Abstract
This study aims to assess whether an asset based index of economic status correlates better with indices of health and social deprivation than do household income and expenditure. Two sets of data were used to investigate this-Matlab Health & Socioeconomic Survey 1996 and Socioeconomic Census Matlab, 1996. Results show that household assets are significantly correlated with household income and consumption expenditure, but the correlation with expenditure (r2 0.32) is higher than with income (r2 0.10). In case of measuring inequality, children’s nutritional status and their educational enrollment are not significantly linked with household economic rank based on income. However, the health and education outcomes are significantly and consistently associated with household socio-economic position grounded on assets and expenditure. Among all the economic indicators, household assets capture the existing inequality more appropriately (measured by poor: rich ratio) regarding health and social outcomes.
1. Background
Measuring household economic status in developing countries receives a lot of attention from both researchers and policymakers, because many studies have shown that the household economic status is highly correlated with household and individual well-being. One of the main objectives of all development projects is to maximize the welfare of the people through extending its services to the most deprived section of the people. A prerequisite for achieving this goal is the accurate measurement of household economic status to identify those who are at highest risk of deprivation in the society.
Defining household economic status in developing countries poses considerable problems. Two frequently used indicators of household economic status are household income and expenditure. However, such data on income and expenditure are often unavailable and unreliable (Houweling 2003). Moreover, in developing countries where the major income source is self-subsistence agriculture or informal sectors, expressing income in monetary terms is both difficult and unreliable (Rahman 1996, Falkingham and Namazie 2002, Houweling 2003). Comparing to the case of income, household expenditure data are easier to gather. But expenditure data are subject to a different set of complications. Income for the majority of people is a regular flow of money. Expenditure, however, may be irregular. In most developing countries, expenditure data are usually collected in surveys on the basis of recall of one week, two weeks or a month. But recall data are susceptible to large measurement errors (Falkingham and Namazie 2002).
Given the problems of measuring income and expenditure, researchers are increasingly trying to identify alternative tools to measure households’ economic rank that are robust but less data intensive and subject to smaller measurement errors. Recent Demographic and Health Surveys (DHS) have used data on ownership of assets and access to services to derive substitute indicators of household economic standing. This idea was mainly developed by the World Bank to create a simple but effective tool to measure the relative economic position of households. Many health equity studies are now adopting World Bank proposed asset based tool in assessing household socio-economic condition (Gwatkin 2000, Kington and Smith 1997).
However, the issue of equating assets indicators as proxies for household income and expenditure has not been intensively explored. Some investigations conducted on this subject have given variable findings. Montgomery et al (2000) assessed the efficacy of proxy measures (commonly used in demographic studies) in exhibiting the consumption expenditure per adult. They found that the proxy variables were feeble predictors of consumption per adult. However, in subsequent evaluates of fertility, child schooling and mortality, the proxy-based coefficient estimates compared satisfactorily to those obtained using consumption. Sahn and Stifel’s (2001) investigation also indicated the correspondence of the asset index with household expenditure to be weak.
In contrast to above findings, Filmer and Pritchett (1999) explored reasonable coherence between asset index and current consumption expenditures. They concluded that asset indices worked as well or better, than traditional expenditure-based measures in predicting educational enrollment. Wagstaff and Watanabe (2002) also found little difference in assessing correlation between child malnutrition and household economic status while using consumption or an asset based wealth index. Using data from the DSS (Demographic Surveillance System) of Adult Mortality and Morbidity Project of three regions in Tanzania Setel et al. (2003) concluded that the proxy variables were good predictors of expenditure. Morris et al. (1999) found that the wealth proxy associated highly with the more complex monetary value of assets (r=.74). But they did not attempt to explore the direct correlation between household wealth and expenditure.
Houweling et al. (2003) compared the World Bank suggested asset index with three other wealth indices (all based on household assets) to see the extent of health inequality by different measures of economic status. Comparing the World Bank index to the alternative indices, they found that the relative positions of households in the national wealth hierarchy varied according to asset index used; observed poor-rich inequalities in under-5 mortality and immunization coverage often changed, in some cases to an important extent; and that the size and direction of this change varied per country, index and health indicator. In a similar attempt, Bollen et al. (2002), focused on how the selection of proxy measures for economic rank influenced the predicted effects of other explanatory variables on fertility. They arrived at the conclusion that if the consideration is on economic status itself, than the choice of proxy can make a difference. If however, focus is on other variables and economic status is being used as a control, then the non-economic status variables are comparatively robust to the choice of proxy.
Hence the existing studies do not indicate consistent findings regarding the validity of various proxy measures in determining household socioeconomic status. Moreover, most of the studies attempted to validate asset based proxy indicators against household expenditure. They could not address the validation of alternative indices against income due to lack of available data on household income. Opportunely, under the Demographic and Health Surveillance system of ICCDDR,B data on household income, expenditure, assets and other social & health variables are available for a large sample households at Matlab, a rural area in Bangladesh.
However, exploring household assets as proxies for household income and expenditure has not been properly addressed in case of Bangladesh. Consequently, this study attempted to address this issue at Matlab, Bangladesh.
Objective of the Study
The main objective of the study is to examine the strength of the asset based alternative indices of household socio-economic status, as proxies for household income and expenditure. This has been explored in two ways- 1) by measuring the association among household income, consumption expenditure and assets, based on correlation & regression coefficient; 2) by examine the extent of inequalities in children school enrollment and their nutritional status by different SES measures.
Methodology
The present investigation used detailed information on household income and expenditure of a large sample of 4364 households, gathered by Matlab Health & Socioeconomic Survey (MHSS), 1996. In addition, Matlab Socioeconomic Census (MSC), 1996 had records on household assets and access to services for a 39895 sample households, where the same households of MHSS Survey were also included. MHSS 1996, also contained detailed statistics of various health and social variables. After matching the two data sets, a sample of 4275 households have been taken as the unit of analysis.
MHSS 1996 gathered thorough information on household income from all possible sources over one year period. The sources of income were divided into five broad categories. These were income from: a) cultivation, b) sale of products, goods and assets, c) rent, d) employment and e) foreign transfer. Following the standard procedure of Bangladesh Bureau of Statistics (BBS), in calculating income, gross revenue from each source has been considered. To get the net household income for the year of 1996, the expenses needed to produce the gross earnings has been subtracted from the gross revenue. Finally, monthly household income has been considered for analysis. The whole sample is then divided into quintiles based on income and ranked from through lowest to highest.
Household consumption expenditure has been calculated by aggregating the value of consumption and certain other outlays. In MHSS survey, three broad sectors of household expenditure were considered. These included expenditure on: a) food item which comprised the value of different items of foods consumed during the past week b) non-food items which included one year spending on clothing, kitchen equipment, household textiles, repair & maintenance of house, gift for ceremonies, charities, dowry, legal expenses and one month expenses on toiletries, fuel, medicine & medical services, conveyances and (c) education which counted in monthly outlays for tuition & pocket money, food & lodging and one year costs for school uniforms, school supplies, registration fees etc. To calculate monthly household expenditure, week-long and annual expenses has been transformed into one month expenditure. Like previous cases, households have been divided into five equal groups and ranked on the basis of consumption expenditure.
Asset score has been calculated through ‘Principal Component Factor Analysis’ developed by the World Bank. Each household is assigned a standardized score for particular asset, where the score differed depending on whether or not the household owned that asset. In case of ownership of land & dwelling, the amount of land and dwelling space have been counted. The asset scores of each family have been summed up to get the ultimate household asset score. Households have been divided into quintiles based on asset score and ranked through lowest to highest.
For the health indicator weight of children under 5 years of age is considered. Normal & mild underweight is viewed as ‘Normal Weight’ and moderate & severe underweight is regarded as ‘Under Weight’. For the schooling indicator, school attendance of the children of 6-14 years of age is taken into consideration.
Result and Discussion
Statistical analysis indicates that household assets correlates better with consumption expenditure (r2 = .32) than with income (r2 = .10). The measurement of inequality in health and social outcome also exhibits the same trend and direction when using assets and expenditure rather than income as socio-economic status indicator. Statistical outcome shows that children nutritional status and school enrollment is not significantly associated with household current income. However, both these outcomes are significantly and consistently correlated with household expenditure and household assets. Again, compare to income and expenditure, household assets seems to be the best indicator to capture the existing poor-rich gap in child nutritional status and their schooling.
The conceivable explanation of the weak correlation between household income and assets is that: income calculated here is the current level of income for a particular year. Income of a particular year may not reflect the long-term household socio-economic status, because household current income tends to vary from year to year. A wealthier household may gain lower earnings than their expected level of returns in a particular year. Such occurrence was also observed in case of MHSS income data. For example in MHSS 1996 data set, there are 110 households which earned negative income for that year. Regarding the negative earnings, these households are treated as poorest one. However, examining other socio-economic variables, they cannot be termed as poor. For example, the average land ownership of the mentioned 110 households is 98 decimal, which is larger than the average possession of the whole sample of 83 decimal of land. The average monthly disbursement (Tk 6947) of these households is also very high than that (Tk. 4767) of the whole sample.
The transitory nature of income was also reflected in possession of various assets by households of different economic categories. The usual pattern of ownership of a particular asset would follow an increasing trend from poorest to richest quintile. However, in case of socioeconomic status (SES) measured by household income, it is found that the households of poorest quintile owned more assets than the households of second quintile. It happened, because the stated 110 households having negative income are classified as poor, who are not truly poor. Thus identifying SES by current income may led to misclassification of household socio-economic belonging.
Contrariwise, when the SES is based on household consumption expenditure, the possession of assets by different economic quintiles followed the usual trend. Comparing to transitory nature of income, assets are the long-term allocation of household resources. While income reflects the flow of capitals over some period of time, wealth captures the stock of assets at a given points of time, and thus the economic reserves (John and MacArthur 2002). Therefore, income of a particular year may not be well related with the long-term assets available to the household (Falkingham and Namazie 2002).
However, household consumption spending tends to be well correlated with long-term stock of household assets. The underlying reason is that the necessity of spending is not always determined by current household income but also on the basis of expected long-term income and of previous accumulation of capital & resources. Schenk (1997-98) explained that people base consumption on what they consider their “normal” income. In doing so, they attempt to maintain a fairly constant standard of living even having inconsistent earnings from month to month or from year to year. As a result, inconsistency in earnings through upsurge or decline have little effect on people’s consumption spending. In other word, consumption depends on what people expect to earn over a considerable period of time that is long-term income, which is accumulated on household wealth & assets. MacArthur (2002) explained wealth as a source of economic security to meet emergencies or absorb economic shocks.
The overall findings strongly suggest household asset index as an alternative and simple indicator to determine household economic status.