Does employment status matter for the wellbeing of rural households in Bangladesh?

Selim Raihan and Fatima Tuz Zohora

In rural Bangladesh, a great challenge is to tackle the low pay, poor-quality jobs that are unrecognized and unprotected by law, widespread underemployment, the absence of rights at work, inadequate social protection, and the lack of representative voice. There remains a big question whether poverty in rural Bangladesh is concentrated in certain employment categories.

Our paper uses the data from the Bangladesh Integrated Household Survey (BIHS) of IFPRI. This data are nationally representative data of rural Bangladesh for the year 2011-2012 where the sample size is 6,500 households in 325 primary sampling units (PSUs). The reason for using the BIHS database for this study is that this is the latest available survey data on rural Bangladesh. Our paper has attempted systematic analysis in understanding the association between employment status and wellbeing of rural households in Bangladesh.

From the BIHS data, our study has used consumption expenditures as the principal indicator of household economic status or wellbeing, and has used per capita consumption expenditure as the proxy for income. The total consumption expenditure is measured as the sum of total food consumption and total non-food expenses excluding lumpy expenditures. Income (expenditure) deciles have been created by dividing the households into ten groups from the lowest to the highest in terms of households’ total income. Employment statuses have been constructed for those household heads who are able and eligible to participate in the labor market. By definition, the labor force consists of everyone above the age of 15 who is employed (including individuals working without pay) or unemployed but actively seeking employment. Household head, not counted in the labor force, includes students, retired people, disabled people, and discouraged workers who are not seeking work.

The distribution of the different employment categories in the labor force is shown in Figure 1. In the x-axis, 10 deciles are organized in ascending order on the basis of monthly consumption expenditure of the rural households. Therefore, first decile is the poorest one and the 10th decile is the richest one. The figure summarizes that, while wage employment is mostly concentrated in the poorer deciles, self-employment is concentrated mostly in the richer deciles. Salaried employed maintains smaller shares among poorer deciles.




Figures 2 and 3 show the educational status of the male and female workers by employment categories in the rural areas. Males with no education seem to be highly concentrated in the wage employment in both farm and nonfarm sector. They are also densely present in the self-employment activities. In the salaried employment category, the dominant share is of the males with less than secondary level but higher than primary education. However, males with HSC and beyond HSC account for around 25% of the salaried employment. Females with no education also seem to be highly concentrated in wage employment (Figure 3). Females with less than primary education has a dominant share in the case of unemployed (55.56 %). In the case of the unpaid family job for the female adults, around 28% of them are with less than secondary but higher than primary education.

In order to investigate the factors affecting wellbeing of rural household in Bangladesh we have used the cross section multinomial logistic regression models. The income status of the household is considered as the dependent variable, where per capita consumption expenditure is used as a proxy for households’ income status. For the explanatory variables, we have used different categories of employment of household head e.g. wage labor in the farm and nonfarm sector, self-employed in the farm and nonfarm sector, salaried worker and unpaid worker. All of these variables are dummy variables, where ‘unemployed’ has been considered as the base employment status. Other explanatory variables are age of household head, years of education of the head, number of dependent members per household, per capita landholding and a dummy variable on whether the household receives international remittance or not.

The major findings from multinomial logistic regressions can be summarized as follows. First, wage employment in the farm sector has statistically significant association with all income declies between 6 and 10. However, such employment status doesn’t have any statistically significant association with income deciles between 2 and 5. For a wage worker in the farm sector, relative probabilities to be in deciles 6, 7, 8, 9 and 10 are respectively 39 percent, 44 percent, 75 percent, 85 percent and 90 percent lower than to be in decile 1. The result depicts the fact that wage employment in the farm sector are more concentrated among the poorer households and doesn’t play any pivotal role in shifting up the status of a household. The result is quite analogous for the wage-employed in the nonfarm sector too: if the household head is employed in nonfarm activities, the relative probability to be in the deciles 9 and 10 are 62 percent and 78 percent lower (respectively) than to be in decile 1.

Second, in case of self-employment, if the household head is engaged in the farm sector, the relative probability of that household to be in decile 10 is 44 percent lower than to be in the base decile 1. This association is insignificant for all other deciles meaning that, self-employment in the farm sector does not necessarily improve the income status. On the contrary, if the household head is self-employed in the nonfarm sector, the relative probabilities to be in deciles 3, 4, 5, 6, 7, and 8 compared to the base category are higher by 90 percent, 86 percent, 124 percent, 84 percent and 72 percent respectively. It shows that, self-employment in nonfarm sector has a strong transitory power to improve household wellbeing.

Third, when considering salaried employment, the study finds no significant influence of salaried employment over shifting the well-being status from income decile 1 to higher income deciles. On the other hand, if the household head is employed as an unpaid worker the relative probability to be in deciles 8, 9 or 10 is more than 80 percent lower than to be in the decile 1.

Among other variables, household characteristics like age of the head, dependent member per household, per capita land holding and remittance status hold significant impact on the nature of economic status of the household. If the age of the household head increases by one additional year, the relative probability to be in the top four deciles compared to the decile 1 increases by around 1.2 percentage points. It is also seen that, with the rise in number of dependents in a family the relative probability of the household to be in a higher decile compared to decile 1 becomes lower. The regression results also suggest that, education and international remittances play a role of pull factor in case of shifting household status from the lowest decile to upper deciles. An increase in the years of education of the household head by one additional year increases the relative probability to be in decile 2 compared to decile 1 by 10 percentage points; whereas, for the same increment, the relative probability to be in decile 10 compared to decile 1 increases by 35 percentage points. In case of remittances, households that receive remittance have more than 3 fold relative probability to be in decile 4 or above. For the remittance receiving households, the relative probability to be in decile 10 compared to the decile 1 is more than 25 times higher than a household that does not receive remittances.  Along with these, per capita land holding is appeared as an important household characteristics that can help a household to be on the higher deciles.

The findings of this paper provide a significant indication that rural nonfarm sector has a crucial role in reducing poverty and increasing the wellbeing of the rural household in Bangladesh. The study also specifies the importance of addressing the concern in the national policy making that poverty in rural Bangladesh is highly linked with certain employment categories.


Cross-country differences in income inequality: Where do South Asian countries stand?


In recent years, there has been a growing interest among general people, researchers and policy makers in income inequality, its causes, and its effects. The most popular index of income inequality is the ‘Gini index’ which measures the inequality among levels of income of the people of any country. A Gini coefficient of zero means perfect equality, where everyone has the same income, and a Gini coefficient of 1 (or 100%) expresses maximum inequality.

For meaningful comparisons among different countries with respect to their levels and trends in income inequality we need comparable data. National surveys on households’ incomes and expenditures in different countries provide data on the Gini index of these countries for some years. However, we are not in a position to use these data for cross-country comparisons due to various reasons. In those surveys there are differences in the population covered, differences in coverage on geography, age and employment status, differences in the definition on welfare (whether to use market income or consumption data), differences in the use of equivalence scale (whether to use household per capita or household adult equivalence), and differences in the treatment of various other items, such as non-monetary income and imputed rents. The Standardized World Income Inequality Database (SWIID), introduced in 2008, provides a dataset on income inequality that facilitates comparability for the largest possible sample of countries and years. A custom missing-data algorithm is used to standardize data on cross-country income inequality using the data from national surveys (Solt, 2016). Using the SWIID database, the World Economy Database (WED) version 9.1 has generated a time series database on the “Gini index” for 207 countries over the period between 1970 and 2015 by filling missing observations with the help of different estimation methods.

Using the WED 9.1, we have produced a scatter plot diagram with data on Gini indices for 207 countries in 1980 in the horizontal axis and data on Gini indices of the same countries in 2015 in the vertical axis. In the scatter plot, dots around the 45 degree line are the countries with ‘no or very small’ changes in Gini indices during 1980-2015; dots above the 45 degree line are the countries which experienced an increase in the Gini index; and finally, dots below the 45 degree line are the countries which experienced a decline in the Gini index. Out of those 207 countries, 18 experienced ‘no or very small’ changes in Gini indices, 109 experienced increases and 80 experienced declines. Among the 8 south Asian countries, 5 countries (Afghanistan, Bangladesh, India, Pakistan and Sri Lanka) observed rises while the rest 3 countries (Bhutan, Maldives and Nepal) experienced declines. We also brought China and South Korea into the picture, and it appears that the Gini index in China increased quite astonishingly, whereas that of South Korea declined.

We have also categorized the values of Gini index as follows: a Gini index value lower than 30 is considered low; an index value between 30 and less than 40 is considered medium; an index value between 40 and less than 50 is considered high; and an index value above 50 is considered very high. Depending on these classifications, we can observe some interesting movements of the South Asian countries during 1980 and 2015. Afghanistan moved from a status of low inequality to medium inequality; Bangladesh moved from medium inequality to high inequality; though Nepal, Pakistan and Sri Lanka remained within the medium inequality range, Sri Lanka was at the border of high inequality; India moved from high inequality to very high inequality; and both Bhutan and Maldives moved from very high inequality to medium inequality. In comparison, China moved from low inequality to very high inequality, whereas South Korea moved from medium inequality to very close to low inequality.

We also explored the factors affecting inequality in the cross-country and over time contexts. Results from a fixed effect panel regression suggest that while rise in the real GDP per capita tends to have a small negative association with the Gini index, an increase in both life expectancy at birth and net secondary school enrollment are strongly associated with the decline in the Gini index. These suggest that, an increase in per capita real GDP is not a guarantee for the reduction in income inequality, whereas investment in social infrastructure with the aim of raising the life expectancy at birth and a rise in secondary school enrollment can be very instrumental in reducing income inequality.

Reference: Solt, F. (2016). “The Standardized World Income Inequality Database”. Social Science Quarterly.

First published at the Thinking Aloud on 1 September 2016

Published at The Daily Star on 1 September 2016