Back to History

Fields Changed

Trial
Field Before After
Abstract Urban workers in Kenya earn twice as much as rural workers with the same level of education. Why don't more rural workers migrate to cities? I use two field experiments to show that low migration is partly due to underestimation of urban incomes, and that this inaccurate information can be sustained by migrants' strategic motives to hide income to minimize remittance obligations. Parents underestimate their migrant children's incomes by nearly half, and underestimation is greater when a migrant's incentive to hide income is higher. Providing information about urban earnings increases migration to the capital city by 33% over two years. In many developing economies, urban workers earn substantially more than rural workers with the same level of education. Why don't more rural workers migrate to cities? I use two field experiments in Kenya to show that low migration is partly due to underestimation of urban incomes, and that this inaccurate information can be sustained by migrants' strategic motives to hide income to minimize remittance obligations. Parents underestimate their migrant children's incomes by nearly half, and underestimation is greater when a migrant's incentive to hide income is higher. Providing information about urban earnings increases migration to the capital city by 39% over two years.
Last Published February 24, 2020 05:18 PM January 24, 2021 10:40 PM
Planned Number of Observations Intervention 1: 497 households. Intervention 2: 340 households. Intervention 1: 497 households. Intervention 2: 340 households, scaled-up to 4,994 households.
Sample size (or number of clusters) by treatment arms Intervention 1: 249 control households, 248 treatment households. Intervention 2: 167 control households, 173 treatment households. Intervention 1: 249 control households, 248 treatment households. Intervention 2: 167 control households, 173 treatment households, scaled up to 1,588 placebo households, 841 pure control households, 2,565 treatment households.
Additional Keyword(s) migration, perceived returns, hidden income, asymmetric information migration, perceived returns, hidden income, urban-rural income gap
Intervention (Hidden) This study includes two interventions. Intervention 1 (Wage Info): Randomized at the household level. Unstratified. Each treatment household is given 3 information sheets: one for Nairobi, Kisumu, and Eldoret. Each information sheet includes the following information: average individual earnings, average individual earnings for men, average individual earnings for women, average individual earnings for workers without a primary education (all expressed as a ratio to the respective average in Bungoma Town; top 3 most common jobs for migrants; average individual earnings for migrants in those occupations (expressed as a ratio with respect to Bungoma Town); formal and total employment percentages for migrants; and the food price ratio in the city with respect to Bungoma Town. These statistics were measured using the Kenya Integrated Household Budget Survey 2005-2006 (the latest round available at time of intervention). Intervention 2 (Remittance Info): Randomized at the household level. Stratified along three dimensions: intervention 1 treatment assignment, an indicator for whether either household head had ever traveled to Nairobi to look for work, and an indicator for whether the female head of household had graduated from secondary school. Each treatment household was told the average share of income that migrants in Nairobi remit back to their origin (village) household. Control households were told the average share of income that Nairobi residents spend on food from kiosks (as a placebo). Information for treatment taken from surveys with migrants (undertaken as part of this study). Information for placebo measured using the Kenya Integrated Household Budget Survey 2005-2006 (the latest round available at time of intervention). Households that had migrants in Nairobi at the time of intervention or at the time of the July 2018 midline surveys were excluded from the treatment. This study includes two interventions. Intervention 1 (Wage Info): Randomized at the household level. Unstratified. Each treatment household is given 3 information sheets: one for Nairobi, Kisumu, and Eldoret. Each information sheet includes the following information: average individual earnings, average individual earnings for men, average individual earnings for women, average individual earnings for workers without a primary education (all expressed as a ratio to the respective average in Bungoma Town; top 3 most common jobs for migrants; average individual earnings for migrants in those occupations (expressed as a ratio with respect to Bungoma Town); formal and total employment percentages for migrants; and the food price ratio in the city with respect to Bungoma Town. These statistics were measured using the Kenya Integrated Household Budget Survey 2005-2006 (the latest round available at time of intervention). Intervention 2 (Remittance Info): Randomized at the household level. Stratified along three dimensions: intervention 1 treatment assignment, an indicator for whether either household head had ever traveled to Nairobi to look for work, and an indicator for whether the female head of household had graduated from secondary school. Each treatment household was told the average share of income that migrants in Nairobi remit back to their origin (village) household. Control households were told the average share of income that Nairobi residents spend on food from kiosks (as a placebo). Information for treatment taken from surveys with migrants (undertaken as part of this study). Information for placebo measured using the Kenya Integrated Household Budget Survey 2005-2006 (the latest round available at time of intervention). Households that had migrants in Nairobi at the time of intervention or at the time of the July 2018 midline surveys were excluded from the treatment. Intervention 2 was scaled up as a separate module in a larger survey of rural households in western Kenya after completion within the smaller sample. Households with migrants in Nairobi were again excluded from the experiment. Treatment was stratified along four dimensions: an indicator for whether the household's non-agricultural earnings were above the sample median, an indicator for whether the household's cultivated farmland was above the sample median, an indicator for whether the female head had graduated from secondary school, and a survey round dummy related to the timing of the survey. A pure control group was included which did not receive any placebo information.
Back to top