REMITTANCES: DETERMINANTS, MOTIVATIONS AND EFFECTS. A Dissertation GEORGES SAMI NAUFAL

Transcription

1 REMITTANCES: DETERMINANTS, MOTIVATIONS AND EFFECTS A Disseraion by GEORGES SAMI NAUFAL Submied o he Office of Graduae Sudies of Texas A&M Universiy in parial fulfillmen of he requiremens for he degree of DOCTOR OF PHILOSOPHY December 2007 Maor Subec: Economics

2 REMITTANCES: DETERMINANTS, MOTIVATIONS AND EFFECTS A Disseraion by GEORGES SAMI NAUFAL Submied o he Office of Graduae Sudies of Texas A&M Universiy in parial fulfillmen of he requiremens for he degree of DOCTOR OF PHILOSOPHY Approved by: Co-Chairs of Commiee, Manuelia Urea Donald Deere Commiee Members, Adalber Mayer Arnold Vedliz Head of Deparmen, Amy Glass December 2007 Maor Subec: Economics

3 iii ABSTRACT Remiances: Deerminans, Moivaions and Effecs. (December 2007) Georges Sami Naufal, B.A., American Universiy of Beiru Co-Chairs of Advisory Commiee: Dr. Manuelia Urea Dr. Donald Deere This disseraion examines he deerminans, moivaions and effecs of remiances. In ha las wo decades remiances have gained ineres due o heir large size. For several developing counries remiances consiue a large porion of heir gross domesic produc and someimes exceed foreign direc invesmen. In he firs essay, I use a unique daa se from Nicaragua o asses he behavior of persons who send money back home. I esimae a heeroskedasic Tobi wih a known form of variance o esimae he correlaion of he remiing decisions of migrans. Working, residing in a developed counry and belonging o he nuclear family posiively affec remiances. The labor saus and he level of educaion of he head of he household boh affec remiances. The decision o paricipae in he remiing process appears o be posiively relaed across migrans wihin he same receiving household. The second essay presens a simple heoreical model of migrans' remiing behavior. I consider wo general moivaions for remiing: alruism and self-ineres. From he same daa se used in he firs chaper, I esimae a heeroskedasic Tobi and a sample selecion equaion o empirically es he findings of he heoreical model.

4 iv Evidence suggess ha migrans from Nicaragua remi for alruisic reasons. Moreover some gender heerogeneiy exiss in he remiing behavior. In he las essay, I sudy he impac of remiances on a small open economy using a sochasic limied paricipaion model wih cash in advance consrains and cosly adusmen of cash holdings. I examine he impac of remiances on he seady sae of he economy and on he dynamic response of variables o money shocks, oupu shocks, and shocks o remiance flows. I also examine he impac on dynamic responses o shocks of alernaive specificaions regarding he iniial impac of a moneary inecion or a remiances shock on he economy. I find ha a posiive remiances shock forces he exchange rae o depreciae and lowers boh oupu and consumpion in he period of he shock, irrespecive of adusmen coss on money balances. Also, he posiive remiance shock lowers uiliy during he period of he shock bu improves i hereafer.

5 v DEDICATION To my parens, Sami Naufal and Wafaa Naufal for heir inconceivable suppor To my brohers, Ziad Naufal and Kamal Naufal for always being here To Lindsey Flasowski, my wife for changing my life, forever.

6 vi ACKNOWLEDGMENTS Firs and foremos, I would like o hank my advisors Dr. Manuelia Urea and Dr. Donald Deere for heir coninuous suppor, ime and creaive inpu. I am hankful o my commiee members, Dr. Adalber Mayer and Dr. Arnold Vedliz for heir insighful commens. I would like o hank Dr. Genevieve Verdier for her ime and effor. I hank Dr. Li Gan for his helpful commens. I also would like o hank Dr. Kishore Gawande and Dr. George Davis. I am grealy indebed o Dr. Timohy Gronberg and Dr. Dennis Jansen for heir menorship, guidance and inellecual conversaions in he las wo years. The deparmen of Economics a Texas A&M Universiy has given me so much during my graduae sudies and I acknowledge everyone in he deparmen from faculy o saff who helped ease my say a Texas A&M. Paricularly, I will always be graeful for Chrisi Ramirez for being so kind and so helpful. My grea admiraion goes o Tyffannee Rowan who manages so many hings in her life ye was always here for help and advice. I also would like o hank Pa Nelson for all her help and Elizabeh Sus for her behind he scene work. For several academic conversaions I hank my colleagues Eric Michem, Diego Vacaflores, Anirban Sengupa, Zeynal Karaca, Winging Wang, and Diego Escobari.

7 vii Many friends made my say in College Saion more ineresing. I hank Daad Abi Ghanem for making me laugh, Azza El Zein for her coninous encouragemen, Iryna Nasadyuk, Salah and Dunia Abi Joumaa, Tarek Abi Fakher and Mohamad Hamid. I would like o express my deep respec and appreciaion for Randy and Rhonda Flasowski for being my American family. I will always be graeful. Jack, Nina and Scooer augh me a lo in heir own special way. Lindsey Flasowski, my wife, has given me so much and has being he ulimae suppor. She gave me he energy ha someimes I desperaly needed. Lindsey, hank you for believing in me. I canno forge all he suppor ha my family in Lebanon has given me. I paricularly hank my grandparens for heir coninuous suppor. Finally, my deepes graiude go o my parens Sami and Wafaa Naufal and my brohers, Ziad and Kamal, I could no have never made i wihou hem.

8 viii TABLE OF CONTENTS Page ABSTRACT DEDICATION ACKNOWLEDGMENTS... TABLE OF CONTENTS LIST OF TABLES.. LIST OF FIGURES. iii v vi viii x xii CHAPTER I INTRODUCTION... II WHO REMITS? THE CASE OF NICARAGUA Inroducion and Moivaion Lieraure Review Model Daa and Esimaion Mehod Resuls Conclusion. 32 III WHY REMIT? THE CASE OF NICARAGUA Inroducion Lieraure Review Theoreical Model Daa and Esimaion Mehod Resuls Conclusion.. 65 IV THE EFFECTS OF REMITTANCES ON A SMALL OPEN ECONOMY Inroducion Lieraure Review.. 7

9 ix CHAPTER Page 4.3 Theoreical Model Resuls Conclusion. 6 V CONCLUSION REFERENCES APPENDIX A APPENDIX B APPENDIX C 36 VITA.. 40

10 x LIST OF TABLES TABLE Page 2. Summary of he Lieraure Findings Characerisics of Naive and Emigran Populaion (percenages) Characerisics of Emigrans by Desinaion (percenages) Emigran Populaion Aged 4 and up: Proporion Remiing by Relaionship o he Head of he Receiving Household and by Group (percenages) Disribuion of Receiving Households and Their Characerisics by Number of Sending Emigrans Aged 4 and up (percenages) Maximum Likelihood Esimaes for a Heeroskedasic Tobi Average Model of he Amoun Remied by Emigrans Aged 4 and up Summary of The Change in Amoun of Remiances and Change in Probabiliy of Remiing Resuls of Model () in Table Maximum Likelihood Esimaes for a Heeroskedasic Tobi Average Model of he Amoun Remied by Emigrans Aged 4 and up: Fewer Conrols Migran Remiing Decisions Among Differen Samples: Relaionship of he Migran o he Head of he Household, Labor Saus and Desinaion 3 3. Characerisics of Receiving Households and Migrans by Remiing Process Characerisics of Receiving Households by Number of Oher Migrans k Disribuion of Households by Reason of Head of he Household Leaving he Las Job Disribuion of Households by Lengh of Job Search... 53

11 xi TABLE Page 3.5 Characerisics of Receiving Households by Measures of Bad Sae versus Unaffeced Households Migrans Characerisics by Measures of Bad Sae versus Unaffeced Households Characerisics of Receiving Households and Migrans for Households wih a mos One Remiing Migran (Limied Sample) versus Full Migran Sample Relaionship of he Migran o he Head of he Receiving Household for Households wih a mos One Remiing Migran, Full Migran Sample and he Remaining Sample Probi Esimaes for Eq. (3.4): All Migrans a Tobi Esimaes for Eq. (3.4) following he Average Model: All Migrans b Sample Selecion Esimaes for Eq. (3.4): Households wih a Mos One Remiing Migran Summary of The Change in Amoun of Remiances and Change in Probabiliy of Remiing Resuls for column () in Table 3.0a Esimaes for Eq. (3.4) wih Differen Specificaions: Male versus Female Model Calibraion Values Seady Sae Values... 9

12 xii LIST OF FIGURES FIGURE Page 2. Top 20 Developing-Counry Recipiens of Workers Remiances by Size of Remiances Top 20 Developing-Counry Recipiens of Workers Remiances by Percenage of GDP 7 4. Trends of FDI and Remiances for a Sample of Lain American Counries Nominal Ineres Rae Dynamics following a Moneary Shock Oupu Dynamics following a Moneary Shock Nominal Exchange Rae Dynamics following a Moneary Shock Consumpion Dynamics following a Moneary Shock Nominal Ineres Rae Dynamics following an Oupu Shock Oupu Dynamics following an Oupu Shock Nominal Exchange Rae Dynamics following an Oupu Shock Consumpion Dynamics following a Technology Shock Nominal Ineres Rae Dynamics following a Remiances Shock Oupu Dynamics following a Remiances Shock Nominal Exchange Rae Dynamics following a Remiances Shock Consumpion Dynamics following a Remiances Shock. 4.4 Uiliy Dynamics following a Remiances Shock Oupu Dynamics following a Remiances Shock wih ξ =

13 CHAPTER I INTRODUCTION This disseraion examines he deerminans, moivaions and effecs of remiances. In ha las wo decades remiances have gained ineres due o heir coninuous increase and heir large size. For several developing counries remiances consiue a large porion of heir gross domesic produc and someimes exceed foreign direc invesmen. Official esimaes show ha remiances averaged around 60 billion U.S. dollars per year in he 990s (World Bank) and reached 67 billion U.S. dollars in 2005 (World Bank s Global Economic Prospecs). In some counries remiances consiue a significan share of gross domesic produc (GDP) (Connell and Brown, 2004; De Haas, 2006; Heilmann, 2006; Chami e al., 2006). For insance, in Nicaragua, he poores counry in he Wesern Hemisphere ( remiances consiue around 6% of is GDP, he fifh highes percenage among he developing counry recipiens of workers remiances in 200. The growing imporance of hese ransfers of money has produced a number of sudies o explore heir dimensions, deerminans, effecs and he governmen policies designed o influence hem. Remiances gain heir significance no us from heir size bu from he poenial and acual effecs of hese money flows on boh he sociey and he individual. Remiances affec labor marke decisions; school reenion levels, expor secor compeiiveness, and can creae moral hazard problems (Funkhouser, 992; This disseraion follows he syle of Journal of Developmen Economics.

14 2 Glysos, 2002; Edwards and Urea, 2003; Amuedo-Doranes and Pozo, 2004; Chami e al., 2005). All hese sudies sress on beer undersanding he remiance behavior in order o develop economic policies ha ake full advanage of hese flows. In he nex chaper, I use a unique daa se from Nicaragua o asses he characerisics of he individuals who remi. Unlike all previous sudies I have informaion on he sender and he receiver from he same source. I esimae a heeroskedasic Tobi wih a known form of heeroskedasiciy which allows me o examine he correlaion of he remiing decisions of migrans in he same original receiving household. The main conribuion of his paper is he abiliy o quanify he correlaion of he remiing decisions beween migrans who belong o he same receiving household. The abiliy o measure his relaionship is crucial since i allows furher undersanding of how inra-family decisions are made regarding he allocaion of resources across households ha are separaed by migraion. The knowledge of he mechanism of inra-family remiing decisions shed ligh on he indirec oucomes of remiance policies. I believe his is he firs paper o address his issue. This paper also adds o he remiance lieraure in compuing changes in boh he likelihood of remiing and he amoun remied. Gender, labor force saus, and desinaion of he migran all have significan effecs on remiances. The relaionship of he migran o he head of he household also affecs he remiing behavior. The labor saus and he level of educaion of he head of The receiving household is he original household in Nicaragua which repored a leas one migran living abroad.

15 3 he receiving household influence he migran s decision o paricipae in he remiing behavior. Evidence suggess ha here is a posiive correlaion beween migrans remiing decisions among migrans belonging o he same receiving household. In Chaper III, I invesigae he moivaion behind remiances. Remiances differ from oher ypes of capial flows. Capial flows such as FDIs are in general profi driven and herefore are posiively relaed o GDP growh. However, his is no always he case for remiances. Remiances are no always profi driven and can be alruisically moivaed. Also, FDIs end o be less sable relaive o remiances (Orozco, 2002). Uncovering he reasons for remiing is crucial for policy implicaion for several reasons. From he original household perspecive, he forces behind remiances can shed some ligh on households migraion sraegies (De La Brière e al., 2002). In fac Hoddino (994) sresses ha remiances should be incorporaed in he model of household migraion decisions. From a macroeconomic look, remiances are hough o be inended o ease he burden of poor economic performance on local recipiens (Chami e al., 2005). Therefore alruisically moivaed remiances are expeced o be counercyclical wih income growh and consequenly can decrease he scope of he governmen inervenion in recession imes. In his paricular case, policies buil on predicions ha remiances behave in he same manner as oher ypes of capial flows migh have unanicipaed consequences. I presen a simple heoreical model of remiance behavior. I consider remiances as unidirecional flows from he migran in a hos counry o he original

16 4 household in he home counry which I refer o in his paper as he receiving household. This allows me o consider he reacion of remiances o a bad sae oucome on he receiving household. This is he firs paper ha looks a he response of remiances o shocks ha perain o he receiving household. This is crucial in erms of invesigaing he remiance behavior since mos remiances consider he migran as a source and he receiving household as he end desinaion and herefore, hey are expeced o reac o any income shocks a he receiving end. This seup gives wo broad moivaions for remiing: alruism where migrans simply care abou he receiving household members welfare and self-ineres where migrans remi for invesmen opporuniies ha are expeced o yield a cerain payoff in he fuure. I es he heoreical predicions of his model using survey daa from Nicaragua. I quanify he resuls of he heeroskedasic Tobi for policy purposes. Alruism seems o be he main moivaion behind he remiing behavior o Nicaragua. Moreover he remiing behavior is no idenical across gender. Female migrans seem o behave more alruisically oward he receiving household. The lieraure on he macroeconomic impac of remiances on he recipien counry is sparse. Chaper IV explores he impac of remiances flows on oupu, consumpion, ineres and exchange raes in he recipien counry. In paricular, I explicily model remiances in a small open economy and analyze he impac of shocks o money, remiances and oupu. I expand a limied paricipaion model ha requires ha money balances be held o finance cerain ypes of purchases and agens incur adusmen coss on money holdings. These wo requiremens generae a large and

17 5 persisen liquidiy effec consisen wih he sylized facs (Hairaul e al., 2004). The impac of he adusmen coss on he predeermined allocaion of money cash available for consumpion is hen analyzed o see how he main real variables of he economy respond o a remiances shock. One of he conribuions of his Chaper is he abiliy o examine he dynamic response of maor macroeconomic variables o remiances shock. Anoher imporan conribuion is he capaciy o observe he impac on he main economic variables when I allow for differen end uses of remiances, and moneary inecions. In fac, many domesic governmens are ineresed in developing policy ools o direc a porion of remiances owards invesmen (The New York Times Magazines April 22, 2007). I disinguish beween he direc effec of remiances on oupu hrough invesmen and he indirec effec hrough consumpion and is muliplier effec. Being able o disinguish he end use of remiances is crucial in looking a he final effec on oupu in he economy (Burgess and Haksar, 2005; Heilmann, 2006; Sayan, 2006). This Chaper also presens a welfare analysis of remiances shock and invesigaes how remiances affec he uiliy of he represenaive agen. Posiive remiances shock forces he exchange rae o depreciae and lowers boh oupu and consumpion in he period of he shock, irrespecive of adusmen coss on money balances. Also, he posiive remiance shock lowers uiliy during he period of he shock bu improves i hereafer. Finally, Chaper V summarizes and concludes.

18 6 CHAPTER II WHO REMITS? THE CASE OF NICARAGUA 2.. Inroducion and Moivaion Inernaional esimaes of official remiances flows sugges ha oal remiances averaged around US$ 59 billion per year during he 990s 2 (World Bank). Fig. 2. ranks he firs 20 developing counry recipiens of workers remiances by size of remiances received in 200. Top 20 Developing-Counry Recipiens of Workers' Remiances 200 Source: IMF Balance of Paymens Yearbook Billions of US Dollars India Mexico Philippines Morocco Egyp Turkey Lebanon Bangladesh Jordan Dominican Republic El Salvador Colombia Yemen Pakisan Brazil Ecuador Yogoslavia Thailand China Sri Lanka Counries Fig. 2. Top 20 Developing-Counry Recipiens of Workers Remiances by Size of Remiances 2 Mos esimaes of remiances are based on he balance of paymens saisics repored o he Inernaional Moneary Fund (IMF). Noe ha hese numbers are generally an underesimae of he acual remiances since hey only include he official flows of his money ino he receiving counries.

19 7 Ou of he firs 20 developing counries receiving remiances six are from Cenral and Souh America wih a oal of 8.5 billion U.S. dollars. In some counries remiances consiue a significan share of gross domesic produc (GDP). Fig. 2.2 ranks he firs 20 developing counry recipiens of workers remiances by percenage of heir GDP. Top 20 Developing-Counry Recipiens of Workers' Remiances 200 Source: IMF Balance of Paymens Yearbook Percenage of GDP Tonga Lesoho Jordan Albania Nicaragua Yemen Moldova Lebanon El Salvador Cape Verde Jamaica Yugoslavia Morocco Dominican Republic Vanuau Philippines Honduras Uganda Ecuador Sri Lanka Counries Fig Top 20 Developing-Counry Recipiens of Workers Remiances by Percenage of GDP Remiances consiue more han 0% of he GDP in welve developing counries. In Nicaragua, he poores counry in he Wesern Hemisphere (World Bank Websie remiances consiue around 6% of is GDP, he fifh highes percenage among he developing counry recipiens of workers remiances in 200. The growing imporance of hese ransfers of money has produced a number of

20 8 sudies o explore heir dimensions, deerminans, effecs and he governmen policies designed o influence hem. Migran remiances affec he performance of he economy. Glysos (2002) shows ha remiances have he poenial o subsiue for foreign aid. Chami e al. (2003) find ha remiing akes place under asymmeric informaion and imply ha remiances have a negaive impac on economic growh. Amuedo-Doranes and Pozo (2004) find ha workers remiances can reduce he inernaional compeiiveness of he receiving counries expor secor by appreciaing he real exchange rae in he receiving economies. Remiances also impac he behavior a he household level. Funkhouser (992) finds opposie effecs of remiances inflows on he Nicaraguan and Salvadorian labor markes. Edwards and Urea (2003) find ha remiances have a large effec on school reenion. All hese sudies sress on beer undersanding he remiance behavior in order o develop economic policies ha ake full advanage of hese flows. In his sudy I use a unique daa se from Nicaragua o asses he characerisics of he individuals who remi. Unlike all previous sudies I have informaion on he sender and he receiver from he same source. I esimae a heeroskedasic Tobi wih a known form of heeroskedasiciy which allows me o examine he correlaion of he remiing decisions of migrans in he same original receiving household. 3 The main conribuion of his paper is he abiliy o quanify he correlaion of he remiing decisions beween migrans who belong o he same receiving household. The abiliy o measure his relaionship is crucial since i allows furher undersanding of how inra-family decisions 3 In his paper he receiving household is he original household in Nicaragua which repored a leas one migran living abroad.

21 9 are made regarding he allocaion of resources across households ha are separaed by migraion. The knowledge of he mechanism of inra-family remiing decisions shed ligh on he indirec oucomes of remiance policies. I believe his is he firs paper o address his issue. This paper also adds o he remiance lieraure in compuing changes in boh he likelihood of remiing and he amoun remied. Gender, labor force saus, and desinaion of he migran all have significan effecs on remiances. The relaionship of he migran o he head of he household also affecs he remiing behavior. The labor saus and he level of educaion of he head of he receiving household influence he migran s decision o paricipae in he remiing behavior. Evidence suggess ha here is a posiive correlaion beween migrans remiing decisions among migrans belonging o he same receiving household. Secion 2.2 below provides an overview of he exising lieraure. Secion 2.3 consiss of a heoreical model. Secion 2.4 describes he daa and he mehodology. Secion 2.5 conains he resuls and secion 2.6 summarizes and concludes Lieraure Review Lucas and Sark (985) discuss several hypoheses for moivaions o remi. The auhors presen hree reasons for remiing ranging from pure alruism o pure selfineres spanning a more empered poin of view combining hese wo exremes. Lucas and Sark explore hese conceps using daa from he Naional Migraion Sudy of Boswana. Remiances are deermined parly by he earnings of he migran and parly by his years of schooling. Lucas and Sark also noe a posiive rend beween hese flows of money and per capia income of he household.

22 0 Oberai and Singh (980) using a household survey in he Ludhiana disric of he Indian Punab repor a posiive relaion beween low income migran and he probabiliy of remiances and a negaive correlaion beween he number of household members working and his same likelihood. In addiion, he auhors find ha he level of educaion does no affec he decision wheher or no o remi. Based on a naionwide survey of households in Kenya, Knowles and Anker (98) presen empirical evidence on issues relaed o remiances. The auhors firs sress ha remiances are primarily limied o members of he nuclear family. Moreover hey conclude ha migran s income of he sending uni, educaion level, sex, ownership of a house back home and he fac of a spouse residing away all posiively affec he probabiliy o remi. Knowles and Anker add ha he lengh of ime a migran has resided away negaively affec hese chances. Also migran s schooling and income negaively influence he level of remiances. Funkhouser (995) uses daa from El Salvador and Nicaragua o invesigae and compare he deerminans of remiances in boh counries. Funkhouser applies a separable uiliy funcion ha values boh absenee s uiliy and he household uiliy. The auhor also follows a linear funcional form in esimaing remiances. Funkhouser presens fairly similar findings for El Salvador and Nicaragua. In boh counries educaion is negaively relaed o he probabiliy of remiances while i is posiively associaed o he level of hese money ransfers. Using he Salvadoran daa Funkhouser noes ha age and gender do no affec he likelihood and he level of remiances. In Nicaragua, age is adversely correlaed wih boh he probabiliy and he amoun of

23 remiances. Furhermore, Funkhouser examines familial relaionship and he period of ime spen abroad and heir effecs on remiances. Rodriguez (996) uses a daa se from he Philippines o noe a posiive connecion beween he age of he migran, ime since migraion and he chance of remiing. However, equally o Knowles and Anker his incidence decreases for long absences. Rodriguez also remarks ha being a member of he nuclear family increases he probabiliy of remiing. Similarly o Oberai and Singh, Rodriguez does no find a clear associaion relaing educaion o remiances. Lianos (997) focuses on he remiances o Greece from Germany for a period of 30 years. Lianos ess he significance of a se of facors in erms of heir effecs on remiances. The auhor finds ha he level of migran s income has a posiive and maor effec on remiances o Greece. Lianos also calculaes he elasiciy of remiances wih respec o income. This elasiciy is greaer han one suggesing a large response of remiances for any small change in income. Furhermore, Lianos finds ha household income in he counry of origin does no significanly influence he level of remiances. Clearly, he empirical evidence on he deerminans of remiances is inconclusive. I summarize hese findings in Table 2.. Boh Oberai and Singh (980) and Rodriguez (996) find ha educaion and remiances are no relaed. Lucas and Sark (985) along wih Knowles and Anker (98) find a relaionship beween hese wo even hough hey do no agree on is direcion. In addiion, Lianos (997) finds ha household income is uncorrelaed wih remiances while Lucas and Sark (985) documen a posiive correlaion. These resuls

24 2 suppor a need for more empirical sudies on he deerminans of remiances. The following secion oulines a heoreical model of remiance behavior. Table 2. Summary of he Lieraure Findings Auhors Daa Findings Oberai and Singh (980) Knowles and Anker (98) Rural Household Survey in he Ludhiana Disric of he Indian Punab (March April 977) Household Survey in Kenya (December 974) Years of schooling do no affec he prob. and he level of remiances. Migran s income has a posiive effec on he probabiliy of remiances. The number of working household members negaively affecs his likelihood. Male, gender and years of schooling have a posiive effec on he probabiliy of remiing. Migran s income and receiving household s income have a negaive effec on his likelihood. Lucas and Sark (985) Household Survey of Migraion in Boswana ( ) Years of schooling, migran s income and he receiving household s income posiively affec he level of remiances. Funkhouser (995) Rodriguez (996) Lianos (997) El Salvador: Survey conduced by Segundo Mones (987) Nicaragua: Household Survey in Managua (December 989) Survey of Overseas Philippians workers (SOW 99) Saisical Daa come from diverse sources El Salvador: Age and gender have no effec on he probabiliy and he level of remiances while years of schooling have an adverse effec on he likelihood of remiing and a posiive effec on he level of remiances. Nicaragua: Age negaively affecs he prob. and he level of remiances. Gender has no significan effec on remiances. Years of schooling have an adverse effec on he likelihood of remiing and a posiive effec on he level of remiances. Age, years since migraion, relaionship o HH posiively affec he probabiliy of remiing. Age and years since migraion negaively affec he level of remiances. Years of schooling have no effec on he level of remiances. Migran s income posiively affecs he level of remiances. The receiving household s income has no effec on he level of remiances.

25 Model This paper builds on he model in Funkhouser (995). A model of remiance behavior considers an emigran s uiliy ha is a funcion of his own uiliy and ha of he receiving household in he home counry. I assume a separable uiliy funcion given by: ( U ; U ) U ( C ) V { U ( C ), Z } U = (2.) i i i ' ' '' '' where U 0; U > 0; U < 0 and U < 0 ; i refers o a paricular migran and refers i > i o a specific receiving household, U is emigran i own uiliy which depends on consumpion C, U refers o he receiving household uiliy which depends on is i consumpion C and Z defines a vecor ha includes he receiving household i characerisics. The emigran chooses remiances level o maximize a separable lifeime uiliy funcion such as: subec o ( ) V U ( Y R N R ), { C } /( ) ( ) { Z } / ( σ ) U i i σ u i v (2.2) C R = I i i i (2.3a) I i = β 0 X i β ε i (2.3b) where C is emigran s consumpion a ime, Y is household income earned by i receiving household in he naive counry a ime, R i refers o remiances received by he receiving household from migran i a ime, N idenifies he number of

26 4 oher household emigrans a ime, R quanifies he average remiances per oher emigran a ime, I i is he income of he emigran i a ime, X i describes a vecor of emigran s characerisics a ime, ( ) uiliy funcion, and /( ) uiliy funcion. ( /( ) ) σ u is a discoun rae applied o emigran σ v is a discoun rae applied o he receiving household Solving his maximizaion problem leads o he firs order condiion for a posiive level of remiances a ime : ' ' [ ( )] ( U ) / σ dr V / ( ) / ( σ )] i u U U [ v dr = 0 (2.4) Solving for he level of remiances yields an implici remiance equaion ( R = R σ, u σ, X, Y, N, i v i R ) (2.5) In a censored regression model Eq. (2.5) deermines boh he probabiliy of remiing and he level of remiances. I use a linear funcional form given by: R = α βx πz u (2.6) in which X is a vecor ha includes emigrans characerisics, Z is a vecor ha consiss of household characerisics in he recipien counry; and u is a normally disribued error erm u ~ 2 N(0, σ ). The obecive of exploring he deerminans of remiances lies in esimaing Eq. (2.6). The domain of he dependen variable is censored since he observed remiances are never negaive. Remiances are zero for a large number of observaions. I rewrie Eq. (2.6) o explicily illusrae his: R i α βx i πz ui if and only if α βx πz = u 0 oherwise i i f 0 (2.7)

27 5 such ha i =,...k ; =,... J and k is he oal number of migrans in household and J is he oal number of receiving households. In a nushell, o explore he deerminans of remiances I need o esimae Eq. (2.6). Ordinary leas squares (OLS) yield biased esimaes because of he naure of he dependen variable. Two alernaive approaches are usually adoped o esimae Eq. (2.6). The firs one is a Heckman (979) wo-sep procedure. This mehod requires ha he decision o remi is a wo-sep decision: he likelihood of remiing and he level of remiances. The second approach is a censored Tobi model. This model assumes ha he decision o remi is a one-sep decision and herefore requires ha all deerminans have he same sign effec on he likelihood and he level of remiances. In his paper he second approach is dicaed by Eq. (2.5) since i deermines boh he probabiliy of remiing and he level of remiances Daa and Esimaion Mehod Daa The daa se is from he 200 Nicaraguan Encuesa Nacional de Hogares sobre Medición de Niveles de Vida (EMNV). This is a naionally represenaive survey ha was adminisered by he Nicaraguan Insiue for Saisics and Census. The naional living sandards measuremen survey (LSMS) was esablished by he World Bank in 980. The LSMS collecs daa on many dimensions of he household well-being including consumpion, income, employmen, educaion and migraion. The daa se conains 400 households including 22,80 household members. The survey was adminisered in 200. Receiving household members were asked abou heir age,

28 6 educaion, propery, income, occupaion, businesses and any agriculural aciviies. This survey includes a remiances model where a knowledgeable member of he receiving household was asked abou oher household members who do no live in he household. All he informaion abou emigrans is exraced from heir primary receiving household. I have informaion on heir desinaion, labor force saus, age, educaion, heir relaionship o he head of he main household and also heir year of migraion. The remiances module documens 897 migrans in oal. Unlike all previous sudies I have informaion on he sender and he recipien from he same source, he original receiving household. One conribuion of his paper is ha I am able o rack informaion on boh sides of he remiance behavior from he same source. This abiliy o idenify each individual allows me o furher undersand how inra-family decisions are made regarding he allocaion of resources across households ha are separaed by he migraion of some of is members (Menívar, 995). Even wih his unique daa se I can only precisely recognize he decision o remi of migrans. I canno idenify he exac amoun of remiances sen by each migran. 4 This lack of informaion causes a problem since I canno idenify he exac amoun remied by each migran. To avoid his problem I separae migrans ino hree caegories based on heir decision o paricipae in he remiing process. The firs caegory includes migrans who do no remi such ha heir remiances are zero. The second caegory has migrans who remi bu also who belong o households wih only 4 A knowledgeable member of he receiving household was asked wheher migran i remis or no. The same member was also asked abou he moneary value of remiances ha he household received in he las 2 monhs.

29 7 one migran remiing ( s = where s is he number of remiing migrans in household ). Again I know he exac amoun ha hese migrans are sending. The hird caegory consiss of migrans who remi and who belong o muliple remiing migrans households ( s > ). In his las caegory I do no observe he exac amoun of remiances for each migran remiing. I average he oal amoun of remiances received by he original household on all he migrans who remi. The following subsecion explains in deails he esimaion mehod Esimaion Mehod I re-wrie Eq. (2.6) as he following: R = α βx πz u (2.8) i i i I ake he average of Eq. (2.8) by summing over migrans who are remiing wihin each receiving household wih muliple migrans remiing and hen dividing by o Eq. (2.9) which, hereafer, I refer o as he average model: s R = R i i s i= s s i= s i= s s s. This leads = α β X πz u (2.9) i where R is he oal amoun of remiances received by household from all remiing migrans belonging o household and s is he number of migrans who remi in household. If he number of remiing migrans s is eiher zero or one hen he model follows Eq. (2.8). Oherwise he model is defined by Eq. (2.9). Also, since u ~ i 2 N( 0, σ ) hen he new error erm s = ui s i= e is no homoskedasic wih

30 8 2 e ~ N(0, σ ). Therefore, Eq. (2.9) defines a heeroskedasic Tobi wih a known form of heeroskdeasiciy. In fac: s ( ) Var e = ( ) = Var ui Var u u2 u3 K us (2.0) s i= s Eq. (2.0) holds for all households and can be rewrien as: s 2 2 ( cov( u ; u ) ( s ) ( cov( u ; u ) = σ ( ( s ) ρ) = σ i i i s (2.) s s where is a migran oher han migran i in household 2 s, u ; u ) Var( u ) = σ cov(, i i = i cov( ui ; us ) cov( u i ; us ) = σ and corr( ui ; us ) = = ρ. The variance of he new sd( u ) * sd( u ) i error erm is a funcion of he variance of he original model in Eq. (2.8), he number of remiing migrans wihin a household and he correlaion of he error erms of differen remiing migrans who belong o he same receiving household. The correlaion coefficien ρ measures he correlaion beween u and u. posiive ρ suggess ha if migran i remis hen migran s also remis and boh remiances amouns move in he same direcion. This suggess some compeiion beween migrans wihin he same receiving household. A less aggressive hypohesis proposes ha migrans coming from he same receiving household share he same background and behave in a similar manner. If migran i sees a need o remi hen s i s 5 A migran s sees he same need and also remis and he laer is condiional on heir 5 The following condiion ρ > applies for s 2 o insure a posiive variance. s

31 9 abiliies o remi. A negaive ρ implies a negaive relaionship beween he error erms of he remiing migrans in he same household. This indirec connecion defines a crowding ou effec. The fac ha migran i is remiing discourages oher migrans in he same receiving household from remiing. This negaive relaionship migh also represen an ex-ane agreemen on he remiing behavior beween all migrans wihin he same receiving household. 6 Boh of hese cases indicae ha remiing decisions among migrans belonging o he same receiving household are inerdependen. Finally, if ρ is equal o zero hen migrans decisions o paricipae in he remiing process are independen. The coefficiens in Eqs. (2.8) and (2.9) α, β and π, are he same as he coefficiens in Eq. (2.6) which insure he same inerpreaion of he resuls. I esimae he average model using maximum likelihood esimaion. The likelihood funcion L = k i= ln L i ln L i ln L for he average model is he following: i ' ' [ Φ( X γ )] = ln[ Φ( X β *θ )] = ln if = 0 (2.2) 2 ' [ ln( θ ) ( θr X γ ) ] 2 i = 0.5* if = (2.3) s s 2 θ ' 2 ln L = 0.5 * ln θr X γ if s i ρ i > (2.4) h h h ρ h ( ) ( ) ( ) 6 A posiive correlaion coefficien can also signal an ex-ane agreemen where migrans have agreed on a remiing schedule.

32 20 where Φ (). is he sandard normal cumulaive disribuion funcion and maximize L β γ = ; θ =. 7 I σ σ wih respec o γ ; θ and ρ. In he following subsecion I describe he daa Descripive Saisics Table 2.2 presens a comparison of characerisics beween emigrans and naive populaion. Migran are wice likely o be in heir 20s relaive o naive populaion. Emigrans end o be male and more educaed and he households hey lef behind are more likely o reside in urban areas. 8 Table 2.2 Characerisics of Naive and Emigran Populaion (percenages) Nicaragua 200 Naive Emigrans Age Beween 2 and Region Urban Gender Male Educaion Less han 4 Years Proporion Remiing Mean Years Since Migraion Mean Remiances per Monh (U.S. dollars) Toal Sample Size 22, The likelihood funcion for he hird case ( >) is derived from he likelihood funcion of he second case ( s = ) wih β γ = ; σ θ σ s = ; ( h ρ( h ) 0. 5 σ = σ and h =. s 8 There is no significan difference in he years of educaion beween differen migrans who belong o he same receiving household.

33 2 The average number of years since migraion is roughly six years. The proporion of migrans remiing is 54% and he mean remiance is around US$ 40 per monh. 9 The average amoun of remiances is similar in magniude o wha Funkhouser (995) found. Table 2.3 presens characerisics of emigrans by desinaion. The main wo desinaions for Nicaraguan migrans are Cosa Rica and The Unied Saes. Table 2.3 Characerisics of Emigrans by Desinaion (percenages) Nicaragua 200 Developing Developed Age Beween 2 and Region Urban Gender Male Educaion Less han 4 Years Proporion Remiing Mean Years Since Migraion Mean Remiances per Monh (U.S. dollars) Toal Sample Size Noes: - The desinaion developing includes he following: Algeria, Argenina, Brazil, China, Colombia, Cosa Rica, Cuba, El Salvador, Guaemala, Guinea, Haii, Honduras, Mexico, Panama and Tunisia. Cosa Rica consiues 86 percen of his sample. 2- The desinaion developed includes Canada, Greece, Sweden and he Unied Saes. Cosa Rica accouns for 58% share of he Nicaraguan emigrans. The Unied Saes accouns for 28%. In Table 2.3 I define wo main subses of desinaions and include all he developing counries under developing and all he developed counries 9 I use 3.44 Nicaraguan Cordobas for one 200 U.S. dollar as an exchange rae from o he Banco Cenral de Nicaragua.

34 22 under developed. 0 Nicaraguan emigrans characerisics in developing counries are differen from hose in developed counries for gender composiion. Emigrans in developing counries come from differen regions, are less educaed and end o be more in heir 20s compared o hose in developed counries. The highes proporion of remiing is by emigrans in developed counries wih a 68%. The proporion of emigrans remiing in developing counries is less han 50%. As expeced, he average amoun remied per monh is higher for migrans living in developed counries. This number is also higher han he mean of he oal sample. This is hardly surprising because in general developed counries offer higher sandards of living, higher wages and sronger currency denominaions han any oher developing counry. Living in he U.S. or Canada for example gives emigrans a sronger remiing power which ranslaes ino higher levels of remiances. I also sress on he significan difference beween he mean years since migraion. One plausible explanaion is ha counries like he U.S. and Canada signal long erm migraion inenions due o availabiliy of opporuniies and more sable economies. In he following I resric he sample size o emigrans older han 3. 2 Table 2.4 describes he proporion of migrans remiing by relaionship o he head of he receiving household and by groups of migrans. A large porion of emigrans are he offspring of he head of he household. Only abou 5.2% of he emigrans are spouses of 0 For a complee lis of counries please refer o Table 2.3. Migrans who belong o he same muliple migran household seem o share he same desinaion counry since 80% of receiving households repor all heir migrans living in he same counry. 2 I limi he sample of emigrans in he household o hose emigrans older han 3 because he minimum age for employmen in he wo mos popular desinaions for Nicaraguan emigrans is 5 in Cosa Rica (The Cosa Rican Consiuion and The Labor Code) and 4 in he Unied Saes (Fair Labor Sandards Ac).

35 23 Table 2.4 Emigran Populaion Aged 4 and up: Proporion Remiing by Relaionship o he Head of he Receiving Household and by Group (percenages) Proporion Remiing Relaionship o he Head of he Receiving Household If Spouse of he Head of he Household If Child of he Head of he Household If Paren of he Head of he Household If Child in law of he Head of he Household If Sibling of he of he Household If Grandchild of he Head of he Household If Oher Relaionship o he Head of he Household If No Relaed o he Head of he Household Emigran Populaion by Groups Working Suden Say a-home Wife Oher Sample Size 823 he head of he household. Siblings o he head of he household form approximaely 0% of he oal number of emigrans. These groups are ranked by he closeness of he relaionship beween he migran and he head of he household from closes o farhes. This ranking also coincides wih he ranking of he fracion of emigrans remiing excep for he las group, no relaed, where I noe a surprising 58% remiing. I suppose eiher a srong friendship or some invesmen opporuniies behind his high proporion.

36 24 Table 2.4 also separaes he emigran populaion ino four subgroups. Approximaely 80% of emigrans have a ob. Around 64% of he working emigrans send money back home. As expeced he proporion of sudens remiing is 30% much smaller han hose migrans working. In Table 2.5 I discuss he characerisics of he receiving households by number of remiing emigrans per household. Ou of 495 households ha have a leas one migran living abroad, 4% have a leas one migran remiing. I seems ha here is a negaive relaionship beween he number of remiing migrans and he percenage of working head of households. Secion 2.5 presens he resuls. Table 2.5 Disribuion of Receiving Households and Their Characerisics by Number of Sending Emigrans Aged 4 and up (percenages) Number of Sending Emigrans per Household Percenage of Households Mean Years of Educaion of he Head of he Household Percenage Residing in Urban Areas Percenage Working Head of Household or more All Households Sample Size of Receiving Households 495

37 Resuls In his secion I discuss he se of independen variables ha form X and Z in Eq. (2.6). A he remiing decision sage a migran s individual characerisics play a maor role. I include migrans age, gender, level of schooling, employmen saus, desinaion, years living abroad and he relaionship o he head of he receiving household. Likewise I expec he receiving household s aribues o have an effec on he migran s remiing decisions. I include he receiving household area of residence, he labor saus and he years of educaion of he head of he household. Table 2.6 ouline hree model specificaions wih differen subse of independen variables for he average model presened in Secion 2.4. Column () shows he maximum likelihood esimaes of a heeroskedasic Tobi on boh emigran and receiving household characerisics. Column (2) includes a se of emigran s characerisics while column (3) includes he receiving household s characerisics. 3 Migran s gender, he labor force saus, desinaion and he relaionship o he head of he receiving household all significanly affec he remiing behavior. The educaion level and he labor saus of he head of he household also affec he remiing process. 4 One of he conribuions of his paper is quanifying he resuls. Table 2.7 decomposes he heeroskedasic Tobi coefficiens ino wo effecs: a change in he probabiliy of remiing and a percenage change in he amoun remied. 3 A lis of definiions and descripive saisics of he variables is in Appendix A. All he independen variables are in discree form. The dependen variable is he logarihm of monhly remiances measured in 200 U.S. dollars. 4 The variables of migran educaion and years since migraion are defined around heir median which is four years. The variable of head of household educaion is defined around four for comparison purpose wih he migran educaion variable (he median is hree for his variable).

38 26 Table 2.6 Maximum Likelihood Esimaes for a Heeroskedasic Tobi Average Model of he Amoun Remied by Emigrans Aged 4 and up Amoun Remied Variables () (2) (3) Inercep ***.993*** (0.648) (0.533) (0.424) Emigran Characerisics if Age beween 2 and (0.269) (0.27) if Male * ** - (0.265) (0.268) if Educaion less han 4 Years (0.29) (0.283) if Years since Migraion less han (0.255) (0.258) if Working 2.200*** 2.25*** - (0.357) (0.364) if Emigran Resides in a Dev. Counry.273***.327*** - (0.294) (0.294) if Spouse of he Head of he Household 2.032*** 2.032*** - (0.55) (0.539) if Paren of he Head of he Household.549**.830*** - (0.654) (0.688) if Child of he Head of he Household 0.866*** 0.905*** - (0.294) (0.297) Household Characerisics if Urban Residence ** (0.284) (0.286) if Educaion of HHH less han ** *** (0.282) (0.286) if Head of he Household Working *** *** (0.239) (.29) Log Likelihood Thea = θ 0.36*** 0.355*** 0.33*** (0.03) (0.02) (0.009) Rho = ρ 0.34* 0.392** 0.392** (0.68) (0.80) (0.72) Sample

39 27 Table 2.7 Summary of The Change in Amoun of Remiances and Change in Probabiliy of Remiing Resuls of Model () in Table 2.6 Average Model Percenage Change in Probabiliy Percenage Change in Amoun Variables Inercep Emigran Characerisics if Age beween 2 and if Male if Educaion less han 4 Years if Years since Migraion less han if Working if Emigran Resides in a Dev. Counry if Spouse of he Head of he Household if Paren of he Head of he Household if Child of he Head of he Household Household Characerisics if Urban Residence if Educaion of HHH less han if Head of he Household Working Male migrans are less likely o remi. The probabiliy of remiing decreases by around 7% for male migrans. These findings srenghen he belief of gender differences in he remiing behavior. Migrans who have a ob are 32% more likely o remi han hose who are no working. Also he percenage change in he level of remiances is a large increase of 77% for working migrans. Living in he U.S. or Canada increases boh he probabiliy (8%) and he percenage change in he amoun of remiances (44%).

40 28 The labor saus and he desinaion of he migran seem o have a significan role in he remiing behavior for Nicaraguans. Togeher hey shape he remiing abiliy of migrans. The probabiliy and amoun of remiances increase for migrans belonging o he nuclear family. The increase in he probabiliy and he amoun is he larges for he migrans who are he spouse or he paren (29% o 22% for he probabiliy and 7% o 54% for he change in he amoun) compared o migrans who are he child of he head of he household (2% o 30%). The difference in hese magniudes is mos likely explained by he responsibiliy ha spouses and parens share oward he receiving household. Spouses and parens share he responsibiliy of providing for he receiving household while his responsibiliy is no ha eviden for child migrans. Also he large difference of he percen change in he amoun beween a migran spouse and a migran paren srenghens his hypohesis since i also illusraes he difference beween he roles of parens and spouses (Menívar e al., 998, page 04). From he household characerisics, he likelihood of remiing and he percenage change in he amoun remied decrease for migrans belonging o a receiving household wih a head of household who repors less han four years of educaion (0% and 25% respecively). The same resuls apply for receiving households wih a working head (4% and 33% respecively). A working head of he receiving household signals a sable source of income and possibly less need for financial help. The main conribuion of he paper is quanifying he ineracion among remiers wihin he same receiving household. Social ineracions and decision making are opics

41 29 of huge ineress for economiss. In fac, he lieraure on he ineracion among members of he same household is exensive (see, for example, Becker, 974; Bergsorm, 989). However, no paper has exended his lieraure ino he heory of migran remiance behavior. I measure he remiance behavior beween migrans belonging o he same household. The correlaion coefficien ρ is posiive for all hree regressions. This coefficien is significanly differen from zero a he 5% significance level and is around 0.39 for wo of he hree regressions. In order o measure ρ wih more precision I reesimae he average model wih fewer conrols. Table 2.8 Maximum Likelihood Esimaes for a Heeroskedasic Tobi Average Model of he Amoun Remied by Emigrans Aged 4 and up: Fewer Conrols A B C D E F Variables Inercep -.673*** (0.380) (0.28) (0.202) (0.206) (0.98) (0.200) if Working 2.327*** (0.369) if in Dev. Counry.483*** (0.260) if Paren of HHH (0.796) if Spouse of HHH.80*** (0.532) if Sibling of HHH -.58** (0.468) Thea = θ 0.333*** 0.329*** 0.323*** 0.330*** 0.325*** 0.322*** Rho = ρ (0.00) (0.009) (0.009) (0.009) (0.009) (0.009) 0.438*** 0.338** 0.475*** 0.593*** 0.505*** 0.470*** (0.60) (0.62) (0.69) (0.96) (0.75) (0.70) Log Likelihood Sample Noe: - * denoes significance a 0% level, ** a 5% level, *** a % level. 2- HHH means head of he receiving household 3-Dev. Counry refers o developed counry.