The Globalization of Household Production 1

Transcription

1 Te Globalization of ouseold Prodution Miael Kremer Stanley Watt Tis draft: May 20, 2009 estritions on migration of low-skilled workers to rier ountries are arguably te largest distortion in te world eonomy and te most ostly to te world s poor. Yet ri ountries seem unlikely to eliminate tese restritions due to onerns about te impat of migration on inequality among natives, publi finanes, and native ulture. A rapidly growing new type of migration may not be subjet to tese onerns. Many new ri ountries issue speial visas for foreigners, women in partiular, to work as private ouseold workers. Old ri ountries often oose low levels of enforement against illegal immigrants working in tis setor. We argue tat by allowing ig-skilled native women to inrease market labor supply, tis type of immigration inreases te wages of low-skilled natives and provides a fisal benefit by orreting tax distortions toward ome prodution. Calibration suggests programs, su as ong Kong s or Singapore s, under wi rougly 7% of te labor fore are foreign private ouseold workers, may inrease te ratio of wages of native low-skilled to wages of ig-skilled workers by 2.9% and inrease native welfare by 0.9% of inome, rougly 60 times te level estimated by Borjas. Paradoxially, owever, even if tese programs are pareto improving, tey may onflit wit etial norms requiring stronger soial obligations to long-term residents tan to oter foreigners. Sort-term programs may be more aeptable. We are grateful to Attila Ambrus, Abijit Banerjee, yan Bubb, Davin Cor, Patriia Cortes, ael Glennerster, Elanan elpman, Ayesa Imtaz, Panwadee Kananusapkul, Maria Petrova, Giovanni Peri, Matew abin, Josep Sapiro, Jose Tessada, ussell Weinstein, Alan Winters, Dan Wood, Dean Yang, and Noam Yutman for omments.

2 . Introdution estritions on movement of low-skilled workers to rier ountries are arguably te most important distortion in te world eonomy and te most armful to te world s poor. Klein and Ventura (2004, for example, estimate tat te removal of immigration restritions in OECD ountries ould inrease world output by up to 72%. 2 Yet despite alls for freer mobility (peraps, most notably from Mexio s President Fox, te elimination of immigration restritions is not under onsideration in ri ountries. Indeed, resistane to low-skill migration seems to be ardening in developed ountries. Many developed ountries are inreasingly fousing on admitting igly skilled immigrants from developing ountries (Kapur, 2004, arguably wit negative brain drain effets on te sending ountry (Bagwati and amada, 974; aque and Kim, 995; Miyagiwa, 99. ig-inome ountries limit migration, partiularly of low-skilled workers, due to onerns tat immigration of low-skilled workers would ( exaerbate inequality among natives, (2 reate a burden on te welfare state, and (3 ange native ulture and inrease rime. In standard models, in wi low and ig-skilled workers are omplements, migration by low-skilled immigrants lowers wages for low-skilled natives, raises wages for ig-skilled natives, and tus inreases inequality among natives. Te estimated welfare gains for te ost ountry are typially small relative to distributional effets. 2 Klein and Ventura assume tat apital is mobile and tat immigrants an take advantage of OECD levels of total fator produtivity. Even witout tese favorable assumptions, Walmsley and Winters (2003 estimate tat an inrease of 3% of labor supply in immigration in developed ountries ould raise world welfare by 0.6%, alf of te gains assoiated wit omplete trade liberalization. Williamson (996 argues tat 9 t Century mass migration led to inome onvergene aross today s ri ountries by diretly raising te wages of te migrants and by reduing labor supply in te sending ountries.

3 Borjas (999 estimates welfare gains to natives on te order of.0 or.02% of GDP 3 and Borjas (995 writes tat te relatively small size of te immigration surplus partiularly wen ompared to te very large wealt transfers aused by immigration probably explains wy te debate over immigration poliy as usually foused on te potentially armful labor market impats rater tan te overall inrease in native inome. We examine a new form of low-skilled immigration and argue it may ave very different effets. Women are inreasingly rossing international borders to work as private ouseold workers. A substantial proportion of new ri ountries ave expliit programs granting temporary visas tat allow foreigners to work as private ouseold workers, su as nannies and maids but restrit tem to tis setor. Table sows tat in Barain, Kuwait, and Saudi Arabia around 0% of te labor fore or more are foreign workers in private ouseolds. In ong Kong and Singapore, 6.8% and 7.0% of te total labor fore, respetively, are foreign domesti elpers. Foreign private ouseold workers make up at least 0.8% of te labor fore in Taiwan, and immigrant workers in private ouseolds are at least 0.8% of te labor fore in Israel. Non-Greek, non-eu employees in private ouseolds onstitute about % of te labor fore in Greee. Women are also traveling from new EU member states to tose older EU member states tat ave not delayed opening teir labor markets: te U.K., Ireland, and Sweden. 427,000 people from Eastern European ountries tat joined te EU just over 2 years ago ave reeived rigts to work in Britain; anedotal evidene suggests tat many of tese 3 Borjas (999 gives a range of values depending on different assumptions about fator prie elastiities. Tese numbers ome from assuming tat te prie of apital is fixed, and are for anges in inome, rater tan welfare. 2

4 are ouseold workers. Te oter EU member states are required to open teir labor markets by 20, and are also likely to see an inrease in tis type of migration. Anedotal evidene suggests tat substantial numbers of foreign women work illegally as ouseold employees in oter old ri ountries of Europe and Nort Ameria. In te US 35% of women illegal immigrants reported tat teir first job was working in a private ouseold (Cortes Autorities are less likely to enfore laws against iring illegal workers wen private ouseolds, rater tan firms, are doing te iring. Italy and Spain, for example, are reported to ave signifiant numbers of foreigners working illegally as private ouseold workers. A limited number of foreigners are working legally as private ouseold workers in ri ountries under Au Pair programs. Wereas mu less skilled migration is bloked by fears tat it will inrease inequality among natives, reate a fisal burden, and ange native ulture. We argue tat migration of foreign private ouseold workers an potentially ( equalize wages among natives, (2 provide a fisal benefit, and (3 limit te pereived impat of immigration on ulture and rime. Wen foreign workers perform servies previously done witin ouseolds, su as ooking, leaning, and are for ildren, te sik, and te elderly, tey free up native labor, partiularly women s labor tat ad been spent on ouseold prodution, for market prodution. (See Cortes and Tessada, Sine ig-skilled natives wit a iger opportunity ost of time are more likely to purase domesti servies from immigrants, native ig-skilled workers will spend more time working in te labor market. 3

5 To te extent tat migration of foreign private ouseold workers leads igskilled natives to inrease labor supply to te market, wage inequality among natives is redued. Te inrease in labor supply of ig-skilled workers leads to a deline in teir relative wage and an inrease in te relative wage of omplementary low-skilled native labor. By allowing women to work more flexible ours, foreign private ouseold workers may also redue gender disparities among ig-skilled natives and elp eliminate te glass eiling. Moreover, wen ig-skilled women ire immigrant private ouseold workers and transfer teir labor from ome prodution to market work, teir output beomes taxable, providing a fisal benefit for te population, even witout onsidering te taxes paid by te migrants temselves. Te long-run impat of ong Kong or Singapore-style foreign private ouseold worker programs on native ulture is limited, sine immigrant private ouseold workers are typially not allowed to bring families wit tem on teir visas. Tese workers are typially female, and rime is, terefore, less likely to be pereived as a problem. We onstrut and alibrate a simple model to examine te eonomi impat of foreign private ouseold workers on natives. We assume tat ig-skilled native families would use dayare enters in te absene of foreign private ouseold workers. owever, dayare enters typially do not ook or lean or admit ildren during olidays or wen tey are ill. Tus, private ouseold workers allow teir employers to supply more time to expand labor market supply on te intensive margin. We report fixed effets estimates in Argentine and Israeli data wi suggest tat ea private ouseold worker leads to a 5% inrease in employer time to market 4

6 prodution. Using tis estimate of 5%, immigration of 7% of te native labor fore an inrease relative wages of te native low-skilled by 2.9% and native welfare by 0.9%. Te effet on welfare is about 60 times te amount found by Borjas (999. We argue tat even if some foreign private ouseold workers leak into oter oupations, native low-skilled are still likely to be better off. Wile immigration by foreign private ouseold workers avoids many of te politial eonomy obstales of oter forms of migration, it raises a new set of etial issues and paradoxes. Under etial norms were obligations depend on loation of residene, rater tan being eiter universal or linked to itizensip, foreign private ouseold worker programs may well be onsidered unaeptable even if tey are Pareto improving. Programs wit temporary non-renewable visas migt make introduing foreign private ouseold workers more palatable. For example, existing Au Pair programs ould be expanded or te requirement tat Au Pair workers be students ould be dropped. Te rest of te paper is strutured as follows: Setion 2 lays out a simple model designed to illustrate te impat of foreign private ouseold workers on wages and welfare in te ost ountry. Setion 3 alibrates te impat of a ong Kong or Singaporestyle program in te US. Setion 4 disusses etial and poliy issues related to te adoption of su programs, and setion 5 onludes. 2. Model We present a model designed to illustrate te impat of foreign private ouseold worker migration. Subsetion 2. sets up te model by desribing ouseolds and produers. Subsetion 2.2 solves te model s equilibrium onditions. We fous on parameter values 5

7 for wi native ouseolds would use dayare enters in te absene of foreign private ouseold workers, and we abstrat from endogenous fertility deisions or labor/leisure tradeoffs. Our model is tus most appropriate for soieties were many women are eduated and an potentially work in te labor market, and were time freed up by tese workers is likely to go into market work rater tan leisure or inreased fertility. A more ompliated model migt be needed in plaes were te fertility response is likely important, as disussed at te end of te setion. 2.. Setup We assume tere are tree types of agents: native ig-skilled workers, native low-skilled workers, and foreign private ouseold workers. We normalize te native population to one and assume tat a fration are ig-skilled, leaving te fration of low-skilled natives. Let m denote te population of foreign private ouseold workers, making te total population in te ost ountry + m. Ea agent is endowed wit unit of labor. Agents onsume two types of private onsumption goods, a general good, denoted good A, and a domesti good like ooking, leaning, or ildare, denoted good C. Consumers utility is linear in good A and is U A v(g were g is te government provision of a publi good to be disussed later. All native onsumers must onsume a required units of te domesti good. Te domesti good an eiter be self-produed, purased from dayare enters, or purased from private ouseold workers. Natives also onsume a publi good paid for by te government as disussed later. 6

8 Te prodution funtion for good A is A L were is ig-skilled native labor, L is low-skilled native labor, and 0 < β <. We assume immigrating foreign private ouseold workers are legally exluded from entering tis setor under te terms of teir visa, as in ong Kong and Singapore. 4 We will onsider te ase were some foreign private ouseold workers leak into te general eonomy and substitute for lowskilled workers in prodution of good A in setion 3. We assume tere are tree tenologies for domesti good prodution: ome prodution, dayare enters, and private ouseold workers. People an ome produe one unit of domesti good wit one unit of labor. Dayare enters produe good C wit prodution funtion k L assume dayare enters ave te same fator intensity as good A. If anyting tis is a C d. We onservative assumption, as Kisker et al. (99 find tat 47% of teaers in US day are enters ave ompleted ollege. Tis is onsiderably more tan te 24.% of workers wit ollege degrees in te general eonomy (Kominski and Adams, 994. Dayare enters only produe ertain types of te domesti good. We terefore assume tere is an upper limit denoted by as to ow mu of te domesti good onsumers an purase from dayare enters. Most dayare enters are open only during limited ours, are losed on olidays, and send ome sik ildren. Consumers 4 Te assumption tat low-skilled immigrants and low-skilled natives are not perfet substitutes and tat tese types of immigrants ave a omparative advantage at produing te domesti good is neessary for our results. Imperfet substitutability of immigrants and native low-skilled workers may arise due to differenes in skills between tese workers, or to government poliies tat restrit immigrants ability to work outside te domesti setor, as in te ong Kong foreign domesti elpers program or te US and UK Au Pair programs, or to differential enforement of immigration restritions aross workers in different setors. Some evidene tat immigrants ave a omparative advantage in domesti work is provided by te fat tat in 998 in te US, non-itizens were almost five times more likely to work in personal servie private ouseold ategory as itizens. 3.6% of non-itizens were employed in tose oupations as opposed to 0.7% of US itizens % of foreign-born US residents (inluding naturalized itizens work in personal servie oupations. Cortes (2006 finds speifially tat 25.8% of low-skilled female immigrants are employed in private ouseolds, a number mu larger tan for te native population. 7

9 employing dayare enters must still drop off and pik up teir ildren, find alternative ildare on olidays, and are for sik ildren temselves. Dayare enters also do not perform oter domesti tasks like ooking and leaning. We also assume a tird type of tenology for good C prodution, private ouseold workers. In most of te developed world, we observe very few natives working as private ouseold workers. In te UK in 990, only 0.05% of te working population were employed as domesti ousekeepers. In te US, only around % of entire employed population (inluding natives and non-natives are employed in te personal servies private ouseold industries. To aount for tis, we also assume tat working onditions and soial stigma assoiated wit private ouseold work ause people to dislike working in te private ouseold setor and/or tat ouseolds prefer or are more effiient at produing teir own domesti good, beause tey know teir own tastes in food, enjoy taking are of teir own ildren, et. Tis will mean tat tere will be a wedge between te after-tax wage of potential employers and te wage in alternative jobs open to potential employees. We model tis by assuming tat workers suffer disutility from doing private ouseold work. Altoug foreign ouseold workers experiene a utility penalty wen working as private ouseold workers, we assume tat teir oter options are even less attrative. Tus, we assume a potentially inelasti supply of foreign private ouseold workers. We assume tat te supply of foreign private ouseold workers is only limited by te number of visas tat te ost government will provide. Foreign private ouseold programs usually restrit immigrants to working for only one native ouseold worker, so we assume tat foreign private ouseold workers 8

10 annot divide teir labor supply among several native ouseolds. (elaxing tis assumption would magnify te impat of foreign private ouseold worker migration. We assume tat one private ouseold worker an provide te full units of domesti good prodution for exatly one ouseold (and tis worker is restrited to work for only one employer, for example beause it is important to be on all all te time for wen a ild gets sik, a parent needs to work late, et. Formal programs often restrit foreign private ouseold workers to a single employer. 5 Te government taxes te labor inome of all natives at tax rate and spends all tax revenue on a publi good tat is only enjoyed by natives. Taxes are paid in units of good A. In order to abstrat from debates about weter immigrants pay more in taxes tan tey reeive from te government in soial servies, we assume tat foreign private Comment [M]: (I ould not add omments to footnote. Maybe te assumption about prodution funtion of native private ouseold workers is wort mentioning in te body of paper; if you ave some justifiation for tis (oter tat tis assumption is not ruial for your results it would be nie to add it ere. I tink tis assumption (about prodution fn is better for native PW tan assuming tat tey (natives are restrited someow to work witin te same ouseold; in su speifiation, tere is no redundant labor. ouseold workers are neiter taxed nor enjoy te benefits of te government good. (In fat, foreign private ouseold workers are most likely to be net ontributors to te welfare system. In ong Kong and Singapore, employers of private ouseold workers are levied a tax, and te workers temselves enjoy very little of government provided goods and transfers. Natives inelastially supply one unit of labor and fae te following budget onstraint: W ( ( L P A P C P C, were W i is te wage of onsumer i( i {, L} i C A w w d d were subsript denotes ig-skilled workers, and L orresponds to low-skilled 5 We assume tat native private ouseold worker provides units of teir own labor to produe units of domesti good for te employing ouseold. Tis implies tat te prodution funtion of native private ouseold workers is C Labor. We, terefore, assume tat native private ouseold workers to employ dayare enters to are for teir own ildren. Tis assumption is not ruial to our results, and it is made for simpliity. 9

11 workers, L C is te amount of domesti good prodution ouseolds self-produe, P A is te prie of good A, P w is te prie of good C produed by ouseold workers, C w is te amount of good C purased from ouseold workers, P d is te prie of dayare, and C d is te amount of good C purased from dayare providers. Good A will be te numeraire, and tus P A =. In addition, te ouseold faes te onstraint tat and tat Cw Cd LC. _ C d ow te ouseold ooses its onsumption bundle of domesti servies depends on te prie of dayare servies, P d, te prie of private ouseold worker servies, P w, and te opportunity ost of ome prodution of te domesti good ( W. ouseolds will oose te ombination of dayare servies, ouseold worker servies, and own prodution tat maximizes teir onsumption of good A (reall tat teir utility funtion is equal to onsumption of good A. If ouseolds ave suffiiently low wages tat Wi ( P d and Wi ( Pw, tey will self-produe, teir entire domesti good i onsumption. If te opportunity ost of working is greater tan te prie of dayare servies, W ( P, but private ouseold workers are too expensive, relative to a i d mix of using dayare and ome prodution, P P W ( (, ten ouseolds will fully utilize dayare enters and self-produe te balane of teir domesti good needs. Finally, if P W ( and P P W ( (, ten te ouseold will ire a private ouseold worker. 6 w i w Te model implies tat ig-skilled natives wit a iger opportunity ost of time will be more likely to utilize private ouseold workers. Tis is onsistent wit empirial d w i d i 6 We ave overed all possible ases exept were te prie of dayare enters is greater tan te ost of iring a private ouseold worker wi we believe is extremely unlikely. 0

12 data. In 990, almost 8% of university eduated moters in Singapore ired domesti workers wile only 5% of polyteni eduated moters, 2% of seondary eduated moters and less tan 0.2% of primary sool eduated moters ired domesti workers (Singapore Census of Population Equilibrium We first onsider te losed eonomy ase were no foreign private ouseold workers are permitted in te ost ountry. Wit perfet labor mobility, for someone of a given skill, wages W i must be te same in bot te good A and dayare labor markets. Furtermore, natives working as private ouseold workers must be paid a iger wage in order to make tem indifferent between working as a private ouseold worker and working in te dayare or good A industries. Tis wage is determined by W w( ( L Cw W i( ( L Ci to make agents wit potential wage W i indifferent between private ouseold work and working in setor A (or dayare provision. ere W ( ( L is te utility of an agent working in setor A, as we i Ci earlier assumed tat it is equal to onsumption of good A, and P A =. is labor LCw supply of natives working as private ouseold workers. Terefore, te prie of te servie provided by native private ouseold workers will be P W ( L w i Ci ( were W i is te wage of workers of type i being ired as private ouseold workers, and LCi is amount of labor wi a worker of type i spends on ome prodution of domesti good. Pw sould be equal to te wage of a private ouseold worker sine te worker is restrited to work only for one employer, and Pw is te maximum amount of money tis worker an earn working for tis employer.

13 Proposition : If Assumption A: Assumption A2: k and ( ( k ( Assumption A3: ( ( k, ( ( ten W WL, Wi ( Pd, and Pw Pd Wi ( ( for i {, L}, and in te absene of immigration, all natives employ dayare enters and do not ire native private ouseold workers. (Proof is in Appendix A. In te subset of te parameter spae defined by Assumptions A A3 7, for ea native te opportunity ost of not working is stritly greater tan te ost of dayare enters. Wit te required wage premium, no native private ouseold workers are ired. Tus, every native fully utilizes dayare servies to proure of te domesti good, and self-produes te remaining. For ea native type, te amount of market labor supplied is. Note tat wit bot low-skilled and ig-skilled making te same domesti good prodution deisions, te ratio of ig-skilled to low-skilled labor in te eonomy is equal to te ratio of ig-skilled to low-skilled workers in te population,. Furtermore, given te idential fator intensities in te prodution funtions for 7 Assumption A ensures tat ig-skilled natives earn iger wages tan low-skilled natives. Assumption A2 ensures tat low-skilled natives (and by impliation ig-skilled natives an afford te use of dayare servie. Assumption A3 ensures tat δ, soial stigma (disutility of natives working as private ouseold workers,is suffiiently ig su tat no natives an afford a native private ouseold worker. 2

14 good A and dayare servies, te ratio of ig-skilled to low-skilled labor supplied in ea industry will equal tis worker population ratio. We, now, onsider te effets of m foreign immigrants wo are given visas to enter te ountry and wo are restrited to working only as private ouseold workers. 8 In order to be as onservative as possible about te benefits to te reeiving eonomy, we assume foreign private ouseold workers ave full property rigts in teir visas and tus extrat all of te surplus in bargaining wit employers, making teir employers indifferent between iring tem and teir next best option. Sine m, only ig-skilled workers will ire foreign private ouseold workers as te reservation prie of ig-skilled workers for purasing private ouseold worker servies is iger tan tat of low-skilled workers. Te key ange indued by te entry of foreign private ouseold workers is te ratio of ig-skilled to low-skilled market labor provided. Sine m ig-skilled workers now ire foreign private ouseold workers to produe te requisite units of te domesti good, tey now supply a full unit of market labor, wereas te rest of te native workers ontinue to supply units of market labor. Te ratio of ig-skill to low-skilled labor provided in bot te good A market and te dayare market is now ( m ( m m(. It is now greater tan te population ( ( ( ( ratio of ig-skilled to low-skilled workers, and onsequently ig-skilled wages derease and low-skilled wages inrease relative to te ase witout immigration. 8 eall tat ea foreign private ouseold worker is restrited to work for exatly one employer. As a result, if m immigrants enter eonomy to work as private ouseold workers, exatly m< ig-skilled workers ire tese immigrants to produe teir domesti good. 3

15 Tis ange in labor pries indued by te release of ig-skilled native labor from ome prodution into te market ould result in a ange in beavior of tose ouseolds tat do not ire a foreign private ouseold worker. owever, under Assumptions A-A3 it was optimal for all ouseolds to ire dayare and not to ire native private ouseold workers. Wit m foreign private ouseold workers working for ig-skilled ouseolds, low-skilled wages ave inreased and ig-skilled wages ave dereased. As long as low-skilled and ig-skilled wages ontinue to bear te same ordinal relation, W W, wit te boost in wages low-skilled workers will still employ L ildare enters and will refuse to work as native private ouseold workers. In order to ensure W W wit m immigrants, we amend Assumption A. Te appropriately L modified Assumption A is ontained in Proposition 2. Proposition 2: If ( m( Assumption A :, Assumption A2: k and ( ( k Assumption A3: ( ( (, k ( ( ten W W, Wi ( Pd, and Pw Pd Wi ( ( for i {, L}, and wit L m immigrants working as private ouseold workers, m ig-skilled ouseolds will ire foreign private ouseold workers, and all oter natives employ dayare enters 4

16 and do not ire native private ouseold workers. Furtermore, ig-skilled wages are dereasing in m and low-skilled wages are inreasing in m: m( W L (, ( ( ( ( W, ( m( W L m m( ( ( ( ( ( 0 ; and W ( ( ( ( ( ( 0 2 ( ( ( (. m m m (Proof is in Appendix A. In equilibrium a fration of ig-skilled natives employ foreign private ouseold workers wile te rest rely on dayare. Terefore, te prie of foreign private ouseold workers, P fw, is set su tat te ig-skilled are indifferent between iring dayare servies and a foreign private ouseold worker. Matematially, tis ondition is given as: W ( ( P W ( P, were te left-and side is te utility d fw of a ig-skilled worker using dayare enters and te rigt-and side is te utility from iring a foreign private ouseold worker. Tis suggests tat te prie of foreign private ouseold worker servie is ( ( Pfw W ( ( ( ( k ( m( k. Proposition 2 states tat te immigration of m foreign private ouseold workers redues wage inequality among natives. Te release of ig-skilled workers 5

17 inreases te wage of te low-skilled and dereases te wage of te ig-skilled. Inome inequality, owever, may not be redued sine te gross inome of te ig-skilled before payments to foreign private ouseold workers may inrease wit teir greater market labor supply. One te population of foreign private ouseold workers reaes te population of ig-skilled natives,, ten all ig-skilled natives will employ foreign private ouseold workers. Any furter inreases in m will ause te prie of private ouseold workers to fall as to persuade low-skilled natives to begin iring tem. Tis will lead to an inrease in native low-skilled labor supply and a deline in teir wages. One te number of immigrants equals te number of natives, one, ten all ig-skilled and all low-skilled workers will employ private ouseold workers and te ratio between te wages of ig and low-skilled natives will be te same as witout any immigration. Note tat beause tax revenue will be iger, as disussed below, all natives will be stritly better off tan tey were before, so tis would be a Pareto improving level of immigration. Te inrease in welfare assoiated wit foreign private ouseold worker migration will be equal to te extra tax revenue olleted from employers, In tis framework, soial welfare is equal to te total amount of good A prodution minus payments to foreign private ouseold workers minus any disutility penalty inurred by natives working as private ouseold workers, and provides a onvenient measure of native welfare if one values a dollar in te ands of all natives equally or assumes ostless redistribution. (To te extent tat one plaes a iger value on te marginal 6

18 dollar of onsumption for low-skill natives, or tat it is ostly to raise revenue, welfare gains will be larger. We an sow te following proposition: Proposition 3. If m< (te level of immigration is relatively small, ten soial welfare (inluding tax revenues is given by: ( m( W ( ( p m( W ( Welfare W L Te ange in welfare aused by a marginal ange in immigration in tis ase is Welfare 0. (See Appendix B for proof. m Tus, for small inflows of migrants, te inrease in native welfare aused by te inflow of foreign private ouseold workers is given by te inrease in tax revenues from te additional market work done by te ig-skilled. To understand te intuition, note tat wen market work is taxed and ome prodution is not, taxation introdues a distortion as ouseolds will purase domesti servies only if te private benefit exeeds te private ost. owever, te soial benefit of working more ours in te market and purasing domesti servies inludes te taxes paid on market labor earnings. Wen immigrants redue te prie of te outsoured good, tis distortion an be ountered, and welfare gains an be aieved. Te extra taxes paid wen someone swites from ome prodution to purasing domesti servies is a pure externality and welfare gain. Altoug we tink te most relevant ase is te one disussed above, we briefly desribe te ase were some natives are employed as private ouseold workers (see disussion in Appendix C. In te absene of immigrants, low-skilled natives will be te only native private ouseold workers. If parameters are su tat dayare enters and native private ouseold workers bot exist in equilibrium, ig-skilled natives will be d 7

19 indifferent between utilizing dayare and iring low-skilled natives as private ouseold workers. Wen te numbers of foreign private ouseold workers are lose to zero, igskilled natives will utilize dayare, ire native private ouseold workers, and ire immigrant private ouseold workers. For tis to our, ig-skilled natives will be indifferent between iring a private ouseold worker and a ombination of dayare and ome prodution. Foreign private ouseold workers will bot displae native low-skilled private ouseold work into good A prodution and inrease te market labor supply of te igskilled by allowing some ig-skilled natives to swit from utilizing dayare enters to iring immigrant private ouseold workers. In equilibrium, te inreases in bot types of native labor will be exatly su tat native ig and low-skilled relative wages do not ange. Te amount of ig-skilled labor freed by natives must be exatly offset by te displaement of low-skilled native private ouseold workers. Oterwise eiter all natives will leave private ouseold work, or ig-skill workers will ease using dayare enters. Inreases in foreign private ouseold workers will not affet relative wages until all native private ouseold workers ave been displaed. One all native private ouseold workers ave been displaed, te analysis of setion 2.2 will apply. 9 To te extent tat private ouseold workers ontribute to inrease fertility or onsumption of leisure, our alibration will overestimate te ig-skilled labor response 9 Te finding tat tis type of immigration as zero effet on wages wen te ig-skilled are using tree types of domesti good prodution is an artifat of te two-type nature of te model, and of te assumption of no diminising returns in ome prodution. More generally, owever, te existene of a domesti setor will stabilize wages in tis range. Low-skilled wages will initially deline wit an inrease in foreign private ouseold workers, until all native private ouseold workers are displaed, and ten will rise as more and more ig-skilled labor is released. In a more realisti model, in wi workers differed in more tan just a single dimension, we onjeture tat tis type of immigration would lower wages of native private ouseold workers and dayare enter employees but raises wages of oter low-skilled workers. 8

20 and will overestimate anges in wages and welfare. If foreign private ouseold workers lead to iger fertility among ig-skilled natives, long-run inome distribution among natives may be equalized for two reasons. First, iger fertility ould inrease te longrun supply of ig-skill labor, at least to te extent tat ig-skilled parents an transmit eduation to teir ildren. Seond, to te extent tat greater fertility among ig-skill workers leads tem to split teir bequests and attention among more ildren te distribution of wealt will be equalized. (elated to te idea tat private ouseold workers may not atually inrease market work is te idea tat in some soieties a large amount of ousework is done by grandparents. To te extent tat foreign private ouseold workers would displae grandparent labor and inrease grandparent leisure, our alibrations would overestimate te effet. Te losed eonomy is subjet to two different distortions. One is a publi finane distortion from taxing market but not ouseold labor. Te oter is a trade distortion due to utting off trade in domesti servies. In teory, an omnisient government ould orret te pure publi finane distortions witout bringing in foreign private ouseold workers. Doing so would involve subsidizing people to ire native private ouseold workers and ten implementing oter taxes and transfers to aieve te desired inome distribution among natives. For example, in our model, te government ould make all expenditure on domesti servies, inluding dayare enters, tax dedutible, or equivalently subsidize dayare enters and private ouseold workers. One ould imagine a government doing tis for dayare enters, but doing it for private ouseold workers seems unlikely sine it would make te tax ode look very regressive, even toug it migt not atually be regressive given general equilibrium effets. Moreover, 9

21 for some parameter values (i.e., a large enoug disutility of doing private ouseold work, even ompletely fixing te tax distortion will not indue natives to work as private ouseold workers. 3. Calibration In tis setion, we estimate te empirial magnitudes of te effets on wages and welfare desribed in te model. Subsetion 3. disusses assumptions about key quantities. Subsetion 3.2 disusses te results, and Subsetion 3.3 onsiders ow te effets of foreign private ouseold workers ould ange under alternative assumptions and if some of tese workers were to leak into te general workfore. 3.. Assumptions To estimate te impat of foreign private ouseold workers programs on relations of wages and on welfare of natives, we will first need an estimate of te amount of ig-skilled labor freed up by iring a private ouseold worker. We will also need te native proportion of ig-skilled workers, te estimated wage of ig-skilled workers, and te marginal tax rate. We present four different soures of information from wi to infer reasonable estimates of te labor supply response of te ig-skilled to foreign private ouseold workers: one from aggregate Israeli labor fore partiipation data, one from Cortes and Tessada (2007, one from Ellwood, Wilde, and Batelder (2004, and one using Argentine and Israeli panel mirodata. We also present some ountry-level time series evidene onsistent wit te ypotesis tat foreign private ouseold workers are assoiated wit ig rates of female labor fore partiipation. 20

22 Information on ow te extensive margin of labor fore partiipation varies wit weter moters ave ouseold elpers/are givers is provided by Israeli data. Table 2 sows Israeli female labor fore partiipation rates in Overall, women wit youngest ildren aged 2 to 4 wo employ ouseold elpers for more tan 6 ours per week ave approximately 2.6 perentage points or 29% iger labor fore partiipation tan tose wo do not ire are givers. Moters wit ildren aged 0 to are almost 50% (30 perentage points more likely to partiipate in te workfore if tey ire a private ouseold worker. On te oter and, women witout ildren wo ire a private ouseold worker for more tan 6 ours per week ave 24% (0 perentage point lower labor fore partiipation. Tese figures ould be eiter larger or smaller tan te ausal effet of iring private ouseold workers on labor supply. To te extent tat some women iring private ouseold workers ave an unobserved taste for work and would ave osen to work regardless, te ausal effet is smaller. On te oter and, to te extent tat tere is variation in te amount of domesti work aross ouseolds, depending on te number and ages of ildren, te number of elderly in te ouseold, and te ability of oter adults to partiipate in ouseold prodution, tese figures are likely to underestimate te impat of private ouseold workers on employers labor supply. ouseolds wit more need for domesti work are more likely to ire private ouseold workers. Te effets of tis bias an be seen omparing te iger gap wen disaggregating women wit te gap wen women are not disaggregated, for example, by age of youngest ild. Tese effets may also be understated if private ouseold workers free up faters, not neessarily to enter te workfore, but to put in longer ours. 0 Table 5 inludes bot foreign ouseold workers and native ouseold workers. 2

23 Cortes and Tessada (2007 find effets of immigration on female labor supply using time use data from Ameria. In order to ontrol for endogeneity in immigrants oie of loation, tey instrument for te sare of te labor fore wit less tan igsool eduation using immigration patterns from 970, arguing tat soial networks are important determinants of future patterns of immigration. Tey preliminarily estimate tat doubling te perentage of workers wit less tan a ig sool eduation will inrease te labor supply of ig-skilled women from our (for tose wit a ollege degree to almost 8 ours a week (for tose wit professional degrees.. Tis point estimate implies tat ea immigrant in an oupation tat arguably largely substitutes for domesti prodution inreases native labor supply to te market by te equivalent of 0.5 workers. 2 Even looking at time use data may lead to an underestimating in te inrease in effetive market labor supply assoiated wit iring a private ouseold worker. igskilled natives using ildare enters may be unable to aept ertain projets or ertain Kananusapkul (2004 also attempts to estimate te labor supply effet of low-skilled immigrants, but, sine se laks time-use data on te intensive margin of labor supply, se is unable to find strong results. 2 Weigting te labor supply response by labor fore partiipation and population sare, we find tat te Cortes and Tessada (2007 estimates imply tat a 0% inrease in low-skilled labor inreases ig-skilled female labor supply by about 0.4%. Only 2.9% of te US labor fore ad less tan a ig sool eduation. A 0% inrease in te number of people witout ig sool eduation due to immigration orresponds to a.29% inrease in te overall labor fore. In te US 20.7% of all immigrants wit less tan a ig sool eduation find employment in industries we lassify as likely to displae substantially native ouseold prodution (private ouseold servies, landsaping servies, ildare servies, restaurant servies, drinking establisments, ar wases, barber sops, beauty salons, nail salons, dry leaning and laundry, and taxi and limousine servies, aording to te Publi Use Mirodata Samples Tis implies tat a 0% inrease in te population of workers witout ig sool eduation due to immigration would orrespond to a.207 x.029 = 0.27% inrease in immigrants in private ouseold worker setors.. Cortes and Tessada (2006 assume tat low-skilled immigrants and low-skilled natives are not perfet substitutes and are aggregated wit a CES aggregator. For simpliity, we assume tat tey are perfet substitutes as teir assumption will not ange te figures by mu. Using Cortes and Tessada s (2007 estimates means tat tis inrease in immigration would lead to about a 0.4% inrease in native, ig-skilled, female labor supply, equivalent to about a 0.04% inrease in total native labor supply, wi suggests tat ea immigrant in oupations tat affet native labor supply inreases native labor supply by 0.04/0.27 = 5.0% of a native worker. 22

24 oupations. It is ard to be a 40 ours a week investment banker or partner in a ig-end law firm. Foreign private ouseold workers may allow some igly eduated natives to funtion as very igly skilled workers. Evidene onsistent wit te ypotesis tat taking are of ildren leads to lower wages is provided by Ellwood, Wilde, and Batelder (2004, wo use a panel data set to estimate tat ig ability women suffer net ourly wage losses of 30% ten years after te birt of a ild relative to ounterparts wo did not give birt. Tis may be an underestimate of te impat of aving a ild on earnings, sine even working moters redue te number of ours worked and some moters drop out of te labor fore. On te oter and, even moters wo employ private ouseold workers will probably experiene some loss of wages and earnings, and te 30% figure above may in part reflet seletion. A fourt soure of evidene omes from Israeli and urban Argentine panel mirodata. 3 We regress maternal labor supply on an indiator for weter te ouseold as a domesti worker, wile in some speifiations inluding individual fixed effets and time-variant ontrols. Desriptive statistis from tese data sow tat te average Argentine moter works 5-6 ours per week, and alf of Argentine moters work (Table 3. Israeli female labor supply is sligtly greater, at 8-20 ours and perent labor fore partiipation. Te Argentine data reord only weter a domesti worker lives in te ouseold, and alf a perent of ouseolds ave a domesti worker. Te Israeli data report weter a ouseold reeives over 40 ours per week of domesti elp, and one perent of Israeli ouseolds do. Essentially all domesti Argentine workers are 3 Te Data Appendix provides details of extrating samples from te Argentine ouseold surveys and te Israeli labor fore surveys. 23

25 female and teir mean age is 30 years. In bot ountries domesti servants work 50 ours per week. Te sample of nannies is small te Argentine balaned panel data ontains only 64 ouseold-waves were a nanny is present and te Israeli balaned panel as only 48 but te regression estimates are fairly robust. Fixed effets estimates sow tat iring a nanny inreases a moter s supply of labor by 5.4 to 5.6 ours. Tis is perent relative to te Israeli and Argentine sample means for all women of 5-20 ours (Table 4 or 5% relative to te sample means for employed women of US Census data. 4 OLS estimates are large, peraps sine te moters wo ire a nanny are also te moters more likely to work, anyway. Fixed effets parameter estimates are invariant to inluding a detailed array of time-variant ontrols, inluding measures of ouseold omposition, dwelling arateristis, and spousal wages. If anyting te parameter sligtly inreases, suggesting tat a survey wit more extensive ontrols ould produe even larger estimates. Sine te average domesti worker in bot ountries works 50 ours (Table 3, tis implies tat ten ours of domesti elp free up one our of a moter s market work. iring a nanny inreases an Argentine s probability of being employed by 25.0% (2 perentage points, wereas in Israel te impat is about 3.4% (2 perentage points, and is statistially insignifiant. Tis differene may arise sine aving a worker live inome (as in Argentina gives a moter more flexibility to seek employment, wile aving daily or ourly ontrats wit domesti workers (as in Israel makes it less feasible to enter te labor fore, toug it is peraps more likely tat te effets of nannies on te extensive margin is simply larger in Argentina tan in Israel. 4 US ensus data is taken from Cortes and Tessada (2007. Tis extrapolation onservatively assumes tat te only impat of iring a nanny is on te intensive margin. 24

26 Te small number of domesti workers makes subsamples impreise, but we present some variety of estimates (Table 6. Table 6, Panel B uses only te observations were te presene of a domesti worker ange during te panel te observations tat drive fixed effets estimates wi gives te same parameter values. Panel B suggests tat an Israeli domesti worker inreases maternal labor supply by 4.9 ours. Tese regressions are identified from 72 ouseold-waves in Argentine and 670 in Israel. Panels C and D sow tat te effet of a nanny grows wit eduation in Israel but dereases in Argentine, toug te estimates ave large onfidene regions. We see no obvious explanation for te divergene. eassuringly, Panels E and F sow tat te effet of domesti workers appears only for ouseolds wit ildren, wi suggests tat tese workers are largely supplying ild are. Using only observations were a domesti worker is present for one or more waves but ten vanises peraps due to a firing, or due to no longer needing te nanny obtains moderate-sized positive estimates in bot ountries. Fixed effets regressions represent an improvement over use of ross-setional or aggregate time-series variation. One ould worry tat unobserved soks like loal eonomi onditions drive bot maternal and nanny employment, toug it is reassuring tat ontrolling for available ouseold arateristis ardly anges and sligtly inreases parameter estimates. Additional evidene tat market labor supply is sensitive to te prie of ildare servies is provided by Blau and obins (988, Connelly (992, ibar (995 and Kimmel (998 wo find tat dereases in ildare pries inrease te likeliood of 25

27 moters entering te workfore and estimate ild-are prie elastiities of employment from to Time series evidene on labor fore partiipation rates aross ountries is onsistent wit te ypotesis of large impats of foreign private ouseold worker programs on female labor fore partiipation, at least in ountries wit signifiant numbers of eduated women and soial norms allowing interation between men and women in te work plae. Table 7 sows labor fore partiipation and fertility rates for a number of ountries, some of wi ave large populations of foreign private ouseold workers. 5 Consistent wit te idea tat tese programs an inrease female labor fore partiipation, ong Kong and Singapore bot ave ig rates of female labor fore partiipation. Tese rates are iger tan in Korea altoug similar to te US and UK. Female labor fore partiipation was not ig in tese ountries in te 970 s, before large numbers of domesti workers from abroad began to enter te labor fore, asting doubt on purely ultural explanations. Fertility rates for ong Kong and Singapore are mu lower and ave fallen more quikly tan in omparable ountries, onsistent wit te view tat women freed up from ouseold work to partiipate in te labor market ave fewer ildren. 6 5 One problem is tat labor fore partiipation statistis for most ountries (exept for Barain inlude foreign private ouseold workers. We estimate te female labor fore partiipation rate witout foreign private ouseold workers for tose ountries by assuming te age distribution from a survey of foreign private ouseold workers in ong Kong in 996 applies to te population of foreign private ouseold workers in ong Kong in 2000, Singapore in 2003, and Kuwait in 995. Te relevant age distribution is multiplied by te number of foreign private ouseold workers in ong Kong in 2000, Singapore in 2003, and Kuwait in 995. Tis number is subtrated from te number of women working in ea ountry and te total number of women in ea ountry wen alulating te female labor fore partiipation rate. (ong Kong Census and Statistis Department Note, owever, tat in te Gulf States, fertility is ig and female labor fore partiipation is low. If soial norms limit female labor fore partiipation or if ouseolds are wealty due in large part to nonwage inome, rater tan ig wages, ten time freed up by foreign private ouseold workers may go into leisure or inreased fertility rater tan into more market labor. If foreign private ouseold workers free up time tat natives devote to leisure or simply lead to more ouseold work being done tan would be 26

28 We onsider te ase in wi iring a private ouseold worker allows one igskilled parent to inrease labor supply by 5% Tis is rougly equivalent, for example, to assuming tat a parent (typially a moter an go from a 40 our a week job wit ildren in dayare to a 46 our a week job wit a private ouseold worker. Tis estimate is lose to te estimates obtained from te four soures desribed above. To estimate te welfare gains and wage effets of foreign private ouseold workers, we also need information on te sare of ig-skilled workers, teir wages, and te deadweigt loss of taxation. If one treats workers aving ompleted a ollege eduation as ig-skilled, ten = 0.24 for te US. ong Kong data imply tat igskilled workers employing private ouseold workers earn twie te average wage in te eonomy. 7 Te Argentine mirodata (Table 3 sow tat moters employing a nanny earn 2.3 (=8.8/3.9 times te mean wage of all moters or all faters. Sine tese data ontain only large ities tis may be a low estimate, toug sine very young and old workers ould ave lower wages it may be a ig estimate. 8 If foreign private ouseold workers were less tan 7% of te labor fore, employer wages would presumably be even greater relative to te average wage as, te igest earning natives are likely te most eager to ire domesti elp. done oterwise, relative wages between different lasses of native workers will be unaffeted by foreign private ouseold workers. 7 Aording to te ong Kong Census and Statistis Department, in January of 996 median montly ouseold inome of ouseolds iring domesti elpers in ong Kong was 40,000 K dollars. In te 4t quarter of 995, median montly ouseold inome of all K ouseolds was 5,700 K dollars 8 Alternative estimates of tis magnitude using national aounts and EP mirodata provide values in te range of 2.36 to In EP, te 397 moters wo ave a stritly positive wage and a domesti worker present report a mean ourly wage of 8.30 pesos (standard deviation 7.22, and te 243,476 individuals reporting a stritly positive wage ave a mean ourly wage of 3.5 pesos (standard deviation Te implied ratio is 2.36 (=8.30/3.5. Using total montly individual inome from all soures rater tan wages, tis ratio is 2.57 (=32.40/ Te average of per apta GDP from INDEC data is 64, so using 534 (=64/2 as te denominator implies a ratio of 2.46 (=32.4/534. All values deflated to 999 pesos. Note tat te mirodata represent only individuals in large ities, wile te GDP data are national. 27