Does information on regional income differentials change migration intentions? A survey experiment in multiple African countries

The intervention consists of an information treatment about regional income or wage differentials in Ghana and Uganda, respectively. The information treatment is embedded in surveys of potential participants in employment and income promoting programs in each country. The survey is conducted face-to-face in Ghana and by phone in Uganda, and implementation of the information treatment depends on the survey mode.

In Ghana, a random subset of participants is presented with a map, which outlines the ten regions of Ghana and depicts the monthly per capita income in each region. To make the income information easily comprehensible, monthly per capita income is illustrated via stacks of coins, with one coin for each 100 GHS. The map is shown by a facilitator who explains the map and the information it contains to the participant. The information on average monthly per capita incomes is based on the seventh round of the Ghana Living Standards Survey (GLSS 7) from 2016/2017 conducted by the Ghana Statistical Service.

In Uganda, a random subset of participants is presented with information on monthly median wage differentials over the telephone. The information provided comprises the monthly median wage in each of Uganda’s four regions plus the capital city Kampala and how many times more or less that wage is relative to the participant’s region of residence. The information on the monthly median wage by region was gathered from the Uganda National Household Survey 2016/2017 Report of the Uganda Bureau of Statistics.

2020-11-12

2021-06-30

(1) Extent of internal migration intentions, measured on a likert scale ranging from “not at all” (0) to “a lot” (3).
(2) Destination choice, measured in terms of the monthly per capita income (Ghana) or median monthly wage (Uganda) of a subject’s top two destinations, the destinations’ income/wage rankings, and whether they are higher-income regions than the subject’s region of residence.

The intervention is expected to affect intentions to migrate internally. We hypothesize that information about how much more or less subjects can earn in different regions of the country will lead subjects to update beliefs and will in turn affect the extent to which they would like to move elsewhere within the country. Further, destination choices are hypothesized to change in response to the information treatment and align with where participants can earn more in comparison to their current region of residence.

(1) Internal migration behavior (incidence and destinations, Ghana only). These outcomes will be analyzed separately, after follow-up data collection in Ghana in 2022.
(2) Internal migration intentions (extent only) conditional on not having a secured job at destination (Uganda only).
(3) Reservation wage (minimum wage) for internal migration (Uganda only).
(4) International migration intentions (extent and destination choice, Uganda only).
(5) Labor market aspirations (confidence in having work and amount of salary, Uganda only).

In response to subjects’ update of their earnings beliefs and internal migration intentions, the intervention may affect actual migration behavior (both in terms of subjects’ decision to migrate internally and their choice of destination), internal migration intentions conditional on not having a secured job at destination, the reservation wage for internal migration, international migration intentions and labor market aspirations.

Experimental arms:
The information treatment is randomly assigned to survey respondents at the individual level. The experiment comprises two experimental arms: (i) one information treatment arm and (ii) one no-intervention control arm. Each participant has a 50% chance to be assigned to the treatment and control group, respectively.

Sampling strategy:
The information treatment is embedded in two distinct surveys, (a) a baseline survey of a randomized controlled trial on the effectiveness of a program for the Professionalization of Artisans (ProfArts) in Ghana and (b) an endline survey of an impact evaluation of a skills training and internship placement program in Uganda.

In Ghana, face-to-face interviews are conducted with up to 20,000 craftsmen in the construction sector in all ten regions of Ghana. Craftsmen are mobilized through a nationwide outreach campaign including flyers, posters, radio jingles and social media and all interested craftsmen can register for program participation. In total, up to 20,000 artisans will be mobilized and registered across Ghana, out of which a random subset of 10,000 people are selected for ProfArts benefits. Additionally, a further set of subjects may be recruited through other programs under the same funding umbrella.

In Uganda, phone interviews will be conducted with approximately 1,400 people who registered their interest in participating in a skills training and internship placement program supported by the German Corporation for International Cooperation (GIZ). The program was advertised over social media and the radio, among other channels. The skills training program was implemented at two training sites in Packwach (Northern region) and Kampala whereas internship placements did not have these regional restrictions.

Implementation procedure:
In both countries respondents are asked the same set of questions in the module for this study. First, we collect information about participants’ migration intentions within their country of residence, including the extent to which they would like to move and their top two destination choices. Second, participants are asked to indicate their perceptions on the average monthly per capita income (Ghana) or median monthly wage (Uganda) in each region before the coronavirus pandemic. Third, a random subset of participants receives the information treatment. Fourth, respondents are asked once again about their migration intentions and preferred destinations.

In both countries the surveys start in mid-November 2020. In Uganda, the survey is scheduled to be completed by the end of 2020 and in Ghana by mid-2021.

Randomization will be conducted on the go using a random draw programmed into the survey tool in SurveyCTO. As a result, randomization will be stratified by survey tool/country.

Participant (individual level randomization)

No

Same as number of observations, no clustering.

Ghana: 20,000 individuals Uganda: 1,400 individuals

Ghana: roughly 10,000 individuals in the control arm and 10,000 in the treatment arm
Uganda: roughly 700 individuals in the control arm and 700 in the treatment arm

In a comparison of group means with 80% power, a significance level of 0.05, and unit standard deviations, the MDE is .038 for the total sample of 21,400 participants, .040 for the sample of 20,000 subjects in Ghana, and .150 for the sample of 1,400 in Uganda.

PAP_wage_differentials_and_migration