The Effects of International Migration on Migrant -Source ... · The Effects of International...

34
1 The Effects of International Migration on Migrant-Source Households: Evidence from Ethiopian Diversity-Visa (DV) Lottery Migrants* Teferi Mergo, UC Berkeley Abstract About a million people have migrated to the US via the DV lottery. Using data from a survey of Ethiopian DV participants, I study the causal effects of emigration. I infer that migration contributes positively to the wellbeing of source families. In particular, families of emigrants spend 22% more on food, and have higher BMI (+0.62). Additionally, winners’ families have better quality durables, drinking water and sanitation facilities. However, they do not have higher savings or physical capital. The positive treatment effects do not diminish with longer stay of emigrants abroad. I find that DV entrants are favorably selected relative to the overall population. JEL: J61, F22; I31 Keywords: International Migration, Welfare, Selectivity, DV Lottery ________________________ * I want to thank Ronald Lee for his guidance and financial assistance. I am also grateful to David Card for his advice on this paper. I thank survey participants; the enumerators who helped me in gathering the best possible data; the CEO of the Ethiopian Postal Service for giving me access to lottery winners data; Maximillian Kasy and Alain de Janvry for their incisive comments and feedback; NSF coordinator at UC Berkeley (Gloria Chun) and the Institute of Business and Economic Research for financial assistance; the president of Addis Ababa University for allowing me to use university facilities free of charge; participants of the Development Seminar at UC Berkeley for useful feedback. All errors are mine.

Transcript of The Effects of International Migration on Migrant -Source ... · The Effects of International...

Page 1: The Effects of International Migration on Migrant -Source ... · The Effects of International Migration on Migrant -Source Households: ... lottery participants in Addis Ababa ...

1

The Effects of International Migration on Migrant-Source Households: Evidence from Ethiopian Diversity-Visa (DV) Lottery Migrants*

Teferi Mergo, UC Berkeley

Abstract

About a million people have migrated to the US via the DV lottery. Using data from a survey of Ethiopian

DV participants, I study the causal effects of emigration. I infer that migration contributes positively to the

wellbeing of source families. In particular, families of emigrants spend 22% more on food, and have higher

BMI (+0.62). Additionally, winners’ families have better quality durables, drinking water and sanitation

facilities. However, they do not have higher savings or physical capital. The positive treatment effects do

not diminish with longer stay of emigrants abroad. I find that DV entrants are favorably selected relative to

the overall population.

JEL: J61, F22; I31

Keywords: International Migration, Welfare, Selectivity, DV Lottery

________________________ * I want to thank Ronald Lee for his guidance and financial assistance. I am also grateful to David Card for

his advice on this paper. I thank survey participants; the enumerators who helped me in gathering the best

possible data; the CEO of the Ethiopian Postal Service for giving me access to lottery winners data;

Maximillian Kasy and Alain de Janvry for their incisive comments and feedback; NSF coordinator at UC

Berkeley (Gloria Chun) and the Institute of Business and Economic Research for financial assistance; the

president of Addis Ababa University for allowing me to use university facilities free of charge; participants of

the Development Seminar at UC Berkeley for useful feedback. All errors are mine.

Page 2: The Effects of International Migration on Migrant -Source ... · The Effects of International Migration on Migrant -Source Households: ... lottery participants in Addis Ababa ...

2

1. Introduction

Although international migration can yield large benefits to individual migrants from poor countries, the net

impacts of migration on the source countries is unclear. In particular, when migrants move away, their

remaining family members lose a share of their income, as well as in-kind contributions to household

production, including the care of elderly parents and younger siblings. These losses can be particularly

large if the most productive members of a family are most likely to emigrate. To the extent that there are

important externalities from human capital, and migrants tend to be relatively young and better-educated,

emigration can also create wider social costs -- the so-called “brain drain” phenomenon.

Remittances are arguably the principal channel through which non-migrants benefit from emigration.1

1 Remittances have become a significant source of income and foreign currency for several developing

countries, in some cases overtaking Official Development Aid and Foreign Direct Investment. According to

the World Bank, official remittances to developing countries are currently in the range of $340 billion per

year.

(Knowledge transfers are another possible channel: see Hilderbrandt and McKenzie, 2006). Some recent

causal studies have reached different conclusions regarding the effects of migration and remittances on

those left behind. In this paper, I add to the literature by focusing on migrants from an extremely poor

country – Ethiopia – who are essentially randomly assigned the possibility of migration through the United

States’ Diversity Visa lottery. The DV lottery, which has been in effect since 1995, attracts tens of millions

of applicants from all corners of the world. Every year, about 50,000 people (not including their immediate

families) migrate to the US by winning the lottery. Roughly three-quarters of the DV migrants are from

Africa; between 8% and 10% of all DV migrants have consistently come from Ethiopia.

Page 3: The Effects of International Migration on Migrant -Source ... · The Effects of International Migration on Migrant -Source Households: ... lottery participants in Addis Ababa ...

3

My analysis is based on a specially designed survey of households of previous DV lottery winners and

lottery participants in Addis Ababa – the Ethiopian capital. I use comparisons between the lottery winners

and the (non-winning) participants to infer the causal effects of having a family member migrate to the U.S.

I find that having a family member win the lottery and migrate has significant positive effects on several

dimensions of the remaining family’s standard of living. Families of DV migrants spend about 22% more on

food, are thus better fed and have higher body mass indexes (+0.62 average BMI). Moreover, families of

lottery winners possess more and better quality consumer durables, which include personal computers,

modern cooking stoves, household furniture and home entertainment appliances. Having a member who

won the DV lottery also gives those remaining behind access to improved sources of drinking water and

sanitation facilities. Winners’ families, however, have about the same savings and physical capital

accumulation as other families.

The positive effects of emigration on the living standard of families remaining behind do not diminish as

migrants spend more time abroad. Rather, the effects appear to grow with longer duration of the lottery

migrants in the US. It is likely that DV migrants, who are typically young adults, earn more as they stay

longer in the US and acquire newer and more marketable skills, allowing them to send more money to their

parents in Ethiopia.

The finding that international migration has no effects on savings and the general business environment in

the migrant-sending countries should be considered tentative. The claims made in this paper are simply

that, when migrants are young adults and the staying family members are their parents and other

dependent siblings, migration has no measurable direct impact on savings and investment behavior and

practices of the latter. There may be other channels (e.g. return migration) through which migration may be

affecting these key variables. After accumulating sufficient physical, human and social capital, earlier

Page 4: The Effects of International Migration on Migrant -Source ... · The Effects of International Migration on Migrant -Source Households: ... lottery participants in Addis Ababa ...

4

migrants could be returning to their countries of origin, thus positively affecting the general business

environment in sending areas.

A final interesting conclusion is that participants in the DV lottery (both winners and losers) have

substantially higher outcomes than non-participants, suggesting that Ethiopian DV migrants are indeed

positively selected. Non-participants have lower food spending, lower variety and value of durables they

own, and less access to clean drinking water and convenient sanitation facilities. They are also the least

likely to use banking facilities and save. Interestingly, however, lottery non-participants spend more on

leisure activities, including movies and watching football at bars.

This paper is organized as follows. Section 2 presents a brief review of earlier work on the effect of

emigration on remaining families. Section 3 describes the data set and the identification issues associated

with the nature of the data collection process. Section 4 presents the study’s main results together with the

underlying empirical frameworks. Section 5 concludes, suggesting where the focus of future research ought

to be in order to more fully understand the consequences of international migration on migrant-sending

nations.

2. Background and Existing Literature

Earlier studies on the effects of emigration, relying on a variety of non-experimental methods, generally find

that migration has positive effects on sending households and countries. The papers can be differentiated

based on claims they make about their conclusions. Some report robust correlations between emigration

and desirable outcomes in sending areas, making no explicit causal claims; others employ a variety of

estimation techniques to tease out the effects of emigration. The methods include instrumental variables

Page 5: The Effects of International Migration on Migrant -Source ... · The Effects of International Migration on Migrant -Source Households: ... lottery participants in Addis Ababa ...

5

estimation (e.g. Mansuri, 2006; Woodruff and Zenteno, 2007; McKenzie and Rappoport 2007, Lopez-

Cordova, 2006), propensity-score matching (e.g, Esquivel and Huerta-Pineda, 2006), and parametric

selection correction models (e.g. Acosta, Fajnzylber and Lopez, 2007). Since migrants are typically

positively selected (see, for instance, Chiswick, 1999; Chiquiar and Hanson, 2005; McKenzie, 2006), non-

experimental estimates of the effects of migration may be biased if there are concerns with the identifying

assumptions. Antman (2012) provides a succinct review of a few of the studies on the effects of

international migration on those left behind, with critical evaluation of their identification strategies.

A few recent papers have tried to substantially address the causality issues in different ways. Yang (2008)

evaluates the effects of remittances made by Filipino migrants on the well-being of their families, exploiting

the depreciation of the Philippine peso as an exogenous source of variation in the amount of money sent

home by migrants. Gibson, et al (2011) and Gibson, et al (forthcoming) exploit lottery migration to New

Zealand of the residents of the Pacific islands of Tonga and Samoa, respectively, to study the effects of

emigration. Interestingly, Yang (2008) argues that remittances have positive effects on family members

who remain at home; whereas, Gibson, et al (2011) find negative overall effects of emigration in the short

run, with Gibson, et al (forthcoming) inferring that migration reduced poverty in Samoa, but the effect may

be short lived.

3. Constructing a New Sample of Families of DV Lottery Winners and Losers

3.1: The Diversity Lottery

The DV was instituted pursuant to the Immigration and Naturalization Act of 1990, Sections 201(d) and

203(c); the latter was amended in Section 131 (Pub. L. 101-649). Section 201 (e) stipulates that the

Page 6: The Effects of International Migration on Migrant -Source ... · The Effects of International Migration on Migrant -Source Households: ... lottery participants in Addis Ababa ...

6

maximum level of diversity immigrants not exceed 55,000 every year. As the title suggests, the purpose of

this congressional Act is to diversify the U.S. population through a lottery made available to people from

countries with historically low rates of immigration. As a result, the majority (about 75%) of diversity

immigrants come from the continent of Africa, with the top five African countries accounting for about 35%

of all diversity immigrants.

A dynamic formula determines how these visas are distributed globally. No diversity visas are granted to

countries which send more than 50,000 immigrants to the United States within a previous five year period.

Accordingly, the natives of Brazil, Canada, China, Colombia, Dominican Republic, Ecuador, El Salvador,

Guatemala, Haiti, India, Jamaica, Mexico, Pakistan, the Philippines, Peru, Poland, South Korea, United

Kingdom (except Northern Ireland) have been deemed not eligible for the DV lottery for the last several

years.

The successful DV applicants have to meet either the education or the work-experience requirement. One

must have “either a high school education or its equivalent, defined as successful completion of a 12-year

course of elementary and secondary education; or two years of work experience within the past five years

in an occupation requiring at least two years of training or experience to perform.” (State Department DV

Immigration Guidelines) Only applicants with formal courses of study are considered eligible; those with

correspondence programs or equivalency certificates (such as the G.E.D.) do not satisfy the education

requirement. The qualifying DV Occupations are those listed on the Department of Labor O*Net Online

Database. None of these requirements is overly burdensome in the sense that a very large segment of the

qualifying countries’ nationals are able to meet them.

In the past, anyone with access to the post-office and satisfying the aforementioned criterion could have

applied for the lottery, but only electronic applications are accepted as of 2003. This limits the pool of

Page 7: The Effects of International Migration on Migrant -Source ... · The Effects of International Migration on Migrant -Source Households: ... lottery participants in Addis Ababa ...

7

potential applicants only to those with access to internet services. Given the low level of computer and

internet penetration rates in some of the DV eligible countries, the online-only application requirement

seems to be more restrictive than either the education or the work experience requirement.

After determining the list of eligible applicants for each qualifying country, the Kentucky Consular Center

selects winners from an applicant pool of millions based on a computer generated, random lottery drawing.

The procedure guarantees each applicant an equal probability of winning the coveted prize as other

applicants from the same country.

DV migrants can be single or married with children. The latter can bring their spouses and dependent

children younger than 21 years of age, but are required to list them at the time of initial DV entry. It is

possible that one’s marital status may change, particularly from single to married, after winning the lottery

and before migrating; when such cases turn up, U.S. embassy staff in each country determine the

legitimacy of these claims on a case-by-case basis, as there seem to be incentives for fraud.

3.2: A Sample of Lottery Winners and Losers

Ideally, I would have liked to obtain a complete enumeration of the entire set of successful and

unsuccessful DV applicants from Ethiopia, from which I could have easily drawn random samples of

winners and losers. However, I was able to get only a complete listing of lottery winners from Addis Ababa

for the years 2006 through 2010. It is not possible to obtain a comparable list of DV lottery applicants from

which to identify lottery losers. Fortunately, given the overwhelming popularity of the DV lottery, the low

threshold requirements needed to enter it, and the length of time the lottery has been in operation, around

50% of Addis’ households are conservatively estimated to have participated in the lottery at one time or

another (see the Appendix), thus allowing me to draw a representative sample for the control group from

Page 8: The Effects of International Migration on Migrant -Source ... · The Effects of International Migration on Migrant -Source Households: ... lottery participants in Addis Ababa ...

8

the city, using a procedure outlined below.2

The lottery rule allowing DV migrants to bring their spouses and dependent children to the US is a potential

source of selection bias if it is not treated properly. To overcome this, I included

In brief, the procedure breaks the entire city into several logical

enumeration areas, from a randomly chosen subset of which a representative sample of control and lottery

non-playing households were drawn, using a simple lottery in conjunction with a set of screening questions.

only those households

whose dependent children took part in the lottery, excluding subjects where the household head(s) were

the participants. Based on the population of lottery winners from Addis Ababa for the years 2006 through

2009, it is estimated that less than 6% of the lottery applicants in the capital are families whose head(s)

entered the lottery.

The Ethiopian Postal Service (“EPS”) maintains a list of DV lottery winners from 2006 onwards. Treatment

subjects were randomly selected from the 2006, 2007, 2008 and 2009 DV lottery winners of Addis Ababa.

Even if data were available for DV lottery winners prior to 2006, however, it might not have been very useful

for the current study. Earlier migrants who have completed their naturalization proceedings can bring their

parents as permanent residents under the Immigration and Naturalization laws of the US. It typically takes

about five years for permanent residents to become US citizens, which entitles them to apply for a Green

3.2.1 Lottery Winners:

2 The approximation in the Appendix is consistent with other estimates in similarly situated countries.

Torres and Pelham (2008) find that upwards of 60 percent of adults in Sierra Leone would like to migrate if

they had the opportunity. The World Bank in its 2007 report had also found that between 50% and 90% of

the young adults in certain developing countries would like to migrate if offered the option.

Page 9: The Effects of International Migration on Migrant -Source ... · The Effects of International Migration on Migrant -Source Households: ... lottery participants in Addis Ababa ...

9

Card on behalf of their parents; it is expected that a non-negligible percentage of DV winners of earlier

years (prior to 2006) have brought their parents to the US as residents. On the other hand, 2010 lottery

winners were excluded from consideration, because at the time of data gathering, some of the winners

were still in the process of migration. It would be safe to assume that even those 2010 winners who had

already migrated before the survey was conducted would not have been in a position to support their kin

back home.

A stratified-systematic-random-sampling strategy was employed to select the treatment group from the

sample frame, since lottery winners are unevenly distributed throughout the different sub-districts

(Kebelles) of Addis Ababa. After the complete list of winners was stratified by the various sub-districts,

winners in each Kebelle were numbered 1 through 𝜔𝐾(total number of DV winners from sub-district K for

the years 2006 through 2009) in ascending order of their Kebelle provided house numbers. A target

number of lottery winners constituting the treatment group from each sub-district are given by (𝜏𝐾).3 The

overall target number of the treatment group was intentionally set higher (at 300) than was justified by

power calculation, which had suggested that 270 DV winner households were sufficient to find effects, if

any. The interval size(𝑖), which is the same for all Kebelles, was then set as follows.4

3 𝜏𝐾 = 𝜔𝐾

𝑊∗ 300;𝑤ℎ𝑒𝑟𝑒,𝑊 = Total number of [2006, 2009] DV winners from all districts

Based on a simple

lottery, the 𝑛𝑡ℎ house (where 𝑛 is any number between 1 and 𝑖) was picked as the first candidate house

for the treatment group from the first interval in each sub-district. The 𝑘𝑡ℎ household (𝑘 is defined in the

4 𝑖 = 𝜔𝐾𝜏𝐾

Page 10: The Effects of International Migration on Migrant -Source ... · The Effects of International Migration on Migrant -Source Households: ... lottery participants in Addis Ababa ...

10

footnote) was then selected from subsequent intervals.5

Not surprisingly, not all randomly pre-selected

lottery winners took part in the study. Because some families were unwilling to participate in the survey, the

aforementioned procedure was repeated until the completed interviews in each sub-district reached as

close as possible to the target number (𝜏𝐾) for each Kebelle. On average, we had to approach about 2.5

DV winners to get one willing participant.

The population of interest for the control group is DV losing households from Addis Ababa for the years

2006 through 2009. Since the complete listing of this population was not available, the following strategy

was employed for selection of a representative sample of the unlucky lottery applicants and lottery non-

participants. It was first roughly estimated that one in two Addis households have at least one member who

played the lottery at least once since the inception of the program in 1995. This estimate is based on

conservative assumptions that are made explicit in the Appendix. The entire set of Addis households were

then divided into several enumeration areas (EA), equaling in number the total count of lottery winners from

the city for the four years. More importantly, since the distribution of lottery applicants can be assumed to

be significantly positively correlated with the distribution of lottery winners, the number of EAs in each

Kebelle is set to be the same as the number of lottery winners in each Kebelle. The number of households

in each EA is inversely proportional to the number of lottery winners in each Kebelle. The main reason for

the variation in the distribution of winners and applicants by the districts is differences in the socio-

3.2.2 The Control Group and Lottery Non-Participants:

5 𝑘 = [(𝑀− 1)𝑖 + 𝑛];where, 𝑀 = {1, 2,… 𝜏𝐾} is the sequence of intervals in a sub-district

Page 11: The Effects of International Migration on Migrant -Source ... · The Effects of International Migration on Migrant -Source Households: ... lottery participants in Addis Ababa ...

11

economic status of the residents, which have roughly sorted into separate neighborhoods by income and

other socio-economic indicators.

Enumeration areas, from which we expected to find control and lottery non-participating families, were

chosen based on the same criteria used to select treatment households for the study, thus guaranteeing

each EA an equal probability of being chosen for the study. Finally, control and lottery non-participating

households in the randomly selected EAs were picked as follows: A household was chosen from the

randomly selected EA based on a simple lottery and contacted to take part in the survey, which had an

incentive for participation.6

If the household expressed willingness to take part in the study, a set of

screening questions was administered to identify whether the family is control or lottery non-participant. If a

household was unwilling or unable to participate in the survey for any reason, the next higher (or lower

numbered) house was invited to take part in the survey, until we found one control and another lottery non-

participating household. Control families and lottery non-participants were asked the same set of questions,

except those dealing with the DV lottery status of the family were disregarded while interviewing the latter.

(The survey questions are available upon request).As anticipated, we were finding both control and lottery

non-participating households in all districts with a reasonable amount of effort.

Differential rates of response of the treatment and control groups may present some risk to the validity of

the empirical estimates if the differences are due to some pre-DV characteristics of the households. If, 6 The incentive was that three members of the family would be invited to attend a concert by prominent

Ethiopian artists at Addis Ababa University. The concert was very successful, thanks to the University

officials, particularly its president, Professor Andrias Eshete, who not only allowed me to use the University-

hall for the event, but also provided security, free of charge.

Page 12: The Effects of International Migration on Migrant -Source ... · The Effects of International Migration on Migrant -Source Households: ... lottery participants in Addis Ababa ...

12

however, lottery losers were less inclined to participate in the study, because their aspirations for better life

through the DV might have been thwarted, resulting in differences in the participation rates, the empirical

estimates would remain unbiased. Fortunately, the groups have roughly similar rates of participation on

average. Most importantly, a randomization check finds that the two groups of households are well

balanced in terms of their pre-DV characteristics (Table 2, Panel A).

A day-long training was given at Addis-Ababa University to survey Team 1 re: the purpose of the survey,

specific guidelines on how to implement it, and most importantly, the appropriateness of the questions

included in the survey.

3.3 Data Gathering and Quality Control

7

The enumerators had very pointed comments and suggestions about what should

be asked, what questions should not be part of the survey, which questions need to be reframed and how,

etc. The Questionnaire was redesigned taking the participants’ comments into account.

After the training, one or two enumerators were assigned to each district to implement a pilot survey,

depending on the anticipated difficulty of finding pre-selected houses in the treatment group, the size of the

district, and the target number of treatment (hence control and lottery non-playing households). The

purpose of the pilot was to get important feedback for the main study. The pilot suggested that we needed

to re-order the sequence of the questions in order to garner the most accurate responses. Some of the

7 The survey team was divided into two groups. Team 1 consisted of 16+ experienced enumerators and a

supervisor, all hired in consultation with the Economics department of Addis Ababa University. The second

team (Team 2) consisted of 4 quality controllers and a Quality Control (QC) head.

Page 13: The Effects of International Migration on Migrant -Source ... · The Effects of International Migration on Migrant -Source Households: ... lottery participants in Addis Ababa ...

13

pilot subjects were unwilling to be interviewed without a letter from Addis-Ababa city administration.8

The

required letter was subsequently obtained for the main study from the city administration, which also

allowed us the use of its electronic data-bases and guides to locate houses easily. In a city like Addis

Ababa, where most neighborhoods were built without any proper planning and most of the streets have no

names, it made the work a lot easier to have access to the data base and the Kebelle assigned guides.

Quality control was undertaken in three phases. The procedures were adopted before the survey was

begun, paying particular attention to the peculiarity of the Ethiopian culture. The first phase was

implemented concurrent with data collection. We phoned about 80% of the interviewed subjects, re-asking

them certain questions. For no particular reason other than the simplicity of the questions, the subjects

were asked to verify their addresses (District, Kebelle and House No.), the gender distribution of household

members, and the family’s monthly food budget. The telephone interviews revealed that less than 3% of the

questionnaires contained some errors: in a few cases, deceased members were recorded as family

members, and certain respondents had initially reported a non-resident member as part of their family.

About 20% of the respondents either could not be reached by telephone despite repeated attempts, or did 8 For the Pilot, the enumerators had a generic letter I obtained from Ethiopia’s Ministry of Foreign Affairs

(MFA) requesting the goodwill and co-operation of all concerned organizations and individuals for the

research. Nevertheless, a copy of the MFA letter was deemed insufficient by some respondents. A number

of them wanted to see a letter from Addis Ababa city Administration, specifically mentioning the names of

the enumerators who also needed to carry a city issued ID of their own. One particular family was so

suspicious that they had the enumerator detained by the police for hours until we were able to clear things

up with higher authorities.

Page 14: The Effects of International Migration on Migrant -Source ... · The Effects of International Migration on Migrant -Source Households: ... lottery participants in Addis Ababa ...

14

not provide their telephone numbers. Questionnaires completed by three enumerators in particular made

up a bulk of this group. Although this could be a cause for concern, it was not entirely alarming that this was

happening, because these results were coming from districts on the lower end of the income distribution.

Nonetheless, we took note of the anomaly in order to properly address it in phases II and III of the QC

procedures. However, even if 100% of the respondents were reachable by phone and the above questions

checked perfectly, additional checks were needed to make sure that the interviews were conducted with

integrity.

In phase II, the enumerators were ranked and divided into two groups – groups A and B - based on the

quality of their work.9

We then randomly selected 10% and 20% of the Questionnaires completed by group

A and B enumerators respectively to check their accuracy in person. We knocked on about 100 doors to do

this. All but four of the randomly selected completed Questionnaires passed this check to our satisfaction.

The only major problem encountered during this phase was that we could not trace one of the non-lottery

playing respondents in Arada district. Although it was likely that this person was living on the fringes of

society, thus may have disappeared for any number of reasons, we took note of this to address the issue in

phase III appropriately.

In phase III, we randomly selected about 25% of the surveys by one enumerator, whose work had turned

up additional errors, such as coding deceased or non-family members as part of the household. We then

launched the survey again to make sure this was not a common occurrence. At the end, we were satisfied

9 Group B enumerators are those whose works have turned up minor errors as well as those with higher

proportion of interviewed subjects with no phone numbers.

Page 15: The Effects of International Migration on Migrant -Source ... · The Effects of International Migration on Migrant -Source Households: ... lottery participants in Addis Ababa ...

15

that the minor errors were not common enough to pursue other methods. Most importantly, we checked,

door to door, 50% of the Questionnaires completed by the enumerator who had gathered information on a

person we could not trace during phase II. Finding that these questionnaires were remarkably accurate, we

were satisfied with the quality of the data gathered and concluded the QC procedures.

Table 1 broadly describes the data using certain key variables by treatment status and for the overall

sample. The summarized variables include estimated monthly family food budget, total estimated value of

durables owned by households, as well as their monthly energy cost, wireless phone bill and quarterly

leisure expense. Summary statistics for some of the important consumer durables (e.g. Sofa, TV) are also

used to further characterize the data. The amounts in the table are all in the Ethiopian currency (Birr).

3.4: Descriptive Statistics

Respondents were asked certain questions to check if the treatment and control subjects were balanced at

baseline. Since the first cohort of DV migrants in the sample frame left Ethiopia in 2006, some of the

questions dealt with household characteristics prior to and including 2005. The variables used for

randomization check include: mean age of households (minors, adults and all members), average family

size, pre-DV household income as well as the education levels of household heads and their spouses.

3.5: Randomization Check

Households in both groups look very similar in terms of their pre-DV characteristics (Table 2, Panel A). The

groups exhibit no systematic differences with respect to their average family size, mean age of household

members, mean age of dependents, mean age of adults, pre-intervention family income, education of

household head and spouse.

Page 16: The Effects of International Migration on Migrant -Source ... · The Effects of International Migration on Migrant -Source Households: ... lottery participants in Addis Ababa ...

16

Some of the information used in these analyses is self-reported (e.g. pre-2006 household income) and thus

may be inaccurate at the margin. Inevitably, certain respondents may have less-than-robust memory and

might have provided erroneous information. However, it would be legitimate to assume that the

inaccuracies are not serious in any way, as the reported figures are well within official figures. Further,

there is no reason to believe that the possible marginal inaccuracies are systematically different across the

two groups.

4. Empirical Results

The effect of the DV lottery can be measured using the reduced form (*). The framework allows me to

compare outcomes of households that won the lottery, that lost the lottery, and that did not participate in the

lottery. The indicator variable 𝐷𝑖 equals one if household 𝑖 won the DV lottery, and zero otherwise. The

dependent variable 𝑦𝑖 measures current outcomes for household (𝑖). These include current household

monthly food budget, anthropomorphic measures of immediate household members (BMI), estimated total

value of consumer durables owned by family, estimated total monthly expenditure on leisure, indicators for

household’s access to clean drinking water, toilets and bathroom facilities, as well as dummies for

household’s ownership of business, bank usage, and savings.

A: Estimation Frameworks

𝑦𝑖 = 𝛽 + 𝛼𝐷𝑖 + 𝜀𝑖 (*)

If all DV lottery winners migrated but none of the DV lottery losers did, 𝛼 would capture the effects of

migration. However, not all DV winners migrate and not all migrants are DV winners. Some DV lottery

winners get disqualified for falsifying their records; others fail to make the final cut due to medical reasons.

Page 17: The Effects of International Migration on Migrant -Source ... · The Effects of International Migration on Migrant -Source Households: ... lottery participants in Addis Ababa ...

17

For these and other reasons, only about sixty three percent of the DV winners actually end up migrating.

On the other hand, not everyone who migrates is a DV winner, as certain people migrate to the United

States via channels other than the DV lottery. An IV-2SLS framework, with the lottery outcomes as the

instrument for migration, will thus be used to estimate the average effects of migration.

Randomness of the lottery does not guarantee that potential outcomes are independent of the instrument.

For the IV estimates to have a causal interpretation, potential outcomes of households have to be

independent of lottery outcomes. For this to hold, it has to be the case that lottery outcomes affect

observed household outcomes only through migration. The only plausible reason for any relation between

household outcomes and the DV lottery is the latter’s effect on migration. Hence, the exclusion restriction is

readily satisfied.

In addition to the average effect of emigration for all the years, it would be interesting to study if the impact

of emigration varies with duration. As has been suggested in the literature, the effect of remittances might

diminish and disappear altogether as migrants spend more time abroad; alternatively, the effects might

increase over time, as migrants adapt to living abroad and perhaps become more successful. I will test

which of the two arguments is borne by the data (at least in the Ethiopian context), using a specification

shown below in (**), which is similar to the one used in Gibson, et al (forthcoming). I instrument for the

interaction between migration status (𝑀𝑖) and duration abroad (𝑡𝑖 ) by the interaction between the dummy

for lottery status (𝐷𝑖) and duration abroad.

𝑦𝑖 = 𝛽 + 𝛼𝑀𝑖 + µ(𝑡𝑖 ∗ 𝑀𝑖) + 𝑢𝑖 (**)

Page 18: The Effects of International Migration on Migrant -Source ... · The Effects of International Migration on Migrant -Source Households: ... lottery participants in Addis Ababa ...

18

Lottery winners have higher food budgets: they spend about 13% more on food than lottery losers (Table 3,

Panel A1). More importantly, they have higher anthropomorphic outcomes (+0.34, average BMI). DV

winners also own more and better quality consumer durables; the level of significance of this effect is

notable, given the valuation of the items is based on self-reported figures, which are noisier than current

market values. In addition, winners are 12% less likely to share latrines with other households than those

in the control group (Table 5, Panel A). The DV lottery also increases the chances of a family having

access to clean drinking water and a modern bathing facility inside its home by about 18%.

B: Reduced Form Estimates

The rate of business ownership is remarkably similar for the two groups of households (Table 4). Lottery

winners, though they have better standards of living in terms of their caloric intake and ownership of

consumer durables, do not start businesses at higher rates than non-lottery winners. Nor does winning the

lottery induce a household to use banking facilities at higher rates. The roughly 4% higher probability of

bank usage by lottery winners is statistically insignificant at traditional levels. More importantly, the

percentage of savers among the two groups is almost indistinguishable.

Both groups of households are not differentiable in terms of how much they spend on leisure (Table 3,

Panel A). Largely, this may be because of two reasons: First, the majority of the population can still afford

the most common leisure activities – movies and watching the English Premier League at bars, which are

generally reasonably priced. The other reason might be cultural. The more expensive leisure activities,

such as taking family vacations, are still rare due perhaps to social norms. In Ethiopia, family vacations are

for ferenji’s (foreigners, typically Europeans), thus few and far between, even for people who might be able

to afford them.

Page 19: The Effects of International Migration on Migrant -Source ... · The Effects of International Migration on Migrant -Source Households: ... lottery participants in Addis Ababa ...

19

Another interesting conclusion is that participants in the DV lottery (both winners and losers) have

substantially higher outcomes than non-participants, suggesting that Ethiopian DV migrants are indeed

positively selected (Tables 3, 4 and 5). Non-participants have lower food spending, lower variety and value

of durables they own, and less access to clean drinking water and convenient sanitation facilities. They are

also the least likely to use banking facilities and save. Interestingly, however, lottery non-participants spend

more on leisure activities.

The impacts of emigration on several dimensions of the remaining family’s standard of living are

significantly positive (See tables 3, 4 and 5). Families of DV migrants spend about 22% more on food, are

thus better fed and have higher body mass indexes (+0.62 average BMI). Migration of a family member

also allows a family to own more and better quality consumer durables, which includes modern household

appliances (e.g. cooking stoves) that increase the productivity of household production and enhance the

working conditions of persons using them. In a developing country like Ethiopia, since household chores

are disproportionately conducted by women and girls, the welfare of young girls and women is bound to

improve as more efficient tools of home production become available. In addition, school-age girls may be

able to focus on their education (e.g. doing their home-work) as a result of the increased efficiency gained

due to ownership of more and better quality home production tools.

C: Instrumental Variable Estimates of Effects of Migration

The gains from migration for staying family members in terms of better access to clean drinking water as

well as sanitation facilities are also remarkable. Migration of a member reduces the chances of a family

sharing a latrine with another household by 20% (Table 5, Panel B); it increases the likelihood of a family

having access to clean drinking water and having a more decent bathing facility by about 30%. By any

Page 20: The Effects of International Migration on Migrant -Source ... · The Effects of International Migration on Migrant -Source Households: ... lottery participants in Addis Ababa ...

20

measure, these are significant improvements with likely affirmative health consequences for those

impacted by migration.

Migration does not seem to have any effect on physical capital accumulation and business ownership in

sending countries (Table 4, Panel B). This may be due to a number of reasons including the following: First

of all, it may be that the positive effects of migration on household income may be just enough to pay for

essential household needs. The other reason could be that, due to the prevailing social norms, families

may have to support their less well-off relatives and even neighbors. In social environments similar to

Ethiopia’s, a family experiencing an exogenous increase in its income often lends a helping hand, typically

to other relatives in need. After taking care of the family’s basic needs and social obligations, therefore,

migrants’ families may not have much left for savings or investment activities. However, the remarkable

similarities in business ownership rates of DV winners, DV losers and lottery non-participants (see Panels

A1 & A2 of Table 4) might suggest that institutional, policy or cultural constraints could be more binding to

productive investment activities than household liquidity constraints.

The duration effects obtained by estimating (**) would be biased if lottery entrants in different years were

differently selected. To check if this is an issue or not, I grouped the subjects into the earlier group (2006

and 2007 lottery winners) and the more recent group (2008 and 2009 lottery winners) and compared them

in terms of certain characteristics. I find that that the two groups are fairly similar in terms of their baseline

characteristics (Table 2, Panel B). The groups exhibit some difference only with respect to the average age

of their members.

D: Duration Effects

The point estimates in the third panels of tables 3, 4 and 5 suggest that the impacts of migration do not

diminish with DV migrants spending more time in the US. There is in fact some evidence to suggest that

Page 21: The Effects of International Migration on Migrant -Source ... · The Effects of International Migration on Migrant -Source Households: ... lottery participants in Addis Ababa ...

21

the treatment effects grow with longer duration of the lottery migrants in the US. Migrants, who are typically

young adults, might earn more as they learn newer and more marketable skills, allowing them to send more

money home to their parents.

5. Conclusion

Much has been done to understand the impacts of international migration; still, more research is needed to

improve our knowledge of how migration affects senders and migrants. In making the case that a new

research agenda is needed to better understand the consequences of emigration, Clemons (2011)

intriguingly argues that allowing a freer global mobility of labor could lead to the doubling of world GDP.

Even traditional research topics on international migration, such as the literature on “brain drain”, have

plenty of room to grow. It is not entirely clear if high skilled emigration is detrimental to the development of

low income countries, as is widely believed to be the case. According to Gibson and McKenzie (2011),

“…we are still some way from a comprehensive global answer on the effect of brain drain on sending

country growth and development outcomes, and further still from knowing the efficacy of policies chosen

with high-skilled migration in mind.” Adding a voice to the call for more research from a different angle,

Yang (2011) argues, “… new data collection and empirical approaches have expanded what we know

about migration, remittances and development in recent years, but many fundamental questions remain

incompletely answered.”

This study has uncovered new findings that will not only be useful for policy makers, but also suggest

possible new areas of inquiry. The paper finds that migration contributes positively to the welfare of family

members remaining behind, by allowing them to increase their consumption expenditure. These positive

Page 22: The Effects of International Migration on Migrant -Source ... · The Effects of International Migration on Migrant -Source Households: ... lottery participants in Addis Ababa ...

22

effects do not decrease with longer duration of migrants abroad. However, emigration does not have any

impact on productive investments in sending countries. Exploring why this is so might be a worthy research

endeavor in the future in order to “maximize the development benefits of migration.” There are a number of

possibilities explaining this, some of which were considered in the last section.

By demonstrating that labor market integration through international migration helps sending households in

regions of the world where decades of development aid has made little difference thus far, this paper may

contribute to the policy debate on international migration in the recipient countries. The conclusion that

emigration helps family members who are left behind, could create a space for policy makers in the aid-

fatigued, migrant-recipient nations, allowing them to pursue creative liberal migration policies, such as the

DV lottery, particularly if these policies benefit the recipient nations as well.

Page 23: The Effects of International Migration on Migrant -Source ... · The Effects of International Migration on Migrant -Source Households: ... lottery participants in Addis Ababa ...

Acosta, P., P. Fajnzylber., and H. Lopez, “The Impact of Remittances on Poverty and Human Capital:

Evidence from Latin American Household Surveys,” In C. Ozden and M. Schiff eds., International Migration,

Economic Development and Policy (Washington DC: The World Bank and Palgrave Macmillan, 2007)

REFERENCES

Adams, R., “Remittances and Poverty in Guatemala,” World Bank Policy Research Working Paper No.

3418 (2004)

Adams, R., J. Page, “International Migration, Remittances and Poverty in Developing Countries,” World

Bank Policy Research Working Paper no. 3179 (2003)

Angrist, J., “Lifetime Earnings and the Vietnam Era Draft Lottery: Evidence from the Social Security

Administrative Records,” American Economic Review 80(3) (1990), 313-336.

Angrist, J., G. Imbens, “Identification and Estimation of Local Average Treatment Effects,” Econometrica

62(2) (1994), 467-75

Angrist, J., G. Imbens, and D. Rubin, “Identification of Causal Effects Using Instrumental Variables,”

Journal of the American Statistical Association 91 (1996), 444-55

Angrist, J., S. Pischke, Mostly Harmless Econometrics (Princeton, NJ: Princeton University Press, 2009).

Antman, F., “The Impact of Migration on Family Left Behind,” Forthcoming In: A. Constant and K.F.

Zimmermann (eds.), International Handbook on the Economics of Migration (2012).

Clemens, M., “Economics and Emigration: Trillion-Dollar Bills on the Sidewalk?” Journal of Economic Perspectives, 25(3) (2011), 83 – 106

Page 24: The Effects of International Migration on Migrant -Source ... · The Effects of International Migration on Migrant -Source Households: ... lottery participants in Addis Ababa ...

Cox, A. and M. Ureta, “International Migration, Remittances, and Schooling: Evidence from El-Salvador,”

NBER Working Paper No. 9766 (2003).

Chiquar, D., G. Hanson, “International Migration, Self-Selection and the Distribution of Wages: Evidence

from Mexico and the United States,” Journal of Political Economy 113(2) (2005), 39-81.

Chiswick, B., “Are Immigrants Favorably Self Selected?” American Economic Review 89(2) (1999), 181-

185.

Dehijha, W., “Causal Effects in Non-Experimental Studies: Re-evaluating the Evolution of Training

Programs,” Journal of the American Statistical Association 94 (1999) 1053-62.

Duflo, E., R. Glennerster, M. Kremer, “Using Randomization in Development Economics Research: A

Toolkit,” NBER Discussion Paper Series No. 6059 (2006).

Easterly, W., “Was Development Assistance a Mistake? “ American Economic Review 97(2) (2007), 328-

332.

Esquivel, G., and A. Huerta-Pineda, “Remittances and Poverty in Mexico: A Propensity Score Matching

Approach,” Integration and Trade No. 27 (2007), 1-28.

Funkhouser, E., “Remittances from international migration: a comparison of El Salvador and Nicaragua,”

The Review of Economics and Statistics 77 (1) (1995), 137–146.

Gibson, J., D. McKenzie, “Eight Questions about Brain Drain” Journal of Economic Perspectives, 25(3) (2011), 107 – 128

Gibson, J., D. Mckenzie, and S. Stillman, “The Imapcts of International Migration on Remaining Household

Members: Omnibus Results from a Migration Lottery Program,” The Review of Economics and Statistics

93(4) (2011) 1297 – 1318.

Page 25: The Effects of International Migration on Migrant -Source ... · The Effects of International Migration on Migrant -Source Households: ... lottery participants in Addis Ababa ...

_______ “Accounting for selection and duration-dependent heterogeneity when estimating the impact of

emigration on incomes and poverty in sending areas,” Economic Development and Cultural Change

(forthcoming)

Kunt, A., E. Lopez-Cordova, M. Martinez-Peria, C. Woodruff, “Remittances and Banking Sector Breadth:

Evidence from Mexico,” The World Bank Policy Research Working Paper No. 4983 (2009).

Lalonde, R., “Evaluating the Econometric Evaluations of Training Programs,” American Economic Review

76(4) (1986) 604-620.

Lopez-Cordova, E., “Globalization, Migration and Development: The Role of Mexican Migrant

Remittances,” Institute for the Integration of Latin America and the Caribbean Working Paper No. 41

(2006).

Lopez-Cordova, E., and A. Olmedo, “International Remittances and Development: Existing Evidence,

Policies and Recommendations,” Institute for the Integration of Latin America and the Caribbean

Occasional Paper No. 41 (2006).

Mansuri, G., “Migration, Sex Bias and Child Growth in Rural Pakistan,” The World Bank Policy Research

Working Paper No. 3946 (2006)

Mckenzie, D., J. Gibson, and S. Stillman, “Experimental Vs. Non-Experimental Measures of the Income

Gains from Migration,” IZA Discussion Paper No. 2087 (2006)

McKenzie, D. and H. Rappoport, “Network Effects and the Dynamics of Migration and Inequality: Theory

and Evidence from Mexico,” Journal of Development Economics 84(1) (2007) 1-24.

Moffitt, R., “Remarks on the Analysis of Causal Relationships in Population Research,” Demography 42(1)

(2005) 91-108.

Page 26: The Effects of International Migration on Migrant -Source ... · The Effects of International Migration on Migrant -Source Households: ... lottery participants in Addis Ababa ...

Torres, G. and B. Pelham, “One-Quarter of World’s Population May Wish to Migrate” http://www.gallup.com/poll/108325/onequarter-worlds-populationmay-wish-migrate.aspx Accessed September, 19, 2011

Woodruff, C. and R. Zenteno, “Remittances and Microenterprises in Mexico,” Mimeo, UCSD (2004)

Yang, D., “International Migration, Remittances and Household Investment: Evidence from the Philippines

Migrants’ Exchange Rate Shocks,” The Economic Journal 118 (April) (2008) 591-630.

Yang, D., “Migrant Remittances” Journal of Economic Perspectives, 25(3) (2011), 129 – 1152 Yang, D., A. Martinez, “Remittances and Poverty in Migrants’ Home Areas: Evidence from the Philippines,”

In Calgar Ozden and Maurice Schiff eds., International Migration, Remittances and the Brain Drain,

(Washington DC: The World Bank, 2005).

Page 27: The Effects of International Migration on Migrant -Source ... · The Effects of International Migration on Migrant -Source Households: ... lottery participants in Addis Ababa ...

Item N Mean SD Min MaxFood Expenditure 494 1,284 694 300 4,500Energy Cost 432 131 87 4 510Mobile Phone Usage Fee 448 147 198 15 2,500House Rent 259 179 370 2 3,000Estimated Value of Durables 494 16,282 47,308 0 861,600Leisure Expense (Quarterly) 494 29 75 0 600

Item N Mean SD Min MaxFood Expenditure 245 1,370 759 300 4,500Energy Cost 216 147 95 4 510Mobile Phone Usage Fee 222 155 222 24 2,500House Rent 115 219 440 3 3,000Estimated Value of Durables 245 21,384 64,411 0 861,600Leisure Expense (Quarterly) 245 27 65 0 420

Item N Mean SD Min MaxFood Expenditure 249 1,199 614 300 3,000Energy Cost 216 116 76 10 500Mobile Phone Usage Fee 226 139 171 15 2,000House Rent 144 146 301 2 2,000Estimated Value of Durables 249 11,262 17,779 0 199,250Leisure Expense (Quarterly) 249 30 85 0 600

Item N Mean SD Min MaxSofa 424 3,354 2,675 200 30,000Stove 353 535 965 25 7,000TV 465 2,777 2,453 100 43,200Mobile Phone 459 1,923 1,861 200 17,000Computer 86 7,312 5,015 400 30,000Car 27 116,796 153,362 4,500 800,000

Panel C: Summary Statistics Of Selected Durables

2) All flow variables except Leisure Expense are monthly estimates

Table 1: Descriptive StatisticsPanel A: Summary Statistics For The Overall Sample

Panel B: Summary Statistics By Treatment Status

DV Winners

DV Losers

Notes: 1) All figures are in the Ethiopian currency (Birr).

Page 28: The Effects of International Migration on Migrant -Source ... · The Effects of International Migration on Migrant -Source Households: ... lottery participants in Addis Ababa ...

Number of Observations

DV Losers (Mean)

DV Winners (Mean)

Mean Difference P-Value

Household Size 494 5 5.09 -0.09 0.59Pre-DV Income (Ethiopian Birr) 428 1,453.93 1,513.80 -59.87 0.6Education of HH Head (Pre-DV) 447 1.9 2.06 -0.16 0.12Spouse's Education 245 1.88 1.88 0 0.98Age 2753 33.22 33.17 0.04 0.94Age of Minors 394 12.54 12.41 0.13 0.8Age of Adults 2359 36.73 36.59 0.15 0.81

Number of Observations

Earlier DV Winners (Mean)

More Recent DV Winners (Mean)

Mean Difference P-Value

Household Size 246 5.25 4.95 0.3 0.23Pre-DV Income (Ethiopian Birr) 222 1,503.13 1,528.18 -25.05 0.87Education of HH Head (Pre-DV) 219 2.16 1.97 0.19 0.23Spouse's Education 116 1.83 1.93 -0.1 0.57Age 1438 33.97 32.42 1.56 0.07Age of Minors 203 12.22 12.56 -0.34 0.62Age of Adults 1235 37.06 36.12 0.93 0.26

3 = Some High School; 4 = Bachelors Degree and AboveNote: 1) Education Indicators: 0 = Illiterate; 1 = Less than High School; 2 = High School;

Panel B: Selectivity Check of Earlier and More Recent Lottery Winners

Table 2: Randomization and Selectivity Checks

Panel A: Overall Randomization Check

Page 29: The Effects of International Migration on Migrant -Source ... · The Effects of International Migration on Migrant -Source Households: ... lottery participants in Addis Ababa ...

Family Food Budget Value of Durables Leisure Expense Energy Expense BMI

Effect of DV Lottery 0.13** 0.26** -0.2 0.24*** 0.34* (2.73) (3.09) (-1.12) (3.75) (2.37)

Number of Observations 494 489 130 432 2412

Family Food Budget Value of Durables Leisure Expense Energy Expense BMI

Effect of DV Lottery 0.36*** 0.73*** -0.57*** 0.39*** 0.68***(7.24) (7.5) (-5.88) (6.17) (4.75)

Number of Observations 520 508 245 432 2482

Family Food Budget Value of Durables Leisure Expense Energy Expense BMI

Effect of Migration 0.21** 0.42** -0.3 0.40*** 0.62* (2.72) (3.07) (-1.10) (3.76) (2.36)

Number of Observations 494 489 130 432 2412

Family Food Budget Value of Durables Leisure Expense Energy Expense BMI

Effect of Migration 0.03 0.19 0.47 0.36* 0.02(0.24) (0.75) (1.07) (1.83) (0.04)

Effect of Each Year in the US 0.07 0.09 -0.3 0.02 0.22(1.43) (1.06) (-1.78) (0.23) (1.24)

Number of Observations 494 489 130 432 2412

Note: BMI regression is level - level; all other coefficient estimates are semi-elasticity.

Panel A1: OLS Estimates Using the Control Group

Panel A2: OLS Estimates Using Lottery Non-Participants

Table 3: Effects of the DV Lottery and Migration on Monthly Expenditure on Selected Items, BMI and Durable Ownership

Panel B: Instrumental Variables Estimates

Panel C: Estimates With Duration Effects

Page 30: The Effects of International Migration on Migrant -Source ... · The Effects of International Migration on Migrant -Source Households: ... lottery participants in Addis Ababa ...

Business Ownership

BankUsage

Savings Account

Business Ownership

BankUsage

Savings Account

Effect of DV Lottery 0.02 0.04 0.02 0.08 0.11 0.06(0.58) (0.97) (0.56) (0.58) (0.98) (0.56)

Number of Observations 493 492 491 493 492 491

Business Ownership

BankUsage

Savings Account

Business Ownership

BankUsage

Savings Account

Effect of DV Lottery 0.03 0.13** 0.07 0.14 0.35** 0.19(0.97) (3.15) (1.7) (0.97) (3.12) (1.7)

Number of Observations 516 516 515 516 516 515

Business Ownership

BankUsage

Savings Account

Business Ownership

BankUsage

Savings Account

Effect of Migration 0.03 0.07 0.04 0.14 0.19 0.11(0.58) (0.97) (0.56) (0.58) (0.99) (0.57)

Number of Observations 493 492 491 493 492 491

Business Ownership

BankUsage

Savings Account

Effect of Migration -0.12 -0.06 -0.27* (-1.22) (-0.44) (-2.02) 0.06 0.05 0.13** (1.8) (1.13) (2.72)

Number of Observations 493 492 491

Effect of Each Year in the US

Panel C: Estimates With Duration Effects

Table 4: Effects of the DV Lottery and Migration on Business Ownership and Bank Usage

Panel A1: OLS Estimates Using the Control Group

OLS Probit

Panel A2: OLS Estimates Using Lottery Non-Participants

OLS Probit

Panel B: Instrumental Variables Estimates

IV IV Probit

Page 31: The Effects of International Migration on Migrant -Source ... · The Effects of International Migration on Migrant -Source Households: ... lottery participants in Addis Ababa ...

Water Bath Toilet Latrine Share

Effect of DV Lottery 0.18*** 0.18** 0.10* -0.12** (3.32) (3.3) (2.23) (-2.73)

Number of Observations 480 486 485 473

Water Bath Toilet Latrine Share

Effect of DV Lottery 0.32*** 0.30*** 0.15*** -0.15***(5.79) (5.54) (3.63) (-3.49)

Number of Observations 497 502 503 489

Water Bath Toilet Latrine Share

Effect of Migration 0.30** 0.30** 0.16* -0.20** (3.28) (3.26) (2.23) (-2.70)

Number of Observations 480 486 485 473

Water Bath Toilet Latrine Share

Effect of Migration 0.06 0.12 0.11 0(0.34) (0.7) (0.81) (0.02)

Effect of Each Year in the US 0.1 0.07 0.02 -0.08(1.63) (1.19) (0.44) (-1.70)

Number of Observations 480 486 485 473

Panel C: Estimates With Duration Effects

Panel B: Instrumental Variables Estimates

Table 5: Effects of the DV Lottery and Migration on Clean Water and Sanitation Facilities

Panel A1: OLS Estimates Using the Control Group

Panel A2: OLS Estimates Using Lottery Non-Participants

Page 32: The Effects of International Migration on Migrant -Source ... · The Effects of International Migration on Migrant -Source Households: ... lottery participants in Addis Ababa ...

In the following, I roughly estimate the total number of applicants for DV lottery from Addis Ababa (TADLA).

TADLA is based on the following assumptions:

Appendix: Estimate of Total Number of Applicants for DV lottery from Addis Ababa (“TADLA”)

1) In a given year, the probability of success for an individual applicant is the same in every country.

Let 𝑊(𝑡), 𝐴(𝑡) and 𝜔(𝑡) denote total DV winners, applicants and the probability of winning the lottery

(𝑡) years ago respectively. Hence, 𝜔(𝑡)𝑗 = 𝜔(𝑡) 𝑘 = 𝑊(𝑡)𝐴(𝑡)

where, j and k are any two DV participating

countries. This assumption can be rationalized by the fact that, at the time of application, citizens of

participating countries have knowledge of roughly how many people are eligible for DV migration from their

home countries.

2) Since the inception of the DV lottery in 1995, 𝜔(𝑡) has declined exponentially over the years. It is

a fact that the lottery has become more popular over the years, as more and more people have been

learning about the program. Over the last four respective years for which data are readily and publicly

available, 5.5, 6.4, 9.1, and 13.6 million people have applied for the lottery from eligible countries; whereas,

the number of people admitted to the US on DV is fixed by law. This, more or less, justifies the exponential

decline of 𝜔(𝑡) assumption.

3) All Ethiopian DV applicants were from Addis Ababa when the program was instituted in 1995, but

the proportion of applicants from the capital is currently about 23% of all Ethiopian applicants. 10 Further,

the rate of decline of the proportion of applicants from Addis Ababa between 1995 and 2009 is assumed to

be constant, hence a yearly rate of decline of 10%. Again, this assumption reflects the growing popularity

over time of the DV lottery spatially. Back in time, the residents of Addis Ababa must have enjoyed

exclusive access to information regarding the lottery, whereas now they would have to increasingly

Page 33: The Effects of International Migration on Migrant -Source ... · The Effects of International Migration on Migrant -Source Households: ... lottery participants in Addis Ababa ...

compete with people from the other urban areas in the country for limited DV spots. Rural residents are the

least likely to attempt to enter the lottery, given the educational and electronic-only-applications

requirements that seem to preclude farming and nomadic households.

Given these reasonable assumptions, the following two formulae give the upper and lower bound estimates

of TADLA. The upper bound assumes that the pool of DV lottery applicants changes every year, whereas

the most conservative estimate allows that people apply for the lottery every year. Supposing 𝐴(𝑡)𝑎

people applied for the DV lottery from Addis Ababa t years ago, the upper and the lower bounds of TADLA

would respectively be approximated by (1) and (2) below: In (1), EDVW stands for total number of DV

winners from Ethiopia every year. This figure has been roughly constant over the years in the range of

4,500 people. 𝜔(0), which is current probability of winning the lottery, is 0.0075.

TADLA =∑ 𝐴(𝑡)𝑎 15𝑡=0 = 𝐾 ∗ ∑ 𝑒−(.06𝑡+1.5) 15

t=0 ; where, K = EDVW/ 𝜔(0) (1)

TADLA =∑ 𝐴(𝑡)𝑎 15𝑡=0 = 𝐾 ∗ {[4

5∗ 𝜔(0) ∗ ∑ 𝑒−.1(𝑡−15)] 15

t=0 + 𝑒1.5} (2)

In all probability, true TADLA is somewhere between these two estimates, as it is safe to assume that the

average resident of Addis would enter the lottery more than once, but likely not every year. Lacking any

information about the number of times the typical aspiring migrant enters the lottery, if we considered an

average of the lower and upper bound estimates as a rough approximation of true TADLA, we get:

TADLA = ∑ 𝐴(𝑡)𝑎 15𝑡=0 = 𝐾

2∗ {∑ 𝑒−(.06𝑡+1.5) 15

t=0 + [45∗ 𝜔(0) ∗ [∑ 𝑒−.1(𝑡−15)] 15

t=0 + 𝑒1.5} (3)

This formula suggests that about 1,500,000 different individuals have so far applied for the lottery from

Addis Ababa. According to the 2007 census of Ethiopia, roughly a third of Addis Ababa’s population is

Page 34: The Effects of International Migration on Migrant -Source ... · The Effects of International Migration on Migrant -Source Households: ... lottery participants in Addis Ababa ...

under 18, which is the age of eligibility for DV lottery. If Addis Ababa’s age structure has roughly been

stable for the last 15 years, (3) suggests that nearly 75% of the city’s adult population have entered the

migration lottery at least once.

However, as a simple average, (3) may not be an accurate approximation of true TADLA. It implicitly

assumes that the number of times every household entered the lottery is roughly the same. In reality,

households in certain socio economic groups are more likely to apply more frequently than those in a

different category. For obvious reasons, those in the extreme ends of the income distribution may not apply

for the lottery as frequently, if any, as those in the middle. Thus (3) overestimates true TADLA since it is

highly likely that those in the middle which constitute the majority of the city’s residents apply more

frequently than those in the other groups. The better estimate of TADLA may thus be anywhere between

the roughly 450,000 adults (the most conservative estimate obtained with assumption of always-repeat

applicants) and the 1.5 million people estimated with (3). Taking the average of these figures, it is

approximated that about 50% of the city’s adult population have entered the migration lottery at least once.

Further stipulating that the number of applicants from a given household is directly proportional to

household size, it is estimated that one in two households in the city have at a minimum one member who

has applied, at least once, for the DV lottery. Note: the most conservative estimate yields that about one in

four households have applied for the lottery.

Given the overwhelming popularity of the lottery, its negligible cost of entry and its expected value, the

estimate is not as unrealistic as it may seem at first glance. This figure is indeed consistent with the World

Bank report of 2007, which finds that between 50% and 90% of the young adults in certain developing

countries, would like to migrate if given the opportunity.