Weather variability, agricultural revenues and internal migration: evidence from Pakistan

Abstract

Migration is a widely used adaptation response to climate and weather variability. In this paper, we investigate how changes and variability in weather may affect internal migration through the agriculture channel. Using panel data for 50 districts of Pakistan, we estimate an instrumental variables regression model that allows us to analyse the impact of weather-driven changes in the crop revenue per hectare on the inter-district migration. Results show that temperature has a nonlinear effect, i.e. as temperature increases, the crop revenue per hectare initially increases and then declines. Furthermore, temperature variability has a negative effect on the expected crop revenue per hectare. A 1% weather-driven decrease in the crop revenue per hectare induces, on average, around 2% (0.02% point) decrease in the in-migration rate into a district. Predicted increases in temperature and its variability during 2016–2035 (relative to 1971–1998) are likely to decrease crop revenues in relatively warm districts and increase them in cooler districts. These effects would decrease the in-migration rate in 18–32 districts and increase the rate in the remaining districts. Thus, the extent and scope of the impacts of weather on migration in Pakistan depend on a district’s geographic location and the variability of temperature in the future.

Keywords:

• migration
• weather change
• weather variability
• climate change
• agriculture
• panel data model
• instrumental variables regression
• Pakistan

Acknowledgements

We would like to thank Jeffrey R. Vincent for his guidance and valuable comments, and Priya Shyamsundar for her support and valuable comments during the research work. We would also like to thank two anonymous referees for their useful comments.

Disclosure statement

No potential conflict of interest was reported by the author.

ORCID

Heman D. Lohano http://orcid.org/0000-0001-7640-1124

Notes

Supplemental information for this article can be accessed at https://doi.org/10.1080/17565529.2017.1372263

1 For this, we first compute using the values of and Equation (4). The variable is defined only if is strictly between zero and one. Since the values of are strictly between zero and one in the data of this study, is defined for all observations in the dataset.

2 The 1981 Census Report provides migration data for two 5-year periods (1971–1976 and 1976–1981) and the 1998 Census Report provides migration data for the two 5-year periods of 1988–1993 and 1993–1998. Migration data are not available for the period 1982–1987 because of the delay in conducting the 1991 Census, which came to be conducted only in 1998. In addition to the Census Reports, migration data are also available in the Labor Force Surveys. Since the Labor Force Surveys however are designed for computing provincial and national level statistics and are based on sampling methods, we use data only from the Census Reports, which provide district level data and are not based on sampling.

3 Pakistan census reports do not contain data on alternative measures of migration such as out-migration and net migration. Out-migration could be computed through a residual approach, however, this requires data on district level birth and death rates, which are not available in Pakistan.

4 We use the growth rate between 1972 and 1981 for estimating population in 1971 and 1976, and the growth rate between 1981 and 1998 for estimating population in 1988 and 1993.

5 Thus, while there were only 73 districts in the year 1981, there were 119 districts in the year 1998.

6 Although there were 73 districts in 1981, merging with consistent boundaries from 1981 to 1998 was possible for 71 districts due to formation of new districts.

7 Appendix A (see Supplementary information) presents a list of these 50 districts or merged districts.

8 Wheat accounted for 51% of total net area sown while these four major crops accounted for 64% of the total cropped area in Pakistan in the year 1997–1998 (Government of Pakistan, 2014a).

9 Appendix C (see Supplementary information) presents the over-time pattern of in-migration rate at the district level.

10 Data show that the wheat revenue per hectare has increased in all districts due to technological change but it increased by only 19% (from PKR 10,414 to 12,429) in these four districts and by 40% (from PKR 9381 to 13,116) in the remaining eight districts of Sindh from 1971–1981 to 1988–1998. Thus, over this time period, the revenue per hectare in these four districts turned into 6% lower than the other districts in 1988–1998 (from 11% higher in 1971–1981).

11 These weather projections are for the winter season (December to February), which is part of the Rabi season (November to April).

Additional information

Funding

This research was supported by the South Asian Network for Development and Environmental Economics [grant number SANDEE/July2011/002].