The Impact of Remittances on Poverty and Human Capital: Evidence from Latin American Household Surveys

Transcription

1 Publc Dsclosure Authorzed The Impact of Remttances on Poverty and Human Captal: Evdence from Latn Amercan Household Surveys WPS4247 Publc Dsclosure Authorzed Publc Dsclosure Authorzed Pablo Acosta, (Corporacón Andna de Fomento) Pablo Fajnzylber and J. Humberto Lopez * (The World Bank) Abstract: Ths paper explores the mpact of remttances on poverty, educaton, and health n eleven Latn Amercan countres usng natonally representatve household surveys and makng an explct attempt to account for one of the nherent costs assocated wth mgraton: the potental ncome that the mgrant may have made at home. The man fndngs of the study are the followng: () regardless of the counterfactual used remttances appear to lower poverty levels n most recpent countres; () yet, despte ths general tendency the estmated mpacts tend to be modest; and () there s sgnfcant country heterogenety n the poverty reducton mpact of remttances flows. Among the aspects that have been dentfed n the paper that may lead to varyng outcomes across countres are the percentage of households reportng remttances ncome, the share of remttances recpent households belongng to the lowest quntles of the ncome dstrbuton, and the relatve mportance of remttances flows wth respect to GDP. Whle remttances tend to have postve effects on educaton and health, ths mpact s often restrcted to specfc groups of the populaton. Keywords: remttances, poverty, human captal. Publc Dsclosure Authorzed World Bank Polcy Research Workng Paper 4247, June 2007 The Polcy Research Workng Paper Seres dssemnates the fndngs of work n progress to encourage the exchange of deas about development ssues. An objectve of the seres s to get the fndngs out quckly, even f the presentatons are less than fully polshed. The papers carry the names of the authors and should be cted accordngly. The fndngs, nterpretatons, and conclusons expressed n ths paper are entrely those of the authors. They do not necessarly represent the vew of the World Bank, ts Executve Drectors, or the countres they represent. Polcy Research Workng Papers are avalable onlne at * The vews expressed n ths paper should not be attrbuted to the Corporacón Andna de Fomento, the World Bank, ts Executve Drectors or the countres they represent. 1

2 I. Introducton Workers remttances to developng countres have experenced a dramatc ncrease over the past few years. Accordng to the World Bank s Global Economc Prospects 2006, n 1990 remttances to mddle and low-ncome countres amounted to about US$30 bllon. Ffteen years later, they are estmated to have reached almost US$170 bllon mplyng annual growth rates that are well above 10 percent. Remttances now account for about 30 percent of total fnancal flows to the developng world and provde sgnfcant foregn exchange earnngs. They are more than twce as large as offcal development assstance flows, and represent the equvalent to 2.5 percent of the gross natonal ncome of the developng world. There s also some evdence ndcatng that remttances flows have a number of postve macroeconomc effects n recpent countres: large remttances flows mprove a country s credtworthness for external borrowng and hence enhance ts access to nternatonal captal markets. Smlarly, there s evdence that some fnancal nsttutons n developng countres have been able to tap nternatonal captal markets at relatvely favorable condtons through securtzaton of future flows. Gven the nature, magntude, and evoluton of these flows t should not be a surprse that remttances are now seen by development practtoners as havng a potentally mportant role to play n supportng the development efforts of recpent countres. There are two man channels n whch remttances can support these efforts. Frst, remttances could flow towards the needest groups of the populaton and therefore drectly contrbute to poverty reducton. Thus t s possble that even f these flows are fully consumed, a concern mentoned by a number of practtoners, they have sgnfcant postve welfare effects. Second, wth mperfect nsurance and fnancal markets, remttances can also contrbute to hgher nvestment n human and physcal captal. For example, these flows can remove some of the fnancal constrants faced by households and small-scale entrepreneurs that prevented them from nvestng. Smlarly, remttances can provde nsurance and therefore allow households and entrepreneurs to pursue more rsky asset accumulaton strateges. Thus n ths regard, remttances may also potentally contrbute to rasng the country s long run growth potental through hgher rates of captal accumulaton. Unfortunately, n practce, these potental postve mpacts may be counterbalanced by other more challengng effects. For example, f there are mportant costs assocated to the act of mgratng t s possble that mgrants do not come from the lowest quntles of the ncome dstrbuton and therefore that remttances do not flow towards the poorest. In fact, as dscussed below n the Latn Amercan context, when one takes nto account both the locaton of recpent households n the ncome dstrbuton, and the magntude of the correspondng flows t s dffcult to defend that remttances are drected manly towards the poor. In those cases, one should not expect remttances to have a large effect on poverty headcounts. Moreover, comparng observed poverty rates wth those calculated on the bass of non-remttances ncome may exaggerate the estmated development mpact of these flows. Indeed, before leavng ther home countres mgrants are lkely to have contrbuted to ther households ncome, whch should be taken nto account when calculatng counterfactual poverty rates n a scenaro of nether mgraton nor remttances. Adams (2006), for example, argues that falure to correct for the reducton n ncome assocated wth the absence of 2

3 mgrants from ther households can dramatcally change the estmated poverty mpact of remttances. Wth data from Guatemala, Adams (2004) llustrates how the ncorporaton or not of ths correcton can lead to ether postve or negatve effects of remttances on poverty. Remttances can also negatvely affect domestc compettveness lowerng the expected returns on captal. For example, remttances can exert pressure on the exchange rate and lead to a real apprecaton somethng that ceters parbus would lower the proftablty of the tradable sector (n a Latn Amercan context see Amuedo-Dorantes and Pozo, 2004, for evdence n ths regard). Smlarly, remttances may rase reservaton wages and negatvely affect labor supply (Rodrguez and Tognson, 2001). In both cases, remttances would affect the nvestment ncentves of households and entrepreneurs and lower the rate of captal accumulaton. In other words, gven the potental counterbalancng effects assocated to a surge n remttances, t may be qute dffcult to determne not only the magntude of the potental development mpacts of remttances but also even the drecton of those mpacts. Thus, emprcal evdence s needed to ascertan the sgns and orders of magntude of the dfferent economc consequences of remttances flows. The exstng emprcal evdence on the varous development mpacts of remttances s, however, stll qute lmted. True, there are a number of studes 1 whch have analyzed the poverty mpact of remttances 2 explotng cross natonal databases (see for example Adams and Page, 2005, and Acosta et al. 2007) or explorng the ssue n specfc country settngs (see for example Adams 2004, 2006 respectvely on Guatemala and Ghana, Lopez-Cordova, 2005; and Taylor, et al., 2005, on Mexco; and Yang and Martnez, 2005, on the Phlppnes) concludng that n general hgher remttances nflows tend to be assocated wth lower poverty ndcators. Yet there are a number of ssues that have to be consdered. Frst, cross country studes tend to offer global results but they do not fully explot dfferences n country characterstcs that may affect the development mpact of remttances. In partcular, ther results can be msleadng n specfc country contexts f there are sgnfcant dfferences n mgraton patterns across countres and regons, and those are not taken nto account n the analyss. Moreover, cross country studes usually take as counterfactual the hypothess of no remttances wthout alterng mgraton patterns rather than the more appealng counterfactual of no mgraton (.e. they underestmate the costs of mgraton and hence may overestmate the benefts of remttances). For example, some of the estmates obtaned usng cross country analyses suggest that that a 10 percent ncrease n per capta remttances may lead to a 3.5 percent declne n poverty. Clearly, ths estmate does not seem very realstc especally when 1 For comprehensve revews see also the Internatonal Monetary Fund s World Economc Outlook 2005, the World Bank s Global Economc Prospects 2006, and the Inter-Amercan Development Bank s Beyond Small Changes: Makng Mgrants Remttances Count all of whch have as ts central topc the economc and development mplcatons of remttances and mgraton. 2 The exstng lterature has also focused on the mpact of remttances on other economc varables and measures of development. See for example the works of Yang (2006) for the mpact of remttances on ntertemporal consumpton smoothng, Ruz Arranz and Gulano (2005) on growth; Amuedo-Dorantes and Pozo (2004) on rsk management; Cox and Ureta (2003) and Hanson and Woodruff (2003) on educaton; Rodrguez and Tognson (2001) on labor supply; Woodruff and Zenteno (2007) on entrepreneurshp; and Amuedo- Dorantes and Pozo (2004) and Rajan and Subramaan (2005) on external compettveness. 3

4 one consders ts mplcatons: wth ths elastcty and the observed growth rates n per capta remttances observed durng the past 15 years (above 10 percent), the exstng poverty level n 1990 should have been almost halved by now just because of the remttances effect. Country specfc studes, on the other hand, allow for rcher analyses but may offer a narrow vew of the problem at hand f the exstng stock of such studes s very lmted. Moreover, the comparson of results may be problematc f the analyses of the varous avalable studes rely on dverse methodologes and assumptons, so that t s dffcult to dsentangle whether the observed cross country dfferences are due to varaton n country characterstcs or to methodologcal dfferences across studes. Aganst ths background, ths paper contrbutes to ths debate by presentng an analyss of the development mpact of remttances n eleven Latn Amercan countres: Bolva, Domncan Republc, Ecuador, El Salvador, Guatemala, Hat, Honduras, Mexco, Ncaragua, Paraguay and Peru. These are the countres for whch the respectve natonally representatve household surveys contan nformaton of the remttances receved by the dfferent households. In ths regard, the paper adds to the exstng lterature along several dmensons. Frst, our results offer a comprehensve vew of the poverty mpact of workers remttances n the Latn Amercan regon. In our analyss we focus on the most commonly used poverty measure -headcount poverty- and explore how the results change when one uses dfferent poverty lnes and when one moves from nternatonally to natonally defned poverty lnes. In ths regard, our analyss extends and completes that n Acosta et al. (2007). Whle due to data lmtatons we do not analyze all the countres n the regon that accordng to Balance of Payments statstcs receve non-trval amounts of remttances flows, our workng sample stll covers more than two-thrds of the man remttances recpent countres n Latn Amerca. Ths s even more mportant when one takes nto account that Latn Amerca s, together wth East Asa and the Pacfc, at the top of the rankng of remttances recevng regons n the world. Second, unlke other studes on the topc whch tend to focus ether on the mpact of remttances on poverty or on the mpact of remttances on measures of household welfare that go beyond the pure ncome dmenson such as educaton or health levels, our analyss looks at both ssues. In the case of the mpact of remttances on educaton we contnue to work wth eleven countres, although data lmtatons now prevent us from ncludng Bolva and nstead allow us ncludng Jamaca. In the case of health, our sample s much more lmted and has only two countres: Guatemala and Ncaragua. Thrd, for each of the development ndcators on whch we focus our analyss s based on a common methodology for all the countres so that dfferences n outcomes have to emerge from dfferences n the pattern and volume of remttances rather than from methodologcal bases. Thus, our results can be used to assess potental heterogenety n the responses to a surge n remttances. To the best of our knowledge ths s the frst exercse of ths type. In ths respect, the paper brngs together the vrtues of cross country and country specfc analyss. It has country specfcty n the sense that each country s analyzed n 4

5 solaton takng nto account all the possble country characterstcs. And t brngs a cross country perspectve because of the relatvely large number of countres beng analyzed usng a common methodology, whch permts an analyss of country heterogenety. Fourth, at a techncal level, we present several exercses that are based on the same assumptons. In some cases we use as a counterfactual the scenaro of mgraton wthout remttances; n others, we take as a counterfactual a stuaton wthout mgraton and hence correct for the potental ncome that mgrants mght have had f they had not left ther country. To ths end, we estmate an econometrc model relatng non remttances ncome to a set of household and household head characterstcs. In addton, n what s our preferred smulaton, we also take nto account that households wth mgrants may not be randomly selected from the populaton and allow for possble selecton bas. Moreover, followng Schff (2006), we also present the results of estmatng the mpact of remttances on the poverty rates prevalng only among the households recevng those flows. Fnally, when estmatng the effects of remttances on human captal formaton, we control for the counterfactual ncome of recpent households estmated for the purpose of assessng the mpact of remttances on poverty. To antcpate some of the results below, we fnd that n general remttances tend to have an overall postve mpact on recpent economes. Remttances seem to reduce poverty levels, ncrease educatonal attanment and contrbute to mprovements n health ndcators. It s thus understandable that polcy makers are ncreasngly nterested n seeng an ncrease n the amount of remttances flowng to ther countres. Yet, n general the estmated mpacts tend to be modest and show sgnfcant country heterogenety. In ths regard, our results ndcate that n about half of the countres analyzed remttances do not have a sgnfcant mpact on poverty. The man two reasons that explan ths fndng are the followng. Frst, n some of the countres we analyze a large share of the remttances recevng populaton belongs to the top quntles of the ncome dstrbuton, whch naturally reduces the potental poverty reducng effects of remttances flows. Second, correctng for the reducton n ncome that would be assocated wth the absence of mgrants from ther households has a sgnfcant mpact. In fact, estmates that do not correct for ths factor suggest sgnfcantly larger declnes n poverty rates n all countres. We also fnd that chldren from households reportng remttances are more lkely to stay at school and that controllng for pre-mgraton ncome chldren from remttances recevng households tend to exhbt hgher health outcomes than those from non-recpents households wth smlar demographc and soco-economc characterstcs (although n ths case the evdence s based on only two countres). These mpacts, however, tend to be restrcted to households wth relatvely low levels of schoolng and ncome. The rest of the paper s structured as follows. In secton II, we revew the data and econometrc methodology used n the paper. Secton III presents the emprcal results regardng the mpact of remttances on headcount poverty. Secton IV moves beyond the ncome dmenson of poverty and explores how remttances affect human captal accumulaton. Fnally, secton V offers concludng remarks. 5

6 II. Data and methodology II.1 Data The dramatc ncrease n remttances observed at the global level over the past few years has been mrrored n Latn Amerca. In fact, offcally recorded remttances flows to the regon have ncreased more than twenty-fold snce 1980 when remttances amounted to about $1.9 bllon to close to US$50 bllon n These trends have placed Latn Amerca (together wth East Asa and the Pacfc) at the top of the rankng of regons by total amount of remttances and more clearly by per capta remttances receved. Moreover, accordng to World Bank 2006 on current trends remttances to the regon could reach about US$75 bllon n Whle ths estmate s lkely to be on the hgh sde of what one could expect, t nevertheless hghlghts the fact that a collapse n remttances does not seem very lkely n the near future. In other words, t s not unrealstc to assume that remttances are here to stay, at least n the short run. Wthn Latn Amerca, remttances are partcularly mportant n some countres. For example, they represent more than 50 percent of GDP n Hat, a rato that s the hghest n the world, and between 15 and 20 percent of GDP n Jamaca, Honduras, and El Salvador (see fgure 1.A). Smlarly, n Guatemala, Ncaragua and the Domncan Republc remttances are between 10 and 12 percent of GDP. The mportance of those flows can also be llustrated by comparson wth other prvate captal flows. Thus, n Guatemala, Honduras, El Salvador and the Domncan Republc, remttances are equvalent to respectvely 14, 4, 3 and 2 tmes FDI flows. Even n Colomba and Ecuador, where n relatve terms remttances are lower, remttances represent respectvely 197 and 112 percent of FDI. On a US$ bass, the Latn Amercan country wth the hghest absolute remttances flows s Mexco, whch receved almost $22 bllon n 2005 (about $18 bllon n 2004, as reported n fgure 1.B). Ths would represent 45 percent of total flows to Latn Amerca n that year ($48.3 bllon) and would make Mexco the largest world recpent n Other countres wth substantal flows are Colomba and Brazl wth flows of respectvely $3.8 and $3.5 bllon n 2005, and Guatemala, El Salvador, the Domncan Republc and Hat wth $2.5 bllon on average. Also worthy of note are Ecuador, Honduras, Jamaca and Peru, where remttances averaged about $1.5 bllon n Fnally, on a per capta bass the country wth the hghest level of remttances would be Jamaca wth approxmately $550 per capta, followed by Barbados wth about $400 per capta and El Salvador wth flows of approxmately $350 (see fgure 1.C). The average for the 28 countres beng consdered here would be $128 per capta per year, but that amount ncreases to $270 among the ten countres wth hghest per capta remttances a group that also ncludes the Domncan Republc, Hat, Antgua and Barbuda, Guatemala, Mexco, Honduras, and Trndad and Tobago. The data just descrbed reveal the mportance of remttances n the dfferent Latn Amercan countres. However, f nterest centers on the characterstcs of those recevng remttances the BOP data descrbed above s not adequate and one has to resort to other sources. In ths regard, the natural canddates are natonally representatve household surveys. A problem wth ths alternatve s that not all the Latn Amercan countres have 6

7 surveys contanng relevant nformaton on remttances. In fact, there are only 11 countres for whch such data are avalable. Table 1 reports the countres for whch the household survey allows an analyss of the poverty mpact of remttances, together wth some nformaton regardng the year of the survey, and the survey used. Overall, remttances to these countres would have amounted n 2004 to US$33.5 bllon, or about 70 percent of total remttances to the regon. Fgure 1. Remttances to Latn Amerca n 2004 Panel A. (% of GDP) % of GDP Hat Jamaca Honduras El Salvador Dom. Rep. Ncaragua Guatemala Ecuador Barbados Colomba Paraguay Mexco Bolva Antgua Costa Rca Domnca Surname Belze Peru Trndad St. Ktts St. Vncent Panama Argentna Brazl St. Luca Chle Uruguay Panel B. (US$ mllons) Mllons US$ Mexco Colomba Brazl Guatemala El Salvador Dom. Rep. Hat Ecuador Jamaca Honduras Peru Argentna Ncaragua Costa Rca Paraguay Trndad Bolva Barbados Panama Uruguay Chle Surname Belze Antgua Domnca St. Ktts St. Vncent St. Luca 7

8 US$ Jamaca Barbados El Salvador Dom. Rep. Hat Antgua Guatemala Mexco Honduras Panel C. Per capta Trndad Ecuador Source: Balance of Payments (macroeconomc) data. St. Ktts Ncaragua Table 1. Household Surveys Used n the Paper Country Year Survey Bolva 2002 Encuesta de Hogares, MECOVI Ecuador 2004 Sstema Integrado de Encuestas de Hogares, INEC El Salvador 2000 Encuesta de Hogares de Propóstos Múltples, MECOVI Guatemala 2000 Encuesta Naconal sobre Condcones de Vda (ENCOVI) Hat 2001 Les Condtons de Ve en Hat, IHSI/Fafo/PNUD Honduras 2002 Encuesta Permanente de Hogares de Propóstos Múltples, MECOVI Mexco 2002 Encuesta Naconal de Ingresos y Gastos de los Hogares, INEGI Ncaragua 2001 Encuesta Naconal de Hogares sobre Medcón de Nvel de Vda, INEC Paraguay 2003 Encuesta Permanente de Hogares, Dreccón General de Estadístca Peru 2002 Encuesta Naconal de Hogares, INEI Rep. Domncana 2004 Encuesta Naconal sobre Condcones de Vda (ENCOVI) One natural queston that may arse at ths stage regards the degree of consstency between the BOP based data and the survey based data. Ths s mportant n ths context because sgnfcant and non-systematc dscrepances between the two data sources would rase a number of questons regardng the relablty of the remttances data and therefore of the results. In practce, however, we fnd that whle there are sgnfcant dfferences between BOP and household survey estmates of remttances flows, those dfferences are at least systematc.e. the orderng of countres on the bass of remttances flows relatve to GDP s very smlar regardless of the data source. Indeed, Acosta et al. (2006) fnd that there s a stable relatonshp across countres between the two sources of data, wth BOP based estmates surpassng household survey data estmates by about 73 percent (wth an R2 of 0.8 obtaned for the correspondng regresson lne lnkng the two seres). That s, household survey based estmates of remttances flows (as a percent of household ncome) are about 40 percent lower than BOP based estmates (as a percent of GDP). Whle n prncple these dfferences are lkely to be motvated by reasons smlar to those behnd the exstng Costa Rca Colomba Belze Domnca Surname Paraguay Peru St. Vncent Panama Argentna Bolva Brazl St. Luca Uruguay Chle 8

9 dscrepances between household survey and natonal accounts based ncome estmates 3 we cannot deny that t s also possble that household surveys are not fully representatve of the populaton of remttances recpents and hence that they may underestmate the relevance of remttances flows. Ths s an mportant caveat that should be kept n mnd when analyzng the results presented n ths paper, because n that case we could be under-estmatng the poverty reducng effects of remttances (although n a smlar way across countres). Wth these elements n mnd, the frst queston whch we would lke to respond on the bass of household survey data has to do wth the characterstcs of those recevng remttances n the sample of Latn Amercan countres under consderaton. Fgure 2. Share of households recevng remttances Percent Hat Dom.Rep. El Salv. Ncaragua Honduras Guatemala Mexco Ecuador Paraguay Bolva Peru Source: Household surveys To start wth, fgure 2 reports the share of households recevng remttances n each of the eleven countres n our sample, suggestng that ths share vares sgnfcantly. For example, n Hat more than 25 percent of the households reported havng receved remttances n At the other extreme, only 3 percent of the Peruvan households would beneft from these flows. In between, remttances reach between 10 and 25 percent of the households n the Domncan Republc, El Salvador, Ncaragua, and Honduras; between 5 and 10 percent n Mexco and Guatemala; and between 3 and 5 percent n Bolva, Ecuador and Paraguay. Thus, other thngs beng equal one should expect that a gven amount of remttances could have qute dfferent effects on the welfare of the populaton of countres such as for nstance Peru and Hat. Cross country dfferences n the welfare effects of remttances become even more lkely once one consders that not all the households recevng remttances are poor, and that the share of remttance recpents lvng n poverty vares consderably across countres. Ths s apparent from Fgure 3 whch plots the percentage of households recevng remttances by quntle of the (non-remttances) ncome dstrbuton. The fgure ndcates that whle n some countres the recpents of remttances are predomnantly poor (61 percent of Mexcan 3 For example, Deaton (2005) estmates that on average survey ncome s less than 60 percent of GDP and dscusses the reasons that may be behnd survey and natonal accounts based estmates dscrepances. 9

10 households recevng remttances fall n the frst ncome quntle; 42 percent of Paraguayan recpents are n the frst quntle of the dstrbuton), ths s not the case n all countres. In Peru and Ncaragua, for example, the dstrbuton of remttances recpents across households s completely dfferent. In Peru less than 6 percent of the households that receve remttances belong to the lowest quntle whle 40 percent belong to the top quntle. Smlarly, n Ncaragua only 12 percent of the recpents are n the frst quntle whle 33 percent belong to the ffth quntle. Thus n these two countres remttances seem to be flowng towards the rchest. Note that the nteracton of these fgures wth the share of total households recevng remttances n the country would mply that less than.2 percent of the Peruvan populaton s made up of relatvely poor households.e. frst quntle of the ncome dstrbuton recevng remttances, and less than 2 percent of Ncaraguan households are remttances recpents and belong to the lowest quntle. Thus, n these two countres t s clear that even f we assume that the remttances receved by each household are enough to take the household above the poverty lne, the potental of remttances to lower poverty levels wll be lmted. Even n Mexco, where a majorty of households recevng remttances belong to the lowest quntle, households recevng remttances and belongng to the lowest quntle would represent only 3 percent of the natonal populaton. Once agan, ths gves an dea of the orders of magntude that we can fnd n the exercses below. In between the groups of Mexco and Paraguay and the group of Peru and Ncaragua, there are three countres (Ecuador, El Salvador, and Guatemala) where recpents tend to be n lower quntles (although n a less dramatc way than Mexco and Paraguay) and four countres (Bolva, Honduras, the Domncan Republc and Hat) where remttances appear to be homogeneously dstrbuted across the dstrbuton of ncome, or exhbt a U-shaped dstrbuton (.e. remttances flow to a larger extent towards the poorest and the rchest quntles). On the whole, the two countres where more than 5 percent of households pertan both to the group of remttances recpents and to the lowest quntle of the populaton are El Salvador and Hat. II. 2 Econometrc Methodology The data descrbed n the prevous sub-secton provdes a frm footng to asses the mpact of remttances on poverty. Frst, ths data allows makng mechancal comparsons of poverty rates calculated wth remttances ether excluded or ncluded n total reported household ncome. These basc comparsons, however, suffer from one mportant shortcomng, namely that remttances are not lkely to be an exogenous transfer but rather a substtute for the home earnngs that mgrants would have had f they had not decded to leave ther countres to work abroad. In fact, the non-remttances ncome reported by households wth mgrants cannot be consdered a good representaton of the stuaton of the famly pror to mgraton. If the mgrant had postve earnngs before leavng the household, t s lkely that the household s total non-remttances ncome s lower after mgraton. Thus, estmatng the effect of mgraton and remttances on poverty would requre takng nto consderaton the counterfactual per capta ncome that the household would have had f the mgrant had stayed at home, otherwse we would be overstatng the true mpact of mgraton and remttances on poverty reducton. 10

11 Fgure 3. Households recevng remttances by quntle of the non-remttances ncome dstrbuton Bolva Ecuador Q1 Q2 Q3 Q4 Q5 Q1 Q2 Q3 Q4 Q5 El Salvador Guatemala Q1 Q2 Q3 Q4 Q5 Q1 Q2 Q3 Q4 Q5 Hat Honduras Q1 Q2 Q3 Q4 Q5 0 Q1 Q2 Q3 Q4 Q5 Mexco Ncaragua Q1 Q2 Q3 Q4 Q Q1 Q2 Q3 Q4 Q5 Paraguay Peru Q1 Q2 Q3 Q4 Q Q1 Q2 Q3 Q4 Q5 Domncan Republc Q1 Q2 Q3 Q4 Q5 The fgure reports the percentage of households recevng remttances that fall n each of the fve quntles of the ncome dstrbuton. Source: Own calculatons usng the last avalable household survey. 11

12 One possble way to address ths ssue s as n Acosta et al. (2007) who mpute per capta household ncome for mgrant households n the counterfactual scenaro of no remttances and no mgraton. Clearly, ths requres nformaton about the ncome of the household before the mgrant left, and ths nformaton s n general not avalable drectly from household surveys. As an alternatve, Acosta et al. (2007) nfer the counterfactual per capta ncome levels for those households wth remttances on the bass of a reduced-form specfcaton for the determnants of ncome among households wthout remttances. More formally, ths approach nvolves estmatng a model lke: log Y = α + βx + γh + μ (1) where Y represents per capta non-remttances ncome, X s a vector of household characterstcs (demographc and locaton covarates), H s a set of characterstcs of the household head, and μ represents random shocks and possble unobserved heterogenety n ncome generaton. Equaton (1) can be estmated usng the sub-sample of households that do not receve remttances. The estmated coeffcents then allow predctng the counterfactual non-mgraton ncome for remttances-recpent households. In the absence of nformaton on mgrant characterstcs whch s the case of all the countres consdered n ths paper except Hat and Ncaragua t s necessary to make some basc assumptons about the number and the demographcs of mgrants. Followng Rodrguez (1998) and Acosta et al. (2007), we assume that, on average, remttances are sent by a sngle adult male famly member. Another needed assumpton has to do wth the educaton level of the mgrant, whch we assume to be equal to the average years of educaton of other adults n the household. One addtonal aspect that needs attenton s that OLS estmates of equaton (1) wll be nconsstent f μ s not ndependently dentcally dstrbuted (..d.). In other words, f mgrants are not randomly selected from the pool of households, estmates of equaton (1) based on the sample of households wthout mgrants could suffer from selecton bas. To control for ths possblty, we add to (1) a varable that represents the household s propensty to not mgrate or not receve remttances (M *). Ths new term s calculated n the context of the two-step estmaton framework proposed by Heckman (1979). In partcular, we adopt the followng model specfcaton: M * α H + ωz + υ (2) = 1 + β1x + γ 1 (no-remttances selecton rule) logy = 2 + β 2 X + γ 2 α H + θλ + ε (earnngs equaton for non-recpent households) (3) In practce, one can only observe the sgn of the varable M *, whch represents the selecton rule for no mgraton or not recevng remttances.e. whether the household receves remttances or not, whch n turns s equvalent to a negatve or postve value for M *. Identfcaton of ths model requres an excluson restrcton: a set of varables Z that are 12

13 related to the mgraton and remttances choce but whch do not drectly affect earnngs for households wth non-mgrants. The term λ s the selecton nverse Mll s rato, defned as: φ λ = 1 Φ ( α1 + β1x + γ 1H + ωz ) ( α + β X + γ H + ωz ) wth θλ ( ν / μ > β X γ H ωz ) = E 1 1 where μ s the error component n the earnng equatons and θ = cov( ν, μ ) / var( μ ). Controllng for λ allows the remanng unexplaned component ε to have the usual desrable..d. propertes. If λ s a sgnfcant predctor of earnngs, t suggests that the selecton nto the non-mgraton status s ndeed correlated wth factors that affect household earnngs and that OLS estmates of equaton (1) would be nconsstent. The excluson restrctons that we employ for the non-remttances selecton equaton are an ndex of household assets, the percentage of households that receve remttances n the respectve county of resdence (a proxy for the presence of mgrant networks), and ther nteracton. When county level ndcators cannot be calculated (Guatemala and Hat), the varable that represents the percentage of households wth mgrants s measured at the department/provnce level. In ths case, only the nteracton of that varable wth the household assets ndex s ncluded. Fnally, when nformaton on household assets s mssng (Bolva, Ecuador and Honduras), the network varable s nteracted wth the number of adult males, whch ensures varablty at the household level. 4 Wth the assumpton stated before concernng mgrant characterstcs, these coeffcents are used n order to mpute the counterfactual non-remttances per capta household ncome for recpent households. Wth ths varable one can proceed to calculate the levels of poverty and nequalty that would have prevaled had mgraton and remttances not taken place. It must be noted, however, that as mentoned by Rodrguez (1998) the 4 The percentage of households recevng remttances n the same county (or provnce) s a proxy for mgraton networks at the county (provnce) levels. Prevous lterature has used mgraton networks as nstruments for mgraton and remttances n order to assess ther mpact on development outcomes (Hanson and Woodruff, 2003; Lopez-Cordova, 2005; Acosta, 2006; Acosta et al., 2006; Woodruff and Zenteno, 2007). In the countres under analyss, the correlaton between the fracton of recpent households n each county or provnce and the per capta ncome of non-recpent households n the same area s non-sgnfcantly dfferent from zero n 9 out of 11 cases (where t vares from to 0.043) and t s stll very low n the other 2 cases (0.14 n Hat and 0.16 n Peru). 13

14 varance of the counterfactual ncome predcted on the bass of observable household characterstcs s artfcally small, because t gnores unobserved determnants of ncome. A potental soluton proposed by Barham and Boucher (1998) s to add to the predcted household ncome a random error component drawn from a dstrbuton wth the same propertes (mean, varance) of the actual estmated errors. We pursue ths approach and obtan 1,000 dfferent estmates of the mputed counterfactual non-remttances ncome for households wth mgrants, and the same number of estmates for the poverty and nequalty levels that would have prevaled n the above counterfactual scenaro. Ths allows us to report not only pont estmates for those varables but also 95% confdent ntervals, based on the 25th and 975th estmates of the varables after sortng them n ascendng order. III. Results We start ths secton by revewng the estmates correspondng to the two-step model descrbed n equatons (2) and (3) whch are reported n Tables 2 and 3. Unlke n Barham and Boucher (1998) and Adams (2006), we fnd that λ s postve and sgnfcant at the 1 percent level n all cases except Ecuador, an ndcaton that recpents of remttances are not randomly selected from the complete pool of households. Moreover, ths result suggests that households wth a lower propensty to mgrate are also more lkely to have hgher per capta ncome levels, whch s consstent wth the standard mgraton argument that potental mgrants compare the returns at home and n ther potental destnaton when decdng whether to work abroad. The rest of the coeffcents have the expected sgns. In the selecton equaton, larger households (partcularly those wth a hgher number of adult males) are more lkely to receve remttances (or mgrate). On the other hand, the number of chldren and adult females s negatvely related to the lkelhood of mgraton and remttances recepts. Hgher educatonal levels are negatvely related to remttances, except n Hat and Honduras (where educaton has the opposte sgn). The presence of mgrant networks, proxed by the percentage of households wth remttances n the area of resdence, and a household wealth ndex also seem postvely and sgnfcantly correlated wth the lkelhood of beng a recpent. In the earnngs equaton, as expected, educaton and demographc characterstcs are strong sgnfcant predctors of ncome n non-recpent households, explanng between 33 and 55 percent of ncome varablty dependng on the case. 14

15 Table 2. Two-Step Estmates of Non-Remttances Household Income Auxlary Regressons - Per Capta Income for HHs Not Recpent of Remttances. Heckman's Two Step Results Dependent Varable: Log of Per Capta HH Income Varables Mexco Guatemala El Salvador Paraguay Honduras Dom. Rep. Bolva Ecuador Hat Peru Ncaragua HH Sze *** *** *** *** *** *** ** *** *** *** *** (0.015) (0.039) (0.021) (0.026) (0.020) (0.035) (0.042) (0.017) (0.043) (0.013) (0.045) Number 0-5 years old chldren *** *** *** * ** *** *** (0.017) (0.042) (0.023) (0.029) (0.021) (0.038) (0.046) (0.019) (0.048) (0.015) (0.049) Number of 6-15 years old boys *** *** ** ** *** 0.161*** *** (0.017) (0.042) (0.023) (0.028) (0.021) (0.038) (0.046) (0.019) (0.047) (0.014) (0.048) Number of 6-15 years old grls *** *** ** ** *** 0.128*** *** (0.017) (0.042) (0.023) (0.028) (0.021) (0.038) (0.046) (0.019) (0.047) (0.014) (0.048) Number of years old males 0.041*** ** 0.061** 0.135*** 0.103*** *** *** 0.073*** 0.147*** (0.016) (0.043) (0.022) (0.026) (0.020) (0.037) (0.043) (0.018) (0.047) (0.013) (0.048) Number of years old females 0.040** 0.154*** 0.228*** *** *** 0.192*** 0.082*** 0.181*** (0.016) (0.041) (0.022) (0.027) (0.020) (0.036) (0.043) (0.018) (0.045) (0.013) (0.047) Rural Area *** *** *** *** *** *** 0.134*** *** (0.031) (0.018) (0.020) (0.015) (0.023) (0.030) (0.015) (0.049) (0.037) Avg. Adult Educaton (16-65 years old) 0.074*** 0.102*** 0.071*** 0.055*** 0.073*** 0.072*** 0.056*** 0.045*** 0.046*** 0.068*** 0.066*** (0.002) (0.006) (0.003) (0.004) (0.003) (0.005) (0.005) (0.003) (0.009) (0.002) (0.007) Age (HH Head) 0.032*** 0.043*** 0.023*** *** 0.042*** 0.018*** 0.011*** 0.019*** 0.011*** 0.017*** (0.002) (0.006) (0.003) (0.004) (0.003) (0.004) (0.005) (0.003) (0.006) (0.002) (0.006) Age Squared x 100 (HH Head) *** *** *** *** *** *** *** ** (0.002) (0.007) (0.003) (0.004) (0.003) (0.005) (0.006) (0.003) (0.006) (0.002) (0.006) Years of Educaton (HH Head) 0.042*** 0.038*** 0.030*** 0.051*** 0.036*** 0.014*** 0.047*** 0.045*** 0.029*** 0.027*** 0.020*** (0.002) (0.005) (0.003) (0.003) (0.003) (0.004) (0.005) (0.002) (0.007) (0.002) (0.006) Lambda 0.379*** 2.245*** 0.915*** 0.457*** 0.300*** 0.600*** 0.382** *** 1.333*** 0.718*** (0.045) (0.184) (0.049) (0.132) (0.060) (0.068) (0.152) (0.090) (0.140) (0.076) (0.114) Department/Provnce Indcators Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Observatons R-Squared *Sgnfcant at 10% level. **Sgnfcant at 5%level. ***Sgnfcant at 1% level. 15

16 Table 3. Two-Step Estmates of Non-Remttances Household Income Auxlary Regressons - Frst Step Results Dependent Varable: HHs Not Recpent of Remttances Indcator Varables Mexco Guatemala El Salvador Paraguay Honduras Dom. Rep. Bolva Ecuador Hat Peru Ncaragua HH Sze *** *** *** *** *** *** *** (0.053) (0.072) (0.031) (0.062) (0.035) (0.051) (0.095) (0.049) (0.047) (0.066) (0.076) Number 0-5 years old chldren 0.105* *** 0.257*** 0.277*** 0.226*** 0.552*** 0.096* 0.281*** 0.172*** (0.060) (0.079) (0.035) (0.072) (0.038) (0.057) (0.112) (0.056) (0.053) (0.081) (0.085) Number of 6-15 years old boys 0.150** *** 0.228*** 0.216*** 0.181*** 0.484*** *** (0.060) (0.078) (0.036) (0.071) (0.038) (0.057) (0.108) (0.055) (0.052) (0.078) (0.082) Number of 6-15 years old grls 0.159*** *** 0.236*** 0.184*** 0.190*** 0.434*** *** (0.060) (0.079) (0.035) (0.070) (0.038) (0.057) (0.107) (0.055) (0.052) (0.077) (0.083) Number of years old males *** *** *** * * *** *** *** *** *** (0.055) (0.074) (0.033) (0.064) (0.042) (0.053) (0.105) (0.054) (0.048) (0.068) (0.080) Number of years old females 0.190*** *** 0.302*** 0.196*** 0.188*** 0.361*** 0.166*** 0.237*** 0.188*** 0.176** (0.056) (0.076) (0.033) (0.067) (0.037) (0.054) (0.098) (0.052) (0.049) (0.068) (0.078) Rural Area * *** 0.174*** 0.168*** *** 0.178*** *** (0.066) (0.034) (0.066) (0.031) (0.040) (0.097) (0.045) (0.055) (0.076) Avg. Adult Educaton (16-65 years old) 0.071*** 0.047*** 0.042*** 0.028*** *** 0.030*** 0.054*** 0.018** *** (0.009) (0.013) (0.005) (0.010) (0.006) (0.007) (0.014) (0.007) (0.009) (0.010) (0.013) Age (HH Head) 0.035*** 0.080*** 0.059*** 0.031*** 0.039*** 0.077*** 0.048*** 0.034*** 0.030*** 0.045*** (0.009) (0.010) (0.005) (0.010) (0.005) (0.007) (0.014) (0.007) (0.006) (0.012) (0.012) Age Squared x 100 (HH Head) *** *** *** *** *** *** *** *** *** *** ** (0.008) (0.011) (0.005) (0.010) (0.005) (0.007) (0.014) (0.007) (0.006) (0.011) (0.011) Years of Educaton (HH Head) 0.040*** 0.036*** 0.015*** *** *** * ** *** (0.007) (0.011) (0.005) (0.010) (0.005) (0.006) (0.011) (0.007) (0.007) (0.008) (0.010) HH Asset Index *** *** *** * *** *** *** *** (0.018) (0.043) (0.017) (0.027) (0.025) (0.043) (0.023) (0.034) % Rem. Recpent HHs n County or Provnc *** *** *** *** *** *** *** *** *** (0.151) (0.199) (0.540) (0.331) (0.220) (1.258) (1.006) (0.659) (0.351) % Rem. Recpent HHs * HH Asset Index ** 0.765*** 0.240* (0.083) (1.385) (0.064) (0.280) (0.093) (0.125) (0.221) (0.138) % Rem. Recpent HHs * Number males *** *** *** (0.163) (0.732) (0.615) Department/Provnce Indcators Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Observatons Pseudo R-Squared *Sgnfcant at 10% level. **Sgnfcant at 5%level. ***Sgnfcant at 1% level. 16

17 Havng revewed the results of the auxlary regressons we are now n a poston to present the results of our poverty smulatons, whch are reported n Tables 4 to 7. As noted above, our analyss s based on headcount poverty ndexes, and t employs four poverty lnes. In tables 4 and 5 we use nternatonally comparable poverty lnes of US$1 and US$2 per person per day, measured at PPP values, whch correspond respectvely to extreme and moderate poverty rates. In tables 6 and 7 we base our analyss on natonally defned extreme and moderate poverty lnes. Readers nterested n country comparsons should probably be more concerned wth the results of tables 4 and 5 whereas readers wth nterest n a specfc country should focus on tables 6 and 7. Before proceedng, t must be noted that some authors (Schff, 2006) have proposed to complement the analyss of the effect of remttances on natonal poverty rates, by estmatng ther mpact on the poverty rate prevalng among the subset of households whch receve remttances. Ther pont s that whle the mpact of remttances on natonal poverty levels may be lmted, the effect on the poverty status of the households wth mgrants could be much larger and therefore ths restrcted analyss can hghlght a number of nterestng facts for the correspondng sub-group of the populaton. Thus n the tables below we also report poverty effects usng Schff s approach. Each table has two panels. Panel A reports: () the natonal headcount poverty rate accordng to the relevant poverty lne; () the poverty rate prevalng among remttances recpent households. Panel B reports: () the dfference between the natonal poverty rate calculated usng non-remttances household ncome and the observed poverty rate; () the same calculaton as n () for the specfc sample of remttances recpent households; () the dfference between the natonal poverty rate calculated usng counterfactual household ncome n the scenaro of no mgraton nor remttances correctng for selectvty and change n household sze; (v) the same calculaton as n () for the specfc sample of remttances recpent households. 5 5 Whle the results presented n the last two columns of panel B n tables 4 to 7 are our preferred estmates, estmated poverty effects of remttances calculated wthout correctng for selectvty or changes n household sze can be found n an appendx avalable from the authors upon request. Those results may be relevant for readers nterested n assessng the separate mpact on the results based on counterfactual ncome estmates, of correctng frst for selectvty and second for changes n household sze. 17

18 Table 4. Impact of remttances on headcount poverty (US$1 PPP) Panel A: Observed Poverty Rates Panel B: Estmated Poverty Effects of Remttances wth and wthout Adjustment for Lost Income Country All Households Recpent HHs Country Wthout adjustment (all HHs) Wthout adjustment (recpent HHs) Wth adjustment (all HHs) Wth adjustment (recpent HHs) (1) (2) (1) (2) (3) (4) Bolva Bolva (-0.4; -0.1) (-11.6; -1.8) Domncan Republc Domncan Republc (0.2; 0.5) (1.0; 2.1) Ecuador Ecuador (-0.4; -0.2) (-5.8; -3.0) El Salvador El Salvador (-0.6; -0.3) (-3.2; -1.7) Guatemala Guatemala (-1.3; -0.7) (-14.0; -7.4) Hat Hat (-4.7; -3.5) (-17.4; -12.6) Honduras Honduras (-0.7; -0.5) (-6.5; -4.5) Mexco Mexco (0.0; 0.2) (0.8; 2.8) Ncaragua Ncaragua (-0.1; 0.1) (-0.7; 0.9) Paraguay Paraguay (-0.1; 0.1) (-2.1; 1.8) Peru Peru (0.0; 0.0) (--0.7; 0.0) 18

19 Table 5. Impact of remttances on headcount poverty (US$2 PPP) Panel A: Observed Poverty Rates Panel B: Estmated Poverty Effects of Remttances wth and wthout Adjustment for Lost Income Country All Households Recpent HHs Country Wthout adjustment (all HHs) Wthout adjustment (recpent HHs) Wth adjustment (all HHs) Wth adjustment (recpent HHs) (1) (2) (1) (2) (3) (4) Bolva Bolva (-0.6; -0.1) (-17.0; -4.2) Domncan Republc Domncan Republc (-0.3; -0.1) (-1.5; 0.6) Ecuador Ecuador (-0.9; -0.6) (-14.5; -9.6) El Salvador El Salvador (-1.6; -1.1) (-8.5; -5.9) Guatemala Guatemala (-2.0; -1.3) (-20.9; -13.2) Hat Hat (-3.6; -2.4) (-13.1; -8.7) Honduras Honduras (-1.7; -1.4) (-15.3; -12.3) Mexco Mexco (-0.1; 0.3) (0.9; 5.1) Ncaragua Ncaragua (-0.5; -0.1) (-3.4; -0.6) Paraguay Paraguay (-0.2; 0.1) (-4.2; 2.8) Peru Peru (-0.1; 0.0) (-1.7; 0.5) 19

20 Table 6. Impact of remttances on headcount poverty (natonally defned extreme poverty lne) Panel A: Observed Poverty Rates Panel B: Estmated Poverty Effects of Remttances wth and wthout Adjustment for Lost Income Country All Households Recpent HHs Country Wthout adjustment (all HHs) Wthout adjustment (recpent HHs) Wth adjustment (all HHs) Wth adjustment (recpent HHs) (1) (2) (1) (2) (3) (4) Bolva Bolva (-0.5; -0.1) (-15.8; -2-6) Domncan Republc Domncan Republc (-0.9; -0.2) (-3.8; -1.0) Ecuador Ecuador (-0.9; -0.6) (-13.4; -8.8) El Salvador El Salvador (-1.5; -1.0) (-7.8; -5.2) Guatemala Guatemala (-1.6; -0.9) (-17.1; -9.5) Hat Hat (-3.1; -2.0) (-11.5; -7.3) Honduras Honduras (-1.8; -1.5) (-16.6; -13.6) Mexco Mexco (0.0; 0.3) (-0.3; 5.2) Ncaragua Ncaragua (-0.5; 0.0) (-3.2; 0.0) Paraguay Paraguay (-0.3; 0.0) (-7.4; 1.1) Peru Peru (-0.1; 0.0) (-2.6; -0.5) 20

21 Table 7. Impact of remttances on headcount poverty (natonally defned moderate poverty lne) Panel A: Observed Poverty Rates Panel B: Estmated Poverty Effects of Remttances wth and wthout Adjustment for Lost Income Country All Households Recpent HHs Country Wthout adjustment (all HHs) Wthout adjustment (recpent HHs) Wth adjustment (all HHs) Wth adjustment (recpent HHs) (1) (2) (1) (2) (3) (4) Bolva Bolva (-0.7; -0.2) (-11.7; -2.8) Domncan Republc Domncan Republc (-1.2; -0.3) (-5.3; -1.3) Ecuador Ecuador (-1.2; -0.8) (-18.5; -12.3) El Salvador El Salvador (-1.6; -1.0) (-8.3; -5.0) Guatemala Guatemala (-1.3; -0.5) (-13.1; -5.3) Hat Hat (-0.8; 0.1) (-3.1; 0.3) Honduras Honduras (-2.3; -1.9) (-20.7; -17.5) Mexco Mexco (-0.1; 0.3) (-1.5; 4.3) Ncaragua Ncaragua (-1.0; -0.4) (-6-2; -2.4) Paraguay Paraguay (-0.4; 0.1) (-9.6; 1.5) Peru Peru (-0.2; 0.0) (-5.9; -0.1) 21