Drivers of of High-skill Emigration during a Crisis: Experimental evidence from graduates in Myanmar

Our study consists of a randomized questionnaire module that will be implemented as part of a large representative survey. Our experiment measures the valuation of emigration opportunities among high-skilled youth (in terms of the rate of wage premium or discount at which the individual is indifferent between staying and migrating) and how this is affected by alternative scenarios of improvements in the political and economic situation of the area of origin, using a randomized survey experiment that assigns alternate scenarios to respondents. Specifically, we examine the value attached to migrating abroad among (i) high-skilled workers in Myanmar when offered a similar job abroad as what they currently are engaged in, and (ii) high-skilled workers in Myanmar when offered a lower-skilled job abroad than what they currently are engaged in – both under alternate scenarios representing the status quo, political stabilization, and economic stabilization. We also propose to measure the value attached to returning to the home country (Myanmar) among Myanmar nationals living abroad when offered a similar job as what they are currently doing abroad, again under alternate scenarios representing the status quo, political stabilization, and economic stabilization, respectively.

2024-02-01

2024-04-30

Our primary measures of interest are:
(1) The (binary) willingness to migrate abroad holding the level of earnings constant
(2) The wage premium (discount) at which the individual becomes indifferent between migrating abroad and staying on (for diaspora respondents in survey C – between returning to Myanmar and staying on in their country of residence).

We will rely on two primary outcomes. The first will measure the willingness to migrate using a binary variable as a response to the following question:

Q1. Imagine a situation: You are offered a job in <Country entered)> which involves the same type of work that you currently do, and which pays you <Current wage>. Your employer at the new job will also help you with relocation, including any travel costs and visas. Will you take the job?

The other primary outcome would be measured as the wage differential (premium or discount) that would make the respondent indifferent between migrating and staying. This would be calculated following a set of iterative questions following Q1 above, depending on the response to Q1.

If the respondent replies to Q1 indicating that s/he would not like to migrate, the following question will be asked:

Q2. E.1.b1. If instead of paying you <Current wage>, the job were to pay you 10% more than that, would you take the job?

If the respondent now replies to say the would accept the job offer abroad, we would record 10 percent as the wage premium at which the respondent has become indifferent, i.e. switched from not wanting to migrate to wanting to migrate. If however they are still not willing to migrate, they will be subsequently asked the same question offering progressively higher wage differentials (in multiples of 10% i..e 20%, 30%, 40 %... 100%, then 150% and finally 200%) to determine the wage premium at which they may become indifferent between migrating and not migrating. Those who refuse to migate even when offered a 200% wage increase will be treated separately as having an exceptionally high preference not to migrate.

While some individuals would, ceteris paribus, prefer not to migrate, as the excess of wages offered abroad over those earned in country (for the same kind of job they are currently doing) increases this preference may switch and at a sufficiently high wage premium, the individual may choose to migrate. The lowest premium at which the individual’s preference switches (from choosing not to migrate to in face migrate) can be seen as the threshold wage premium at which the individual becomes indifferent.

In other cases, some individuals would, ceteris paribus, want to migrate abroad in response to Q1. To such individuals we would pose the following question:

Q3: If instead of paying you <Current wage>, the job were to pay you 10% less than that, would you take the job?

If they do not accept to migrate when offered a 10% wage cut (discount or negative premium) we treat a 10% wage cut (-10% wage premium) as the threshold wage differential at which the respondent’s migration preference switches. If the respondent replies by indicating “yes” to migrating, they will subsequently be asked about their decision to migrate with progressively higher wage cuts (in multiples of 10%) until their preference switches to not to migrate. Those who would like to migrate even after at 90% wage cut will be treated separately as having an exceptionally high preference to migrate.

Similar to the primary outcomes but for (1) in Survey 1 - the specification offering a job abroad that the respondent is overqualified for, and (2) in Survey 2 - for Myanmar migrants living abroad to consider wage premia(discounts) that would induce them to return to Myanmar.

As with the primary outcomes described above

Our experiment measures the valuation of emigration opportunities among high-skilled youth (in terms of the rate of wage premium or discount at which the individual is indifferent between staying and migrating) and how this is affected by alternative scenarios of improvements in the political and economic situation of the area of origin, using a randomized survey experiment that assigns alternate scenarios (political stabilization, economic stabilization, status quo) to respondents.

Not available

Computer-based during the survey

Individual

No

1,600

1,600 The experiment is part of a CATI survey administer using random digit dialling and using screening filters to ensure that respondents meet the sample criteria: high-skilled, between the ages of 18 and 45, and currently or recently employed full-time. The survey will be nationally representative of the target population (high-skilled recently/ currently employed youth) and will be roughly equally allocated across provinces and by gender. The survey includes other modules on individual and household demographic, labor market and wealth characteristics, the ability to migrate, skills, risk appetite, recent shocks and coping strategies, and migration intentions and hypothesis. As we interview only one member household who meets the target criteria, we do not cluster our standard errors at the household level and avoid any spillover effects. In an ideal scenario, the number of experimental observations will match the number of individuals in the sample.

For Survey 1 of high-skilled youth in Myanmar, the whole sample is split into 3 equal treatment arms. When disregarding attrition and non-response, we expect to have at least 530 respondents per treatment arm among the target respondents in Myanmar.
For the online survey (Survey 2) of the contacts of respondents of Survey 1 living abroad, we expect to have a minimum of 1 referred respondent living abroad per respondent and therefore a total sample of 1600 respondents. In case respondents of Survey 1 refer more than one respondent for Survey 2, we will cluster standard errors by the refereeing respondent from Survey 1; this can be done easily as we plan to send unique survey links to each Survey 1 respondent through which we can track the refereeing respondent in Survey 1 for each completed online survey in Survey 2.

The analysis will be conducted separately for Surveys 1 and 2 on high-skilled youth in Myanmar (Survey 1) and their referred contacts living abroad (Survey 2), and will enable us to separately examine the five distinct hypotheses. Consequently, we have provided below the Minimum Detectable Effect size across the five hypotheses. Survey 1 respondents: High-skilled Youth in Myanmar (N=1600) Survey 2 respondents: Contacts of Survey 1 respondents (N=1600) H1, H5: n1=1066 (treatment group) n2=533 (control group) Power=0.8 MDE=0.0696 n1=1066 (treatment group) n2=533 (control group) Power=0.8 MDE=0.0696 H2, H3, H4: n1=533 (treatment group) n2=533 (control group) Power=0.8 MDE=0.0804 n1=533 (treatment group) n2=533 (control group) Power=0.8 MDE=0.0804 These MDEs are realistic, given the results of similar strands of literature (e.g. Batista and McKenzie (2023), Ghorpade (2024)).