Remittances and Vulnerability to Poverty in Rural Mexico

World Development Vol. 38, No. 6, pp. 828–839, 2010� 2010 Elsevier Ltd. All rights reserved

0305-750X/$ - see front matter

www.elsevier.com/locate/worlddevdoi:10.1016/j.worlddev.2010.02.002

Remittances and Vulnerability to Poverty in Rural Mexico

ALEJANDRO DE LA FUENTE *

World Bank, Washington DC, USA

Summary. — Remittances have been portrayed as the human face of globalization given their potential to alleviate poverty by directlyincreasing household income. Using a panel of rural households in Mexico from October 1998 to November 2000 this study assesseswhether this is in fact the case. However, rather than examining whether remittances income would reduce future consumption povertywe asked if remittances are likely to reach people whose conditions are prone to worsen in the future. We found a negative and statis-tically significant relationship between the disbursement of remittances and the threat to future poverty that rural households could expe-rience.� 2010 Elsevier Ltd. All rights reserved.

Key words — globalization, risk, vulnerability, poverty, private transfers, Mexico

* I am indebted to Stefan Dercon and Valpy Fitzgerald for their helpful

comments on earlier versions of this draft. This paper has also benefited

from comments by Machiko Nissanke, Erik Thorbecke, Sabina Alkire,

Adrian Wood, Proochista Ariana, and two anonymous referees. The views

expressed in this article do not necessarily reflect the views of the World

Bank. Final revision accepted: January 23, 2009.

1. GLOBALIZATION, REMITTANCES, AND POVERTY

One of the most salient consequences of globalization hasbeen the surge of remittances from developed to developingareas mediated through migratory processes (remittances areunderstood as transfers of money by foreign workers to theirhome countries hereafter). According to the World Bank’sGlobal Economic Prospects, 2006, in 1990 workers’ remit-tances to middle and low income countries amounted to aboutUS$31 billion. Fifteen years later, they are estimated to havereached US$167 billion. They have become a major sourceof financing for developing countries and are especially impor-tant in Latin America and the Caribbean (LAC), where theyrepresent about 70% of foreign direct investment (FDI) andare five times larger than Official Development Assistance(Fajnzylber & Humberto, 2007).

In Mexico, migration to the United States has gained prom-inence during the last couple of decades leading to unprece-dented flows of remittances to the extent that now thecountry tops the ranking of remittances in the world. 1 Workerremittances have continued to exceed revenues from tourismsince 1998. In 2009, tourism garnered US$8.6 billion, whileremittances hovered around US$21 billion. Likewise, remit-tances have already surpassed FDI in the recent past. In2009, remittances almost doubled the US$1.4 billion sent inFDI (see Figure 1).

If remittances have become the human face of globalizationas migrants commit portions of their income to help theirhouseholds in the host and the home country (Orozco,2002a; Orozco, 2002b), then—the argument goes—given thatthey often go to poor households and directly increase thesehousehold’s level of income, they could become a massive re-source transfer mechanism, able to bring down poverty lev-els. 2 Following this logic it has become commonplace toassume that a higher and more regular reliance on remittancesrepresents an opportunity for agricultural households to be-come less vulnerable to poverty and risks.

Nevertheless, those in more need are not always the onesthat receive remittances (Escobar Latapı, 2006), either becausethey go to better-off households or simply because the demandfor help from poor households is turned down. When there isno immediate need to cope with risks, the activation of remit-tances would also not necessarily be aligned to the needs facedby the recipient even if there is a permanent risk of staying

828

poor or becoming poorer due to the household’s own struc-tural characteristics, but rather depend on who is the net pro-vider of insurance, relative risk preferences, and relativebargaining power (Lucas & Stark, 1985). According to Lucasand Stark (1985) when the migrant is a net provider and per-ceives no reason to continue doing so (i.e., if the recipienthousehold is poorly endowed, has limited command over themigrant and thus weak negotiating power to maintain him/her in the contract) then he/she might be tempted to stop send-ing remittances. We therefore anticipate that the existence oftransnational support networks cannot be taken for grantedas an effective tool to mitigate poverty since their activationdepends in part on the type of risks faced, the constraints im-posed on the donor by its own circumstances and the presenceof nonaltruistic motivations.

Based on these counter-arguments we reconsider the pre-sumption that remittances in rural areas reach the “right peo-ple” from a poverty-reduction perspective and ask whetherremittances actually reach those rural households in Mexicowhose conditions are likely to worsen in the future.

We consider the relationship between poverty and remit-tances within a dynamic context that takes into account theimpact that risk-induced fluctuations and other more perma-nent disadvantages at the household and community levelcan have on the future welfare status of households (as as-sessed through consumption) and the subsequent implicationsthis could have for the reception of remittances. Householdscan send money because we assume implicitly that their mem-bers, including the permanent migrants living abroad as for-mer members, have a good knowledge of the process thatleads to consumption outcomes. Even if they ignore the stateof the world that will prevail in each point of time they canrecognize welfare-deteriorating conditions or events whichcould lead to future poverty and respond accordingly. Oncethe household-specific estimates for future poverty are ob-tained we will look at the main features of remittances and

http://dx.doi.org/10.1016/j.worlddev.2010.02.002

0

5

10

15

20

25

30

35

1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009Year

U.S

. Bill

ion

dolla

rs

Foreign Direct Investment Workers Remittances Revenues from Tourism

Figure 1. Foreign currency flows into Mexico. Source: Bank of Mexico.

REMITTANCES AND VULNERABILITY TO POVERTY IN RURAL MEXICO 829

those engaged in them in our sample. We then bring bothpieces of information together to create various models onthe incidence of remittances and their levels. The results ofthese models are then discussed before concluding.

2. VULNERABILITY

(a) Definition

Vulnerability in this work is understood as the magnitude ofthe threat of future poverty that a household experiences at agiven point in time (Calvo & Dercon, 2005). This definition ofvulnerability makes it different, but closely related to the con-cept of poverty. Poverty measures are generally fixed in time,making poverty itself essentially a static concept. By contrast,vulnerability here aims to identify households at risk of wel-fare shortfalls in the future based on their current standing.In this sense, it is an ex ante forward-looking measure. Thethreat component in the definition alerts us about the central-ity of risk. Given that risks refer to possibly occurring eventsthat can damage well-being, it is their presence that is associ-ated with a threat to welfare until they materialize or vanish.Finally, the threat of poverty is experienced in very differentways by each household, according to its own human andphysical stock as well as the type of risks it faces. This leadsto varying welfare circumstances across households that willbe represented in the magnitude of the threat. Hence this mag-nitude is compounded by the likelihood of becoming poor(er)and the depth of such impact. The measure takes the followingform:

V ðp; c; zÞ ¼Xn

h¼1

phvðxhÞ; where xh ¼~ch

z; ð1Þ

where v(�) is a monotonic decreasing and convex function, z isa standard poverty line, c denotes the vector containing differ-ent states of the world for the relevant welfare outcome (in thiscase consumption per capita) and p is the probability of eachoccurring. ~ch ¼ minðz; cÞ represents the standard characteristicof poverty measures under which values above the poverty linedo not get any weight, in order to satisfy the focus axiom. 3 Byusing ~ch we then ensure that ch has no effect on vulnerability

when c > z. Therefore, xh becomes a ratio of coverage of min-imal needs (assuming that z conveys a reliable monetary valueof the cost of obtaining a basket of goods and services consid-ered adequate to satisfy a group of basic needs) that takes val-ues between 0 and 1 and v(x) is its transformation to make itappropriate for the desired feature of an increase (decrease) inrisk with its corresponding upward (downward) shift in vul-nerability. Given the need to stress the role of risk in the vul-nerability function one can rewrite (1) in the following way:

V aðp; c; zÞ ¼ 1� E½xah�; where 0 < a < 1: ð2Þ

Here a can be interpreted as the coefficient of relative riskaversion, where xh is defined as before and takes the form ofan exponential function to fulfill the condition of risk increas-ing when the mean value spreads from the middle to the tails.Therefore this normalized ratio of minimal needs obtained foreach household by dividing total consumption per capita bythe poverty line becomes adjusted by the risk aversion param-eter. The expectations operator will summarize the weightedratios by attaching probabilities to their occurrence. Finally,one minus this adjusted needs ratio transforms it into a mono-tonically decreasing convex function which achieves again thedesired nature of risk-sensitivity in the measure of vulnerabil-ity. 4

(b) Measurement

From (2) first, and most importantly, one has to estimate anex ante probability distribution of future consumption foreach household E(Fc). This is because at the core of the notionof vulnerability lies the possibility that from the current stand-ing of a household in time t, one could say something about itswelfare prospects in t + 1. This E(Fc) distribution will beformed by a series of potential consumption values (~ch) whichwill provide the raw material for coming up with the desiredoutcome against which one could assess the welfare level ofeach household.

We have suggested elsewhere (de la Fuente, 2008, chap. 4)that consumption for the group of households analyzed in thispaper could be broke down into fixed and time-varying com-ponents. In the first case, this would reflect seasonal traits orpatterns that are permanent or slowly changing, such as the

830 WORLD DEVELOPMENT

endowments of the village where households reside. The tran-sient component, for its part, would express the imperfectionsof coping mechanisms within households (their ability tosmooth consumption) in the face of shocks. Bearing these fea-tures in mind, one can express the process that generates con-sumption in the following way:

~cth ¼ lh þ bc�t

h þ acth þ et

h: ð3ÞIn (3) the superscript �t expresses a time-average and t standsfor time-varying. Thus, total consumption would result fromthe sum of a fixed component over time and a time-specifictransient component. Based on (3), the consumption-generat-ing process for our group of households took the following lin-ear specification:

cth ¼ uh þ

X

e

deDe þX

h

bhSth þ cX t

h þ eth; ð4Þ

Where per capita consumption (as left-hand-side variable)was run over three rounds of a longitudinal household survey(see Section 3) spaced one year apart and thus controlling forany seasonality effects (i.e., October 1998, November 1999,and November 2000), based on a fixed-effects regression mod-el. The need to rely on a Fixed Effects estimation was reas-serted given the presence of endogeneity (i.e., the correlationbetween unobserved effects and our vector of controls wasconfirmed), and the presence of unobserved household-specifictime-invariant characteristics (the Hausman test was rejected)(Greene, 2003). The subscript e refers to state entities whereour group of rural households h reside, and the superscript tdenotes time periods; uh captures the influence of unobservedtime-invariant household-specific characteristics on predictedconsumption; D is a vector of dummy variables, identifyingeach state separately by survey round (entity-round interac-tion terms) to account for fixed effect state-level conditions;S are observed locally covariate shocks (drought, flood, andso on) and idiosyncratic shocks at the household level overtime (illness of the main income earner and children); X is avector of household (e.g., household size, dependency ratio,head’s age, gender and occupation) and community character-istics (availability of health clinic and number of productiveactivities in the village), and asset endowments (e.g., accessto land and share of working members at home); and e is ahousehold-specific error term capturing changes in unobserv-able components of household preferences (time-varying dis-turbance term). The vector of parameters to be estimated isu, d, b, and c.

For space reasons, the above specification and a discussionof its results are not reported. Yet, it is worth mentioning thatthe model was statistically significant (at a 99 confidence levelaccording to the F test). More importantly, the indices of vul-nerability per household that will be obtained later, and arebased on this estimates of consumption, were robust to differ-ent sub-samples and specifications. 5

Once the relevant parameters were estimated we proceed toobtain a distribution of future consumption per household. 6

With the scope of our panel dataset (three rounds), we foundinsufficient the three recorded expenditures per household asenough observations (i.e., “states of the world”) to come upwith a sensible statistic (i.e., mean) of what the expected con-sumption outcome for each household would be. For this rea-son, we opted to use a simulation methodology that involvedadditional parameterization as an alternative for coming upwith household-specific distributions of consumption. Thissimulation followed a technique similar in spirit to that de-scribed by Kamanou and Morduch (2005) in their study ofvulnerability for a panel of households in Cote d’Ivoire in

the period 1985–88 and by Kuhl (2005) in his study of vulner-ability for a three-round panel of rural households in Ethiopia.Our methodology implied performing a Monte Carlo boot-strap simulation in the following way:

(i) From the sample of households, we obtained the time-average values of consumption and its covariates over thethree rounds ð�ch; �Dh; �Sh; �X hÞ and then predicted the time-invariant component of consumption for each householdbased on such values and the parameters obtained in regres-sion (4):

~c�th ¼ uh þ

X

e

d�DþX

i

bi�Si

h þ c�X h þ eh: ð5Þ

(ii) We then predicted the stochastic component of con-sumption vt

h (which contains a time-varying part c^

th along with

residuals eth) by subtracting the predicted fixed part of con-

sumption for every household across time cth from the ob-

served level of consumption for that same household in eachperiod of time:

vth ¼ ct

h � c�th; where vt

h ¼ c^

th þ et

h: ð6Þ

(iii) We next drew B bootstrap samples from the resultingstochastic consumption distributions per round to formMonte Carlo estimates of total consumption. Each new sam-ple B was formed by taking the values of stochastic consump-tion across households and re-sampling them withreplacement. This procedure was repeated n number of timesand for the three rounds of observations. We decided toresample 250 times per round ending up with 750 new sam-ples. 7 For illustrative purposes, sample 1 in round 1 wouldlook like below:

Bðvt¼1h Þ ¼ vt¼1;B¼1

1 ; vt¼1;B¼12 ; . . . ; vt¼1;B¼1

H

� �: ð7Þ

(iv) Each sample served to carry out a simulation where thestochastic component of consumption vt

h was randomly as-signed to all households. This was possible precisely due toits stochastic nature. The simulations were carried out at thelocality level to keep consistency with the fact that the stochas-tic distribution of consumption whence we draw the samplesremains homogenous across the group of households in thevillage. Hence, after each re-sampling was done we took fixedconsumption in each sample and added it to the relocatedtime-varying consumption values until arriving at the simu-lated value of total consumption for every household in eachround:

~ctBh ¼ c�t

h þ vtBh 8h; t: ð8Þ

(v) Once the simulated vectors of consumption values wereobtained (B = 250 values for every household per round)and thus the distribution of future consumption, we obtainedan expected consumption per household for two points in time(i.e., November 1999 and November 2000).

With the welfare benchmarks (adjusted poverty lines z foreach period of time) and parameter of risk aversion a alreadyavailable, we were then in a position to obtain the rate of cov-erage of minimal needs for each household. Recalling from thedefinition in (1), this ratio by construction will be bounded be-tween 0 and 1 as all levels of consumption beyond z are cen-sored to fulfill the focus axiom. Once we had the risk-weighted minimal needs ratio for each household we pro-ceeded to compute the vulnerability estimates by subtractingthem from one.

Considering we want to say something about future con-sumption starting from our baseline in October 1998, theabove procedure allowed us to end up with estimations for

Table 1. Levels of household vulnerability. Source: Own calculations usingdata from ENCEL99N, ENCEL00N

Percentiles November 1999 November 2000

5% 0.131 0.13525% 0.248 0.26150% 0.343 0.35675% 0.445 0.46295% 0.620 0.63599% 0.991 0.956Mean 0.354 0.367Standard deviation 0.149 0.152

Note: Difference in means statistically significant at 0.01%.


every household in both November 1999 and 2000. In Table 1we observe that the levels of vulnerability rose steadilythroughout this interlude for most households, including thosewith relatively low levels of exposure to poverty threats, froman average level of vulnerability of 0.353 in November 1999 to0.367 by the end of 2000.

3. DATA

The main source of information for this paper comes from agroup of 18,893 households located in 506 localities across se-ven states of Mexico: Guerrero, Hidalgo, Michoacan, Puebla,Queretaro, San Luis Potosı and Veracruz. They comprisesome of the most deprived rural communities in the countryand were set to be the prime target of the poverty alleviationProgram in Education, Health, and Nutrition (Progresa andsince 2002 Oportunidades). 8 A broader sample was initiallysurveyed in August 1997 (Household Socioeconomic Charac-teristics Survey—ENCASEH for its Spanish acronym) at theonset of the program to determine its beneficiaries. Subse-quently, a series of socio-economic household surveys (House-hold Evaluation Surveys—ENCEL for its Spanish acronym)were carried out approximately every six months for a periodof three years (from March 1998 to November 2000) with thepurpose of providing inputs for assessing the program’s per-formance. These successive questionnaires plus two newrounds undertaken in 2003 and 2007 make up a total of eightrounds of information for the same group of households.These surveys were supplemented with questionnaires fromthe localities where households reside (Household EvaluationSurveys Locality Questionnaires—ENCELLO for its Spanishacronym).

The ENCEL looks at households and their communities withgreat detail. For households, it provides information on expen-diture and income levels, labor market participation and socio-demographic characteristics, asset ownership and familywealth, and even information on private transfers, includingremittances between co-nationals. At the community level, it in-cludes the socio-economic characteristics of localities, such astheir access and distance to physical, health, and educationalpublic infrastructure. Furthermore, one can find informationon the incidence and impact of health and weather-relatedshocks as well as the responses to them, both at the householdand community levels. This level of comprehensiveness alongwith the fact that the same group of households are tracked overtime make the ENCEL especially attractive for studying pat-terns of vulnerability in these communities.

It was also decided to obtain and process a detailed accountof rainfall records expressed on a monthly basis for a period of40 years (from 1961 to 2000) coming from meteorological sta-tions spread across the localities under study. We make use ofthe full rainfall distribution obtained to instrument our main

variable of interest (i.e., vulnerability to poverty) as the prone-ness to future poverty might present a reciprocal causationwith transfers—for empirical applications of rainfall as aninstrumental variable for economic conditions at household le-vel, see Dercon and Christiaensen (2007). This is explained indetail in Section 4.

Despite its ample coverage of many topics, the ENCEL didnot apply the same questions in each round of the survey.Each contains a common group of enquiries on the demo-graphic composition of households and their socio-economicstatus, but many other questions were replaced by round-spe-cific modules designed to capture other aspects of interest topolicymakers. As a result, most of the eight waves of thehousehold questionnaires have a core set of identical questionsand a series of modules on different topics. This situation re-stricts comparison across rounds. In the end, the main datasource for this paper will be three rounds of the ENCELspaced one year apart (i.e., October 1998, November 1999,and November 2000) that contain detailed modules on privatetransfers, including information on remittances.

(a) Remittances

Each of the three rounds in the ENCEL contains data ondifferent types of inter-household transfers in two separate sec-tions. The first asks if the household has received and/or givenany type of support (cash, food, clothing, or work) in the lastmonth, from and/or to friends and relatives, including thoseliving abroad. A second section is devoted exclusively to per-manent migrants. It enquires about any form of support re-ceived, but over the last six months, and only from thosewho had lived with the family and left within the five yearsprior to the survey date. In both sections, only cash transfersare given a monetary value. Likewise, both sections containinformation on some characteristics of donors, includingwhether they live inside or outside the country. Hence, onecan gather information on remittances from both modules.

It should be said at this stage that while the focus of this pa-per is on remittances and vulnerability, we choose to explainthe construction and report the incidence of private transfers(of which remittances are always part) in this section. Thiswould be especially relevant to assess whether the presenceof a system of reciprocity extending beyond the family withincommunities (and captured by the mobilization of transfers)could make up for the absence of remittance flows to the mostvulnerable households if this were to happen.

For the construction of the remittance (transfer) variables,we aggregated all individual information into the household.Since any remittance (transfer) comprises both its activationand the actual level mobilized, we first created a dummy thattakes a value of one if any remittance (transfer) had been re-ceived. Second, we calculated the net level of remittances(monetary transfers) exchanged. This is almost equivalent tothe amount of remittances received, as the percentage ofhouseholds giving support back is negligible. To construct thisvariable we summed the information regarding inward andoutward remittances (monetary transfers) given by or to out-siders over the previous month, adjusting the permanent mi-grant’s remittances to a monthly basis for comparability.Some double counting of remittances (monetary transfers) ispossible since those permanent migrants who provided cashcould be the same relatives that donated to the household inthe last month. The remittance (monetary transfer) levels var-iable was thus constructed as an aggregate of the two sectionsreferred to above, netting out double counting as others havedone previously (Teruel & Davis, 2000).

Table 2. Incidence and levels of private transfers and remittances

October 1998 November 1999 November 2000

Total T C Total T C Total T C

No. of observations 18,892 11,389 7,503 18,893 11,388 7,505 18,888 11,385 7,503% households that

Received transfer from anyone 10.62 10.47 10.9 8.63 8.45 8.9 7.88 7.95 7.78From family 9.96 9.86 10.1 7.99 7.78 8.31 7.15 7.33 6.89From friends 0.59 0.55 0.65 0.56 0.58 0.53 0.74 0.58 0.97

Gave transfer to anyone 0.79 0.75 0.85 0.16 0.13 0.2 0.26 0.29 0.21

% of households that

Received from abroadAny type of transfer 2.57 2.54 2.62 2.13 1.99 2.35 2.81 2.82 2.8Monetary transfer 2.52 2.49 2.55 2.05 1.94 2.21 2.71 2.71 2.71Nonmonetary transfer 0.14 0.14 0.13 0.16 0.11 0.24 0.15 0.16 0.13

Gave transfer to anyone abroad 0 0 0 0 0 0 0.02 0.02 0

No. of observations 18,892 11,389 7,503 18,893 11,388 7,505 18,888 11,385 7,503Transfers received (as % of average consumption)

Over all householdsa 2.92 4.75 7.37 3.06 4.88 8.03 3.95 6.52 10.15Conditional on transfersb 34.27 31.84 37.93 50.88 47.04 57.01 62.41 58.46 68.76

Remittances (as % of average consumption)

Over all householdsb 1.60 2.61 4.12 1.86 2.97 4.93 2.26 3.63 5.65Conditional on remittancesc 61.38 58.65 65.60 92.78 83.67 106.46 84.07 80.99 88.97

T = treatment, C = control.a All changes significant at 1%, except between October 1998 and November 2000.b All changes significant at 1%, except between October 1998 and November 1999.c Only change significant at 1% is between October 1998 and November 2000.

Table 3. Vulnerability to poverty (VTP) by quartiles

VTP November 1999 VTP November 2000

Unit 1 2 3 4 1 2 3 4

Consumption per capita $ 327.22 225.85 176.98 136.47 323.55 227.49 178.03 138.44Household size # 2.98 4.37 5.69 7.58 3.16 4.36 5.63 7.47Dependency ratio % 0.70 0.91 1.19 1.46 0.71 0.91 1.18 1.46Female head % 0.21 0.12 0.08 0.05 0.20 0.12 0.08 0.06Head is wage laborer % 0.48 0.56 0.60 0.61 0.48 0.56 0.60 0.61Working members at home % 0.41 0.33 0.28 0.24 0.41 0.33 0.28 0.24Asset index 0.29 0.27 0.24 0.20 0.30 0.26 0.23 0.19Cattle # 0.93 0.66 0.55 0.48 1.04 0.60 0.55 0.44Avg. schooling of household migrants # 8.98 9.28 9.21 9.14 9.17 9.21 9.11 9.08Has received a transfer % 0.17 0.08 0.06 0.05 0.16 0.09 0.06 0.05Has received domestic transfer % 0.12 0.06 0.04 0.04 0.11 0.07 0.04 0.04Has received foreign transfer % 0.05 0.02 0.02 0.01 0.05 0.02 0.02 0.01Net level of transfer $ 65.72 26.79 19.16 16.89 65.67 28.17 19.14 15.59Net level of domestic transfer $ 30.96 10.38 6.66 5.98 29.77 12.25 5.98 5.97Net level of foreign transfer $ 34.29 16.24 12.20 10.78 35.46 15.57 13.13 9.35Household experienced a drought % 0.15 0.17 0.18 0.20 0.17 0.17 0.17 0.18Household experienced a flood % 0.04 0.04 0.04 0.04 0.04 0.04 0.04 0.04Rainfall % �.078 .008 .045 .077 �.074 .007 .035 .083

Notes: (1) All monetary amounts expressed in August 1998 pesos. (2) VTP is a score index that ranges from 0 (least vulnerable) to 1 (most vulnerable). (3)The first (last) quartile comprises the least (more) vulnerable. (4) Asset index is a composite index that goes from 0 to 1 and includes the following items:blender, fridge, gas-heater, water-heater, radio, record-player and television. (5) Net levels of transfers reported are the average amounts over allhouseholds. (6) Drought and flood are self-reported shocks. (7) Rainfall is the percentage deviation from the 40-year mean average rainfall for that group.Hence, 0 is normal rainfall and �.078 is 7.8% below an average year.


(i) Going to the neediest?If the rates of transfers exchanged between households

were indicative of the domestic and international supportnetworks available in these communities, these would be sur-prisingly low. As Table 2 illustrates, only about one in every10 households reported giving or receiving any type of trans-fer during October 1998 and this decreased in the followingtwo years. This low figure could be a consequence of the

questionnaire design, as the report period for transfers fromfriends, neighbors and relatives considers only one monthprior to the survey. However, it could also reflect that notall poor households are involved in support networks.

From the household’s welfare point of view, it also seemsthat remittances are not going to the poorest members in thesecommunities. Table 3 disaggregates our vulnerability estimatesby quartiles to show that those facing the least threat of future


poverty (lower quartiles) are better endowed, and receive moreand higher levels of remittances. In both years, for instance,the share of households receiving foreign monetary transfersin the first quartile is more than four times those in the topquartile.

The lack of remittance flows to those in most need wouldnot necessarily preclude them from responding effectively toadverse income shocks as long as a system of wide reciprocitywere in place within localities, enabling the flow of resourcesfrom better to worse off households. However, as Table 2shows, most of the exchanges reported by friends, neighbors,and relatives are one way transactions with less than 1% ofhouseholds giving any type of transfer in October 1998 andan even smaller proportion in the following two years.

(ii) Where do remittances go and where do they come from?In Table 4 we observe that remittances tend to go to smaller

households with fewer children and more elders and to fami-lies with few working adult members but more permanent mi-grants.

On the other hand, having a permanent migrant seems tohave great relevance for recipient families. By far the maincontingent that leave home are the offspring. As Table 5proves, most of them leave unmarried and for working pur-poses. Migrating daughters stay closer to their homes thansons, who more often migrate to the United States. Finally,male foreign migrants are more likely to provide support thanfemales, and this is done mostly in cash and in higheramounts. There is also a gradual increase in the amount ofremittances sent back home, which corresponds with the na-tional trends observed between 1998 and 2000.

4. ECONOMETRIC ESTIMATION

The relationship between an increase in the risk of futurepoverty (VTP) for households and their prospects for receivingsupport from others living abroad will be addressed formallyby adopting the following specification:

rh;t ¼ aþ bX h;t þ b2Sh;t þ b3ðVTPÞh;tþ1 þ eh;t; ð9Þ

where R is a latent variable that results from an unobservedcontinuous process that determines the propensity to receive

Table 4. Household characterist

October 1998

Variable Unit Received Not received

Consumption per capita $ 329.83 230.77Household size # 4.74 5.44Children 0–5 # 0.58 1.00Male 6–14 # 0.66 0.77Female 6–14 # 0.65 0.74Male 15–64 # 1.35 1.50Female 15–64 # 1.52 1.50Male >65 # 0.27 0.15Female >65 # 0.24 0.14Head age # 61.43 51.23Head female % 0.26 0.16Head is agricultural laborer % 0.33 0.57Head in nonagriculture % 0.08 0.07Working members at home % 0.31 0.32Asset index 0.39 0.21Household has a migrant % 0.73 0.15

Note: Monetary amounts expressed in August 1998 pesos.

remittances and takes positive values if the remittance is re-ceived. In (9) if R > 0 then the remittances r received take avalue of one whereas if R 6 0, the remittances r observed takethe value of zero. X is a vector of household controls thatcould affect the incidence and level of grants obtained. Wecontemplate several features of the household head; the ratiobetween dependent and working age members, the percentageof working household members and if the household reportedto have any permanent migrants; some asset endowments suchas land hectares possessed and cattle units; and finally, publictransfers from different government programs. S is a vector ofdummies that contains multiple impacts from weather-relatedshocks (e.g., destruction or loss of physical and human capitalin the form of land, crops, animals, items or home, or if anymember from the family died or got injured) and health shocks(e.g., the number of productive days lost by the breadwinnerdue to illness) as well as the number of sick children reportedby households. As usual, the error term e is assumed to be nor-mally distributed. Our main regressor is the vulnerability scoreto future consumption poverty of each household (VTP). On apurely normative ground, if our coefficient of interest b3 turnsout positive (b3 > 0), it would indicate that the need for mu-tual insurance which an increase in vulnerability would requirewould be compensated for by a higher probability of receivingremittances. Given our current dataset, this specificationtranslates into a model of the incidence of remittances forthose already poor in October 1998 given the possibility ofstaying poor or becoming poor(er) in November 1999, as wellas the incidence for November 1999 given the likelihood ofstaying poor or becoming poor(er) a year after.

However, such a specification could pose endogeneity prob-lems between the dependent variable and our main regressorVTP. By definition, the future welfare outcome used for thecomputation of vulnerability takes into account both negativeand positive shocks, including public and private transfers. Itwould therefore be difficult to detect what effect increasing therisk of future poverty could have on receiving more remit-tances.

The proposed solution is to obtain the relevant parametersthrough an Instrumental Maximum Likelihood EstimationProbit model on the probability of receiving remittances. Wecircumvent the problem of endogeneity through the identifica-tion of proper instruments for VTP that would purge its

ics by receipt of remittances

November 1999 November 2000

Received Not received Received Not received

325.29 198.08 331.17 207.544.28 5.38 4.55 5.090.41 0.79 1.23 1.150.54 0.75 0.67 0.760.54 0.71 0.62 0.740.89 1.41 1.76 1.571.41 1.42 1.84 1.580.27 0.15 0.23 0.180.23 0.14 0.21 0.1764.26 52.40 58.11 53.790.28 0.18 0.33 0.190.27 0.56 0.22 0.530.05 0.07 0.03 0.080.23 0.31 0.23 0.310.44 0.23 0.49 0.250.69 0.16 0.81 0.12

Table 5. Characteristics of permanent migrants abroad

October 1998 November 1999 November 2000

Variable Unit Total Male Female Total Male Female Total Male Female

Households w/foreign migrants (over those w/permanent migrants) % 16.48 15.69 25.83Households w/foreign migrants (over all households) % 2.68 2.65 3.61Migrants abroad # 749 563 186 747 555 192 1027 774 253Civil status at departure Obs. 738 553 184 720 539 181 1017 767 250Unmarried % 80.5 82.3 75.0 72.1 72.0 72.4 71.6 71.8 70.8Married 13.7 12.3 17.9 21.3 21.9 19.3 22.8 23.2 21.6Cohabitation 3.0 3.1 2.7 3.8 3.7 3.9 3.9 3.7 4.8Separated 1.4 1.3 1.6 2.1 1.9 2.8 0.8 0.5 1.6Widowed 1.5 1.1 2.7 0.8 0.6 1.7 0.8 0.8 0.8Kinship with head Obs. 726 541 185 721 533 188 1022 769 253Son/daughter % 92.2 91.7 93.5 86.6 86.7 86.2 75.0 76.7 69.6Husband/wife or partner 2.9 3.7 0.5 4.6 5.6 1.6 10.4 13.5 0.8Brother/sister 1.5 1.9 0.5 0.6 0.8 0.0 1.2 0.8 2.4Son/daughter-in-law 1.1 0.7 2.2 2.5 1.5 0.0 3.7 1.4 10.7Niece 1.0 0.9 1.1 2.8 3.0 5.3 6.2 4.7 10.7Nephew 0.6 0.7 0.0 0.8 0.9 2.1 0.7 0.5 1.2Reason for leaving home Obs. 747 562 185 744 554 190 1014 764 250Work % 81.8 88.8 60.5 82.3 89.7 60.5 79.5 88.4 52.4Got married 13.8 6.8 35.1 9.5 4.2 25.3 9.1 3.4 26.4Study 2.3 2.3 2.2 2.4 1.8 4.2 2.4 2.8 1.2Problems 0.5 0.5 0.5 0.3 0.4 0.0 0.4 0.5 0.0Other 1.6 1.6 1.6 5.5 4.0 10.0 8.7 5.0 20.0Type of support Obs. 738 553 185 723 540 183 1016 764 252No support provided % 45.5 40.3 61.1 57.4 53.7 68.3 53.4 45.7 77.0Cash 53.4 59.0 36.8 42.3 46.3 30.6 44.4 52.2 20.6In kind 0.3 0.2 0.5 0.0 0.0 0.0 2.0 2.0 2.0

Amount of remittance Obs. 391 324 67 303 248 55 437 387 50(over those persons that sent) pesos 1892 1973 1499 3924 3637 5216 4085 4217 3058

Amount of remittance Obs. 749 563 186 742 551 191 1027 774 253(Over all foreign migrants) pesos 998 1136 540 1592 1625 1494 1738 2109 604


correlation with the residual eh. The endogenous variable VTPwould be regressed upon such set of instruments and fit intothe full Probit model; in other words, once instrumentedVTP will be used as an explanatory variable in the remittanceequation to compute the relevant parameters.

The selected instruments consist of the bottom 20th percentileof the aforementioned absolute rainfall distribution alone andinteracted with the amount of cattle and land possessed by thehousehold and the percentage of working members at home.The measure of central tendency derived from the absolute rain-fall distribution could inform about the dispersion of future con-sumption provided shocks materialize. Rainfall-related events,for example a drought, would lead to fluctuations away fromthe mean given their high level of aggregation, as well as the rel-atively extended lack of insurance against them in rural areas.On the other hand, the cattle and land household controls areparticularly sensitive to weather-related shocks and the laborforce potentially helpful to cope with them. All three controlsare included because they can reveal part of the more stablecomponent of the future welfare outcome employed (i.e., futureconsumption). Overall, this group of variables should inform usabout the distribution of vulnerability. After some testing re-ported in the following section we are able to confirm that theyfulfill the conditions that make for a good instrument, that is,they are correlated with the endogenous regressor (VTP) anduncorrelated with the residuals from Eqn. (9) (Deaton, 1997)(i.e., cannot reject they are valid instruments using a 1% or 5%significance test), and they are needed.

While other studies dealing with Mexican migrants in theUnited States have employed rainfall as an instrument for

migration (finding a strong correlation between rainfall at ori-gin and immediate migration to the United States) (Munshi,2003), in our case, we found no evidence of migration respond-ing to rainfall shocks, which could create selectivity problemsand thus affect our rainfall-related instruments for vulnerabil-ity. The ENCEL survey asks households if as a consequence ofany natural hazard (mostly droughts and floods) any familymember migrated elsewhere to look for a job. In all threerounds considered, less than 1% of households reported themigration of family members as a response. Also, the possibil-ity that the poorest households reside in hazard prone areas(where rainfall is most unfavorable) making them more vul-nerable and thus less able to put aside resources for sendinghousehold members abroad was not present for our groupof rural families. As Table 3 shows rainfall-related self-re-ported shocks are almost equally distributed across vulnerabil-ity quartiles. Moreover, if anything, rainfall is poorer (belowlocality average precipitation) for those households with lowersusceptibility to future poverty.

Other potential problems with the proposed specificationcould be the endogeneity of remittances and the allocationof household members abroad with other relevant characteris-tics of the household, for instance the education of its mem-bers. For example, richer families might obtain higher flowsof remittances due to their better capacity to invest in the edu-cation of the would-be migrants and to send them to the Uni-ted States. In contrast, remittances to more vulnerable familiescould be lower given the lower levels of human and social cap-ital of their migrant members (which as a result earn less andhence had less to remit). However, we found no ground to


substantiate this concern in our sample. Table 3 shows that theaverage years of education for migrant household membersacross vulnerability quartiles are similar.

Given that a remittance indeed occurs, we next consider ifthe amount received by the household increases or decreasesgiven a rise in the possibility of it becoming poorer in the fu-ture. This involves a model specification where the latent var-iable R still takes discrete choice values of zero for thoseunobserved remittances, but positive values for those house-holds that actually receive remittances, making r P 0 in Equa-tion (10) below. X and S are the same control vectors asbefore, and v is the truncated distribution of residuals.

rh;t ¼ aþ bX h;t þ b2Sh;t þ b3ðVTPÞh;tþ1 þ E½vh;tjrh > 0�: ð10Þ

We believe there are no reasons to find differentiated effects onthe probability of both receiving grants from abroad as well as

Table 6. Effect of correlates on

t O

IV(1)

Instrumented vulnerability to poverty in t + 1

Rainfall distribution 20th p% 0.002(0.000)***

# Land ha * rainfall 20th p% 0.000(0.000)***

# Cattle * rainfall 20th p% 0.000(0.000)

Working household members * rainfall 20th p% �0.001(0.000)***

Household size 0.043(0.000)***

Dependency ratio 0.019(0.001)***

% Working household members �0.044(0.005)***

Head age �0.003(0.000)***

Female head �0.029(0.003)***

Head schooling �0.001(0.000)***

Head is agricultural wage laborer 0.008(0.002)***

Head works in nonagricultural activity �0.016(0.003)***

Land hectares �0.000(0.000)***

# Cattle �0.002(0.000)***

Household has a migrant �0.003(0.002)

Noninstrumented vulnerability to poverty in t + 1

Observations 15,579Amemiya–Lee–Newey test P-vWald test of exogeneity P-v

Notes: (Standard errors). Marginal effects (dF/dx) reported at the 75th percentvarious government programs and dummies for health and weather-related shNoninstrumented vulnerability to poverty is the same Probit model without conot reported in the table.* Significant at 10%.** Significant at 5%.*** Significant at 1%.

the amount obtained. Hence, we considered a Tobin’s Probitmodel to be an appropriate estimation strategy to account forthe potential bias that arises from the truncation of t at zero.

5. RESULTS

Our statistical analysis suggested that the remission ofmoney from abroad does not appear to end up with thosewho are more likely to be needy in the future. Our econometricapproach also reported an inverse and significant relationshipbetween the probability of receiving a remittance and experi-encing an increase in the risk of future poverty. To arrive atthis conclusion, we first carried out a Probit on the incidenceof monetary transfers from abroad. As columns (2) and (4)in Table 6 show at the bottom (noninstrumented vulnerability

the incidence of remittances

ctober 1998 November 1999

Probit (dF/dx) IV Probit (dF/dx)(2) (3) (4)

�0.153 �0.286(0.037)*** (0.070)***

0.002(0.000)***

0.000(0.000)***

0.000(0.000)*

�0.001(0.000)***

0.006 0.044 0.012(0.002)*** (0.000)*** (0.003)***

0.003 0.017 0.004(0.001)*** (0.001)*** (0.001)***

�0.014 �0.032 �0.026(0.003)*** (0.007)*** (0.006)***

�0.000 �0.002 0.000(0.001) (0.000)*** (0.000)�0.003 �0.022 �0.004

(0.001) *** (0.003)*** (0.002)*

�0.000 �0.002 �0.000(0.000)* (0.000)*** (0.000)*

0.000 0.020 0.003(0.001) (0.002)*** (0.002)|**

�0.002 �0.000 0.001(0.001) (0.003) (0.002)0.000 �0.004 �0.001

(0.000) (0.000)*** (0.000)***

0.000 �0.002 �0.000(0.000) (0.000)*** (0.000)*

0.024 0.019 0.045(0.003)*** (0.002)*** (0.005)***

�0.072 �0.029(0.007)*** (0.005)***

15,579 15,379 15,379alue = 0.9849 P-value = 0.0544alue = 0.0000 P-value = 0.0000

ile. Regressions also include whether households obtained resources fromock impacts, but coefficients are not reported.

rrecting for endogeneity. The rest of the marginal effects for this Probit are


to poverty in t + 1), there is a significant and negative associ-ation between the increase in vulnerability of the rural pooracross our sample of households and the disbursement ofmoney from foreign migrants to them. Given that those atthe upper end of the distribution face a higher risk of futurepoverty we chose to report the marginal effects at the 75th per-centile.

To reinforce our conclusion and strengthen our econometricestimation, we accounted for possible sources of endogeneityby instrumenting the variable that could have a reciprocal cau-sation with the outcome variable (i.e., the relationship betweenvulnerability to poverty and the reception of remittances). Therejection of exogeneity at 99% confidence level in both yearssuggested that we should proceed along these lines to try toget some consistent estimates. Columns (1) and (3) displaythe results of the instrumental variables regression. We con-firmed that the instruments are all individually and jointly sig-nificant and behave as expected. For instance, records at thebottom tail of the rainfall distribution (i.e., poor rainfall) forour group of localities increase the magnitude of the house-hold threat of future poverty. In a similar vein, the interactionof negative fluctuations on locality precipitation with rainfall-related risk-sensitive assets such as land and cattle also tendsto augment such exposure. On the contrary, the possible im-pact on vulnerability of droughts might be attenuated by anincreased labor availability at home. The results in the firstrow of columns (2) and (4) represent our Probit model onthe incidence of remittances after correcting for endogeneity.We obtained the same result as our noninstrumented probitdisplayed in the bottom row of Table 6, that is, our fundamen-tal claim regarding the negative relationship between remit-tances and vulnerability to poverty in this sample of ruralhousehold’s remains.

Finally, in accordance with our statistical exploration,households with more dependent members—especially elderpeople—and with permanent migrants are more likely to re-ceive remittances. In contrast, those where the head is moreeducated and with more working members at home are lesslikely to obtain inflows from relatives living abroad.

02

46

810

Perc

ent o

f hou

seho

lds

0 .2 .4Vulnera

November 1999

Note: Difference in means test significant at 0.0

Figure 2. Vulnerab

The OLS and Tobit estimates on remittance levels conveythe same conclusion: an increase in the risk of future povertytranslates into a reduction in the level of remittances received(detailed results are available upon request). Figure 3 illus-trates the negative trend of fitted values for remittances (madein October 1998 and November 1999) for a selected range ofpercentiles for the instrumented index of vulnerability to pov-erty (for November 1999 and 2000, respectively). In bothyears, most values are comprised between the first and thirdquartiles as the indices follow a slightly positively skewed nor-mal distribution as shown in Figure 2. As households move tohigher levels of risk the decline in the level of grants observedbecomes steeper. A nonmarginal shift from the 25th percentileof the distribution to the 75th percentile, though unlikely,would translate into a decline of approximately $900 pesos(US$90) in October 1998 and $1300 pesos (US$130) inNovember 1999.

6. CONCLUSION

International remittances have been perceived as the humanface of globalization especially through their potential effectson poverty as they directly increase the household’s level of in-come and therefore should become a massive poverty-reduc-tion mechanism if targeted on deprived households.However, even though transnational support networks aremeant to be one of the most important resources that poorrural households can draw upon to cope with problems byproviding risk-pooling mechanisms in the absence or limitedaccess to formal credit and insurance markets (Bando & Lo-pez-Calva, 2004), their activation might not necessarily bealigned to the needs and risks faced by recipients.

Given that remittances have reached historically high flowsin Mexico in the recent past becoming one of the most salientfeatures that have taken place in the country under the mantraof globalization we therefore considered of most relevance tounderstand whether remittances indeed reach those whoseconditions are likely to worsen in the future.

.6 .8 1bility Index

November 2000

1% . O means No Vulnerability.

ility to poverty.

Figure 3. Fitted values after Tobit in remittances.


We addressed this question by creating a model that wouldallow us to predict future consumption expenditures based oncurrent household socio-demographic characteristics, as wellas risk-mitigating assets. This resulted in differentiated levelsof vulnerability to poverty for a panel of rural households inMexico that extends from October 1998 to November 2000.After coming up with such levels for each household, webrought into a series of Probit and Tobit models to discernif an increase in the risk of future consumption poverty auto-matically translated into a higher probability for receivingmonetary help from relatives or friends living abroad. Caution

was paid to reciprocal causation by instrumenting throughmonthly rainfall data the possible source of endogeneity con-tained in our vulnerability regressor.

We found with statistical significance that an increase in thethreat to future poverty that rural families could experience re-duces their likelihood of receiving remittances. We did not en-quire as to why this could be the case as this would demandmore information about migrant households themselves whichwas not available, such as the migrants’ earnings and otherimportant personal characteristics. Certainly this remains aline of research worth exploring in the future.


The absence of remittances flowing to the most vulnerablehouseholds would be less of a problem if our rural communi-ties were characterized by vertical arrangements of solidarityand/or reciprocity. In such case, an injection of remittancesanywhere within the village would trickle down to those inmost need. Unfortunately, the prevalence of vertical social ex-changes proved almost inexistent for our group of households.

Remittances can certainly result in higher levels of welfarefor their recipients and they should be facilitated in this spirit.However, their reception appears highly unlikely for those

households at higher risk of future poverty (similar resultsare reported by the Human Development Report for Mexico2006–2007). For this reason, the eradication of poverty cannotrest exclusively on the activation of this type of flows. Remit-tances should not be considered a substitute for the introduc-tion of publicly-funded schemes of social protection coupledwith the gradual improvement of economic conditions in thehome villages that could generate viable alternatives to thedecision to migrate, especially when migration occurs due tothe shortage of job options at the rural community of origin.

NOTES

1. Migration from Mexico to the United States has always been adouble-edged sword. Some perceive the increasing flows of people asevidence of an increased interdependence between complementary econ-omies in an era of economic integration while others interpret the sameexperience as a painful reminder of the exclusionary nature of theglobalization friendly development model embraced by the country sincethe eighties, which has forced many to the other side of the border.

2. This argument has led to perceive remittances as one of the factorsbehind the recent drop in poverty rates in Mexico (Esquivel & Huerta-Pineda, 2006; Lanjouw, 1998) and other LAC. According to the WorldBank’s Development Impact of Remittances in Latin America report worldcross-country and microbased estimates indicate that, on average, anincrease in one percent point in the ratio of remittances to GDP would beassociated with a decline in poverty of about 0.4% (Fajnzylber & Humberto,2007). In a similar vein, a recent study by the same institution on south-eastern Mexico (Quentin, Gonzalez Konig, Angel-Urdinola, Ojeda Revah,& Siaens, 2002) attributed a two point reduction in the income poverty ratesof Guerrero and Oaxaca to local and foreign remittances by a pure incometransfer effect, an impact of similar scope to that of the public transferprograms Procampo and Oportunidades (Progresa).

3. The focus axiom property requires concentrating only on thoseoutcomes that are likely to be poverty threats.

4. The measure of vulnerability presented at the individual level can besummed over all households to convey a measure of aggregate vulnera-bility in the following way: V ðp; y; zÞ ¼

Pni ¼ 1V iðp; y; zÞ , where n is the

total number of households.

5. We retained the specification that maximized the predictive power ofconsumption, but tested the robustness of our proposed vulnerabilitymeasure with the addition of different control variables in the specificationof consumption. The resulting vulnerability scores were highly correlated(q � 0.79).

6. An observed practice when having panels with several rounds of datahas been to assume that the distribution of consumption for eachhousehold in one period of time is similar to any other period. Given sucha claim, the levels of consumption observed at different points in time forthe same household could be taken as potential “states of the world” ofwhat consumption might have been in the initial period, and thus, byextension, in any other period of the panel, allowing the researcher to treatthem as the distribution of future consumption (Ligon & Schechter, 2003).

7. The measures of vulnerability obtained with different number ofreplications (100, 150, 250, and 500) were very similar. The correlationcoefficients among them hovered at around 0.80.

8. Progresa was introduced in 1997 and re-launched as Oportunidades in2002. Its basic objective is to improve the education, health, and nutritionof poor families, particularly children and their mothers. In addition tonutrition supplements, it provides cash transfers to families in exchangefor regular school attendance and visits to health posts. The payments areprovided directly to mothers or the female head of households. The othermajor innovation of the program is the creation of an evaluation samplethrough a randomization of villages into treatment and control groups,given the impossibility to start the provision of benefits everywheresimultaneously.

REFERENCES

Bando, R. & Lopez-Calva, L. F. (2004). The rural sector and informal safetynets: Evidence from Mexico. Institute of Public Policy and DevelopmentStudies. Working Papers, Universidad de las Americas, Puebla.

Calvo, C., & Dercon, S. (2005). Measuring individual vulnerability.Working Paper 229, University of Oxford, Oxford.

Deaton, A. (1997). The analysis of household surveys: A microeconomicapproach. Baltimore: Johns Hopkins University Press.

Dercon, S., & Christiaensen, L. (2007). Vulnerability, seasonality andpoverty in Ethiopia. Policy Research Working Paper 4257, WorldBank, Washington, DC.

de la Fuente, A. (2008). Vulnerability to poverty in Rural Mexico. Doctoraldissertation, University of Oxford, Oxford.

Escobar Latapı, A. (2006). Migracion internacional, pobreza y desigualdad enMexico. Documento de apoyo del informe sobre desarrollo humano Mexico2006–2007. Mexico: Programa de las Naciones Unidas para el Desarrollo.

Esquivel, G., & Huerta-Pineda, A. (2006). Remittances and poverty inMexico: A propensity score matching approach. Working Paper,Colegio de Mexico, Mexico.

Fajnzylber, P., & Humberto, L. J. (2007). Close to home: The developmentimpact of remittances in Latin America. Washington, DC: World Bank.

Greene, W. H. (2003). Econometric analysis (fifth low price edition). Delhi:Prentice Hall.

Human Development Report Mexico 2006–2007: Migration and HumanDevelopment (2007). United Nations Development Programme, Mexico.

Kamanou, G., & Morduch, J. (2005). Measuring vulnerability to poverty inStefan Dercon, Insurance against poverty (pp. 155–175). Oxford:Oxford University Press.

Kuhl, J. (2005). A household vulnerability profile – Analyzing fluctuations inconsumption using a simulation approach (pp. 1–31). Copenhagen:Institute of Economics, University of Copenhagen.

Lanjouw, P. (1998). Poverty and the non-farm economy in Mexico’s Ejidos:1994–1997. Background paper prepared for Economic Adjustment andInstitutional Reform: Mexico’s Ejido Sector Responds. Washington,DC: World Bank.

Ligon, E., & Schechter, L. (2003). Measuring vulnerability. EconomicJournal, 113(486), 95–102.

Lucas, R. E. B., & Stark, O. (1985). Motivations to remit: Evidence fromBotswana. Journal of Political Economy, 93(5), 901–918.

Munshi, K. (2003). Networks in the modern economy: Mexican migrantsin the US labor market. The Quarterly Journal of Economics, 118(2),549–599.

Orozco, M. (2002a). Globalization and migration: The impact of familyremittances in Latin America. Latin American Politics and Society,44(2), 41–66.


Orozco, M. (2002b). Worker remittances: The human face of globalization.Inter-American Dialogue. Working Paper commissioned by the Mul-tilateral Investment Fund of the Inter-American Development Bank,Washington, DC, IDB.

Teruel, G., & Davis, B. (2000). An evaluation of impact of progresa cashpayments on private inter-household transfers. Washington, DC: Inter-national Food Policy Research Institute.

Quentin, W., Gonzalez Konig, G. E., Angel-Urdinola, D., Ojeda Revah,D., & Siaens, C. (2002). Migration and poverty in Mexico’s southernstates. <http://mpra.ub.uni-muenchen.de/10574/>.

World Bank (2006). Global economic prospects: Economic implications ofremittances and migration. World Bank, Washington, DC.

Available online at www.sciencedirect.com

http://mpra.ub.uni-muenchen.de/10574/

Remittances and Vulnerability to Poverty in Rural Mexico

Documents

Transcript of Remittances and Vulnerability to Poverty in Rural Mexico