Do Migrants Really Save More? Understanding the Impact of Remittances on Savings in Rural China

Abstract

This article studies the impact of remittances on the savings behaviour of rural households in China, using a primary survey undertaken by the authors in 2006. Allowing for endogeneity and left-censoring of remittances, we find that the marginal propensity to save out of remittances is well below half of that out of other sources of incomes. Moreover, we find no evidence of any direct effect of remittances on either capital input or gross output of farm production. These findings are robust with respect to alternative definition of savings and are in line with recent studies which conclude that remittances are largely used for consumption purposes by rural Chinese households.

Acknowledgements

Many thanks to the British Academy for funding this research and to the Chinese Academy of Social Sciences (CASS) for providing the data. We are grateful to the Ford Foundation Beijing Program Office for supporting the data collection efforts. We also thank participants of the Work Pensions and Labour Economics (WPEG) Conference 2008 at University of Sheffield, and in particular Stanley Siebert and Yasheng Maimaiti, for comments. We have also benefited from very valuable comments from two anonymous referees. The views expressed in this article are those of the authors, and should not be interpreted as those of the institutions with which the authors are affiliated.

Notes

1. Roberts (1997) shows that there are striking similarities between Chinese internal migration and undocumented Mexican migration to the United States – in such key respects as the dominance of circular and repeat migration, large income differentials between sending and receiving areas, legal obstacles that prevent permanent settlements and surplus labour in agriculture.

2. This implies that households who have migrated as a whole on a permanent basis are not included in the survey. NBS (2005: 75) documents that out of the 118.23 million rural–urban migrants, only 24.70 million, or 21 per cent, migrated with all their family members according to the 2004 Rural Household Survey.

3. Migrants are included in the denominator when calculating per capita rural net income, in so far as they are regarded as permanent members of the households who are surveyed at the source of migration.

4. According to the 2000 census (NBS, 2000), while all four counties had inflows of migrants well in excess of 10 per cent ofthe local hukou population, the net inflows of migrants (that is inflows less outflows) as a share of the populationwith local hukou are 19.5 per cent, 16.9 per cent, 7.0 per cent and 4.3 per cent for Xishan, Wuzhong, Danyang and Jiangning respectively.

5. Migrant households and households with remittances are used interchangeably in this article.

6. The theoretical justification is that most of these expenditures should be regarded as investment rather than consumption. Moreover, our survey only asks the year (but not the month) of construction so it is difficult to assign housing costs into the right budget year. There were only 42 and 15 households which built new houses in 2005 and 2006 respectively in the data.

7. This is also the definition of a migrant in the 2000 census.

8. Our findings are robust with respect to the dropping of these demographic controls.

9. It is also possible that real net income and hence real savings are under-reported for this group.

10. To save space, we only report the coefficients of the instrumental variables in the first-stage. A complete set of results are available from the authors upon request.

11. This assumption is more reasonable than it appears given the low marginal product of labour in farming and the prevalence of seasonal migration which allows migrants to work on the farm during the busy seasons (see Zhu et al., 2009).

12. An indicator for weak instruments in the case of a single endogenous regressor is F < 10.

13. To save space, we do not report these results which are available from the authors upon request. In most cases, the differences from 2SLS estimates are negligible.

14. We are very grateful to an anonymous referee for this idea.