rep eal and 0 otherwise. If there is an y selectiv e migration conditional on the state ab ortion legalit y status, this migration should not aect men. Therefore I estimate the follo wing equation for eac h of the migration outcomes M i(b) = 1955 X b=1930 b BC i(b) + 1955 X b=1930 b Female i BC i(b) whereBC i is a birth cohort indicator andFemale i is a gender indicator of individuali. In case of p ositiv e selectiv e migration among w omen of certain birth cohorts, co ecien ts of these cohorts should b e p ositiv e and signican t. Estimation results are rep orted in the table b elo w for b oth migration outcomes. No serious selectiv e migration is detected. Except for the 1946 to 1951 birth cohorts w ere female w ere found to b e signican tly less lik ely to migrate from non-rep eal to rep eal states. These birth cohorts are of college attendance age in 1970, this suggest that men from non-rep eal state w ere m uc h more lik ely to attend colleges in rep eal states compared to their state birth cohort p eer w omen. 100 T able A.9: Evidence on Selectiv e Migration Birth Cohort Migr ation to R ep e al States Migr ation to Non-R ep e al States 1955 -0.000 -0.001 (0.001) (0.001) 1954 -0.001 -0.002 (0.001) (0.001) 1953 0.000 -0.002 (0.001) (0.001) 1952 0.000 0.001 (0.002) (0.001) 1951 -0.007 -0.002 (0.002) (0.001) 1950 -0.018 -0.008 (0.004) (0.009) 1949 -0.026 -0.015 (0.005) (0.013) 1948 -0.020 -0.015 (0.010) (0.010) 1947 -0.011 -0.012 (0.003) (0.007) 1946 -0.005 -0.010 (0.003) (0.006) 1945 -0.002 -0.011 (0.003) (0.009) 1944 -0.000 -0.013 (0.003) (0.008) 1943 -0.011 -0.014 (0.003) (0.008) 1942 -0.007 -0.003 (0.003) (0.004) 1941 -0.005 -0.008 (0.003) (0.004) 1940 -0.011 -0.004 (0.003) (0.005) 1939 -0.006 -0.003 (0.003) (0.001) 1938 -0.007 -0.004 (0.003) (0.003) 1937 -0.003 -0.002 (0.002) (0.001) 1936 -0.005 -0.004 (0.002) (0.002) 1935 -0.004 -0.009 (0.002) (0.005) 1934 -0.005 -0.009 (0.002) (0.005) 1933 -0.005 -0.004 (0.002) (0.004) 1932 -0.003 -0.010 (0.002) (0.005) 1931 -0.006 -0.008 (0.002) (0.004) 1930 -0.005 -0.009 (0.002) (0.004) 101 Appendix B Appendix to Chapter 2 B.1 Likelihood F unction The lik eliho o d function for the problem prop osed in section 5 can b e written as follo ws L = N Y i=1 Pr(a i = 1;w i =w;h e i =h) a i Pr(a i = 0;w i =w;h e i =h) 1a i The log-lik eliho o d is therefore l = N X i=1 a i log(Pr(a i = 1;w i =w;h e i =h)) + N X i=1 (1a i )log(Pr(a i = 0;w i =w;h e i =h)) It is straigh t forw ard to see that Pr(a i = 1jw i =w;h e i =h) =Pr(w i C(X i ;h e i ) i 0jw i ;h e i ) =Pr( i w i C(X i ;h e i )jw i ;h e i ) = (w i C(X i ;h e i )jw i ;h e i ) Therefore Pr(a i = 0jw i =w;h e i =h) = 1 (w i C(X i ;h e i )jw i ;h e i ) Giv en that the w age and hours w ork ed are observ ed in the case where the individual participates in the lab or mark et, maximizing the lik eliho o d function ab o v e is equiv alen t to maximizing N X i=1 a i log(w i c(X i ;h i )jw i ;h i )+ N X i=1 (1a i )log Z 1 0 Z 1 0 [1(w i c(X i ;h e i )jw;h)]f(w;h)dwdh Where the la w of conditional probabilit y ha v e b een applied to replace join t probabilities b y conditional probabilities. 102 Appendix C Appendix to Chapter 3 C.1 Solution to the dynamic problem Optimization problem (P) is a m ulti-stage optimal con trol problem with endogenous switc hing p oin ts. F orm ulation of this t yp e of problem and necessary conditions for the solution w ere outlined in T omiy ama (1985) and T omiy ama and Rossana (1989). An extension of the result to the innite time horizon case is pro vided in Grass et al. (2008). The curren t v alue Hamiltonians for the dieren t life p erio ds are: H 1 =u(C(t))H(t)S(t;) + 1 (t) C(t)S(t;) +rA(t) H 2 =fu(C(t))LgH(t)S(t;) + 2 (t) fw(H(t);t s ;t)C(t)gS(t;) +rA(t) H 3 =fu(C(t)) +RgH(t)S(t;) + 3 (t) fB(t)C(t)gS(t;) +rA(t) where i i = 1; 2; 3 are the adjoin t v ariables relating to the assets la w of motion. Optimal consumption in ev ery life p erio d is determined b y the follo wing conditions H(t)u c (C(t)) = i i = 1; 2; 3 (A1) The adjoin t v ariables should satisfy the follo wing conditions at the optim um _ i = i r i i = 1; 2; 3 (A2) The solution to these dieren tial equation is i (t) = i e (r)t i = 1; 2; 3 103 Optimal switc hing time should satisfy the follo wing matc hing conditions for the adjoin t v ariables and the Hamiltonian 1 (t S ) = 2 (t S ) := t S 2 (t R ) = 3 (t R ) := t R H 1 (t S ) + Z t R t S e (tt S ) @H 2 @t S dt =H 2 (t S ) H 2 (t R ) =H 3 (t R ) Giv en the adjoin t v ariables matc hing conditions and the adjoin t v ariables equations w e infer 1 = 2 = 3 =. Whic h imply that the adjoin t v ariable can b e represen ted b y a con tin uous function o v er all life phases: (t) =e (r)t The necessary condition for optimal consumption prole can then b e written H(t)u c (C(t)) =e (r)t 8t2 [0;1) The marginal rate of substitution b et w een an y t w o p erio d t and t 0 is: H(t)u c (C(t)) H(t 0 )u c (C(t 0 )) =e (r)(tt 0 ) Applying the matc hing conditions to the curren t v alue Hamiltonians dened ab o v e, w e obtain the follo wing system of equations that determines the optimal age of transition from sc ho oling to lab or mark et and optimal age of retiremen t. Z t R t S (t) @w @t S S(t;)e (tt S ) dt +LH(t S )S(t S ;) = t S w(H(t S );t S ;t S )S(t S ;) (A3) t R w(H(t R );t S ;t R )B(t R ) = (L +R)H(t R ) (A4) The deriv ation ab o v e pro vides necessary condition for the optimal solution of problem (P). F ollo wing Kuhn et al. (2015) w e pro vide suciency results in t w o steps. Lemma A1 Given arbitr ary switching times ft S ;t R g. The c onsumption pr ole C (t) that satisfy the F.O.C (A1) and (A2), c onstitutes an optimal solution of (P) Pro of. Suciency requires that the Hamiltonians and salv age v alue functions are con- ca v e in C(t) and A(t) (Grass et al. , 2008) 1 . Conca vit y of the Hamiltonians in A(t) is ob vious since it en ters all of them linearly . Moreo v er, conca vit y of the utilit y function implies conca vit y of the Hamiltonians in C(t). The salv age v alue of the rst phase equals 1 Salv age v alue is dened in p.106, the statemen t and the pro of of the suciency condition can b e found in Theorem 3.29 p.120. 104 the sum of the v alue functions of the second and third phase. The salv age v alue of the second phase is the v alue function of the third phase; w e resp ectiv ely denote these salv age functions V 1 (A(t s );t S ;t R ) and V 2 (A(t R );t R ). Linearit y of the v alue functions in A giv e us @ 2 V 1 @A 2 = @ 2 V 2 @A 2 = 0 Lemma A2 The system of e quations (A3) and (A4) has an interior solution if 1. @w @ts > 0 andj @w @ts jM for some M > 0 2. w(H(0); 0; 0)B(0)> L+R Uc(C(0)) 3. lim t!1 w(H(t);t s ;t) = 0 Pro of. Giv en the optimal consumption path fC (t)gthe system of equation can b e written as Z t R t S H(t) H(t S ) U c (C (t)) @w @t S S(t;) S(t S ;) e (tt S ) dt +L =U c (C(t S ))w(H(t S );t S ;t S ) (A3’) U c (C(t R )) w(H(t R );t S ;t R )B(t R ) =L +R (A4’) Con tin uit y of health, surviv al, w age and utilit y function in addition to the b oundedness of @w @ts imply that the in tegral on the left hand side of (A3’) is w ell dened. Hence, there exists a function (t) suc h that (t) = Z t 0 H() H(t S ) U c (C ()) @w @t S S(;) S(t S ;) e (t S ) d The in tegral is w ell dened for ev ery t2 [0;1), hence (t) is surjectiv e. Moreo v er (t) is increasing, since _ (t) = H(t) H(t S ) U c (C (t)) @w @t S S(t;) S(t S ;) e (tt S ) > 0 Hence (t) is in v ertible. (A3’) can then b e written (t R ) (t S ) =U c (t S )w(H(t S );t S ;t S )L )t R = 1 f(t S ) +U c (C(t S ))w(H(t S );t S ;t S )Lg := (t S ) (t S ) is a one to one mapping from t S on to t R , meaning that for ev ery giv en t S there exists a unique t R that w ould solv e equation (A3’). Equation (A4’) b ecomes U c (C((t S ))[w(h((t S ));t S ; (t S ))B((t S ))] =L +R The left hand side of the equation is a con tin uous function of t S . Giv en assumption 2, this function is larger than L +R at the b eginning of the planning horizon, while assumption 3 implies that the inequalit y is in v erted later in the agen t life. This guaran tees the 105 existence of at least one t S for whic h the equalit y is satised, to whic h w e could asso ciate a uniquet R = (t S ). Hence, there exists at least one tuple (t S ;t R ) that solv e the system of equations (A3’) and (A4’) 106 Bibliography Aaronson, D. , Dehejia, R. , Jordan, A. , Pop-Eleches, C. , Samii, C. and Schulze, K. (2018). The eect of fertilit y on mothers’ lab or supply o v er the last t w o cen turies, NBER W orking P ap er No 23717. Anana t, E. O. , Gruber, J. , Levine, P. B. and St aiger, D. (2009). Ab ortion and selection. The R eview of Ec onomics and Statistics , 91 (1), 124136. Angrist, J. D. and Ev ans, W. N. (1998). Children and their paren ts’ lab or supply: Evidence from exogenous v ariation in family size. The A meric an Ec onomic R eview , 88 (3), 450477. and (2000). Sc holing and lab or mark et consequences of the 1970 state ab ortion reforms. R ese ar ch in L ab or Ec onomics , 18, 75113. , Imbens, G. W. and Rubin, D. B. (1996). Iden tication of causal eects using instrumen tal v ariables. Journal of the A meric an Statistic al Asso ciation , 91 (434), 444 455. Askenazy, P. (2013). W orking time regulation in france from 1996 to 2002. Cambridge Journal of Ec onomics , 37 (2), 323347. Banerjee, S. and Blau, D. (2016). Emplo ymen t trends b y age in the united states. wh y are older w ork ers dieren t? The Journal of Human R esour c es , 51 (1), 163199. Bearak, J. , Burke, K. L. and Jones, R. K. (2017). Disparities and c hange o v er time in distance w omen w ould need to tra v el to ha v e an ab ortion in the usa: a spatial analysis. L anc et Public He alth , 2, 493500. Becker, G. S. (1993). A T r e atise on the F amily , Harv ard Univ ersit y Press, c hap. 2, pp. 3079. Belman, D. and Heywood, J. S. (1997). Sheepskin eects b y cohort: Implication of job matc hing in a signaling mo del. Oxfor d Ec onomic Pap ers , 49 (4), 623637. Ben-Pora th, Y. (1967). The pro duction of h uman capital and the life cycle of earnings. Journal of Politic al Ec onomy , 75 (4), 352365. 107 Blundell, R. , Dias, M. C. , Meghir, C. and Sha w, J. (2016). F emale lab or supply , h uman capital, and w elfare reform. Ec onometric a , 84 (5), 17051753. Bronars, S. G. and Grogger, J. (1994). The economic consequences of un w ed moth- erho o d: Using t wins as a natural exp erimen t. A meric an Ec onomic R eview , 84 (5), 11411156. Casanov a, M. (2012). W age and earnings proles at older ages, w orking P ap er. Cer vella ti, M. and Sunde, U. (2013). Life exp ectancy , sc ho oling, and lifetime lab or supply: Theory and evidence revisited. Ec onometric a , 81 (5), 20552086. and (2015). The eect of life exp ectancy on education and p opulation dynamics. Empiric al Ec onomics , 48, 14451478. Chung, Y. , Downs, B. , Sandler, D. H. and Sienkiewicz, R. (2017). The paren tal gender earnings gap in the united states, cen ter of Economic Studies (CES) Researc h P ap ers. Clarke, D. and Muhlrad, H. (2018). Ab ortion la ws and w omen’s health, iZA Discus- sion P ap er Series No. 11890. Cor tes, P. and P an, J. (2013). Outsourcing household pro duction: F oreign domestic w ork ers and nativ e lab or supply in hong k ong. Journal of L ab or Ec onomics , 31 (2), 327371. and (2017). When time binds: Returns to w orking long hours and the gender w age gap among the highly skilled, w orking P ap er. and Tessada, J. (2011). Lo w skilled immigration and the lab or supply of highly skilled w omen. A meric an Ec onomic Journal: Applie d Ec onomics , 3 (3), 88123. Cunningham, S. , Lindo, J. M. , Myers, C. and Schlosser, A. (2018). Ho w far is to o far? new evidence on ab ortion clinic closures, access, and ab ortions, NBER W orking P ap er No. 23366. Donohue, J. J. and Levitt, S. D. (2001). The impact of legalized ab ortion on crime. The Quarterly Journal of Ec onomics , 116 (2), 379420. Duflo, E. (2001). Sc ho oling and lab or mark et consequences of sc ho ol construction in indonesia: Evidence from an un usual p olicy exp erimen t. A meric an Ec onomic R eview , 91 (4), 795813. Finer, L. B. and Zolna, M. R. (2016). Decline in unin tended pregnancy in the united states, 2008-2011. The New England journal of Me dicine , 374 (9), 843852. 108 Fischer, S. , Royer, H. and White, C. (2018). The impacts of reduced access to ab or- tion and family planning services on ab ortions, births, and con traceptiv e purc hases, NBER W orking P ap er No. 23634. Galama, T. , Kapteyn, A. , Fonseca, R. and Michaud, P.-C. (2013). A health pro duction mo del with endogenous retiremen t. He alth Ec onomics , 22 (8), 883902. Garcia-Gomez, P. , Galama, T. , Doorslaer, E. V. and Lopez-Nicholas, A. (2017). In teraction b et w een nancial incen tiv es and health in the early retiremen t decision, hCEO W orking P ap er Series. Goldin, C. (2014). A grand gender con v ergence: Its last c hapter. A meric an Ec onomic R eview , 104 (4), 130. Goux, D. , Maurin, E. and Petrongolo, B. (2014). W orktime regulations and sp ousal lab or supply . The A meric an Ec onomic R eview , 104 (1), 252276. Grass, D. , Caulkins, J. , Feichtinger, G. , Tragler, G. and Behrens, D. (2008). Optimal Contr ol of Nonline ar Pr o c esses: With Applic ation in Drugs, Corruption, and T err or. Springer. Grossman, D. , White, K. , Hopkins, K. and Potter, J. E. (2017). Change in dis- tance to nearest facilit y and ab ortion in texas, 2012 to 2008-2014. Journal of A meric an Me dic al Asso ciation , 317 (4), 437439. Gruber, J. , Levine, P. and St aiger, D. (1999). Ab ortion legalization and c hild living circumstances: Who is the marginal c hild? The Quarterly Journal of Ec onomics , 114 (1), 263291. Hazan, M. (2009). Longevit y and lifetime lab or supply: Evidence and implications. Ec onometric a , 77 (6), 18291863. Heckman, J. (1974). Shado w prices, mark et w ages, and lab or supply . Ec onometric a , 42 (4), 679694. Hensha w, S. K. (1998). Unin tended pregnancy in the united states. F amily Planning Persp e ctives , 30 (1), 2429. Herr, J. L. (2016). Measuring the eect of the timing of rst birth on w ages. Journal of Population Ec onomics , 29, 3972. Hotz, J. , McElroy, S. W. and Sanders, S. G. (2005). T eenage c hildb earing and its life cycle consequences: Exploiting a natural exp erimen t. The Journal of Human R esour c es , 40 (3), 683715. and Miller, R. A. (1988). An empirical analysis of life cycle fertilit y and female lab or supply . Ec onometric a , 56 (1), 91118. 109 Hotz, V. J. , Mullin, C. H. and Sanders, S. G. (1997). Bounding causal eects using data from a con taminated natural exp erimen t: Analysis the eects of teenage c hildb earing. The R eview of Ec onomic Studies , 64 (4), 575603. Jaeger, D. A. and P age, M. E. (1996). Degrees matter: New evidence on sheepskin eects in the returns to education. The R eview of Ec onomics and Statistics , 78 (4), 733740. Ja y achandran, S. and Lleras-Muney, A. (2009). Life exp ectancy and h uman capital in v estmen ts: Evidence from maternal mortalit y declines. The Quarterly Journal of Ec onomics , 124 (1), 349397. Johnson, R. W. and Neumark, D. (1996). W age decline among older men. The R eview of Ec onomics and Statistics , 78 (4), 740748. Jones, R. K. , Ingerick, M. and Jerman, J. (2018). Dierences in ab ortion service deliv ery in hostile, middle-ground and supp ortiv e states in 2008-2014. W omen ’s He alth Issues , pp. 17. and Jerman, J. (2014). Ab ortion incidence and service a v ailabilit y in the united states, 2011. Persp e ctives on Sexual and R epr o ductive He alth , 46 (1), 314. and (2017). Ab ortion incidence and service a v ailabilit y in the united states, 2014. Persp e ctives on Sexual and R epr o ductive He alth , 49 (1), 1727. and Kooistra, K. (2011). Ab ortion incidence and access to services in the united states, 2008. Persp e ctives on Sexual and R epr o ductive He alth , 43 (1), 4150. Kalemli-Oczan, S. (2003). A sto c hastic mo del of mortalit y , fertilit y , and h uman capital in v estmen ts. Journal of Development Ec onomics , 20 (1), 103118. , R yder, H. and Weil, D. (2000). Mortalit y decline, h uman capital in v estmen t, and economic gro wth. Journal of Development Ec onomics , 62 (1), 123. Kane, T. J. and St aiger, D. (1996). T een motherho o d and ab ortion access. The Quarterly Journal of Ec onomics , 111 (2), 467506. Killingswor th, M. R. and Heckman, J. J. (1986). F emale lab or supply: A surv ey . Handb o ok of L ab or Ec onomics , 1. Klepinger, D. , Lundberg, S. and Plotnick, R. (1999). Ho w do es adolescen t fer- tilit y aect the h uman capital and w ages of y oung w omen? The Journal of Human R esour c es , 34 (3), 421448. Kleven, H. , Landais, C. and Sogaard, J. E. (F orthcoming). Children and gender inequalit y: Evidence from denmark. A meric an Ec onomic Journal: Applie d Ec onomics . 110 Kohler, H.-P. , Behrman, J. R. and W a tkins, S. C. (2001). The densit y of so cial net- w ork and fertilit y decisions: Evidence from south n y anza district, k en y a. Demo gr aphy , 38 (1), 4358. Kuhn, M. , Wrzaczek, S. , Prska wetz, A. and Feichtinger, G. (2015). Optimal c hoice of health and retiremen t in a life-cycle mo del. Journal of Ec onomic The ory , 158 , 186212. Kuziemko, I. , P an, J. , Shen, J. and W ashington, E. (2018). The momm y eect: Do w omen an ticipate the emplo ymen t eects of motherho o d?, w orking P ap er. Levine, P. B. and St aiger, D. (2004). Ab ortion p olicy and fertilit y outcomes: The eastern europ ean exp erience. The Journal of L aw and Ec onomics , 47 (1), 223243. , , Kane, T. J. and Zimmerman, D. J. (1999). Ro e v w ade and american fertilit y . A meric an Journal of Public He alth , 89 (2), 199203. Lundberg, S. and Rose, E. (2000). P aren tho o d and the earnings of married men and w omen. L ab our Ec onomics , 7 (6), 689710. Lundborg, P. , Plug, E. and Rasmussen, A. W. (2017). Can w omen ha v e c hil- dren and a career? iv evidence from ivf treatmen ts. The A meric an Ec onomic R eview , 107 (6), 16111637. Ma thews, T. J. and Hamil ton, B. E. (2002). Mean age of mothers. Natio anl Vital Statistics R ep orts , 51 (1). Miller, A. R. (2011). The eects of motherho o d timing on career path. Journal of Population Ec onomics , 24 (3), 10711100. Mosher, W. D. , Jones, J. and Abma, J. (2012). Intende d and Unintende d Births in the Unite d States: 1982-2010 . T ec h. Rep. 55, Cen ters for Disease Con trol and Prev en tion National Cen ter for Health Statistics. Murphy, K. and Topel, R. (2006). The v alue of health and longevit y . Journal of Politic al Ec onomy , 114 (5), 871904. Oster, E. , Shoulson, I. and Dorsey, E. R. (2013). Limited life exp ectancy , h uman capital and health in v estmen ts. The A meric an Ec onomic R eview , 103 (5), 19772002. P asseron, V. (2000). L es 35 Heur es, L’emploi et les Salair es . T ec h. Rep. 50.2, D ARES. (2002). 35 Heur es: 3 ans de Mise en Oeuvr es du disp ositif A ubry I . T ec h. Rep. 06.2, D ARES. Pop-Eleches, C. (2006). The impact of an ab ortion ban on so cieconomic outcomes of c hildren: Evidence from romania. Journal of Politic al Ec onomy , 114 (4), 744773. 111 Quast, T. , Gonzalez, F. and Ziemba, R. (2017). Ab ortion facilit y closings and ab or- tion rates in texas. INQUIR Y: The journal of He atlth Car e Or ganization, Pr ovision, and Financing , 54, 17. Rosenzweig, M. R. and Schul tz, T. P. (1985). The demand for and supply of births: F ertilit y and its life cycle consequences. The A meric an Ec onomic R eview , 75 (5), 992 1015. and Wolpin, K. I. (1980). Life-cycle lab or supply and fertilit y: Causal inferences from household mo dels. Journal of Politic al Ec onomy , 88 (2), 328348. Roser, M. (2019). Life exp ectancy . Our W orld in Data , h ttps://ourw orldindata.org/life- exp ectancy . Sigle-Rushton, W. and W aldfogel, J. (2007). Motherho o d and w omen’s earnings in anglo-american, con tinen tal europ ean, and nordic coun tries. F eminist Ec onomics , 13 (2), 5591. Sklar, J. and Berkov, B. (1974). Ab ortion, illegitimacy , and the american birth rate. Scienc e , 185 (4155), 909915. Soares, R. (2005). Mortalit y reductions education attainmen t and fertilit y c hoice. The A meric an Ec onomic R eview , 95 (3), 580601. Tomiy ama, K. (1985). T w o-stage optimal con trol problems and optimalit y conditions. Journal of Ec onomic Dynamics and Contr ol , 9 (3), 317337. and Rossana, R. (1989). T w o-stage optimal con trol problems with an explicit switc h p oin t dep endence. Journal of Ec onomic Dynamics and Contr ol , 13 (3), 319337. W ard, M. P. and Butz, W. P. (1980). Completed fertilit y and its timing. Journal of Politic al Ec onomy , 88 (5), 917940. Wilde, E. T. , Ba tchelder, L. and Ell wood, D. T. (2010). The momm y trac k divides: The impact of c hildb earing on w ages of w omen of diering skill lev els, NBER W orking P ap er No 16582. 112

Abstract (if available)

Abstract This dissertation comprises three essays in labor economics. The First paper studies the consequences of unplanned births on women's careers. To answer this question, I exploit variation in access to legal abortion that could potentially affect the rates of unplanned births. I then use this random variations in fertility realizations to evaluate the effect of fertility shocks on women's earnings. The early legalization of abortion in five US states led to variation in access to abortion across states and birth cohorts, which allows the estimation of the effect of accessing abortion at a certain age on women's fertility. The evidence suggests that early access to abortion led to a significant delay in the age of start of motherhood. I also document an increase in completed fertility among black women who received access to abortion early in their fertility cycle. I then find that labor earnings increase by 13% as a result of the delay of an unplanned start of motherhood. Results from the effect of age of start of motherhood on labor supply and occupation status suggest that most earnings gains are due to better occupations rather than increase in hours worked. The second paper explores how labor hours flexibility affect the willingness of married women to participate in the labor market. I exploit a law change in France reducing full time workweek hours from 39 to 35 hours. I show that the labor force participation among married women increases by 3 percentage points as a result of the law change. I then set up a discrete choice model of labor force participation. Participation costs vary with the total number of hours spent in the workplace. Moreover, I allow for flexible correlation between wages and hours spent in the workplace. Estimates of the structural model show a positive correlation between wages and hours spent in the workplace and that longer hours are more taxing for women than men. The last paper proposes a theoretical framework that links longevity and quality of life to individual choices of human capital investment and retirement. The model predicts an increase in schooling years and a delay in retirement in response to an increase in life expectancy and health quality. These findings rationalize recent trends in employment, where employment has been decreasing a younger ages and increasing at older ages.

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Essays on family and labor economics

PDF

The EITC, labor supply, and child development

PDF

Three essays on the microeconometric analysis of the labor market

PDF

Essays in development economics

PDF

Essays on labor and development economics

PDF

Essays in the economics of education and conflict

PDF

The impact of minimum wage on labor market dynamics in Germany

PDF

Essays on family planning policies

PDF

Essays on applied microeconomics

PDF

Essays on political economy and corruption

PDF

Against the wind: labor force participation of women in Iran

PDF

Essays on macroeconomics of health and labor

PDF

Essays on urban and real estate economics

PDF

Essays on environmental economics: education, employment, and experiments

PDF

Two essays on major macroeconomic shocks in the Japanese economy: demographic shocks and financial shocks

PDF

Diverging fertility patterns? Racial and educational differences in fertility behaviors and their implications for socioeconomic mobility

PDF

Three essays on behavioral economics approaches to understanding the implications of mental health stigma

PDF

Intergenerational transfers & human capital investments in children in the era of aging

PDF

Design and testing of SRAMs resilient to bias temperature instability (BTI) aging

PDF

Essays on health insurance programs and policies

Asset Metadata

Creator Abboud, Ali (author)

Core Title Essays in labor economics: demographic determinants of labor supply

School College of Letters, Arts and Sciences

Degree Doctor of Philosophy

Degree Program Economics

Publication Date 04/24/2021

Defense Date 04/24/2019

Publisher University of Southern California (original), University of Southern California. Libraries (digital)

Tag ageing,demographics,fertility,labor supply,longevity,OAI-PMH Harvest

Format application/pdf (imt)

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor Kahn, Matt (committee chair), Bassi, Vittorio (committee member), Nugent, Jeff (committee member), Rozo, Sandra (committee member)

Creator Email aabboud@usc.edu,ali.abboud87@gmail.com

Permanent Link (DOI) https://doi.org/10.25549/usctheses-c89-144438

Unique identifier UC11675302

Identifier etd-AbboudAli-7253.pdf (filename),usctheses-c89-144438 (legacy record id)

Legacy Identifier etd-AbboudAli-7253.pdf

Dmrecord 144438

Document Type Dissertation

Format application/pdf (imt)

Rights Abboud, Ali

Type texts

Source University of Southern California (contributing entity), University of Southern California Dissertations and Theses (collection)

Access Conditions The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the a...

Repository Name University of Southern California Digital Library

Repository Location USC Digital Library, University of Southern California, University Park Campus MC 2810, 3434 South Grand Avenue, 2nd Floor, Los Angeles, California 90089-2810, USA