The Short- and Long-term Effects of Rainfall on Migration: A Case Study of Chitwan, Nepal Nathalie Williams and Clark Gray 18 October, 2012 Introduction In the past decade, both policymakers and academics have focused serious interest and concern on the relationship between climate change and migration. Conventional wisdom and general theory leads us to believe that changes in weather or natural disasters will influence large out-migration flows from rural areas. This is primarily due to the belief that changes in weather will ruin agricultural harvests and force people to migrate in search of economic gain elsewhere. However, there is a growing body of empirical literature that shows this not to be the case; studies find small migration responses after large or even drastic weather changes. Although there is consistent evidence in this regard, it is unclear precisely why this is the case. More theoretical development and empirical detail need to assess not just if climate influences migration, but how and why. In this study, our primary objective is to assess the extent to which detrimental weather affects out-migration. However, given that there is a reasonable body of literature on this subject, but still a need to understand the details of the relationship better, we seek to delve further into the climatemigration relationship here. Two further research questions which we investigate are as follows. If there is a migration response, to what extent is it an immediate response to immediate circumstances? In other words, are households using migration as a short-term solution to make up for a poor harvest in one season? The second question is: to what extent is migration a response to longer-term weather patterns that make agriculture unsustainable as a primary livelihood strategy? Given current predictions regarding climate change and the large and growing role of migration in global population distribution, insights into these questions, about the short- and long-term migration responses to climate, will likely be integral to predicting and understanding global population change and distribution. We use a detailed case study approach, investigating these questions in a rural agricultural area of Nepal. We are able to investigate these questions through using exceptionally temporally precise data on migration and weather patterns. Our data come from the longitudinal Chitwan Valley Family Study in Nepal and feature monthly records of migration for all individuals. Paired with data from a local weather station that provide daily records of temperature and rainfall for the past 50 years, we are able to investigate the long- and short-term dynamics of the climate-migration relationship in this setting. Setting The empirical data in this study come from the Chitwan Valley, located in south-central Nepal. The administrative district of Chitwan borders India and is about 100 miles from Kathmandu. There is one large city, Narayanghat, and the rest of Chitwan s population, like much of Nepal, lives in small, rural villages. Agriculture is the dominant occupation, with about 80% of households in the study area involved in farming or animal husbandry. Most of these households operate on a subsistence level, owning or farming small amounts of land and livestock. Many of the households involved in agriculture in this study area own their land, but sharecropping, mortgaging, or rental agreements are also common. In many cases, a family is involved in a mix of these, owning some of the land they farm and renting or sharecropping additional land. The main subsistence crops in this area include rice, wheat, buckwheat, maize, pulses, and a variety of vegetables. Common livestock, which are kept for meat, milk, or as draft animals, include water buffaloes, cows, goats, pigs, and chickens. Recently, some households have begun to engage in market agriculture. Small fruit orchards and chicken farms are the most common enterprises.
As Chitwan is subject to annual monsoons, weather patterns vary dramatically throughout the year and heavily determine farming seasons. In turn, farming seasons, or more precisely, planting and harvest times, determine when labor is most needed. As shown in Figure 1, the monsoon generally starts around May, with heavy rains of over 600 mm per month and hot weather reaching 30 degrees Celsius. After September, the rains subside to almost nothing each month and the temperature usually decreases to 20 or 25 degrees Celsius. Accordingly, most farming households plant rice, which is the most important crop of the year, in June and harvest around November. Wheat, pulses, and vegetables are grown from November through April, and maize is grown February through June. [Figure 1 about here.] Historically, there has been significant migration from the Chitwan Valley to other areas of Nepal and nearby areas of India. More recently, with new government legislation and encouragement of labor contracting organizations, migration to more distant parts of the world has increased dramatically. Estimates suggest that perhaps 1.5 million Nepalis live and work in India, although two to three million might work there seasonally (NIDS 2004; Thieme 2006). At least 800,000 Nepalis live and work in the Persian Gulf (NIDS 2011). Other common destinations include Malaysia, South Korea, and over 100 other countries around the world. Much of the migration is impermanent and viewed as a strategy to supplement regular farm and household incomes (Kollmair et al 2006; Thieme and Wyss 2005). As shown in Figure 2, migration varies by month and by year. [Figure 2 about here.] Data and Measures Data for this study come from two sources individual survey data from the Chitwan Valley Family Study (CVFS) and a weather data from a local weather station in Chitwan. The CVFS is a large-scale multidisciplinary study of the western Chitwan Valley of Nepal, designed to investigate the impact of macro-level socioeconomic changes on micro-level behavior (Axinn, Barber, and Ghimire 1997; Axinn, Pearce, and Ghimire 1999; Barber et al. 1997). Amongst other data, the CVFS includes an individual interview and life history calendar that were collected in 1996, a prospective demographic event registry that has been collected monthly since 1997, and household agriculture and consumption surveys in 1996 and 2001. The primary data on migration comes from the prospective registry. The prospective nature of this data set makes it ideal for studying migration, by providing information on a representative sample of all people exposed to the possibility of migration and following them until the present to record who migrates, who returns, and who migrates again. The demographic event registry includes 151 separate neighborhoods that were selected with an equal probability, systematic sample. All individuals between the ages of 15 and 59 within these neighborhoods were included in the survey. Weather data come from a local weather station run by the National Maize Research Program and obtained from the Department of Hydrology and Meteorology of the Government of Nepal. They include daily records of rainfall and high and low temperatures since the 1960 s. In this study, we use weather records from 1990 through 2006. Measures of migration Migration is measured from the CVFS prospective demographic event registry which provides residence records on a regular basis. Migration is defined as a one month or longer absence from an individual s original 1996 residence. We separate short- and long-term migration by defining shortterm migration as being absent from the original residence from one to 11 months and long-term migration as being absent for 12 months or more. The use of a prospective demographic event registry collected on a monthly basis allows for this precise recording of migration.
Measures of weather Our primary measure of rainfall is the cumulative amount of rain that fell in a year. We use six different variables to measure the amount of rain that fell in the past year, between one and two years ago, between two and three years ago, between three and four years ago, between four and five years ago, and between five and six years ago. Analytical Strategy We use discrete-time event-history models to predict out-migration from the Chitwan Valley in any given month. Person-months are the unit of exposure to risk. The logistic regression models test the monthly hazard of moving out of the Chitwan Valley neighborhood after June 1997, contingent upon rainfall and individual, household, and neighborhood characteristics. Multinomial models are used to disaggregate short-term migration (between 1 and 11 months) and long-term migration (12 months or more). Both of these outcomes are compared to the reference of no migration. Notably, this is a study of any migration, including first and higher order migrations. This is implemented by creating a data set that excludes a migrant for the period of time they were living out of the origin neighborhood. When a migrant returned to their original neighborhood, they are again included in the data set. As noted above, an extensive battery of measures is used to control for migration experience. Results Preliminary results are shown in Table 1. In general, we find that rainfall has significant, but opposite, effects on short- and long-term migration. In addition, rainfall in previous years can affect migration, to the extent that rainfall up to six years previous has larger effects on migration than rainfall in the past one or two years. As shown in Table 1, rainfall in the past three to six years has negative effects on short-term migration, with coefficients from 0.91 to 0.98. This means that the more rain that fell, the lower was the likelihood of short-term migration. Alternately, this could be stated as the less rain that fell (or the closer to a drought that was experienced), the higher the likelihood of migration. This is consistent with predictions that low rainfall or drought would cause people in rural areas to migrate in search of alternate sources of income. With long-term migration, we find results that are consistently in the opposite direction. Odds ratios for rainfall from two to four years previous range between 1.02 and 1.05. This indicates that the more rain that fell, the higher was the likelihood of long-term migration; or the less rain that fell (or closer to a drought), the lower the likelihood of migration. Paired with the negative results for shortterm migration, these positive results for long-term migration suggest that more rain could result in better harvests, greater incomes, and thus the possibility to finance a longer-term migration that might be more lucrative than a short-term migration. Turning to the differences in effects of recent and more distant rainfall, we find the surprising result that there are no statistically significant effects of rainfall in the past year on short- or long-term migration. Rainfall in the previous two through five years has significant effects on long-term migration, and it is not until three years previous that we find significant effects on short-term migration. In addition, the strongest effects we find are from five to six years previous for short-term migration and three to four years previous for long-term migration. In summary, we find no evidence that migration is an immediate response to immediate weather changes; instead, these results suggest that migration is more commonly a response to longer-term detrimental weather in rural areas. [Table 1 about here.]
Table 1. Effect of rainfall in the past six years on short- and long-term migration Multinomial model, outcomes of migration 1-11 months and migration 12+ months are compared to the reference no migration Multilevel models used Migration 1-11 months Migration 12+ months Rainfall (in 100 mm units) in past 12 months 0.99 0.99 in past 13-24 months 0.99 1.02 * in past 25-36 months 0.93 *** 1.05 *** in past 37-48 months 0.98 * 1.05 *** in past 49-60 months 0.91 *** 1.02 in past 60-72 months 0.92 *** 1.03 Individual and household characteristics Age 0.98 *** 0.95 *** Education 1.04 *** 1.04 ** Owns business 0.73 ** 0.79 * Land (natural log of kattha) 0.91 ** 0.90 ** Livestock (livestock units) 0.98 0.98 Distance to Narayanghat 1.05 *** 1.06 *** Services in the community 0.95 ** 1.00 Ethnicity Bahun/Chetri Reference Reference Dalit 0.91 1.11 Terai indigenous 0.75 ** 0.72 ** Hill indigenous 0.88 1.30 ^ Newar 0.80 0.98 Marital status Never married 0.95 0.69 *** Married, living with spouse Reference Reference Married, not living with spouse 0.37 *** 0.58 ** Divorced, separated, or widowed 1.09 0.79 Migration experience Number of previous migrations 1.33 *** 1.15 *** Number of migrants in the household 1.08 *** 1.08 *** Number of migrants in the neighborhood 1.01 ** 1.01 * Study year 0.91 * 0.78 *** Months of the year January Reference Reference February 1.34 * 1.01 March 1.24 ^ 1.27 ^ April 1.15 1.01 May 1.36 ** 1.21 June 1.26 * 1.11 July 0.96 0.99 August 1.67 *** 1.16 September 1.90 *** 0.81 October 0.82 0.63 ** November 1.46 ** 1.08 December 1.71 *** 1.07 Odds ratios shown. ^ p<0.10 * p<0.05 ** p<0.01 *** p<0.001
mm rainfall degrees Celsius Figure 1. Average monthly rainfall, temperature, and crop seasons, Chitwan Nepal 1990-2006 600 35 500 400 30 25 Rice season Maize season 300 20 15 Wheat, pulses, vegetables Monthly rainfall 200 10 Mean monthly temperature 100 5 0 Jan Feb Mar Apr May June July Aug Sep Oct Nov Dec 0 Figure 2. Percent of resident population migrated out of Chitwan, 1997-2006 10% 9% 8% 7% 6% 5% WOMEN MEN 4% 3% 2% 1% 0%