Bank of Uganda Working Paper Series Working Paper No. 03/2014 Worker s remittances and household capital accumulation boon in Uganda

Bank of Uganda Working Paper Series Working Paper No. 03/2014 Worker s remittances and household capital accumulation boon in Uganda Kenneth Alpha Egesa Statistics Department Bank of Uganda January 2014 Working papers describe on-going research by the author(s) and are published to elicit comments and to further debate. The views expressed in the working paper series are those of the author(s) and do not in any way represent the official position of the Bank of Uganda. This paper should not therefore be reported as representing the views of the Bank of Uganda or its management

WORKER S REMITTANCES AND HOUSEHOLD CAPITAL ACCUMULATION BOON IN UGANDA Kenneth Alpha Egesa Bank of Uganda January 2014 Abstract This paper analyses the developmental effect of workers remittances using Probit regression and the two-step Heckman selection techniques. Empirical evidence, based on micro level data, indicates that households which received higher amounts of worker s remittances were more likely to spend on capital accumulation. Results also show that household characteristics such as asset ownership and the level of education of the household head are important determinants of spending of worker s remittances on capital goods. The findings support the need for promoting inflows of worker s remittances to supplement private capital accumulation. Key words: Worker s remittances, household capital accumulation, probit regression, economic development. Journal Classification: I0, I2, I3, N3, O3 Contact E-mail Address: kegesa@bou.or.ug or ike_iek@yahoo.com 1

1. Introduction Worker s remittances from the Diaspora also commonly referred to as workers remittances are defined as all current transfers in cash or in-kind sent from households abroad to households in the local economy. Over time, worker s remittances have increasingly gained prominence as a key source of foreign funding for poor countries. Part of their prominence is due to their direct welfare effects to the beneficiaries. This is mainly because worker s remittances are sent directly to the beneficiaries who then spend the funds according to their needs often utilizing the funds to address basic consumption needs before spending on other things. Over time, their value has grown rapidly with estimates of receipts in developing countries during 2012, placed at $401 billion, which is an increase of 5.3 percent compared with 2011 1. Compared to other types of flows, worker s remittances were the second most important flows to developing countries after foreign direct investment (FDI) estimated at about 600 billion during 2012. They exceeded both private debt and portfolio equity and overseas development assistance (ODA) estimated at about US$260 billion and US$140 billion respectively in 2011 (see Figure 1). The evolution of worker s remittances also depicts resilience as shown by the smooth upward trend compared to FDI and private debt and equity flows which faltered during the global financial crisis and ODA which leveled off. The resilience of worker s remittances is attributed to the altruism nature of senders (Vanwey, 2004) and the strong attachment of the senders to the beneficiaries (Orozco et al. 2006). 1 World Bank Migration Development Brief, 2012. 2

The most recent estimates of worker s remittances to Uganda indicate that they exceeded US$0.9 billion during 2012 which is the highest that has been reported so far (Bank of Uganda, 2013). Similar to the global scene, worker s remittances exceeded estimated ODA at US$677 million and private debt and portfolio equity estimated at US$ 278 million. Previous reports of the annual personal transfer survey conducted by Bank of Uganda 2 also indicated that the bulk of personal transfer receipts are spent on consumption estimated at a share of over 55 percent of the total. Generally, household expenses and education are the most common expenditure categories, with about half of the received cash spent on items in these groups. The key issue to note from this spending pattern of personal transfer resources is one of sustainability. In a number of studies, it has been noted that migrant s attachment to original households weakens over time, resulting in a reduction and sometimes discontinuation of worker s remittances (Lillard and Willis 1997; Agarwal and Horowitz, 2002). Where much of such resources are put to consumption, it is likely that there will be immediate welfare improvements although over the long-term, such improvements may not be sustainable. It is for this reason that the need to channel worker s remittances towards more developmental use especially since they constitute a large share of foreign inflows has been emphasized by recipient countries. Some initiatives have been undertaken to tap worker s remittances for developmental use including introduction of dedicated bonds by governments; municipals and businesses; specialized real estate projects; banking products etc. However, it is not clear how successful the initiatives have been. This notwithstanding, it is important to note that at the household level, the remittances are a key source of income for poor households. This study investigates how worker s remittances influence the spending choices of recipient households and whether or not individual household characteristics matter when it comes to spending worker s remittances. This is important for distinguishing the different effects and determining what policy choices need to be taken to enhance the usefulness of worker s remittances. The main objective of the study was therefore to establish the importance of remittances in explaining the expenditures of recipient households on capital goods. The paper proceeds in four parts. Section II provides a review of the literature on worker s remittances, section III provides a description of the data and methodology used and Section IV provides the main findings of the study. Section V concludes. 2. Literature Review There are two main opposing views in the literature on the use of remittances and their respective effects on recipient households. One view contends that remittances are fungible and, as such, a dollar of remittance income is treated by the household just like income from any other source (de Vreyer & Lambert, 2006). The second view posits that remittances induce behavioral change in the recipient household which may alter its spending pattern (Taylor & Mora, 2006; and Adams & Cuecuecha, 2010). 2 Inflows of Annual Personal Transfers, 2010/2011/2012. 3

The notion that remittances are fungible and as such their effect on household spending does not significantly alter their welfare in a different way compared to other sources of income stems from the argument that they simply comprise additional income to the household (Yang & Choi, 2007 and Taylor,2006). From this perspective, to the extent that they are not migrants own savings, the household retains the spending decision. A case is made for this view taking into account that households comprise several members with different access to resources, both earned and unearned. Consequently, according to the unitary household model, if all resources are pooled to form the household total budget and if allocation disregards contribution of each source, then the origin of the income ceases to matter (de Vreyer & Lambert, 2006). The other consideration is to do with decision making in the use of the remittances by the receiving household. Where the decision of spending is left to the recipient, remittances can be argued to be fungible with other sources. Looked at this way, spending of remittances on any item would not be different from spending on the same item from any other source. There are important developments however, that negate the two arguments. First with the advent of improved technology, it is possible for the sender to earmark remittances to specific expenditures. This could be achieved either through direct worker s remittances to the intended beneficiaries within each household who then proceeds to spend according to the earmarked item without necessarily making the resources available to other household members through pooling. Alternatively, worker s remittances could be earmarked to a specific benefit of a recipient through the sender directing such remittances to the provider of the good or service. Common examples include direct payments to schools of education fees or landlords for rent. The second view which has gained prominence argues that remittances can cause behavioral changes at the household level and that they may sway household expenditure towards more consumption or towards more investment goods. There is however no clear consensus on the nature of expenditures that tend to be influenced by remittances as different studies have tended to contradict each others. Taylor (2006) noted that in the case of Mexico, remittances can be viewed as an outcome of household integration with outside labor markets via migrants. In this setting, migrants may link households to new markets, societies and culture influencing the recipient households spending behavior. For instance household spending behavior may be altered through provision of information by migrants which broadens the consumption set. Households may also spend remittances less conservatively where they are considered to be a less risky source of income compared to other forms. In addition, where remittances are considered to be a transitory source of income, spending may be more on investment goods while where it s deemed permanent, spending could be on items whose use and upkeep requires additional spending such as fuel for a car. 4

Adams and Ceuceucha (2010) found for Guatemala evidence of less consumption spending by recipient households when compared to non-recipient households. The finding was also supported by the result that recipient households spent at the margin more on education and housing. Their study lends support to the view that remittances positively influence development through human and physical capital accumulation affirming the hypothesis that remittances are a transitory income stream that is spent more on investment than consumption. Acosta (2006) found similar results for household expenditure on education although the effect was found to be significant among children between 11 and 14 years (primary school) and not significant for those between 15 and 17 years (secondary school). In the case of housing investment, evidence was found for Nigeria of a positive effect of remittances on the financing of housing investments (Osili, 2004). However, the study also showed that such investment was responsive to socioeconomic factors such as home town amenities and social networks and macroeconomic shocks. For instance, the study showed that risks that are usually associated with such investment could be mitigated by home family and home town association networks. Be they fungible or not, remittances constitute an important source of resources for recipient households, without which their spending would be curtailed. In terms of the desirable use of remittances, both consumption and investment goods constitute important items for all households and contribute to expansion of output for the economy as a whole. Obviously, owing to the nature of worker s remittances, it is not possible to guarantee them as a permanent source of income. For this reason it would be preferable to use worker s remittances for the acquisition of investment goods as opposed to consumption goods. This notwithstanding, the evidence available in the literature is not conclusive on whether worker s remittances contribute to capital accumulation. This study therefore contributes to the debate on the effect of worker s remittances on capital accumulation in the case of Uganda. The analysis addresses some key methodological concerns of endogeneity of explanatory variables and selection bias associated with samples of remittances recipients used in such studies. 5 3. Data and Methodology 3.1 Data The sample data used for analysis consisted of 985 households that received worker s remittances during 2012. The households included in the analysis were extracted from a larger sample of 1118 households covered during the annual personal transfer s survey of 2012 conducted by Bank of Uganda and Uganda Bureau of Statistics due to consideration of incomplete observations for some of the households. The survey covered a total of 500 enumeration areas spread across the country. A two-stage sampling design was used with the

first stage comprising enumeration areas which were selected using Probability Proportional to Size on the basis of the 2002 Population and Housing Census data as a sampling frame. Households with a recipient of worker s remittances were selected at the second stage from each of the selected enumeration areas. The dependent variable used had a value of 1 when a household spent part of its worker s remittances received on a capital good and 0 otherwise. Spending on capital goods was defined for the purpose of the study as any spending on education, health, business capital and physical capital. The explanatory variables comprised worker s remittances received, household head s employment status, and education level, and household ownership of assets such as a car and house. The value of worker s remittances received was measured as the natural log of the amount in Uganda shillings received by the household. The employment variable was 1 if the household head was employed and 0 otherwise. Employment for purposes of this study encompassed self employees, paid employees and unpaid family work while unemployment covered full time students and those who were out of work be they in active search of employment or not. Education status had two dummies: no education which took on a value of 1 for household heads who had no formal education and 0 otherwise and secondary education which took a value of 1 for household heads who had secondary education and 0 otherwise. Household ownership of assets was represented by two variables; one based on car ownership and the other based on whether they owned or rented the house they lived in. 3.2 Methodology The analysis was done using a Probit model for household expenditure behavior on capital goods. The model used took the form: The function F(.) is the cumulative distribution function of a Standard Normal distribution. In the model y i represents spending on capital goods by household i, X i is a vector of demographic characteristics of the household head, W i is a vector of household characteristics, Z i refers to remittance receipts and δ i is the error term. There are two key problems associated with the estimation of Equation 1 using survey data resulting from endogeneity and sample selection bias. Endogeneity arises from the fact that remittances are not predetermined so they are explained in part by some of the same variables that may influence expenditures (Taylor and Moira,2006). Selection bias problem arises from the use of a sample that is comprised of only households that receive remittances. The possibility 6

that households receiving remittances might have certain unmeasured characteristics that are different from those of households which do not receive remittances results in a biased sample. To deal with the two problems, a three stage estimation approach was adopted. In the first stage, instrumental variables were identified for worker s remittances receipts which were used for estimation in the second stage. The instrument used was the number of family members who live abroad. In the second stage, a two-step Heckman selection model was estimated from which estimates of the mills ratio were obtained. The estimated indicated significant selection bias. The final stage involved estimating a Probit model with various households characteristics expected to influence spending, instruments for worker s remittances receipts and the mills ratio to deal with the endogeneity and sample selection problems respectively. The mills ratio was included in the probit to control for the selection bias. 4. Findings 4.1 Descriptive analysis Table 1 shows some descriptive statistics of the sample used for the analysis. The table shows comparative statistics for households that spent on all capital goods, business capital and physical capital with those that did not. As shown, 606 households spent part of their worker s remittances on capital goods while 379 did not acquire any capital goods. Of those that acquired capital goods using their personal transfer proceeds, 102 spent on businesses and 83 spent on physical capital. Grouped by household characteristics, Table 1 shows that most of the households with employed, male and literate heads spent on capital goods. Similarly, the likelihood of spending on business and physical capital was highest among households with employed, male and literate heads. Table 1: Household spending of worker s remittances and characteristics of household heads Unemp loyed h/h head Emplo yed h/h head Female h/h head Male h/h head Illitera te h/head Literat e h/h head Total % Did not spend on capital goods 45 334 158 221 31 348 379 38 Spent capital goods 84 522 250 356 39 567 606 62 Did not spend on business capital goods 122 761 366 517 61 822 883 90 Spent on business capital goods 7 95 42 60 9 93 102 10 Did not spend on physical capital goods 123 779 374 528 68 834 902 92 Spent on physical capital goods 6 77 34 49 2 81 83 8 7

Total 129 856 408 577 70 915 985 100 % 13 87 41 59 7 93 100 Source: Author s computations Table 2 shows a cross tabulation of household characteristics with spending on capital goods, business capital and physical capital. The majority of households that did not own motor vehicles, owned houses and had family abroad spent part of their worker s remittances on capital goods. However, only a small share of households with employed household heads spent on business and physical capital. Similarly, most of the male headed households and households with literate heads spent part of their worker s remittances on capital goods although spending on business capital and physical capital was only among a few households. In addition, the majority of households that spent part of their worker s remittances on capital goods had at least one family member living abroad. Households that acquired capital goods had a higher average share of their total income from worker s remittances compared to those that did not acquire capital goods. However, the reverse was true for households that acquired business capital while those that acquired physical capital had almost a similar share of their income from worker s remittances to those that did not acquire physical goods. Table 2: Household spending of worker s remittances and household characteristics Live in own house Live in rented house Do not own a car Own a car Do not have family abroad Have family broad Total % Didn t acquire capital goods 187 192 321 58 2 377 379 38 Acquired capital goods 309 297 478 128 4 602 606 62 Didn t acquire business capital 452 431 727 156 5 878 883 90 Acquired business capital 44 58 72 30 1 101 102 10 Didn t acquire physical capita1 442 460 740 162 6 896 902 92 Acquired physical capital 54 29 59 24 0 83 83 8 Total 496 489 799 186 6 979 985 100 % 50 50 81 19 1 99 100 Source: Author s computations In terms of the amount of worker s remittances received, households that acquired capital goods generally received higher worker s remittances irrespective of the type of capital goods. In addition, those that acquired physical goods received higher remittances as shown in Table 3. 8

Table 3: Household spending, share of worker s remittances to total income and average worker s remittances received Average share of worker s remittances to total income Average worker s remittances received Didn t acquire capital goods 23 1,467,048 Acquired capital goods 33 5,420,786 Didn t acquire business capital 30 3,691,108 Acquired business capital 24 5,703,516 Didn t acquire physical capita 29 3,351,846 Acquired physical capital 31 9,851,113 Total sample 29 3,899,500 Source: Author s computations 4.2 Probit regression results As is well known, in a Probit model, the dependent (often a dummy) variable is an outcome of the Bernoulli trial so the derived coefficients do not give the size of the effect of each of the explanatory variables on the dependent variable. Thus, these are fed back into the normal cumulative density function (CDF) to give the marginal effects, i.e. the average changes in the conditional expectation function (CEF) implied by a non-linear Bernoulli model. Following is the interpretation of the model results. The regressions results indicate that the amount of personal transfer received significantly affect the likelihood of spending on capital goods (see Table 4). An increase in transfer receipts increases the probability of spending on capital goods by 28 percent. Households which own cars are more likely to spend on capital goods. Car ownership increases the probability of spending on capital goods by 13 percent. Education of the household head is also an important determinant of household expenditures on capital goods. The probability for household spending on capital goods reduces by 13 percent in households where the heads do not have any formal education. Table 4: Probit regression results on the determinants of household spending on capital Independent variables Index Coefficient Marginal effects Ln (transfer receipts) 0.76(0.0752)*** 0.28(0.0271)*** H/h ownership of dwelling -0.017(0.0848) -0.01(0.0319) H/h ownership of a car 0.36(0.1127)*** 0.13(0.0377)*** Employment status of h/h head -0.07(0.1098) -0.03(0.0405) No education -0.34(0.1978)* -0.13(0.0786)* Attended secondary education -0.08(0.0924) -0.03(0.035) 9

Mills ratio 0.59(0.1703)*** 0.22(0.0629)*** Number of observations 985 985 Percent correctly predicted 71.07 Log likelihood value -572 Source: Author s computations Notes: ***, ** and * indicate 1%, 5% and 10% levels of significance respectively; standard errors in parenthesis.. 5. Summary and policy considerations The study found evidence of a developmental effect of worker s remittances on households through their positive influence on spending on capital goods. As shown, households that received higher amounts of worker s remittances were more likely to spend on capital accumulation. The finding is consistent with similar findings by Taylor and Mora (2006) for Mexico, and Acosta (2006) for El Salvador. However, although rising worker s remittances are a necessity for increased household spending on capital accumulation, they are not sufficient in explaining household investment behaviour. Formal education of household heads and household ownership of a car are both important for spending on capital goods. The findings support the need for promoting inflows of worker s remittances to supplement private capital accumulation. 10

References Acosta, P. (2006). Labor supply, school attendance, and remittances from international migration: the case of El Salvador. World Bank Policy Research Working Paper, (3903). Adams Jr, R. H., & Cuecuecha, A. (2010). Remittances, household expenditure and investment in Guatemala. World Development, 38(11), 1626-1641. Adams Jr, R., Cuecuecha, A., & Page, J. (2008). Remittances, consumption and investment in Ghana. World Bank Policy Research Working Paper Series, Vol. Agarwal, R., & Horowitz, A. W. (2002). Are international remittances altruism or insurance? Evidence from Guyana using multiple-migrant households. World Development, 30(11), 2033-2044. Alderman, H. (1996). Saving and economic shocks in rural Pakistan. Journal of Development Economics, 51(2), 343-365. Arellano, M. (2007). Binary Models with Endogenous Explanatory Variables. Bank of Uganda Annual Personal Transfers Report, 2011. Bank of Uganda Annual Personal Transfers Report, 2012. De la Briere, B., Sadoulet, E., De Janvry, A., & Lambert, S. (2002). The roles of destination, gender, and household composition in explaining remittances an analysis for the Dominican Sierra. Journal of Development Economics. De Vreyer, P., Lambert, S., & Safir, A. (2008). Remittances and Poverty: Who benefits in the Household?. Paris School of Economics mimeograph. Lillard, L. A., & Willis, R. J. (1997). Motives for intergenerational transfers: Evidence from Malaysia. Demography, 34(1), 115-134. Orozco, M., Lowell, B. L., & Schneider, J. (2006). Gender-specific determinants of remittances differences in structure and motivation. Report to the World Bank Group, mimeo. 11

Osili, U. O. (2004). Migrants and Housing Investments: Theory and Evidence from Nigeria*. Economic Development and Cultural Change, 52(4), 821-849. Ratha, D., Mohapatra, S., Silwal, A., Irving, J., & Plaza, S. (2009). Migration and development brief. Outlook. Taylor, J. E., & Mora, J. (2006). Does migration reshape expenditures in rural households?: evidence from Mexico (Vol. 3842). World Bank Publications. Vanwey, L. K. (2004). Altruistic and contractual remittances between male and female migrants and households in rural Thailand. Demography, 41(4), 739-756. Wagh, S., & Pattillo, C. (2007). Impact of remittances on poverty and financial development in sub-saharan Africa. IMF Working Papers, 1-43. Yang, D., & Choi, H. (2007). Are remittances insurance? Evidence from rainfall shocks in the Philippines. The World Bank Economic Review, 21(2), 219-248. 12