An Inquiry into the Heterogeneous Outcomes of International Migration: Evidence from Rural Households in Bangladesh

Abstract

There are considerable differences in the duration, intensity, and history of migration experiences among migrant households. The variation can have a significant effect on their investment behaviour. We classify our sample households in Bangladesh into four groups depending on the history and stage of migration and estimate its differential effect on physical and human capital investment. The results show that the patterns of investment are distinct between the groups and even of opposite directions in some cases. These heterogeneous effects often cancel each other out, leading to insignificant effects of migration when the average effect on all migrant households is estimated.

Acknowledgements

We thank late Dr. Mahabub Hossain for granting access to data for this research. We also wish to thank Kazushi Takahashi and two anonymous referees for their insightful comments and suggestions. All errors remain ours.

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

1. For example, Böhme (2015), Mendola (2008), and Taylor and Lopez-Feldman (2009) use a migrant binary indicator for estimation.

2. In recent years, the proportion is changing; woman migrants constituted 18 per cent of total departures in 2015.

3. More detailed information and discussion on household characteristics of migrant households in Bangladesh are found in Kikkawa and Otsuka (2016) and Sharma and Zaman (2013).

4. The cost typically includes visa/passport fees, agent service fee, and travel costs.

5. The survey module defines household head as the one who lives in the household and makes day to day decisions. Migrants are not given this status. Therefore, the change in migration status sometimes involves change in household head.

6. Consumption data on food intake (or its monetary value) was not available.

7. Because our sample are rural households, it is likely that household members juggle rural business and agricultural tasks. It is not apparent from data as to how tasks are shared at individual or household level.

8. Annual income from business activities amounted to 20598 BDT on average in 2014 (2010 price level) and 92910 BDT for those who reported having this income. Average income of the surveyed households was 116262 BDT on average and 79093 BDT at median. Total income of households with business income was 160030 BDT on average and 111956 BDT at median.

9. A similar observation is made by Gibson et al. (2013), where they find that agricultural income of new migrant households actually increased in the first year of migration. They explain that migrants likely exerted more effort on agricultural production before their departure to overseas jobs.

10. Taking the first-order time difference of Equation (1)(1) $Y_{ijt}={\alpha }_i+{\beta }_{1}MI{G}_{it}+{\beta }_{2}H{H}_{it}+{\rho }_{jt}+{\epsilon }_{ijt}$(1) over the two periods, we obtain that $\mathrm{\Delta }{y}_{ij}={\beta }_{1,new}NE{W}_i+{\beta }_{1,cnt}CN{T}_i+{\beta }_{1,post}POS{T}_i+{\beta }_2\mathrm{\Delta }H{H}_{ij}+\mathrm{\Delta }{\rho }_j+\mathrm{\Delta }{\epsilon }_{ij},\mathrm{w}\mathrm{h}\mathrm{e}\mathrm{r}\mathrm{e}\mathrm{\Delta }{x}_{ij}\equiv x_{ij2}-x_{ij1}\mathrm{f}\mathrm{o}\mathrm{r}x=y,HH,\rho ,\mathrm{o}\mathrm{r}\epsilon ;\mathrm{a}\mathrm{n}\mathrm{d}{\beta }_1=\left[{\beta }_{1,new},{\beta }_{1,cnt},{\beta }_{1,post}\right].$ So each coefficient of the migration status dummies corresponds to the average additive change in the outcome variable over the two periods among households with each of the migration statuses relative to the BASE households.

11. We also implemented an alternative specification without a domestic migration dummy as a part of sensitivity analysis and the findings remain largely unchanged.

12. This variable, representing non-farm labour participation, may be endogenous because a migrant household may receive remittances, which could help to create more non-farm jobs for the household members. To reduce this bias, we construct the variable representing village-level non-farm sector participation by subtracting own household (i).

13. There are a total of eight administrative divisions in Bangladesh.

14. Another source of endogeneity is simultaneity because remittances may lead to the general equilibrium effects on village economies, which then influence the level of household investment.

15. Taking the first-order time difference of Equation (1)(1) $Y_{ijt}={\alpha }_i+{\beta }_{1}MI{G}_{it}+{\beta }_{2}H{H}_{it}+{\rho }_{jt}+{\epsilon }_{ijt}$(1) over the two period under this specification (3), we obtain the following relationship; $y_{ij}={\beta }_1\mathrm{\Delta }Mi{g}_{ij}+{\beta }_2\mathrm{\Delta }H{H}_{ij}+\mathrm{\Delta }{\rho }_j+\mathrm{\Delta }{\epsilon }_{ij},\mathrm{w}\mathrm{h}\mathrm{e}\mathrm{r}\mathrm{e}\mathrm{\Delta }{x}_{ij}\equiv x_{ij2}-x_{ij1}\mathrm{f}\mathrm{o}\mathrm{r}x=y,HH,\rho ,\mathrm{o}\mathrm{r}\epsilon ;and\mathrm{\Delta }Mi{g}_{ij}=\mathrm{\Delta }NE{W}_{ij}+\mathrm{\Delta }CN{T}_{ij}+\mathrm{\Delta }POS{T}_{ij},\mathrm{w}\mathrm{h}\mathrm{e}\mathrm{r}\mathrm{e}\mathrm{\Delta }NE{W}_{ij}=1-0=1;\mathrm{\Delta }CN{T}_{ij}=1-1=0;and\mathrm{\Delta }$ $POS{T}_{ij}=0-1=-1$.

16. Equation used can be written as $Y_{ijt}={\alpha }_i+{\beta }_{IV}\widehat{POS{T}_{ijt}}+{\beta }_{2}H{H}_{it}+{\beta }_3{\rho }_{jt}+{\epsilon }_{ijt}$ where ${\widehat{POST}}_{ijt}={\pi }_0+{\pi }_1{z}_{jt-1}+{\pi }_{3}H{H}_{it}+{\pi }_4{\rho }_{jt}+v_{ijt}$ and z is instrumental variables (=historical rate of migration per village) drawn from previous survey (t-1) which is of approximately five years prior to the survey year. For ρ, our preferred model (Vill-dum) uses village dummies that control for all village specific time-variant unobservable factors. But the past migration rate used as IV is a village level variable, so we instead use a Division dummies (Div-dum) model that uses a set of time-variant village level variables along with division dummies.

17. The difference indicates that housing construction may have an endogenous relationship with the decision to return.

18. Some of the households included in UAE sub-samples are also influenced by other county effects. For example, among a total of 139 households in our sample which have sent workers to UAE sometime between 2008 to 2014, as many as 19 (14%) households also sent worker(s) to countries other than UAE.

19. It is possible therefore that households classified as NEW may have prior history of migration in 1999 or earlier.

20. Being POST household does not guarantee that the household stopped sending family members abroad permanently (i.e. they may resume migration in the next period).

21. The survey contains household member data, but this information does not come with unique and traceable ID code for individual.

Additional information

Funding

This work was supported by the Japan Society for the Promotion of Science (JSPS) KAKENHI [JP15J11506, 25101002].