Split Decisions: Household Finance when a Policy Discontinuity allocates Overseas Work Michael Clemens and Erwin Tiongson Review of Economics and Statistics (Forthcoming) Marian Atallah Presented by: Mohamed Issa Nov. 22, 2016
Outline I. Introduction & Motivation II. Aim (Research Question) III. Context IV. Contribution V. Theoretical Framework VI. Dataset VII. Empirical Methodology (Estimation strategy) VIII. Main Results IX. Robustness checks X. Conclusion & Critique
Introduction & Motivation Temporary migration programs are widely used around the world, but there is little evidence on their development impacts. Measuring the household impacts of migration is challenging because: 1. Households that self-select into migration can differ unobservably from other households. 2. Migration affects households by various mechanisms. It can raise a family s income but also split the household geographically, with ambiguous net effects on financial decisions. This ambiguity arises because migration can influence the household through 3 different channels:
Introduction & Motivation Migration Remittance Income Non-remittance Income Decision-making power Alleviating capital Absence of migrant Giving non-migrants or credit constraints (forgone income if greater control over main breadwinner) HH decisions Changing labor supply of non-migrants
Aim (Research Question) What are the effects of a temporary migration program on the financial behavior of households in the Philippines via these 3 mechanisms (remittance income, non-remittance income and decision-making power)?
Context Why the Philippines? In 2004, the government of the Philippines signed a bilateral agreement with the Republic of Korea allowing participation of Filipino workers in Korea s Employment Permit System (EPS), a temporary migration program. Starting 2005, all Filipino applicants to EPS jobs were required to pass a Korean Language Test (KLT). The maximum score was 200 points, and a score of 120 points or greater was required to secure a work permit. This setting is well-suited for a regression discontinuity design (RDD) analysis.
Contribution to the literature 1. Uses a natural policy discontinuity to minimize selection bias with new data from the Philippines 2. Compares the results of the RDD quasi-experimental approach to observational estimators to show that substantial bias could arise from using the latter to measure the effect of migration on households 3. Presents evidence that migration affects households not only through changes in remittance and non-remittance income, but also in household decision-making power (the last being underexplored in the literature)
Theoretical Framework Migration is modeled as a family decision by a collective household engaged in dynamic optimization. The collective household model is chosen as it is more suited to studying households split between different countries than unitary household models. HH members (1, 2) get utility from consumption (c1, c2) and leisure (L1, L2) over lifetime T, and have an available investment that requires no labor input. The household solves: Ø = Disutility of migration
Theoretical Framework Migration can affect household investment through different channels: Increased investment as borrowing constrained HH uses migration earnings to raise home production. F = r if no borrowing constraints Migrant s absence can directly change home production technology by altering its productivity (θ) Effect on labor income net of consumption (ψw*) by altering the wage w & non-migrant labor supply Decrease in the influence of migrant in household decisions/change in balance of power between migrant and non-migrants N.B.: Home production could represent a family business, a farm or human capital investment.
Dataset In 2010, the authors constructed a purposebuilt survey of the households of EPS-Korea job applicants who have scored near the passing threshold of the language test. The outcome of this survey was a sample of the households of 899 workers in the Philippines from a universe of 25,320 who applied to high-wage temporary jobs in Korea between 2005 and 2007.
Policy discontinuity allowing RDD implementation
Checking the validity of RDD assumptions Lack of applicants ability to manipulate scores: 1. McCrary non-parametric test (no significant difference in density around cutoff) 2. Test score used is exclusively the score from each worker s first attempt (to exclude self-selection from re-taking). 3. No reports of cheating, corruption or other irregularities
Estimation Methodology Regression Discontinuity Design As shown before, the context and data offer a natural quasiexperiment. This setting is well-suited for RDD which (as a LATE) can approximate the causal identification randomized experiments in real settings. offered by Sharp Parametric RDD (Intent-to-Treat): Forcing variable (test score) perfectly predicts treatment (migration). where y = outcome of interest, d = 1 if the applicant passed the test and 0 otherwise, s is the applicant s score minus the cutoff score (thus s = 0 barely passes), X is a vector of predetermined covariates, α is a constant, e is an error term, and βitt, η, and vector ρ are to be estimated.
Estimation Methodology Fuzzy Parametric RDD (Treatment-on-the-Treated): 2SLS Estimation Test score highly influences the likelihood of migration but does not perfectly predict it. where treatment τ = 1 if the applicant is a migrant and 0 otherwise, α and ζ are constant terms, ν is an error term, and βtot, η, µ, and vector ρ are to be estimated. Non-parametric sharp RDD (ITT) Non-parametric fuzzy RDD (TOT)
Main results Impact of migration on: Household income and spending Labor supply of adult nonmigrants Children s Health and Education Disentangling the relative importance of the 3 channels (remittances, non-remittance income, decision-making power)
Main results: Effect on household income Rise in remittance income offset by decline in non-remittance income
Main Results: Effects on spending, saving, borrowing and investment Migration has a positive significant effect on quality of life, education and health expenditures, and reduces borrowing from family for non-business purposes.
Main results: Effect on adult (non-migrant) labor supply No significant effect on non-migrant labor supply, Consistent with the program s temporary nature
Main results: Impact of migration on children
Main results: Impact of migration on applicants role in household decision-making Migration significantly reduces the migrant s involvement in household decision-making for applicants who were initially married.
Assessing the relative importance of each channel The remittance and non-remittance income channels are relevant for all households, but the decision-making channel appears to be at work only in the case where migration separates married couples.
Robustness checks Displaying both parametric and non-parametric RDD results (precision-bias tradeoff) Re-estimating the impact on decision-making (Table 7) in the parametric regressions controlling for the gender of survey-respondent to ensure that the shift in reported decision-making power isn t reflecting the mere opinion of the person filling the survey Significance unaffected
Showing the importance of RDD in the estimation /identification strategy The different results obtained in OLS and PSM reveal the value of the RDD quasi-experimental approach.
Conclusion A natural quasi-experiment has shown that having a member temporarily work in Korea has important effects on households in the Philippines within 3 5 years. These effects arise not only from the remittance channel but also from changes in non remittance income and shifts in household decision making power. The study has shown that one has to be cautious when interpreting the estimated effects of migration in observational studies.
Critique: Limitations & the way forward External validity concerns: 1. The study only captures the medium-term effects of migration; the long-run impact could be different. 2. Lack of representativeness to average Filipino HHs 3. Inability to study female or return migrants separately due to their small representation in the sample 4. The paper focuses on a temporary migration program, so its results need not apply to permanent migration.
Thank you! Comments? Questions?