International remittances and household welfare in vietnam from VHLSS 2006 to VHLSS 2008

This paper examines the impact of international remittances, which increases over time, on the household welfare of receiving households in Vietnam.. When studying separately for urban a

UNIVERSITY OF ECONOMICS INSTITUTE OF SOCIAL STUDIES

VIETNAM - NETHERLANDS PROJECT FOR M.A IN DEVELOPMENT ECONOMICS

INTERNATIONAL REMITTANCES AND

HOUSEHOLD WELFARE IN VIETNAM

FROM VHLSS 2006 AND VHLSS 2008

“This paper was submitted in partial fulfillment of the requirements for the

Masters of Development Economics (MDE) degree at the Vietnam - The Netherlands Programme (VNP), August/2012”

BY

NGUYEN VAN PHUC

Ho Chi Minh City, August 2012

ABSTRACT

International remittances has more important role in progress of economy and society in Vietnam With US$ 9 billions in 2011, Vietnam is one of 10 top countries received the remittances in the world This paper examines the impact of international remittances, which increases over time, on the household welfare of receiving households in Vietnam The thesis has combined the propensity score matching and difference-in-differences methods with panel data taken from Vietnam Household Living Standard Surveys 2006 and 2008 It is found that international remittances increases income and expenditures for recipients, but the effect of remittances on expenses of healthcare and education is not statistically significant although the expenses also rise over time When studying separately for urban and rural areas, the thesis found that the impact of international remittance on income and expenditures are positive and statistically significant for rural areas and positive and insignificant for urban areas; meanwhile, the paper has not detected the affects of foreign remittances on education and healthcare in urban or rural areas

Key words: international remittances, household welfare, household surveys, Vietnam

TABLE OF CONTENTS

ABSTRACT i

TABLE OF CONTENTS ii

CHAPTER 1: INTRODUCTION 04

1.1 Problem statement

1.2 Research objectives

1.3 Research question

1.4 Thesis structure ………

CHAPTER 2: LITERATURE REVIEW 08

2.1 The key concepts

2.2 Empirical literature

2.3 Empirical framework

2.4 Overview of theory on impact evaluation

2.5 Summary of PSM and DD methods

2.5.1 Propensity Score Matching (PSM)

2.5.2 Difference-in-difference (DD)

CHAPTER 3: METHODOLOGY AND DATA 22

3.1 The research model

3.2 Variable introduction

3.3 Data …

3.4 Estimation Strategy

CHAPTER 4: INTERNATIONAL REMITTANCES IN VIETNAM 33

4.1 General view of migration remittances in Vietnam ………….………

4.2 Role of international remittances on economy………

4.3 International remittances and household welfare in Vietnam………

CHAPTER 5: EMPIRICAL ANALYSIS……… 42

5.1 Data description

5.2 Estimation results

5.3 Interpretation of results

CHAPTER 6: CONCLUSIONS AND RECOMMENDATIONS 55

6.1 Conclusions

6.2 Recommendations

6.3 Limitations of the paper

REFERENCES

APPENDIX

CHAPTER 1: INTRODUCTION

1.1 Problem statement

After foreign direct investment (FDI), remittances by international migrants to their home countries contribute the largest source of external finance to developing countries, around

$300 billion/year in stage of 1995-2005 The funds are used for consumption and investment

in migrants’ home countries Remittances are found to give significant impacts on receiving households, especially on low-income families It may help households to establish or expand their small business (Woodruff and Zenteno, 2007 and Amuedo-Dorantes and Pozo, 2006), increase of expenditures in family (Ahmed and et al 2010), or reduction on poverty (Adams,

1991 and Lopez-Cordova, 2005)

Some other studies examine the impact of remittances on household welfare, such as savings, consumption, health care and education However, the empirical studies exposed the different results about the impact of remittances on household welfare For instance, Adams (2005), Adams and Cuecuecha (2010) found that the remittances has positive impacts on health-care and educational expenditures in Guatemala Ahmed and et al (2010) showed the significant impacts of remittances on food, education, clothing and recreation in Pakistan Conversely, in the study of McKenzie and Rapoport (2006), the negative impact of migration detected on schooling ratio of children; or Hildebrandt and McKenzie (2005) uncovered that the preventative health-care of children in receipt households lower than in non-receipt households in Mexico

According to Vietnamese Oversea Committee (2009), there are about 4 millions Vietnamese living, working and studying permanently in 102 nations and territories in the world The international remittances have been rising over time in Vietnam The average of remittances from oversea Vietnamese sent to the home country in 2008-2010 by formal channel is bigger than $7 billion/year, which accounts more 7% of GDP in that period So the fact indicates the important of international remittances on economic progress in general and improve the living standard of receiving households in particular There are many studies on the impact of migration in Vietnam, such as Dang (2001), Djamba and et al (1999), Andrew T Pham (2010)

Nonetheless, there are only a few of studies that examined the impact of international remittances on household welfare in Vietnam, such as the studies of Pfau and Long (2008) and Nguyen (2009) Pfau and Long (2008) studied the impact of international remittances on household welfare in term of economic inequality and poverty By using the Vietnam (Household) Living Standards Survey in 1992/93, 1997/98, 2002 and 2004, this study found that foreign remittances reduce inequality and poverty in Vietnam with regards to per-capita household expenditures and income Nguyen (2009) examined the impact of international and internal remittances on welfare of receiving households By using the panel data from VHLSS 2002 and VHLSS 2004, he concluded that the international remittances have positive and significant impacts on income, expenditures of non-food consumption (excluding health-care and educational spending); but not significant impacts on expenditures of food, health-care and education He also showed the effect of remittances in urban area is much more than

in rural area Differently from the paper of Nguyen (2009), the thesis of Toan (2010) and Ha (2010) showed that the foreign remittances have positive and significant impacts on health-care and educational expenditures, but not significant between urban and rural areas through using data from VHLSS 2006

Therefore, the objective of the paper is to re-examine the impact of international remittances

on the welfare of receiving household in Vietnam by some reasons The first, the last studies used the series of data only until 2004, for instant in the period 2002-2004 the average of international remittances is about more US$ 2.8 billion per year However, in our paper, we use the data from VHLSS 2006 and 2008 with the average of foreign remittances is double more to US$ 5.8 billon per year in 2006-2008 compared to the period of 2002-2004 The increasing trend of remittances over time, which may be change the relationship between remittances and household welfare, compared the last studies The second reason is the results of Toan (2010) and Ha (2010), based on VHLSS 2006, about the impact of international remittances, which is difference from Nguyen (2009)’s results The last reason, the result of empirical studies is important referential sources for policy-makers on building suitable policies with different stages, because of the implementation of non-suitable policy may lead to waste social resources

Relying on that, the thesis is to contribute some new findings to the debate about relationships between international remittances and household welfare in Vietnam and

suggest some suitable policies for the government to explore efficiently international remittances on improving the household welfare in Vietnam

1.2 Research objectives

The thesis has three objectives:

1) To determine whether international remittances significantly influence the household welfare in Vietnam through total income, total expenditures, health-care expenditures and education expenditures

2) To examine whether there is the differential impact of remittances on household welfare between urban and rural area

3) To make recommendations to policy-makers for efficiently exploiting remittances, this improves the household welfare

1.3 Research question

To meet the objectives, the study has three research questions:

1) Does international remittance significantly influence on the household welfare

concepts, the results of empirical studies, and summarizes the methods of impact evaluation

The third chapter presents the research model, describes the variables and the dataset, and estimation strategy The fourth chapter illustrates an overview of international remittances in

Vietnam, the role of the remittance for the economy in general and statistical descriptions of

the relationship between the remittance and household welfare in particular The fifth chapter

shows statistical description of the variables, the results of difference-in-difference (DD)

regression with propensity score matching (PSM), and interpretation of researching results

The last chapter presents a summary of main findings, policy recommendations and the

limitation of the thesis

CHAPTER 2: LITERATURE REVIEW

2.1 The key concepts

International Monetary Fund defined the economic concept of remittances as follows:

“Remittances represent household income from foreign economies arising mainly from the temporary or permanent movement of people to those economies Remittances include cash and non-cash items that flow through formal channels, such as across electric wire, or through informal channels, such money or goods carried across borders They largely consist of funds and non-cash items sent or given by individuals who have migrated to a new economy and become residents there, and the net compensation of border, seasonal, or other short-term workers who are temporarily employed in an economy in which they are not resident” (Appendix 5 on remittances to the “Balance of Payments and Investment Position

Manual” (2008)

In this research, we only examine international remittances by formal channels For informal channels, it is very difficult to define how flows in amount of informal remittances and what the effects on the economy are It needs further deeply studies, which is excluded in the research

Welfare is physical and mental health and happiness, especially of a person It relates to the

income and consumption of a person Household is a group of people, often a family, who

live together (Cambridge Advanced Learner's Dictionary) However, the economic concept

of the household-welfare is very complex and wide In this paper, we only to analyze the impacts of international remittances on household welfare by examining the impact on indicators of household welfare: income and expenditures of the receipt households

2.2 Empirical literature

There are many researches, which describe the role of remittances in the economic progress

of the nations The motivations of overseas migrants are altruism (Lucas and Stark, 1985), helping family members to improve their house’s infrastructure (Duryea et al 2005), financing for household business (Woodruff and Zenteno 2007, and Amuedo Dorantes and Pozo 2006), covering medical expenses (Amuedo Dorantes and Pozo, 2006), and contributing

in education investment of children (Edward and Ureta, 2003) Many empirical studies explore the impact of international remittances on welfare of receiving households, such as Quartey (2006) in Ghana, Soraya (2007) in Philippines, Subedi (2009) in Nepal, Adams and Cuecuecha (2010) in Guatemala, Ahmed and et al in Pakistan (2010) and Raihan (2009) in Bangladesh

Quartey (2006) used Ghana Living Standards Survey (GLSS) to examine the impact of international remittances on household welfare in Ghana He concluded that international remittances is the important sources of income for consumption smoothing, improving household welfare and decreasing negative effects of economic shocks, and households which own land can withstand economic shocks and have better welfare than those without land Soraya (2009) found the positive and significant impact of international remittances on education and recreation expenditures in Philippines

By using Nepal Living Standards Survey (NLSS) for analyzing in Nepal, Subedi (2009) realized that remittances from India increase income and decrease inequality on receiving households At the micro-level, Nepal reached a significant reduction in poverty over the period of 1996-2006 (from 42% in 1995-1996 to 2003-2204), despite a low economic growth and political instability in that period Therefore, the remittance is one of key factors for declining the poverty in Nepal

Meanwhile, Raihan et al (2009) carried out the examining effects of international remittances

on household consumption and poverty at macro and household levels in Bangladesh, where the remittances accounted 10% GDP in 2008 At macro level, Raihan found the positive impact of foreign remittances on economy and reducing poverty At micro level, he discovered the different effects of remittances on indicators of household welfare: positive and significant impact on food and housing, positive but insignificant on education and health-care, and negative and significant on durable good The positive and statistically significant impacts on health-care of remittances also were found in Guatemala (Adams and Cuecuecha, A 2010) In Pakistan, Ahmed and et al (2010) showed the significant impacts of remittances on food, education, clothing and recreation

In Vietnam, by using the Vietnam (Household) Living Standards Survey in 1992/93, 1997/98, 2002 and 2004, Pfau and Long (2008) included that foreign remittances come from

throughout the world, but the United State is a main source The destinations of international remittances have become more diversification when they move away from Ho Chi Minh City and other urban areas to other regions and rural areas over time However, the percentage of household receiving external remittances held at around 5 to 7 percent of population The elderly, female-headed and the head not working households received disproportionately foreign remittances As a result, these remittances help improving equality in Vietnam in term of per-capita household expenditures, although the improvements are small Furthermore, international remittances are also to help to reduce poverty in Vietnam In the paper of Nguyen (2009), it was examined the impact of international and internal remittances

on welfare of receiving households Namely, the paper focuses on direct welfare indicators: income, consumption expenditure, food and non-food expenditures, education and health-care By using the panel data from VHLSS 2002 and VHLSS 2004, he concluded that the international remittances have positive and significant impacts on income, expenditures of non-food consumption (excluding health-care and educational spending); but not significant impacts on expenditures of food, health-care and education

2.3 Empirical framework

As stated above there are various empirical studies analyzing the impact of international remittances on household welfare The most relevant to research questions mentioned in this paper including the researches of Raihan (2009) in Bangladesh, Adams and Cuecuecha (2010) in Guatemala, Nguyen (2009) in Vietnam and Quartey (2006) in Ghana

Raihan and et al (2009) examined the relationship between remittances and expenditures in Bangladesh by using the multivariate regression model, which summarized as follows:

Household exp = ß 0 + ∑ ß j X ij + ε

Where Household exp is the household expenditures as dependent variable (i.e., housing,

medication, education, durable good and food) X ij is the explanatory variables including

international remittances, household and geographic characteristic, for instant household size, education level of household head, religion, marital status, urban or rural etc The paper revealed that an international remittance has different impact on indicators of household welfare: positive and significant impact on food and housing, positive but insignificant on education and health-care, and negative and significant on durable good

Quartey (2006) ascertained the impact of remittances on household welfare by using the semi-log functions:

Log (Ui) = α + ∑ ßjXij + εi

Where Ui is real per capita expenditures Xij are the vector of explanatory variables in cluding foreign remittances and economic shocks (as measured by the food and non-food prices) The results of the study which showed international remittances is the important sources of income for consumption smoothing, improving household welfare and decreasing negative effects of economic shocks The trend of using the semi-log functional form of income and expenditures, because income and expenditures frequent follow log-normal distribution, as mentioned above, can be found from empirical studies, e.g Glewwe (1991) and Nguyen (2009)

Nguyen (2009) examined the impact of remittances on household welfare in Viet Nam by using the method of Average Treatment Effect on the Treated (ATT), which summarized as

followings The impact of receiving remittances on a household i: ∆ i = Y i1 – Y i0 , where Y 1

and Y 0 denote potential outcomes (observed values of household income and expenditures)

with remittances and no remittances Therefore, Average Treatment Effects (ATT) are:

ATT = E(∆ i │D i = 1) = E(Y i1 │D i = 1) – E(Y i0 │D i = 1)

Where D is a dummy variable (equal to 1 for the receiving household and zero otherwise)

E(Y i0 │D i = 1) is not observed and has to be estimated This is called a counterfactual

outcome, which is the outcome of the recipients if they had not received remittances

E(Y i1 │D i = 1) is the observed outcome, which is followed the semi-log functional form

The application of Average Treatment Effects (ATT) on impact valuation of income on expenditures, which also can be found at the empirical studies (e.g Heckman et al., 1999 and Adams and Cuecuecha, 2010)

In the thesis we co-ordinate a difference-in-differences or double-difference (DD) method with propensity score matching (PSM) for the impact evaluation of international remittance

on household welfare The basis content of PSM is to find a control group or comparison group (without remittances), which has the similar distribution of the variables X as the

treatment group (with remittances) By applying the technique of PSM, we have created the control group (2,603 households not receiving international remittances) and the treatment group (135 households receiving international remittances) (Appendix 5.1) from VHLSSs

2006 and 2008 Then, we use the DD method to evaluate impacts of foreign remittances by running a regression on the panel data, which is created from data of control and treatment groups in two years 2006, 2008 We will present detailed the methodology in the chapter 3

There are a lot of documents and empirical studies, which state the applying PSM and DD methods on impact evaluation as follows Such as, the impacts of rural road rehabilitation on market development in rural Vietnam (Mu and Walle, 2007) The effects of improving rural roads on poverty reduction in Bangladesh (Khandker et al., 2009), or impacts of increasing minimum wage on average employment in New Jersey and Pennsylvania States of United States (Angrist and Pischke, 2009) The presentation the methods of impact evaluation (Khandker et al 2010); estimating the rate of return to schooling in Vietnam (Nguyen Xuan Thanh, 2006) or examining impact of tap water on welfare of households in Vietnam (Nguyen et al., 2011) and so on

However, we have not yet discovered any papers, which apply DD with PSM in impact evaluation of foreign remittances on household welfare in Vietnam This is the reason, which accompanies with the using the data taken from VHLSS 2006 and VHLSS 2008, as presented

in the section of Introduction, for re-examining the effects of international remittances on household welfares and expecting to contribute some new findings on the debate

2.4 Overview of theory on impact evaluation

2.4.1 The concept and purpose of impact evaluation

In order to determine the influence of any economic-social factors or a program (project) of the government, which is on standard of living of individuals, households and communities, the researchers often use the methods of impact evaluation The results of impact evaluation are essential in economic research, as well as for policy makers to make appropriate economic policies to improve living standards for households and communities

In term of purpose of impact evaluation, Paul J Gertler et al (2011) said that the impact assessment is needed to inform policy makers on a range of decisions, from reducing

inefficient programs, to expansion of intervention work, to adjust the program's benefits and choosing the alternative programs to meet the policy objectives more effectively

2.4.2 Classification of impact evaluation

There are two methods of impact assessment are qualitative and quantitative ones Qualitative method based on information such as socio-cultural context, details about the program and participants to determine the mechanism that produces effects Nevertheless, the qualitative approach cannot identify the relevant counterfactual outcome that is it not indicates what happens in the lack of the interventions So people often use quantitative methods based on evidence from survey data to valuate impacts However, in economic researching, a combination of qualitative and quantitative methods may be helpful in achieving a comprehensive view of the impact of economic-social factors or programs (the project)

According to Khandker et al (2010) there are two kinds of quantitative impact assessment is prospective and retrospective Prospective impact assessment is to quantify the expected impact of the programs and policies in the future based on the state of a given target area In contrast, the retrospective impact assessments measure actual impacts on the beneficiaries after the intervention of the program Most of the impact evaluation study, is quoted in this thesis belongs retrospective quantitative impact assessment

2.4.3 The basic problem of the impact evaluation

Khandker et al (2010) said that the basis problem of impact assessment is to resolve the lack

of data through building counterfactual situation or compare outcomes of treated individuals

or households with individuals, households in the comparison group without intervention It means that we must select a comparison group, which is very similar to the treatment group,

to ensure that the participants will have the same outcomes as non-participants in absence of intervention

The success of the impact evaluation depends on finding a good comparison group The researchers often simulate the counterfactual situation of the treatment group in two ways: (1) create a comparison group by statistical design or (2) adjusting the way of selecting of the program to eliminate the differences possible between the treatment and comparison groups

before comparing outcomes between the two groups Equation (2.1) presents a direct effect of

T on the program Y when comparing outcomes of individuals i with and without

intervention:

i i i

Where T is a dummy variable equal 1 with participants and 0 with non-participants; X is the observed characteristics of individuals, households and the local environment, and i is error term representing unobserved characteristics that affect Y

The problem of estimation equation (2.1) is that intervention assignment often does not made randomly by the selection of locations in program depend on the needs of communities and individuals The selection based on the observed or unobserved factors, or both In the case of unobserved characteristics, the error term contains the variables correlated with T variables,

which could lead to unobserved selection bias It means that, cov(T,)  0 indicates a violation of the assumption on the independent of regression to error term ε, which is one of key assumptions of OLS on unbiased estimations The correlation between T and ε typically will lead to errors in other estimates of the equation, including estimates of program’s efficiency β

In general, suppose we are evaluating the impact of the credit support programs to increase

household income Let Yi is the income per capita of household i Denote Ti = 1 for participants, and the value of Yi under intervention is replaced by Yi (1) Denote Ti = 0 for non-participants, and Yi is replaced by Yi (0) Therefore, the average effect of the program when comparison between Yi (1) and Yi (0) is:

D = E(Yi(1) | Ti = 1) – E(Yi(0) | Ti = 0) (2.2) However, the problem is that the situation of the treatment and comparison groups before intervention may be different, so the expected difference between these groups may not be entirely the result of the program intervention If adding counterfactual or expected results of the non-participated group if they participate in the program E (Yi (0) / Ti = 1) to (2.2), we get:

D = E (Yi (1) | Ti = 1) – E (Yi (0) | Ti = 0) + [E (Yi (0) | Ti = 1) – E (Yi (0) | Ti = 1)]

= [E (Yi (1) | Ti = 1) – E (Yi (0) | Ti = 1)] + [E (Yi (0) | Ti = 1) – E (Yi (0) | Ti = 0)] (2.3)

D = ATE + B (2.4)

In equation (2.4), ATE = [E (Yi (1) | Ti = 1) - E (Yi (0) | Ti = 1)] is the average treatment effect, that is the average increase in results in the participant corresponding to the non-participation, as in the case of non-receiving households were also got intervention ATE corresponds to a situation, in which households were randomly selected from the specified population participating in the program, so the treated and non-treated households have a similar probability of the intervention T

The term B = [E (Yi (0) | Ti = 1) - E (Yi (0) | Ti = 0)] is the selection bias when using D as an estimation of the ATE Because we do not know E (Yi (0) | Ti = 1) so we cannot calculate B Therefore, if one does not identify selection bias in D, then one would not know the exact difference in results between the treatment and control groups

In sum up, the basic objective of impact evaluation is seeking to eliminate selection bias B or find a way to calculate this factor It is suggested that selection bias will disappear if we assume that whether households or individuals receiving the intervention or not (conditional depend on a variety of variables, X) is also not dependent to get results This assumption is

called the assumption of un-confoundedness, or also known as conditional independence

assumption (Khandker et al., 2010 cited in Lechner 1999, Rosenbaum and Rubin 1983):

(Yi (1), Yi (0))  Ti | Xi (2.6) One also builds a weaker assumption of conditional exogenous factors in the selection of

program placement Thus, the accuracy of the impact estimations depend on the level of reasonable assumptions when compared between participants and comparison group, as well

as exogenous factors of the selected object of the program between the intervention and no intervention areas Obviously, if there are no methods or assumptions, then one cannot evaluate the extent of bias B These different assumptions associated with different impact evaluation methods, which are outlined below

2.4.4 Methods for retrospective impact evaluation

Khandker et al (2010) has presented a variety of methods retrospective impact assessment and accompanying assumptions about how to address selection bias when calculating the effect of interventions:

1 Randomized evaluation method

2 Propensity score matching (PSM) methods

3 Difference-indifferences or double difference methods (DD)

4 Instrument variable (IV) methods

5 Regression discontinuity (RD) design and pipeline methods

6 The distributional impacts

7 Structural and other modeling approaches

Randomized evaluation method seeks to identify the effectiveness of program by identifying

a group of subjects that have same observed characteristics and assigned randomly interference Meantime, a group without intervention is comparison group that will simulate the counterfactual outcomes This method has the advantage of avoiding selection bias in the random level

However, randomized evaluation is not always feasible Then researchers will apply the experimental methods, such as propensity score matching method (PSM) PSM methods compare intervention effects between the treatment and control groups based on a variety of observed characteristics The method assumes that selection bias only based on observed characteristics and not accounts the unobserved heterogeneity, which affects the participation

non-on program

Double difference method (DD) is used in both the experimental and non-experimental evaluations This method examines the effect of intervention by taking the difference in outcomes between the treatment and control units before and after intervention The DD methods assume that unobserved heterogeneity does not change over time The thesis will discusses further PSM and DD methods in section 2.5

Instrument variable (IV) method allows the selection bias on unobserved characteristics vary over time One can corrects the bias by finding a variable (or instrument) associated with participation status but no relationship with the unobserved characteristics that affect the outcomes The instrument is used to forecast the participation The IV method eliminates assumptions of the time-invariant of unobserved heterogeneity

Regression discontinuity (RD) design and pipeline methods are the extended forms of the IV methods, which exploit the exogenous rule of programs (such as suitable conditions for program) for comparison between participants and non-participants, which have closed location around the criteria for participation In particularly, the pipeline methods to build a comparison group who is suitable to join the program but not receiving intervention

Finally, distributional impacts and modeling approaches highlight mechanisms (such as the intermediary market forces), in which the program have an effect These methods can assess the welfare level of households or individuals, or geographic area level such as village, province and the whole country They can also help distinguish the smaller impact by gender, percentiles of income or economic, social, and demographical characteristics As a result, it can direct the levels of policy makers in the fields have the potential to reach the higher impact

However, according to Paul J Gertler et al (2011), all methods of impact assessment have

risks for bias, and “by combining methods, we can often off set the limitations of a single

method and thus increase the robustness of the estimated counterfactual” (page 119) The

combination of PSM with DD is an applied combination, as presented in Section 2.3 on the empirical studies In this topic, we use PSM technique to find the control group with similar characteristics to the treatment, then using DD regression to find the results

2.5 Summary of PSM and DD methods

2.5.1 Propensity Score Matching (PSM)

The PSM method builds a statistical comparison (or control) group relied on a model of the probability of joining in the treatment T depending on observed characteristics X, or propensity score P(X) = Pr(T=1|X) Rosenbaum and Rubin (1983) proved that matching on P(X) is almost like matching on X under some assumptions The needed assumptions are: (i) conditional independent and (ii) having a region of common support

The assumption of Conditional Independent is also known as un-confoundedness, which

shows that participation of the program completely based on observed characteristics So that,

if unobserved characteristics determining the participation status then the conditional independent will no longer true and PSM will not appropriate The combining PSM with DD can resolve the problem

Heckman, LaLonde and Smith (1999) stated that the second assumption (the common

support: 0 < P(Ti = 1|Xi) < 1) ensuring treatment observations have comparison observation

“nearby” in the distribution of propensity score It means that the control group who have the

X variables similar to the treatment group However, in calculating TOT (treatment effect on

the treated), the second assumption can be mitigated to P(Ti = 1|Xi) < 1

Khandker et al (2010) presented the technique of PSM through three steps in order to defining treatment and control groups, which have overlapping distributions of propensity score in the region of common support Then depending on the number of comparison units are matched with one of treatment unit, the authors offer some different matching estimators, such as: nearest-neighbor matching, caliper or radius matching, stratification or interval matching, Kernel and local linear matching and difference-in-difference matching Khandker

et al (2010) [cited in Heckman, Ichimura and Todd, 1997, 1998] also show that the bias in

PSM estimators can be low by three conditions First, using the same survey instrument or source of data for participants and non-participants; second, a representative sample survey is

eligible for participants and non-participants can improve the accuracy of the propensity

score, and the bigger sample of non-participants is the more accurate matching; third, both

participants and non-participants have same economic motives that leads accessing to the same markets

2.5.2 Difference-in-difference (DD)

According to Angrist and Pischke (2009), the physician John Snow (1855) was the pioneer exploring the DD idea when he studied cholera epidemics in London in 19th century Snow wanted to prove that cholera is spread by polluted drinking water (as opposed to “bad air”, the theory in vogue at that time) Snow compared differences in mortality rates from cholera

in districts supplied by two water companies, Southwark and Vauxhall Company and Lambeth Company In 1849, both companies supplied water from the dirty Thames River in

central London Then, in 1852, Lambeth Company moved their works upriver to an area comparatively free of wastes Snow discovered that, mortality rates in districts supplied by Lambeth fell strongly in comparison to the difference in mortality rates in districts supplied

by Southwark and Vauxhall

Nowadays, difference-in-differences is the popular estimator in impact evaluation The DD

approach bases on a comparison between participants and non-participants before and after the interference Such as, we can make the surveys about the impact of international remittances on income and expenditures of recipients (treatment) and non-recipients (comparison) before and after receiving remittances From this information, we can calculate the differences between the observed outcomes for the treatment and comparison groups before and after the intervention

The DD models can be generally presented as follows Dividing the analyzing objects into two groups: one group received the intervention (treatment group) and the other group received no intervention (control group) Denote D as a dummy variable indicating receipt the intervention, i.e D equals 1 if a unit belong to treatment group, and D equals 0 if a unit belong to control group And let T as a dummy variable indicating the present of the intervention, i.e T equals 1 denotes treatment or after intervention, whereas T equals 0 denotes untreated areas or before intervention Farther let Yit as observed outcomes of two groups before and after influence

For the treatment group, denote outcomes before and after receiving the intervention are Y10

(D=1, T=0) and Y11 (D=1, T=1) So, the change on outcomes of the treatment group between two periods is: (Y11 – Y10 ) For the control group, denote outcomes before and after receiving the intervention are Y00 (D=0, T=0) and Y01 (D=0, T=1) The change on outcomes of the control group between two periods is: (Y01 – Y00 ) Therefore, the impact of intervention on outcomes of treatment and control groups between two periods by DD approach is:

DD = (Y11 – Y10 ) – (Y01 – Y00 ) (2.1)

It is illustrated in figure 2.1 Under the DD approach, the needed important assumption is both the treatment and control groups, which must have relatively similar characteristics at the time before receiving the intervention for ensuring that the outcomes of two groups will

have the same trend of changes over the period From the figure 2.1, this assumption is written as follows:

(Y01 – Y1’0 ) = (Y00 – Y10 ) (2.2)

Figure 2.1 An Example of DD

Income

T=0 T=1 Time

Source : Khandker et al (2010)

Where Y1’0 is the outcomes of the treatment group (D=1) at T=1 without receiving the intervention, or called the counterfactual outcomes Applying the equality (2.2) in the DD equation (2.1), we get:

This is the impact of the intervention on the treatment and comparison groups between two periods, which is shown in Figure 2.1 The preceding assumption also requires unobserved characteristics, which affected the two groups, to be time-invariant It means that the unobserved heterogeneity is time invariant, so the bias will cancel by differencing If these unobserved variables vary over the period, the DD approach would lead underestimates or overestimates of intervention

In the other side, according to Khandker et al (2010), and Angrist and Pischke (2009), it can calculate the DD estimate in a regression framework Namely, the regression equation would

be defined as follows:

it t it t it

Where Yit is the outcome of individual i at time t If using the notation from equality (2.1), the coefficientof interaction between D and T, which will equals DD Precisely, one can prove it by rewriting the regression equation (2.3) in expectations form (temporally rejecting the subscript i) as following:

Y1’0(D=1) Y00(D=0) Y10(D=1)

Y11(D=1) Y01(D=0)

Participants

Control Impact

Intervention

E (Y11 – Y10) = (α + β + ρ + γ) – (α + ρ) (2.4a)

E (Y01 – Y00) = (α + γ) – α (2.4b) Subtracting (2.4b) from (2.4a) is the quite equation (2.1) in expectation form:

DD = E (Y11 – Y10) – E (Y01 – Y00) = β (2.5)

So β is the average effect of the intervention on outcomes of treatment and control groups before and after intervention Khandker et al (2010) asserted that, the DD estimator is only correct with some following conditions:

(i) The outcome equation should be constructed correctly (such as, the additive

structure must be correct) (ii) The error term is not correlated with other variables:

Cov(ε it, Dit) = 0

Cov(ε it, Tt) = 0

Cov(ε it, DitTt) = 0

The last assumption is most important, also called as the parallel-trend assumption It said

that unobserved characteristics, which influences program participant, do not vary over time with treatment status

CHAPTER 3: METHODOLOGY AND DATA

3.1 The research model

In order to answering the research questions about the impact of international remittances on household welfare in Vietnam and based on empirical literatures and framework mentioned

above, the paper suggests the combined research model of propensity score matching (PSM) and difference-in-difference (DD) methods as follows The first, we apply the technique of

PSM method to construct one group of households receiving foreign remittances (called the treatment group) and one another group of households non-receiving foreign remittances (called the control group), in which a number of comparison units are matched with one of treatment unit Then calculating the impact of remittances (or DD) based on the regression equation:

Y it = ß 0 + ß 1 D it + ß 2 G t + ß 3 D it *G t + ß 4 X it + ε it (3.1)

Where Y it are income and expenditures per capita (outcomes) of household i at time t belong

to treatment or control groups; Dit are the dummy variables represent receipt of international remittances; Gt are the time dummy, which equals to 1 for the year 2008 and 0 for the year 2006; Dit*Gt is the dummy variable for international remittances and times Xit are the explanatory variables, which indicate the characteristics of household head, households,

geographical and populated areas; ε it are the error terms

Implementation the ordinary least square (OLS) regression allows to getting the differential effect on outcomes of the households with international remittances between two years is:

DD = { E(Yit│Dit = 1, t=2008) – E(Yit│Dit =1, t=2006 )}

– { E(Yit│Dit = 0, t=2008) – E(Yit│Dit =0, t=2006)} (3.4)

In the equality (3.4), the effect of international remittances (DD) is measured by ß 3 , which is the coefficient of interaction between Dit and Gt The DD estimator is only correct when accompanied with the most assumption, which is unobserved characteristics affecting to two

groups to be time-invariant with treatment status (presented in chapter 2)

we use four of indicators to analyze that are income per capita, expenditures per capita,

education expenditures per capita and health-care expenditures per capita, denoted Y it in the equation (3.1)

The first indicator (income per capita) is always used to measure the standard of living of families Nevertheless, the data of income in surveys, which is usually lower than in reality and instability over year by year, so it cannot show a good picture about household welfare Therefore, the second indicator (expenditures per capita) should be applied in analyzing, because the data of expenditures is often closer to reality than income and relatively stability over years Two of first indicators with two of last ones will help to get a clearer painting in analyzing the effects of international remittances on household welfare

Since the average values of these indicators or dependent variables in equation (3.1), which have the deviation distribution (Appendix 3), so that the transformation them to the form of logarithm is needed to lead them follow normal distribution Thus, the equation (3.1) is rewritten to the semi-log function as follows:

Ln (Y it ) = ß 0 + ß 1 D it + ß 2 G t + ß 3 D it *G t + ß 4 X it + ε it (3.5)

The dependent variables are: lognepe of income per capita (ln_income), lognepe of total expenditures per capita (ln_texpen), lognepe of education expenditures per capita (ln_education) and lognepe of health-care expenditures per capita (ln_healthcare)

3.2.2 Independent variables

The independent variables are the same in equations (3.5) and (3.1) The dummy variables

consist of International remittances Dit (inremit), Time effect Gt (year), Interaction of

foreign remittances and time Dit *G t (inremit*year), and Local remittances (loremit) The

variable (D it *G t) is the most important variable, because the coefficient ß 3 of them specifies the impact of international remittances on household welfare

Based on the empirical studies in the world and Vietnam, it is shown that there are many factors (or variables) affecting to household welfare Generally, these variables can be divided in three categories: group of household head, group of household characteristics and group of geographical and populated areas In the equation (3.5), three of groups are denoted

Xit, which is also presented in detail in the table 3.1

Table 3.1: Summary of variables

Variable Description Measurement Signification in the

research model

ln_income total income per capita The values of total income, To specify the impact of

of households total expenditures, international remittances

Napierian logarithm of total expenditures per capita of households

education expenditures and other factors on total

expenditures

income, total expenditures (measured from data of a , total education

Napierian logarithm of household in 12 months expenditures and total ln_education education expenditures with unit of currency is healthcare expenditures per capita of households 1,000 VND), which is per capita of households Napierian logarithm of divided by household size (or outcomes)

ln_healthcare healthcare expenditures

per capita of households

Independent variable

Dummy variable indicates Evaluation the difference whether a household Equals 1 with received in effect of foreign inremit received international households; equals 0 with remittances on outcomes remittances or not not received households between receiving and

non-receiving households

year Time dummy variable Equals 1 for the year 2008; of dependent variables

equals 0 for the year 2006 (or outcomes) under

Table 3.1: (continued)

Variable Description Measurement Signification in the

research model Household head characteristics

headage The age of household Number of calendar years Effect of Age of house-

Binary variable shows Equals 1 for male and Influence of the gender of gender the sex of household head equals 0 for female household head on

Household characteristics

hhsize The size of households Number of household Effect of size of

The square of household The square of number of Effect of the square of hhsize 2

size household members size of household on

Ratio of members younger

child16 than 16 years old, divided Number of calendar years Influence of the

by household size dependent members

Ratio of non-working Number of calendar years in household on outcomes elder age's members, divided by (over 60 for male, over 55

household size for female)

Ratio of members with Impact of the technical technic technical degree, divided The highest technical degree (or labor

by household size degree obtained productivity) on outcomes

secondary secondary education,

divided by household size

Ratio of members with The highest diploma Effect of education on highschool highschool education, obtained dependent variables

divided by household size

college college education,

divided by household size

Table 3.1: (continued)

Variable Description Measurement Signification in the

research model

pcannuland land per capita of

pcpereland crop land per capita of The unit is square Influence of agricultural households meter (m 2 ) lands on outcomes

pcforexland per capita of households

pcwatersuface water surface per capita

Binary variable indicates Equals 1 for Red River

reg_1 population living in Red Delta area; equals 0 for

River Delta or other areas other areas in Vietnam

Binary variable indicates Equals 1 for North East

reg_2 population living in North area; equals 0 for other

East or other areas areas in Vietnam

Binary variable indicates Equals 1 for North West

reg_3 population living in North area; equals 0 for other

West or other areas areas in Vietnam Effect of geographical Binary variable indicates Equals 1 for North Central areas on dependent reg_4 population living in North Coast area; equals 0 for variables

Central Coast or other other areas in Vietnam

Binary variable indicates Equals 1 for South Central

reg_5 population living in South Coast area; equals 0 for

Central Coast or other other areas in Vietnam

Table 3.1: (continued)

Variable Description Measurement Signification in the

research model Geographical and populated areas (continued)

Binary variable indicates Equals 1 for Central

reg_6 population living in Central Highlands area; equals 0 for

Highlands or other areas other areas in Vietnam

Binary variable indicates Equals 1 for South East

reg_7 population living in South area; equals 0 for other Effect of geographical East or other areas areas in Vietnam areas on dependent Binary variable indicates Equals 1 for Mekong River variables

reg_8 population living in Delta area; equals 0 for

Mekong River Delta or other areas in Vietnam

2008 was collected by two times in the year of 2008: first time on May and June, second time

on September and October 2008

The thesis use information of income and expenditure surveys that consist of 9,189 households in VHLSS 2006 and 2008 The surveys include the detailed information on total income, total expenditures, education and healthcare expenditures, foods and non-foods expenditures, characteristics of household and householder, characteristics of geographical and populated areas and so on, which reflect the living standards of households in the different areas and all the country

As was presented above, the needed sample size of VHLSS in 2006 and 2008 is the same of 9,189 observations After setting up the panel data of 2006 and 2008 with the households receiving and non-receiving international remittances, one gets 2,754 non-receipt households and 135 receipt households, who were presented in both VHLSS 2006 and 2008 Applying the Propensity Score Matching Method, we have constructed the panel data of 2,738 observations that are representative for geographical and populated areas in the whole country in 2006 and 2008 (Appendix 5.1) The observations include two groups: control (or comparison) group with 2,603 observations and treatment group with 135 observations

3.3.2 Data of variables

The source of data for the variables in the regression equation (3.5) accompanied with the panel data, which is collected from VHLSS 2006 and 2008 In particularly, the information were colleted from section 1 “List of household members, data on main demography”, section 2 “ Education level and expenses”, section 3 “Health and related issues”, section 4

“Employment and Income”, section 5 “Expenditures”, section 6 “Fixed assets and durable things”, section 7 “Housing” and Section “Balance of income and expenditures”

(1) Dependent variables

In VHLSS 2006 and 2008, household income includes income from agricultural and agricultural production, salary, pension, income from house rental and loan interest, remittances and subsidies This data is collected from sections 2, 3, 4 and section 7 (or section “Balance of income and expenditures”) that are the income in a year of families Then

non-it can calculate the income per capnon-ita of households

The household expenditures include all of expenses on foods, education, healthcare, house, power, water supply and so on in 12 months The information can be obtained from sections

2, 3, 4, 5, 6 and section 7 (or section “Balance of income and expenditures”) After that, we

can calculate the total expenditures per capita of households

The expenses of education consist of school fees, uniforms, equipments, extra costs on learning of computers or foreign languages, etc Moreover, the expenses of health include medical treatments, medicines, health equipments, and medical insurance and so on

Similarly, it can be got the educational expenditures per capita from section 2 and the healthcare expenditures per capita from section 3

In order to eliminating the inflation effect, the data of income and expenditures in 2006 are adjusted to the price in 2008

The data of international and local remittances is the dummy variables, which can be taken from section 4D1 “Other Income” - lines 101 and 102 respectively International remittance

(inremit) variable equals 1 with households receiving foreign remittances and equals 0 with households non-receiving Local remittance (loremit) variable is code 1 with households

receiving local remittances and vise verse coded 0

The variable of time effect dummy (year) equals 1 for households examined in 2008 and equals 0 for households in 2006 The interaction term (inremit*year) is the most important variable in the thesis Since under the DD approach, the interaction between inremit and year variables specifies the impact of international remittances on household welfare This

interact variable equals 1 if households received foreign remittance and examined in 2008, and equals 0 in remaining cases

The information of householder’s characteristics is obtained from section 1 in VHLSS 2006,

2008 The age of household head (headage) is calculated on number of calendar years The binary variable of householder’s sex (gender) equals 1 with male and equals 0 with female

The data of household characteristics is collected from section 1, 2 and 4 in VHLSS 2006 and

2008 The size of household (hhsize) is got from section 1, which is measured by the number

of members in household The variable of squared household size (hhsize 2) is used in the

paper, because the variation of hhsize variable on income and expenditures over time which

is not only a linear function The information of dependent members is also obtained from

section 1, detailed as follows The ratio of younger less than 16 years old (child16) is

calculated on the number of members below 16 years olds, divided by household size The

ratio of non-working’s age (elder) is calculated on the number of members higher than 60

years olds for male and 55 for female, which divided by household size

The technical degree and educational diploma of members in households are collected from section 1 and 2, which is the highest degree or diploma obtained The ratio of members with

technical degree (technic) equals the number of members with technical degree divided by

household size The variables of educational diploma include three types: ratio of members

with diploma of secondary education (secondary), high-school education (highschool) and college education (college) equals the number of members with secondary, high-school and

college diploma respectively divided by household size

The kinds of land using in agricultural production of households which can be drawn from section 4 A member of one household can receive fixed amount of land, so the area of received land of a household equals total areas of land from all of members received in a year The unit is the squared meter (m2) There are four types of land variables: per capita of

annual crop land (pcannuland), perennial crop land (pcpereland), forestry land (pcforexland) and aquaculture water surface (pcwatersuface) equals the total of land area of

household divided by household size

The last group of independent variables is the one of geographical and populated areas, whose information is collected from files hhexpe06.dta in VHLSS 2006 and hhexpe08.dta in VHLSS 2008 The geographical variables are the binary variables, which specify eight of

areas in whole country as follows Reg_1 equals 1 with households living in the Delta of Red River, equals 0 with households living in other areas Similarly, reg_2, reg_3, reg_4, reg_5, reg_6, reg_7 and reg_8 are the binary variables for households living in the areas of North

East, North West, North Central Coast, South Central Coast, Central Highlands, South East and Mekong River Delta respectively The variable of populated areas is the dummy variable, which equals 1 if households living in urban, and equals 0 in case of households in rural

(3) From the restricted models, estimation DD and the effect of other factors on household welfare Interpretation of the estimation results

CHAPTER 4: INTERNATIONAL REMITTANCES IN VIETNAM

4.1 General view of migration remittances in Vietnam

According to a newest World Bank Report (2012) on international migration and remittances, the developing countries are expected receiving $351 billion in the year 2011 Vietnam is estimated in the top ten recipients of official remittances, including India ($58 billion), China ($57 billion), Mexico ($24 billion), Philippines ($23 billion), Pakistan, Bangladesh, Nigeria, Vietnam, Egypt and Lebanon in 2011 The empirical studies also proved that foreign remittances is the important resource of external finance, and so the migrant national play a role as external resources of capital, human and know-how which flow to developing countries

It should pay attention to the special case of Vietnam in term of remittance and the motive of migration Pfau and Long (2008) noted that most studies of foreign remittances based on the assumption that international migration is motivated by economic factors, meanwhile in the case of Vietnam, it has driven by economic and non-economic factors Barbieri et al (1996) specified that there are 1.2 million people leaving Vietnam between 1975 and 1993, in which

60 percent were illegal refugees and 40 percent were part of the Orderly Departure Programme (ODP) set up by Vietnam’s government Meantime, the research of Andrew T Pham (2010) identified the Vietnamese refugees can be divided three waves of migration The first wave appeared in the last days of April 1975 when the Republic of Vietnam (RVN)

in the South collapsed The second wave of refugees associated the matter known as the “boat people” and the ODP program from after 1977 to the early of 1980s So the non-economic factors (such as political or religious) are the mainly motivations of the refugees in these waves The final wave of migration occurred in the late of 1980s accompanied with the economic reasons Consequently, Andrew T Pham (2010) [cited in the UNHCR, 1996] shows that there were 1.3 million Vietnam refugees resettled in Western countries from 1975-

1995

There are also other flows of migration, which are the labor export programs from Vietnam to other communist countries, such as form Union of Soviet Socialist Republics (USSR) and European Eastern Countries, and Asian countries from the early 1980s However, the quantity of export worker to the former Allies decreased strongly by the collapse of socialism

system in the early 1990s From that period, the exporting of Vietnamese workers changes to explore new markets, such as Taiwan, Japan, South Korea, Malaysia, in which the number of export labors rapidly increases and make up big ratio of total export workers Since the late 1990s, Vietnamese workers begin to go to Middle East and African countries with small number

According to Vietnamese Oversea Committee (2009), there are about 4 millions Vietnamese living, working and studying permanently in 102 nations and territories in the world Table 4.1 shows the distribution of Vietnamese overseas in 2005, and top destination countries are United States, Russia and Eastern Europe, France, Australia, Canada

Table 4.1: Vietnamese Overseas in 2005

Source: Andrew T Pham (2010) [cited in Sidel, 2007]

The table 4.2 presents the interesting information on quantities of Vietnam export labor in period of 2001-2008 Vietnam has 36,168 export workers in 2001, but increased to 94,988 workers in 2008 (or rose more two and haft times) By the late 2008, under data compiled by the Department of Overseas Labor Management, total export labors of Vietnam in all markets around the world is more than 550,000 persons Four main markets are Japan, South Korea, Taiwan and Malaysia, which accounted for more 80% of total Vietnamese export workers from 2001 to 2008

Table 4.2: Number of Vietnamese Labor overseas from 2001-2008, classified by countries

In general, Vietnamese export labor are mostly of manual labor and skill is not high, who concentrate in industries: construction, textile, maritime transport, crewmember of fishing vessel, family maid, etc Meanwhile, industries need skilled and qualified labor, such as finance, banking, and information technology and so on, the amount of Vietnamese workers

is still very little According to report of Department of Overseas Labor Management, the ratio of trained workers for exporting to other countries reached only 15%

4.2 Role of international remittances on economy

In the early 1980s, the recipients have to sell foreign currencies and get Vietnam dong, so they preferred to receive goods (medicines, clothes, electric devices, etc) and sell in the free market The estimated value of remittances in the 1980s was about $100-200 million per year

Along with economic reforms process, the central government has promulgated some policies

to attract remittances from overseas Vietnamese and encourage Viet Kieu (1) going back

Vietnam for investment Relying on that, remittances in 2001 is only US$2 billion, but increasing trend is steadiness in subsequent years, reaching US$8 billion in 2010 and estimated values is more than US$ 8.6 billion in 2011 Figure 4.1 illustrates the rapidly increasing of overseas remittances from 2001 to 2008, decreasing a little in 2009 because of economic recession, and bouncing back in 2010, 2011 Therefore, foreign remittances have more and more remarkable contribute to economic growth and improve the living standards

of received households in Vietnam

(1) This term is officially referred by the GVN as “Persons of Vietnamese origin residing abroad” and defined as “Vietnamese people who used to have Vietnamese nationality which has been determined at the time of their birth on the consanguinity principle and their offspring’s

and grandchildren are permanently residing foreign countries.” (2008 Law of Vietnamese Nationality, 1:3) (Andrew T Pham, 2010)

Figure 4.1: International Remittances from Vietnamese overseas

million USD

2,000 2,714 2,700

3,200 4,000 4,800 5,500

7,200 6,840 8,000 8,600

Source: World Bank (2012) (more detailed data in Appendix 4.1.)

Table 4.3 presents international remittances and GDP of Vietnam in the first decade of 21stcentury According to GSO, the average growth rate of GDP in this period is 7.25 percent, so Vietnam is the one of countries having the highest growth rate in the world Meanwhile, remittance made up approximately 7.34 percent of GDP in 2001-2010, and relatively stability over years In terms of GDP, this represents a significant contribution of overseas remittance for the total GDP of the economy

Table 4.3: Ratio of International Remittances (IR) on Gross Domestic Product (GDP)

Figure 4.2 shows a general picture about international remittances, FDI and ODA, which are important financial resources for economic growth, in decade of 2000s In the period 2001 –

2008 remittance increases rapidly from US$2.0 billion to US$7.2 billion, meantime ODA rises only lightly from US$1.4 billion (2001) to US$2.6 billion (2008) Especially to FDI, in 2001-2006 it has the same increased trend (like ODA) from US$1.3 billion to US$2.4 billion, but in the phase 2007-2008 FDI has bounced strongly to US$6.7 billion (2007) and US$9.6 billion (2008) because of Vietnam have joined in World Trade Organization (WTO) in 2006

In 2009, both overseas remittances and FDI decrease and increase back in 2010 Conversely, ODA rises remarkable in 2009 and decreases lightly in 2010

Generally, in the ten years (2001-2010) average amount of ODA is around US$2.2 billion/year, meanwhile FDI is US$4.2 billion/year and International Remittance is more than double, US$4.7 billion/year

Figure 4.2: Comparison between foreign remittance, FDI and ODA of Vietnam in 2001-2010

million USD

0 1,000

Source: World Bank (2012) and MPI (more detailed data in Appendix 4.1.)

Figure 4.3 illustrated the comparison between foreign remittances and trade balance in ten years 2001-2010 Excluding 2001, the trade balance of Vietnam are always negative and increasingly from 2002 to 2008 and decreasingly in 2009, 2010 by economic recession Therefore, overseas remittances have helped considerably on improving the balance of payment