1
Gender Bias and its Effects on Remittances in Armenia
ByMasha Bontje1 10003995
MSc Thesis Development Economics Supervisor: Erik Plug
July 2017
Abstract
This study focuses on Armenian households and the influences of having a daughter as the oldest-child on the likelihood of receiving remittances, the monetary value of the received remittance and the remittance allocation decisions made by the household. The household survey: Integrated Survey of the Migration of the Republic of Armenia Population 2007-2013 is used in this research. There is a significant effect, with the inclusion of extensive control variables, of oldest-daughters on the likelihood of receiving remittances. Oldest-daughters decrease the chance to receive remittances by -0.03. A negative, non-significant relationship is found between having daughter as the oldest-child and the logarithm of the monetary value of the received remittances. Both groups of households most often dedicate remittances to the allocation category: “consumption”. Oldest-daughter households allocated remittances significantly more often to the spending categories “education” and “migration” than oldest-son households.
1
Masha Bontje, MSc Student Development Economics, University of Amsterdam. I would like to thank my supervisor Erik Plug (University of Amsterdam) for his supervision and feedback. I am grateful to Moritz Meyer (Economist, World Bank Group) for his inspirational thoughts and his specific knowledge about Armenia.
2
Content
1. Introduction 3
2. Literature 3
2.1 Reasons for migrants to remit 4
2.2 Remittances allocation decisions by households 4
2.3 Influence of the gender-composition of households on resource allocation decisions 5
2.4 Remittances and gender 5
2.5 Armenia and migration 6
3. Background on Armenia 7
3.1 Economy 7
3.2 Migration 8
3.3 Sex ratio at birth 8
4. Data: Migration Survey 2007-2013 9
4.1 Introduction and main findings 9
4.2 Collection of the data and data sampling 10
4.3 Problems with the data 10
5. Descriptive statistics 11
5.1 Explanation 11
5.2 Descriptive statistics: households with children 12
5.3 Descriptive statistics: households with children receiving remittances 14
6. Methodology 15
6.1 Hypothesis 16
6.2 Empirical methodology 17
7. Results 18
7.1 Likelihood of receiving remittances 18
7.2 Monetary value of received remittances 19
7.3 Allocation of remittances by spending category 19
7.4 Critical note 21
8. Conclusion and discussion 22
Appendix 1. Explanation of the variables 23
Appendix 2. Results: reduced data sample 25
3 1. Introduction
Remittances sent by migrants generate an enormous transfer of money within and between countries. Trends in remittance flows are particularly noticeable in The Republic of Armenia
(“Armenia”), which has a long history of migration. Remittance payments as a percentage of GDP are a significant part of Armenia’s economy, as evidenced by the World Bank. Personal remittances comprised 17.9% of the country’s GDP in 2007. This percentage fluctuated a lot the last decade, decreased to 16.6% in 2009, rose to 19.7% in 2013 and decreased again to 14.2% in 2015 (World Bank, 2016). These remittances are for many households a significant (additional) source of income. This phenomenon raises several interesting questions: Which households receive these remittances? What is the monetary value of these remittances? And how do households allocate these additional resources?
Armenia is not only a country that experiences high migration movements, but also the needs to deal with the phenomenon of son-preferences: around 115 boys are born per 100 girls. This difference is related to sex selection, future parents prefer having boys over girls, and therefore some expecting parents decide to undergo an abortion if they expect to conceive a daughter. The
implications of this phenomenon raises a key question: does this preferences to have sons continue to effect decisions made by households after family’s children are born? Differences in expectations for girls and boys could encourage households receiving remittances to focus their attention on one sex over the other. Scholars conclude that remittances, besides pure monetary gains, are also associated with greater human development outcomes such as gender equality (Ratha, 2013). This research will link the preference of Armenian families to have sons (“son-preferences”) with the likelihood of receiving remittances and the allocation decisions made in regard to remittances. To research these questions there is a focus on the oldest-child of households, because the oldest-child tends to influence the decisions that parents make. The following question is analyzed in this thesis:
Does the gender of the oldest-child in a household influence the likelihood of receiving remittances, the monetary value of the remittances and decisions that are made in regard to remittance allocation?
This research will add to the extensive literature on remittances because of the children’s gender component. This approach is particularly relevant and interesting in the case of Armenia because of the phenomenon of son-preference in this country.
The main finding of this thesis is that having an oldest-daughter significantly decreases the chance of receiving remittances with -0.03, by including extensive control variables. A negative, non-significant relationship, is found with the inclusion of control variables, for oldest-daughter
households on the logarithm of the monetary value of the received remittances. Moreover, oldest-daughter households allocated remittances significantly more often to the spending categories “education” and “migration” than oldest-son households. The spending category migration refers to dedicating money to the cause of going abroad.
Section 2 provides a literature overview about (Armenian) migration, remittances and resources allocation decisions. Section 3 gives a short introduction into the economy, migration and the sex ratio by birth in Armenia. Section 4 describes the data and the main findings of the Integrated
Survey of the Migration of the Republic of Armenia Population 2007-2013. Furthermore, the problems
encountered with the data are discussed. Section 5 presents the descriptive statistics of households with children and households with children that receive remittances. Section 6 introduces the
methodology and outlines several hypotheses which answer the research question. Section 7 presents the results and answers the research question. Section 8 concludes and provides some thoughts for discussion.
2. Literature
Firstly, this section focuses on research related to the reasons why migrants send remittances to their household of origin. Secondly, several studies that focus on remittance allocation decisions by households are discussed. Thirdly, allocations of resources by households in relation to the gender composition of a household are highlighted. Fourthly, research about remittances in relation to gender
4
(in)equality is discussed. To conclude, literature that focuses on migration and remittances in Armenia is summarized.
2.1 Reasons for migrants to remit
This research does not try to understand the motivations of the migrants themselves, but focuses on households receiving remittances. A short overview of the motivations of migrants sending remittances is also given to get a broader understanding of this concept.
A number of empirical studies analyze the motivations of migrants to send remittances to their household of origin. According to a comprehensive review by Rapoport and Docquier (2006), two main motivations to remit can be identified:
1) Altruistic relationship: the desire of the migrant to help family members at home;
2) Contractual relationship: migrants remit to insure their family against adverse risks, to invest in the household or to receive a potential family bequest.
Altruistic theories of remittances argue that members of households act to improve the welfare of every member of the family (Agarwal and Horowitz, 2002). This altruistic behavior can be
influenced by coercive social milieu and ingrained social norms (Vanwey, 2004). The degree of altruism, varies across families, and could be stronger in cohesive, traditional families and weaker in nontraditional families with unstable bonds (Sana and Douglas, 2005). Altruistic behavior could play an important role in Armenia because of the strong family ties that exist in Armenian culture.
Migrants can also decide to remit to the household of origin to repay an earlier (migration related) loan or to pay back for their education. Moreover, the origin household and the migrant can try to insure each other against potential shortfalls in income. In the absence of formal unemployment insurance or health insurance, household members provide insurance for each other (Poirine, 1999). Gigorian and Melkonyan (2011) find that in Armenia it is less likely for large households to receive remittances. Explanations for this finding could be that large Armenian households are perceived (by the migrants or the household itself) as less vulnerable and therefore depend less on the “insurance” of remittances. Another explanation is that the bequest-related motives are weakened as the migrant faces competition from other members of the household for an inheritance.
Literature shows that it is difficult for empirical studies to discriminate between these various reasons for remitting. Most empirical studies therefore find that remittances are motivated by some combination of altruistic and contractual motivations (Adams, 2011).
Vanwey (2004) examines the patterns of remittances between rural Thai households and female and male migrants in both rural and urban destinations and analyzes how these factors are related to altruistic and contractual remittance patterns. The author finds support for both approaches: women and migrants from poorer households behave more altruistically, while men and migrants from richer households behave more contractually. Furthermore, Vanwey concludes that the results also show that there should be a more complex view of remittance behavior: It is for instance important to also consider the role of gender and social class of the migrant.
2.2 Remittances allocation decisions by households
The literature does not uniformly agree on the question how households allocate received remittances. It is an interesting question to analyzehow and in which manner remittances receiving households benefit from remittances. Scholars focus in their research on the impact of remittances on different allocation categories such as: household consumption (Chami, Fullenkamp and Jahjah, 2003; Adams, 2010), investment (Amuedo-Dorantes and Pozo, 2006; Woodruff and Zenteno, 2007), health (Kroeger and Anderson, 2014) and education (Yang, 2006; Grigorian and Melkonyan, 2011; Kroeger and Anderson, 2014). Differences in expectations for daughters and sons may encourage household receiving remittances to allocate more remittances to one sex over the other (Kroeger and Anderson, 2014). This section gives a brief overview of the literature related to the allocation of remittances on spending categories by households.
Chami et al. (2003) focus in their research on consumption goods, and state that a “significant proportion, and often the majority of remittances is spent on status-oriented consumption goods.” The
5
authors also find that the typical goods in which remittances are invested (such as housing, land and jewelry) are not productive to the economy as a whole.
Adams and Cuecuecha (2010) analyze nationally-representative household data from
Guatemala to examine the spending patterns of remittances. The authors use the Duban and McFadden method and an instrumental variable approach focusing on historic distance to railroads and rainfall shocks. Adams and Cuecuecha conclude that households receiving international remittances spend less at the margin on one key consumption goods but more at the margin on two investment goodsm education and housing, compared to households that do not receive remittances.
Studies show a mixed outcome in regard to what extent remittances are used to stimulate investments and entrepreneurial activities. Amuedo-Dorantes and Pozo (2006) analyze the impact of remittances on business formation by using a small household survey from the Dominican Republic. The authors find that households receiving international remittances are not more likely to own a family business than households that do not receive remittances. An explanation could be that remittances increase the reservation wage of the head of the household which makes them less likely to invest in a business. The authors do not control for selection bias for the households that receive remittances.
To address this issue, Woodruff and Zenteno (2007) use an instrumental variables approach focusing on historic distance to railroads to analyze migration and business formation in Mexico. They find that international remittances help migrant households in Mexico gain the capital that is needed to expand their small enterprises.
Yang (2008) uses a ‘natural’ experiment, exchange rate shocks, to analyze panel household data from the Philippines to examine how remittances affect household expenditures on investment and education. Results suggest that a 10% increase in the exchange rate leads to a 5.5% increase in a migrant household’s expenditure on education.
2.3 Influence of the gender-composition of households on resource allocation decisions
Studies have addressed the influence of the gender of children on resource decision making by parents. Whiles scholars have researched this topic, existing literature does not specifically focus on (additional) resources retrieved from remittances.
Spending patterns related to a child’s gender in households changed substantially in the period between the 1970s to the late 2000s in the United States. Households with only daughters spent significantly less than households with only sons in the early 1970s; but by the 1990s spending equalized; and by the late 2000s households which consists of daughters spent more (Kornrich and Furstenberg 2013).
Lundberg and Rose (2004) find that housing expenditures in the United States are
substantially and significantly higher for families with a son, with housing as the primary public good purchased by American families. The authors conclude that their results are in line with the theory that son-households try to increase marital surplus and thus invest in the family. Lundberg and Rose did not find any other significant effects of the gender composition of children on purchases of other durable items.
Wei and Zhang (2011) also did research on investments made by parents and explain China’s high household savings rate with a competitive saving motive by parents of sons. The authors state that the competitive savings in China could be related to the rising sex ratio imbalance. Due to intensified competition in the marriage market, son-households ratchet up their savings rate in hopes of improving their sons’ chance to find a wife.
2.4 Remittances and gender
Most research focused on remittances and gender, focus on the effect of the gender of the household head on the receiving of remittances (Guzmán, Morrison and Sjöblom, 2007; Antman, 2011).
Guzman et al. (2007) explore the relation between gender of household members and remittances in Ghana by using the Living Standard Survey 1998/1999. This research concludes that when the remitter is the husband, the share of expenditure on education increases when the wife is in
6
charge of managing the family resources. However, when the wife is the one sending remittances and the head of the household is a male, the allocated to education decreases.
Antman (2011) examines the allocation of resources to children in a household that (recently) experienced migration through a difference-in-difference strategy. Antman compared the behavior of Mexican households in which the head currently lives in the United States to households of which the migrant head just came back to Mexico. A pattern is observed of switching resources towards girls while the household head migrates to the United States, but shifting resources back to boys once the household head returns. In Mexico, men are far more likely to migrate, and migration of the head of household typically implies a father’s absence from the home, and thereby allows for an increase in women’s decision-making power.
The receipt of remittances can also effect education spending. Kroeger and Anderson (2013) analyze the effect of receipt of remittances on the education and health of children in Kyrgyzstan. Fixed effects and instrumental variables estimation are used to isolate the effect of remittances and other events. The authors find that boys aged 14-18 in households that receive remittances are less likely to be enrolled in school than other children.
Lopez-Ekra, Aghazarm, Kötter and Mollard (2011) summarize research focused on the
influence of migration and remittances on changing gender roles and responsibilities within the family. Lopez-Ekra et al. give a mixed overview of the influences of migration and the receiving of
remittances. The authors stress the importance that further research and policies on migration should take additional dynamics in account, including changing gender roles, work division between men and women and the influence on children of growing up without one of their parents.
Research is often limited by describing how women behave differently from men in remittance activities. However, research should also take into consideration how gender
representations and norms influence the behavior of women and men in remittance activities and how remittances influence gender relations (Kunz 2008). In countries which highly depend on remittances, it is essential to take gender components into account when analyzing the impact of remittances. This thesis pursues a holistic approach by taking gender issues and household composition into account by analyzing remittances.
2.5 Armenia and migration
This section discusses literature that specifically focuses on migration and receiving remittances in Armenia.
Armenia experienced a more moderate outflow of (male) migrants compared to other former Soviet Union countries after the collapse of the Soviet Union. The most common destination for migrants was and still is the Russian Federation (“Russia”) with the main purpose of employment (Dermendzhieva, 2011). Most Armenian migrants are in the age category 36-65. Households with younger children encounter less temporary migration movements which might be explained by strong traditional family structures (Danzer and Dietz, 2009). There is a significant correlation between migration and the presence of a family business, which could suggest that remittances are used to provide capital for a business (Dermendzhiva, 2011)
Makaryan and Galstyan (2013) state that 80% of the households that receive remittances spent between 80% and 100% of the received remittances on current consumption needs. Only 8% of the remittances are saved for a specific purpose such as future consumption, education, savings and investments. The OECD/CRRC-Armenia report (2017) comes to the conclusion that most remittances are used to pay back loans.
Grigorian and Melkonvan (2011) examine the implications of migration-driven remittances on the households of origin by analyzing 6,800 households included in the 2004 Armenian Integrated
Living Conditions Survey. Armenian households receiving remittance work fewer hours and spend less
on the education of their children. These households also save more, but the households do not leverage their savings to borrow more money to expand their business activities.
Households with a larger share of women are more likely to receive remittances, and receive remittances in larger amounts. Furthermore, households that consist of more women spend more on the education of their children. Grigorian and Melkonyan hypothesize that this occurs because women
7
could pay more attention to the needs of children, or because female students are more expensive than male students as they are more likely to attend extracurricular activities such as music and arts.3
Armenian households headed by females are more likely to invest in schooling than those headed by males (14% versus 8% respectively). Households receiving remittances with children of school age spend on average 3.2% of their yearly budget on education, compared to 2.7% of
households that do not receive remittances. Having a migrant in the household is negatively associated with school attendance of girls. This indicates that even though remittances can stimulate more
investments in education, migration may have disruptive effects on youth schooling, especially for girls (OECD/CRRC-Armenia report, 2017).
Some households receive remittances without having a direct migrant in the household. These households receive remittances from relatives, friends and for the repayment of migration loans (Adams, 2006). Relatively strong family ties in Armenia could explain transfers going to households of which the migrant is not an immediate member. Remittances received in Armenia also originate from a large diaspora, which do not necessarily have immediate family in Armenia but they do remit money to their distant relatives or friends (Grigorian and Melkonyan, 2011).
3. Background on Armenia
This section gives a short introduction to the Armenia economy and migration movements in the country, to place the outcomes of the Migration Survey 2007-2013 in a wider social-economic context. Moreover, the concept of sex ratio by birth is explained in relation to Armenia.
3.1 Economy
Armenia’s independence, after the collapse of the Soviet Union in 1991, resulted in a huge downturn in its economy. The economy underwent a transformation, from being a centrally planned state to being market-oriented. The fall of the Soviet Union influenced the strong trade links with Russia and other former Soviet Union countries. These structural economic deficiencies were compounded by an armed conflict with Azerbaijan over Nagorno Karabakh4 during 1991-1994 (Gevorkyan, 2015). In the 1990s, Armenia was facing lost output, deteriorating capacities, unemployment, poverty and outgoing migration.
In the early 2000s, factors such as the construction boom, an influx of foreign aid, significant financial involvement by the diaspora, remittance transfers from the migrant workers abroad, and increasing consumer spending elevated growth (Gevorkyan, 2015). In 2006, the World Bank even introduced the term “Caucasian Tiger” to describe the economy’s impressive growth in the first half of the 2000s (World Bank, 2006). However, rising remittances levels have spending effects that lead to real exchange appreciation, creating a Dutch disease effect which makes the tradable sector less competitive (Lartey, Mandelman and Acosta, 2008).
In 2009, the world entered a major recession. The International Monetary Fund stated that former Soviet Union countries would experience a large reversal of economic fortune because of three major shocks:
1) Financial turbulence, which has greatly curtailed access to external funding; 2) Slumping demand from advanced economics;
3) Fall in commodity prices, notably for energy (IMF, 2009).
Armenia was highly affected by the economic crisis: Armenia’s export earnings decreased because of the steep fall in copper prices, foreign direct investment in fell by nearly two-thirds in 2009 compared to 2008 levels and the GDP collapsed by 14.2% in 2009.
The Armenia economy has still not fully recovered from the financial crisis. In 2013, growth decelerated again. The post recovery period was shaped by a shrinking construction sector, rising unemployment and an increase of migration to Russia. The international sanctions against Russia and the decline of the international oil prices led to a decrease of remittances received in Armenia.
3 The authors do not provide quantitative evidence to support this explanation.
8
In 2015, the total population of Armenia consisted of 3 million people. The unemployment rate in 2014 was 17.1% and the youth unemployment was 35.1%5 (United Nations, Department of Economics and Social Affairs, Population Division, 2014).
3.2 Migration
Armenia has a long-standing history of migration. Seasonal labor migration to Russia and other parts of the Soviet Union was a common phenomenon in Armenia before its independence. The economic hardship and unemployment of the early 1990s stimulated international migration (Sevoyan and Agadjanian, 2010).
Today, Armenia still has one of the highest migration rates in the world, with about 30% of the population living outside the country. Unemployment is the dominant push factor for migration, but other reasons include geopolitical threats, negative perceptions of economic governance, social injustice and development uncertainty.
The main destination for migrants is Russia, which attracts 90% of the total Armenian migration population. The remaining 10% migrates internally or goes abroad to the United States, the European Union, the Ukraine and Georgia (International Labor Organization, 2009). Russia is geographically close to Armenia and has a visa-free border entry policy. Furthermore, there is a large wage gap between Russia and Armenia: in 2013 the average monthly salary in Russia was US$ 863 and the average Armenian monthly income was US$ 352 (ILO, 2016). The main sectors employing migrant workers are construction and trade/wholesale.
Remittances sent home by migrants constitute an important source of income for many households in Armenia (United Nations, Department of Economics and Social Affairs, Population Division, 2014). Remittances not only influence the economy in general, but also affect human development, drive up education and health care spending and contribute to buying real estate (OECD/CRRC-Armenia, 2017).
A fall of remittances was experienced during the financial crisis, mainly due to the downturn of the economy in Russia: Over 75% of the remittances received in Armenia are originated from Russia (Mitra, 2010). Table 1 displays the remittances flows into Armenia between 2007 and 2015.
Table 1: Remittances received in Armenia in US$ millions
2007 2008 2009 2010 2011 2012 2013 2014 2015 Inward
remittances flows
1,644 1,904 1,440 1,669 1,799 1,915 2,192* 2,079 1,491 * To compare: net foreign direct investment inflows was US$ 0.38 billion and net official development assistant received was US$ 0.29 billion (Ratha et al., 2016).
3.3 Sex ratio at birth
This study focuses on the differences between (oldest) son and daughter-households in Armenia, which is especially interesting in the case of Armenia because of the high sex ratio at birth6 in the country, which is one of the highest in the world.7 Recent research estimating that 114.5 to 116 boys are born for every 100 girls in Armenia (Guilmoto, 2013; Das Gupta, 2015). The sex ratio at birth in a population with no sex preferences is around 105 boys for 100 girls. For example, the sex ratio at birth in the United States varied from 104.6 to 105.9 in the period 1940-2002 (Mathews and Hamilton, 2005).
This high sex ratio at birth occurs because parents are increasingly focused on giving birth to a son (Das Gupta, 2015). As a result, women use prenatal technology to know the gender of their unborn child and sometimes undergo an abortion if they are pregnant from a daughter. This phenomenon explains the high sex ratio at birth in Armenia.
5 Average unemployment in the OECD countries is 7.3% and the average youth unemployment (age category 15-24) is
16.4% (United Nations, Department of Economics and Social Affairs, Population Division, 2014).
6 A ‘sex ratio’ refers to the total number of males for every hundred females in a population. Naturally more boys than girls
are born.
9
This sons-preference can be explained by different factors. First, the patriarchal structures of Armenian society create a social imbalance among children and tend to strongly favor sons over daughters. Second, prenatal technology is widely available and abortions are easily accessible across Armenia. Third, family size has decreased over the last decade and therefore, repeated pregnancies are no longer the preferred solution to ensure the birth of a son (Guilmoto, 2013).
The official figure of the abortion rate in 2011 in Armenia was 18.9 per 1,000 women aged 15-448 (United Nations, Department of Economics and Social Affairs, Population Division, 2014). The unofficial figure is predicted to be higher because of underreported abortions and a lack of systemic statistical registration of abortions (Guilmoto, 2013). Data from the latest Demographic and
Health Survey shows that 29% of the pregnancies may end in an abortion and that the number of
abortions per woman is 0.8 (Guilmoto, 2013). Another study concludes that half of the women with migrant partners in rural Armenia had at least one abortion in her lifetime (Sevoyan, 2011).
The Armenian de facto9 census data of 2001 and 2011 are used to calculate the sex ratio in Armenia by administrative territory (“marz”). The sex ratio for children between the ages of 0-4 is used as a proxy measure for sex ratios at birth because of the limitation of the available Census data and to reduce the influence of incomplete birth registration.10 The sex ratio for this age category is presented in Figure 1. Based on this calculations, the average sex ratio in the country is 115.6. The marzes that have a ratio above the country average are Aragatzotn, Armavir, Gegharkunik, Kotyak, Shirak and Vajots Dzor.
Figure 1: Sex ratio by marz in Armenia age category 0-4, Census data 2001 and 201114. 4. Data: Migration Survey 2007-2013
The characteristics, main findings, data sampling and problems of the Integrated Survey of the
Migration of the Republic of Armenia Population 2007-2013 (“Migration Survey 2007-2013) are
explained in this section.
4.1 Introduction and main findings
The Migration Survey 2007-2013 was a joint project between the Armenian National
Statistical Service, the International Organization for Migration and the Russian-Armenian (Slavonic)
8 The abortion rate in Georgia is 32.3 per 1,000 women (2010), in the Netherlands the abortion rate is 9.7 per 1,000 women
(2012) and the United States has a rate of 12.1 abortions per 1,000 women (2012) (United Nations, Department of Economics and Social Affairs, Population Division, 2014).
9 The de facto data is used because this data expresses the population more accurately.
10 The sex ratio for young children (0-6) are also used as a proxy measure for research about sex ratios at birth in India
because of incomplete birth registration (Mathews and Halmilton, 2005).
14 There is only Census data available from 2001 for the marz Kotayk.
100 105 110 115 120 125
10
University. The survey includes questions about the socioeconomics background of households, migration experience within households, the motivation of migrants and migration destinations.
The Migration Survey 2007-2013 has the scope to retrieve information about the following household members:
1) Current household members;
2) Usual household members that are temporarily absent;
3) Internal and international migrants who are still considered part of the household as of January 2007;
4) Household members who migrated before 2007 and have not since returned to Armenia. The Migration Survey 2007-2013 consists of data on 7,890 households and incorporates information of 31,118 household members. The majority of the survey respondents are ethnically Armenian (99.1%). The survey also included other ethnic groups like Yazidis, Russians, Kurds, Ukrainians and Assyrians. The size of the household varies between one and fourteen household members (IOM, 2014).
Around 12% of the household members undertook one or more migration trips after 2007. Two thirds of the migrants are male, suggesting a bias for males regarding migration movements. There is a concentration of migration trips among the active age group (20-54 years old). The main migration destination is Russia. Other destinations are Nagorno-Karabakh, the United States, the European Union and internal migration within the country (IOM, 2014).
Of the surveyed household members, 8.2% have plans to move abroad. The reasons to migrate are overwhelmingly economic, the most frequent answers are:
1) Absence of jobs in Armenia;
2) Unsatisfactory amount of remuneration received; 3) Earning money for the household;
4) Uncertainty towards the future (IOM, 2014). 4.2 Collection of the data and data sampling
The Migration Survey 2007-2013 is a household survey, which has the advantages of being more cost efficient than a Census survey and asks specific, in-depth questions about migration. Specialized surveys focused on migrants have the problem that international migration is a relatively rare event. The design of the Migration Survey 2007-2013 took this into consideration and adjusts the research design to these specifics (McKenzie and Mistiaen, 2007; IOM, 2014).
The data is collected through face-to-face interviews in July and August 2013, with an average duration of 45 minutes. The respondents’ participation in the survey was voluntary. The most
informed household member was interviewed, and answered the questions on behalf of all the
household members. The household member tried to recall and answers questions related to the 2007- July 2013 period. Questions on financial matters cover the twelve months leading up to when the interview was conducted (IOM, 2014).
The households included in the survey were selected based on the National Statistical Service
Sampling Frame with marzes as stratum and includes urban and rural areas. This sampling frame was
based on the 2001 population census. The sampling of the households was conducted in a two-stage process. The numeration areas were identified and in each numeration area, twenty households were selected to be interviewed. An ultimate sampling unit was replaced if it could not be identified, could not be contacted or refused to participate.
The dataset of the survey include weights, which are used as expansion factors to make the survey representative for the total population (IOM, 2014). This research takes the weights of the survey into account by the analysis of the data (section 7).
4.3 Problems with the data
The most informed household member answered the questions for household members that were not present at the moment of the interview, which influenced the accuracy of the answers given. Moreover, because the survey covers the period from 2007-2013, problems of recall bias arise. The financial prosperity questions cover a period of twelve months, which is a long period to accurately
11
remember financial details. The respondents answered the questions by using their own judgement and did not have to provide additional documents, such as bank statements.
The survey response rate was approximately 52%. The selected households can give a (partially) biased overview because an unknown number of migrants do not have close relatives in their place of origin. Two main issues that were encountered during the field work can explain (a part) of the non-response:
1) The interviewers used an old address list15 to identify the household. The old address list included addresses of households that were abandoned. The abandoned dwellings could have belonged to entire families who have migrated. Marz with a high migration related non-response rate are Gegharkunik, Kotayk, Syunik and Tavush (IOM, 2014).
2) The survey contained sensitive questions and asked a lot of in-depth questions (IOM, 2014).
It is also worth noting that migration and remittance indicators may be subject to measurement errors. Households could underreport remittances to understate their true income. In addition,
households may choose to underreport the number or existence of migrants in the household to avoid compulsory military service (Grigorian and Melkonyan, 2011).
Several survey design problems were encountered during this research. For instance, a couple of relevant and important variables were not included in the survey, such as the characteristics of the migrant that sends the remittances, a variable that directly links children to their parents, and the total number of children of a mother.16 Furthermore, several of the answer possibilities in the Migration Survey 2007-2013 were not specific or accurate enough.
5. Descriptive statistics
The main emphasis of the Migration Survey 2007-2013 are Armenian households and the survey does not provide specific, information about the characteristics of the migrants who send remittances to the households. This study focuses on a specific subset of the Migration Survey 2007-2013. First, this subset is explained. Second, the descriptive statistics for households with children are presented. Third, the descriptive statistics for households with children that are receiving remittances are discussed.
5.1 Explanation
This research compares two different household groups: oldest-son and oldest-daughter household. The division between these two groups is not fully exogenous, some Armenian parents make decisions to influence the gender of their child by having an abortion. The characteristics of these two groups can therefore differ, for instance traditionally oriented parents are more influenced by son-preferences. Control variables, which are explained in section 6, are used to take account for these differences.
The focus of this study are households that include children within the age category 0-18 years old. This age category is, in general, part of the household and are financially dependent on the other household members. Having children influences household decisions made by the other household members. Households that only include children that are nineteen years and older but still live with their parents are not included in this research because young adults are often less financially dependent on their parents. Moreover, households that include a daughter-in-law aged seventeen or eighteen, which is the only household member that was indicated as a “child” are excluded because the assumption is made that the daughter-in-law is not financially dependent on her parents-in-law.
Households who have a twin as the oldest children with different genders are also not included because for these households it is impossible to identify them as an oldest-son or daughter-households. This research only includes households with children for which it was possible to generate the age of the mother when she gave birth to her oldest-child. The way the survey was conducted made it impossible to generate this variable for households in which the mother is deceased and households
15 Based on the 2001 Population Census.
16 The survey did include a question about the relationship between the head of the household and the other household
members. However, because in Armenia multiple generations and families often live together the link between children and parents was not always traceable.
12
that consist of multiple families and generations. Therefore, these households are not include in the analysis, which can lead to selectivity problems.
The survey also included questions about the allocation of remittances. Respondents had several answer possibilities and could indicate as many categories as they wished. They did not need to provide any preference over the categories. The questions with possible answers are displayed in Table 2 and are grouped together in different categories. The categories “leisure” and “other” are not discussed in this research because the frequency that households allocate resources to these categories was very low.
The respondents could answer monetarry questions in Armenian Dram, United States Dollars, Russian Rubles or Euros. The answers are all converted to US dollars to create uniformity.17 The remittances values are converted to a monthly monetary value to compare the values with other (monthly) monetary values. It should be noted that remittances are not usually received on a monthly basis but rather on a generally unscheduled, periodic basis.18
Table 2: Spending decisions made about received remittances Question 7 Section 7B, Migration Survey 2007-2013 Armenia Did the money received within the last twelve months give you or
another member of your household an opportunity to:
Category: 1. Buy food and/or clothes for the household consumption
2. Buy other household items consumption
3. Pay for professional education of a household member education
4. Pay for medical services health
5. Pay debt debt
6. Pay for ceremonies (wedding party, birthday party, funeral) ceremonies 7. Pay for going abroad (migration) migration 8. Pay for rest/tourism (travel) and for leisure time leisure
9. Buy land, house, apartment investment
10. Rent (more) land investment
11. Improve land investment
12. Acquire property, equipment for the farm investment 13. Invest in non-agricultural business investment 14. Make other financial investments investment 15. Pay for apartment and repair investment 16. Make savings (in bank and so on) investment
17. Other other
5.2 Descriptive statistics: households with children
Table 3 displays the characteristics of the households with children which are divided in two groups: oldest-son households (1,789 households) and oldest-daughter households (1,737 households). The characteristics of the two groups are analyzed and the potential differences are examined by using an Adjusted Wald test.
Around 28% of the oldest-son households received remittances in the last year, as compared to 26% of the oldest-daughter households. The difference between the two groups is not significant.
Oldest-daughter households have on average significantly more household members and children than oldest-son households. Furthermore, oldest-son households (27%) are significantly more often one-child households than oldest-daughter households (22%). These observations are in line with the conclusion of Guilmoto (2013). He examined the probability that Armenian families decide to have an additional child by looking at the size and gender composition of the family. Guilmoto
concludes that if parents already have two children, the decision to have a third child is highly
dependent on the gender composition of the family: only 19% of the households that have at least one son will have an additional third child as opposed to 37% of the household that do not have a son.
17
The average exchange rates over the month July 2013 (retrieved from Oanda.com) are used to convert the currencies.
18 The International Labor Organization estimated that 80% of the migrants sends remittances once every quarter
13
Table 3: Descriptive statistics: households with children Migration Survey 2007-2013
Oldest-son household
Oldest-daughter household
Total households included 1,789 1,737
Total household members included 7,677 7,559 Receiving remittances
mean (sd) mean (sd) differences Receive remittances during the last 12 months (%) 28.06 % 25.47 % 0.108
Household characteristics Number of members in a household 4.99
(0.04)
5.15 (0.04)
0.002 *** Number of children in the household 1.69
(0.03) 1.83 (0.03) 0.000 *** One-child household (%) 27.13 % 21.73 % 0.017 ** Rural settlement (%) 38.31 % 34.31 % 0.047 **
One or more household members are retired (%) 37.31 % 38.30 % 0.602 Total household income retrieved from salaries in
USD per month
$ 411.36 (14.50)
$ 384.67 (13.71)
0.146 Household do not receive income from salaries (%) 21.21 % 22.39 % 0.463
Migration
Household member migrated between 2007-2013 (%) 35.27 % 34.99 % 0.875 Household member migrated during any year (%) 38.77 % 38.85 % 0.965
Characteristics of the head of the household
Age of the household head 50.32
(0.44) 49.40 (0.44) 0.090 * Female (%) 51.27 % 55.53 % 0.022 ** Higher education (%) 27.57 % 24.64 % 0.076 *
Unemployed and looking for a job (%) 12.11 % 12.46 % 0.775 Characteristics of the oldest-child
Mother’s age when she gave birth to her first child 23.80 (0.15)
23.22 (0.14)
0.001 *** Age oldest-child in the household
Enrolled in an educational institution (%) Not enrolled in an educational institution (%) Wants to continue higher education (%)
10.17 (0.15) 62.57 % 11.08 % 13.25 % 10.54 (0.18) 66.11 % 8.443 % 16.97 % 0.127 0.075 * 0.018 ** 0.006 *** Standard errors are mentioned in parentheses.
* significant at the 0.10 level ** significant at the 0.05 level *** significant at the 0.01 level
The age of the mother when she gave birth to her first child diverges significantly for oldest-son and daughter-households. When a woman gives birth to a oldest-son, she is on average 23.80 years-old, compared to 23.22 years-old for woman that give birth to a daughter. These differences could be related to the son-preferences in Armenia. Abortion rates in Armenia are high (Sevoyan, 2011; Guilmoto, 2013; United Nations, Department of Economics and Social Affairs, Populations Division, 2014) and women that are pregnant with a daughter more often decide to have an abortion. These women will try to get pregnant again with the aim to give birth to a son. This implies that these women are older when they (finally) give birth to a son.
Another striking outcome is that oldest-daughter households have significantly more often a female head of the household (56%) than oldest- son households (51%). The higher percentage of female headed households for oldest-daughter households can be explained by this principle: female household heads have the power to influence household decisions, such as having an abortion, and are probably less in favor of this option due to their gender.
Oldest-son households live more often in a rural environment than oldest-daughter households. This could be an indicator that son-preferences are more present in rural settlements.
14
The head of oldest-son households are significantly higher educated than household heads of oldest-daughter households. To the contrary, oldest-daughters are significantly more often enrolled in an educational institution, and more often want to continue higher education as compared to oldest-sons.
The financial situation of the two household groups is very similar based on the characteristics analyzed. The average household income retrieved from salaries is higher for oldest-son households but this difference is not significant.
5.3 Descriptive statistics: households with children receiving remittances
The subsequent section focuses on households that received remittances in the last twelve months and expressed the monetary value of these remittances. This group is again divided into two groups: oldest-son households (441 households) and oldest-daughter households (362 households). Table 4 displays the characteristics of the two groups.
The average amount of remittances received per month is US$ 197.84 for oldest-son households and oldest-daughter households receive US$ 179.08 on average. The average income received from salaries is also higher for oldest-son households (US$ 759.22) than for oldest-daughter households (US$ 698.05). This difference is not significant. The high share of remittances in relation to the total income received from salaries by households indicates that remittances are an important source of household’s disposable income. For both groups, a relatively high percentage of the households do not receive any salaries but only receive remittances (16% oldest-son households and 19% oldest-daughter households) which makes these households highly dependent on remittances.
Both groups of households most often dedicate remittances to the allocation categories: “consumption”, “health” and “debt”. “Education” stands out as a category where significantly different allocation decisions are made: 24% of oldest-daughter households allocate remittances to “education” compared to 18% of oldest-son households.
The average number of household members in a household is almost the same for both groups, however oldest-daughter households have significantly more children. Around 83% of the households has a household member that migrated during any year. Grigorian and Melkonyan find that 62% of the households with migrants receive remittances, but the authors recognize that this number is likely to be on the low side. Households receiving remittances may choose to underreport the number (or existence) of migrants in the household, if the latter left the country for reasons that are illegal, such as avoiding compulsory military service (Grigorian and Melkonyan, 2011).
Almost 16% of the households receive remittances without ever having a migrating household member. This implies that households receive remittances from indirect family and friends. This finding is in accordance with the literature (Adams, 2006; Grigorian and Melkonyan, 2011).
Around 63% of the oldest-sons and 65% of the oldest-daughters are enrolled in an educational institution. Oldest-daughters want more often to continue higher education in the upcoming year, and proceed to college, university or postgraduate studies, however this difference is not significant.
15
Table 4: Descriptive statistics: households with children receiving remittances Migration Survey 2007-2013
Oldest-son household
Oldest-daughter household
Total households included 441 362
Total household members included 2,250 1,899 Receiving remittances
mean (sd) mean (sd) differences Remittances received in USD per month
Remittances allocation decision1
Consumption Education Health Debt Ceremony Migration Investment 197.84 88.13 % 18.21 % 39.74 % 28.42 % 16.76 % 12.02 % 30.76 % 179.08 88.47 % 24.17 % 36.54 % 30.36 % 17.98 % 15.08 % 31.09 % 0.275 0.896 0.060 * 0.403 0.596 0.668 0.235 0.930 Household characteristics
Number of members in a household 5.08 (0.07)
5.17 (0.08)
0.400 Number of children in the household 1.66
(0.04) 1.85 (0.05) 0.001 *** One-child household (%) 25.60 % 21.39 % 0.271 Rural settlement (%) 48.26 % 42.95 % 0.196
One or more household members are retired (%) 34.85 % 38.93 % 0.315 Total household income retrieved from salaries in
USD per month
$ 759.22 (42.08)
$ 698.05 (46.70)
0.347 Household do not receive income from salaries (%) 15.64 % 18.54 % 0.381
Migration
Household member migrated between 2007-2013 (%) 80.27 % 78.00 % 0.498 Household member migrated during any year (%) 84.48 % 82.58 % 0.552 Remittance receiving and no migrant household
member (%)
15.52 % 17.42 % 0.552
Characteristics of the head of the household
Age of the household head 51.14
(0.81) 51.57 (0.91) 0.717 Female (%) 47.58 % 55.99 % 0.029 ** Higher education (%) 22.67 % 19.01 % 0.238
Unemployed and looking for a job (%) 11.19 % 9.16 % 0.381 Characteristics of the oldest-child
Mother’s age when she gave birth to her first child 23.38 (0.26)
22.66 (0.25)
0.047 ** Age oldest-child in the household
Enrolled in an educational institution (%) Not enrolled in an educational institution (%) Wants to continue higher education (%)
10.22 (0.30) 63.02 % 10.18 % 14.56 % 10.48 (0.37) 65.38 % 9.07 % 17.46 % 0.600 0.545 0.642 0.346 1 Households can identify multiple categories, therefore the total percentage adds up to more than 100 %. Standard errors are mentioned in parentheses.
* significant at the 0.10 level ** significant at the 0.05 level *** significant at the 0.01 level
6. Methodology
This methodology section consists of two parts. The first part focuses on the hypotheses which can answer the research question. The second part explains the equations that are used to answer the research question.
16 6.1 Hypothesis
This thesis focuses on the following research question:
Does the gender of the oldest-child in a household influence the likelihood of receiving remittances, the monetary value of the remittances and decisions that are made in regard to remittance allocation?
The hypotheses for the receiving of remittances, the monetary value of the remittances and the allocation of the remittances are based on previous research and are summarized in Table 5.
Table 5: Hypotheses of allocation decisions for remittances to different categories for oldest-son and daughter households.
Oldest-son households Oldest-daughter households Receiving remittances Ambiguous Ambiguous
Monetary value Ambiguous Ambiguous
Allocation categories
Consumption same same
Education - +
Health no prediction no prediction
Debt no prediction no prediction
Ceremony - +
Migration + -
Investment + -
Households that have a child will spend part of the additional resources that they receive from remittances on their child. The expectation is that oldest-son and daughter households make different allocation decisions. Grigorian and Melkonyan (2011) conclude that Armenian households with a greater share of women are more likely to receive remittances and receive a higher monetary value of remittances. The authors take into consideration all the females in the household. This research however focuses on the gender of the oldest-child, so no direct comparison can be made to previous work done by Grigorian and Melkonyan.
The son-preferences at birth could not only influence prenatal decision but could also
influence decisions made by the household after the child is born. Households with migrants who send remittances could for instance unconsciously discriminate and send no remittances or less remittances to oldest-daughter households. On the contrary, oldest-daughter households could also be more in need for remittances. The average total household income for oldest-daughter households is for instance lower than for oldest-son households, which could be a reason that oldest-daughter households would receive more remittances. Since the effects are not clear, empirical research can shed some light on these questions.
The prediction is that both household groups will most often dedicate resources to the
spending category “consumption” (Chami et al., 2003; Makaryan and Galstyan, 2013). No significant differences is expected because this category covers essential expenditures for the households.
Based on existing literature, there are no hypotheses about how the gender composition of the household influences the allocation decisions for the categories: “debt” and “health”.
It is expected that oldest-daugther households will more often allocate resources to the
category “education” because based on research of Kroeger and Anderson (2013) boys20 are less likely to be enrolled in school. This implicates that less resources need to be allocated to education.
Households that have a larger share of women spend more on the education of the children in the household (Grigorian and Melkonyan, 2011). Moreover, educational expenditures are higher for girls if the household head migrated (Antman, 2011).
The prediction is that households with an oldest-daughter spend more often remittances on the category “ceremony” because of marriage related expenditures. There is a higher demand imposed on
20 In the age group between 14-18 years old. Even though this research focused on Kyrgyzstan a comparison can be made
17
families by costs associated with the presence of women in families, such as wedding expenditures (Grigorian and Melkonyan, 2011).
The hypothesis is that oldest-son households dedicate more remittances to the category
“migration” since migration is mostly male-dominated and therefore sons are more likely to migrate in the future than daughters (IOM, 2014). Resources are needed to be able to migrate, therefore
households could decide to (already) allocate more resources to this spending category to be able to fulfill future costs that are associated with migration.
It is expected that son-households will make the decision to allocate more remittances to the category “investment”. The prediction is that son-households will invest more in boys because of the social norm of patrilocality, which implicates that sons provide care for their elderly parents while daughters leave the house of their parents to leave with her in-laws. This social norm could influence monetary decisions (Ebenstein, 2014). Furthermore, the son-preference in Armenia can decrease the likelihood to find a wife for males because of the high sex ratio at birth. Son-households can decide to allocate remittances to the category “investment” because it is important to improve the son’s position on the marriage-market (Lundberg and Rose, 2004; Wei and Zhang, 2011).
6.2 Empirical methodology
This thesis focuses on three, different outcomes: the reception of remittances, the monetary value of the remittances and the allocation categories to which the remittances are dedicated by the households.
A linear probability model with binary variable outcomes is conducted to analyze if a household receives remittances or not. In equation (1), 𝑌𝑟𝑒𝑚𝑖𝑡𝑡𝑎𝑛𝑐𝑒𝑠,𝑖 indicates if the household 𝑖 receives remittances “yes” (1) or “no” (0), 𝛼1 is a constant, 𝐷𝑖 is 1 if the oldest-child of the household is a daughter, 𝑥𝑖 represents the limited control variables and 𝑧𝑖 embodies the extensive control
variables, which are defined in Table 6.
𝑌𝑟𝑒𝑚𝑖𝑡𝑡𝑎𝑛𝑐𝑒𝑠,𝑖 = {
0 𝑖𝑓 𝑛𝑜 1 𝑖𝑓 𝑦𝑒𝑠
𝑌𝑟𝑒𝑚𝑖𝑡𝑡𝑎𝑛𝑐𝑒𝑠,𝑖 = 𝛼1+ 𝛽1𝐷𝑖+ 𝛾𝑥𝑖+ 𝛿𝑧𝑖+ 𝑒𝑖 (1) An Ordinary Least Square regression is used to analyze the logarithm of the monetary value of the received remittances. Equation (2) represents the monetary value that a household receives if
𝑌𝑟𝑒𝑚𝑖𝑡𝑡𝑎𝑛𝑐𝑒𝑠,𝑖 = 1. In (2), 𝑌𝑣𝑎𝑙𝑢𝑒,𝑖 is the logarithm of the monetary value in dollars of the remittances
that the household receives, 𝛼2 is a constant, 𝐷𝑖 is 1 if the oldest-child of the household is a daughter, 𝑥𝑖 represents the limited control variables and 𝑧𝑖 embodies the extensive control variables, which are also defined in Table 6.
𝑌𝑣𝑎𝑙𝑢𝑒,𝑖= 𝛼2+ 𝛽2𝐷𝑖+ 𝛾𝑥𝑖+ 𝛿𝑧𝑖+ 𝑒𝑖 (2) A linear probability model with binary variable outcomes is conducted to analyze if a
household allocates remittances to a specific spending category. Equation (3) represent the categories to which households can allocated remittances, if a household allocates remittances to a category 𝑌𝑎𝑙𝑙𝑜𝑐𝑎𝑡𝑖𝑜𝑛,𝑛,𝑖 = 1. In total, seven categories are included so 𝑁 ranges from 1 till 7. In (3), 𝛼3 is a constant, 𝐷𝑖,𝑛 is 1 if the oldest-child of the household is a daughter, 𝑥𝑖,𝑛 represents the limited control variables and 𝑧𝑖,𝑛 embodies the extensive control variables. The control variables are explained in Table 12. 𝑌𝑎𝑙𝑙𝑜𝑐𝑎𝑡𝑖𝑜𝑛,𝑛,𝑖 = { 0 𝑖𝑓 𝑛𝑜 1 𝑖𝑓 𝑦𝑒𝑠 𝑌𝑎𝑙𝑙𝑜𝑐𝑎𝑡𝑖𝑜𝑛,𝑛,𝑖 = 𝛼3+ 𝛽3𝐷𝑖,𝑛+ 𝛾𝑥𝑖,𝑛+ 𝛿𝑧𝑖,𝑛+ 𝑒𝑖 (3) 𝑛 = 𝑐𝑎𝑡𝑒𝑔𝑜𝑟𝑦 1 𝑡𝑖𝑙𝑙 7
18 7. Results
This section and the subsequent sections focus on answering the research question. 7.1 Likelihood of receiving remittances
In this research, not only the gender of the oldest-child is taken into consideration. The gender of younger brothers and sisters is included by using limited control variables. In addition, this research also controlled for other characteristics of the households, these limited and extensive control variables are outlined in Table 6.
Table 6: Control variables: receiving remittances and the monetary value of remittances Limited control
variables:
Total household members, household is located in a rural area, marz, age of the household head, total members of the household that are pensioners, age of the mother when she gave birth to her first child, age of the oldest-child (0-18 years), total children that are 19 years and older, total children minus the oldest-child that are 18 years or younger, percentage of female children minus the oldest-child (18 years and younger)
Extensive control variables:
Total household members, household is located in a rural area, marz, age of the household head, total members of the household that are pensioners, age of the mother when she gave birth to her first child, age of the oldest-child (0-18 years), total children that are 19 years and older, total children minus the oldest-child that are 18 years or younger, percentage of female children minus the oldest-child (18 years and younger)
Head of the household is female, the head of the household is higher educated, the head of the household is unemployed and looking for a job, the household included a member that migrated between 2007-2013, logarithm of the total income of the household retrieved from salaries per month
Equation (1) in Table 7 and 8 does not include any control variables, equation (2) includes limited control variables and equation (3) includes extensive control variables. Table 7 displays, that with inclusion of extensive control variables, there is a significant effect of the gender-composition of the household on the receiving of remittances. Having a daughter as the oldest-child of a household decreases the chance to receive remittances by -0.03. This could be evidence for monetary
discriminatory decision making against oldest-daughter households.
Table 7: The effect of having a daughter as the oldest-child of the household on receiving remittances
Migration Survey 2007-2013
Receiving remittances (yes or no)
(1) (2) (3) Oldest-child is a daughter -0.03 (0.02) -0.02 (0.02) -0.03 ** (0.01) Constant 0.28 *** (0.01) 0.16 (0.08) 0.05 (0.07) R2 0.001 0.037 0.314 Number of observations 3526 3525 3525 No control variables X
Limited control variables X
Extensive control variables X
Standard errors are mentioned in parentheses. * significant at the 0.10 level
** significant at the 0.05 level *** significant at the 0.01 level
19 7.2 Monetary value of received remittances
Table 8 shows that having a daughter as the oldest-child of a households has a negative effect on the on the monetary value of received remittance. The logarithm of the received remittances decreases with -0.04 for oldest-daughter households, by using extensive control variables. This impact is not significant. This result could point towards discriminatory decision making in regard to oldest-daughter households.
Table 8: The effect of having a daughter as the oldest-child of the household on the (logarithm) monetary value of the receive remittances
Migration Survey 2007-2013 (1) (2) (3) Oldest-child is a daughter -0.02 (0.11) -0.06 (0.10) -0.04 (0.10) Constant 6.99 *** (0.09) 7.00 *** (0.51) 6.28 *** (0.57) R2 0.000 0.107 0.154 Number of observations 803 803 803 No control variables X
Limited control variables X
Extensive control variables X
Standard errors are mentioned in parentheses. * significant at the 0.10 level
** significant at the 0.05 level *** significant at the 0.01 level
7.3 Allocation of remittances by spending category
The different allocation decisions made by oldest-son and daughter households are already discussed in section 5.3. The results in Table 10 control for the differences in characteristics between the two groups by using limited control variables and Table 11 includes extensive control variable. The control variables are defined in Table 12. The results in Tables 9,10 and 11 are not representative for the total population because the analysis only takes a very specific sub-sample, households with children receiving remittances, into consideration.
Table 9: The effect of having a daughter as the oldest-child on the allocation of remittances No control variables
Consumption Education Health Debt Ceremony Migration Investment Daughter-household 0.00 (0.03) 0.06 * (0.03) -0.03 (0.04) 0.02 (0.04) 0.01 (0.03) 0.03 (0.03) 0.00 (0.04) Constant 0.88 *** (0.02) 0.18 *** (0.02) 0.40 *** (0.03) 0.28 *** (0.02) 0.17 *** (0.02) 0.11 *** (0.03) 0.31 *** (0.04) R2 0.000 0.005 0.001 0.001 0.000 0.002 0.000 Number of observations 803 803 803 803 803 803 803
20
Standard errors are mentioned in parentheses. * significant at the 0.10 level
** significant at the 0.05 level *** significant at the 0.01 level
Table 12: Control variables: allocation of remittances Limited control
variables:
Total household members, household is located in a rural area, marz, age of the household head, total members of the household that are pensioners, age of the mother when she gave birth to her first child, age of the oldest-child (0-18 years), total children that are 19 years and older, total children minus the oldest-child that are 18 years or younger, percentage of female children minus the oldest-child (18 years and younger)
Extensive control variables:
Total household members, household is located in a rural area, marz, age of the household head, total members of the household that are pensioners, age of the mother when she gave birth to her first child, age of the oldest-child (0-18 years), total children that are 19 years and older, total children minus the oldest-child that are 18 years or younger, percentage of female children minus the oldest-child (18 years and younger)
Head of the household is female, the head of the household is higher educated, the head of the household is unemployed and looking for a job, the household included a member that migrated between 2007-2013, logarithm of the total income of the household retrieved from salaries per month, logarithm of the total remittances received per month
Both groups allocate remittances to the allocation category “consumption” as often as one another. This is in line with the hypothesis made in Section 6.1.
Oldest-daughter households spend significantly moreremittances on the allocation category “education” than oldest-son households. If a household has a daughter, the chance that household allocates resources to the spending category “education” increases with 0.06 by including extensive control variables. This finding is in accordance with the hypothesis made in Section 6 and previous research (Grigorian and Melkonyan, 2011, Kroeger and Anderson, 2013, Antman 2011).
Table 10: The effect of having a daughter as the oldest-child on the allocation of remittances Limited control variables
Consumption Education Health Debt Ceremony Migration Investment Daughter-household 0.00 (0.03) 0.07 ** (0.03) -0.04 (0.04) 0.01 (0.04) 0.00 (0.03) 0.03 (0.02) 0.02 (0.03) Constant 0.92 *** (0.14) -0.03 (0.16) 0.44 ** (0.19) 0.43 ** (0.19) 0.18 (0.14) -0.03 (0.12) 0.47 (0.17) R2 0.070 0.148 0.069 0.094 0.088 0.142 0.238 Number of observations 803 803 803 803 803 803 803
Table 11: The effect of having a daughter as the oldest-child on the allocation of remittances Extensive control variables
Consumption Education Health Debt Ceremony Migration Investment Daughter-household 0.00 (0.03) 0.06 ** (0.03) -0.04 (0.04) 0.02 (0.04) 0.00 (0.03) 0.04 * (0.02) 0.02 (0.03) Constant 0.80 *** (0.14) -0.03 (0.16) 0.38 * (0.22) 0.24 (0.21) 0.02 (0.17) -0.34 (0.14) 0.00 (0.17) R2 0.092 0.167 0.085 0.108 0.102 0.207 0.303 Number of observations 803 803 803 803 803 803 803
