Credit Availability and Internal Migration: Evidence from Thailand

Abstract

This study investigates the relationship between borrowing and internal migration. Exploiting the micro-finance scheme Village Fund in Thailand, it investigates the short- and medium-term effects of borrowing since the policy’s introduction. Employing an instrumental variables approach, borrowing is instrumented using the inverse number of households per village at the beginning of the policy to capture potential credit availability. The findings suggest that internal migration in Thailand is not credit constrained and its interaction with credit is time-related. Borrowing reduces internal migration in the medium-term, but is found not to affect migration when the policy is first introduced.

Acknowledgements

The author is grateful to two anonymous referees for their helpful comments. Discussions with Barry Reilly, Michael Lipton, Richard Tol, Yashodhan Ghoparde, Eva-Maria Egger, Hector Rufrancos, Alan Winters, and the participants of many conferences and seminars are also appreciated. The datasets used are available from the Townsend Thai Data Project at http://dvn.iq.harvard.edu/dvn/dv/rtownsend which the author gratefully acknowledges. The Stata code employed in the analysis can be made available by the author upon request. Usual disclaimer applies.

Disclosure statement

No potential conflict of interest was reported by the author.

SUPPLEMENTAL DATA

Supplementary Materials are available for this article which can be accessed via the online version of this journal available at https://doi.org/10.1080/00220388.2018.1498969

Notes

1. For example, Abramitzky et al. (2013) report for Norway that higher parental wealth reduces the likelihood of international migration and that its effect becomes stronger when expected inheritance, by birth order and sibling composition, is inspected. Bazzi (2017) looks at the income elasticity of international migration in the presence of shocks (rainfall and rice prices) in Indonesia. The study finds that international migration increases with positive shocks but this effect is heterogeneous across village characteristics. Hirvonen (2016), using data for Tanzania, finds that temperature-induced income shocks reduce internal migration, and finds that male migration is liquidity constrained.

2. The average official exchange rate for 2001 is 44.43 Baht per one US$, used throughout the article.

3. Borrowing in Thailand takes the form of individual or joint-liability loans, from state-owned or commercial enterprises, and from micro-finance institutions such as production credit groups or targeted women’s groups (Ahlin & Townsend, 2007; Kaboski & Townsend, 2005). Moreover, informal borrowing from moneylenders or kin networks is widely used (Kinnan & Townsend, 2012; Kislat, 2015). Research shows no substitution of VFP credit with other types of borrowing (Kislat & Menkhoff, 2012), but actual increases in other types of credit transactions due to this scheme (Kaboski & Townsend, 2012).

4. The maximum limit per transaction is US$1,125 (50,000 Baht) with higher values provided in case of an emergency (De La Huerta, 2011).

5. The sampling strategy accounts for income differentials among provinces and ecologically balanced characteristics among sub-counties, such as forested against non forested areas, to then leave the choice of villages and households at random (Binford, Lee, & Townsend, 2004).

6. Other longitudinal surveys for Thailand find similar trends, see for example Junge, Revilla Diez, and Schätzl (2015).

7. For further evidence on the regional characteristics of Thai internal migration see Guest et al. (1994) and Huguet and Chamratrithirong (2011).

8. A broader definition of formal credit may limit the investigation if there is endogeneity in the location of institutions (Coleman, 2006). For example, confounding effects may affect the empirical analysis if some areas have a specific type of institution that closes down over time or never actually existed.

9. The use of the LPM avoids the non-trivial incidental parameter problem when using fixed effects in non-linear response models (see for a discussion Greene, 2004). Preliminary tests reveal seven LPM predicted probabilities outside the unit-value range. Applying the Horrace and Oaxaca (2006) trimmed LPM method, consistent estimates are obtained (available on request).

10. Wealth stock is also interpreted here as control for potential past borrowing behaviour. It is constructed using the household history of asset ownership collected at baseline (Townsend, 2009) and depreciated yearly by 10 per cent in addition to land (not depreciated). As wealth accumulation may be done in preparation for future migration, a preliminary analysis excludes the wealth stock variable, ensuring that it does not change any outcome found (available on request).

11. If total borrowing from other institutions was used in addition to VFP, it would require a more sophisticated econometric modelling, going beyond the scope of the present analysis.