Original source
Labeled Remittances: A Field Experiment among Filipino Migrant Workers in the UAE
Discussion paper · Research Seminar in International Economics, Gerald R. Ford School of Public Policy, University of Michigan · Discussion Paper 684 · Version date: February 2022
Search for this source ↗Permission for School’s use is unestablished.
Used in these cards
Implement structural financial constraints or targeted vouchers rather than voluntary communication signals when addressing principal-agent misalignment in transnational families.
"A key barrier to maximizing the development potential of remittances is that migrants who are sending these funds have limited ability to monitor or control how they are used by beneficiaries. Migrants' preferences on how remittances should be spent often differ from the beneficiaries' preferences."
Target financial behavior labeling features toward higher-income migrant segments who possess unallocated discretionary income capacity.
"There is also heterogeneity in the labeling treatment effect with respect to baseline income levels. The migrants for which the labeling feature triggers a significant increase are those with relatively higher income (i.e. income above the median)."
Segment financial policy evaluations by pre-treatment transaction volume to detect intervention impacts specific to under-performing user segments.
"The estimated treatment coefficients indicate a consistent positive effect of labeling on all remitting dimensions when the sample included migrants that were sending below the median in the first 6-9 weeks before the labeling feature was activated in Padalapp."
Evaluate whether digital financial tools drive behavioral adoption through sender personal mental accounting rather than enforcement mechanisms over recipients.
"Alternately, it could mean that labeling positively affects migrant remittances via a different channel or mechanism (other than enhancing migrant control over household expenditure patterns). For example, migrants with access to the labeling feature may do a better job budgeting and tracking how much they are sending in remittances, and perhaps become more likely to send more frequent and higher levels of remittances as a result."
Embed automated transaction notification messaging into financial apps to reduce personal communication expenditures between migrants and recipients.
"There is a reduction in phone and communication expenditures, which may be due to the labeling messaging reducing the need for phone or communication expenditures."
Implement balanced dual-channel sampling across commercial MTO branches and community intercept locations to minimize recruitment selection bias in migrant economic studies.
"Half of our study sample was recruited via face-to-face intercepts of Filipino migrant workers (OFW) in the UAE at locations frequented by Filipino migrants. The other half of the sample was recruited in the context of a partnership between Innovations for Poverty Action (IPA) and UAE Exchange... The screening criteria of this study has not led to the exclusion of potential participants of certain observable characteristics..."
Combine remittance labeling with dedicated direct tuition disbursement mechanisms rather than soft messaging when attempting to promote educational spending.
"There is a scattering of effects on other specific expenditure items, including (surprisingly) negative impacts on educational expenditures. We can propose no obvious reason why the treatment may have reduced educational expenditures."
Measure transaction frequency on the extensive margin alongside total transfer volume when testing financial technology innovations.
"We find a significant effect on the probability of sending remittances (column (1) in Panel (a)), whereas for all the other outcomes the coefficient on treatment is small in magnitude and not statistically significantly different from zero."
Publish photos of participants holding verified completion certificates on official social media channels to establish peer trust and boost recruitment in diaspora field research.
"We also maintained a Facebook page for this study and asked respondents to like the page. We regularly posted pictures of respondents holding their certificates of participation on the Facebook page and announced the current number of participants. Seeing that a large number of fellow Filipinos are participating gave people more confidence in the study."
Cap skewed financial outcome distributions at the 99th percentile to prevent extreme top-tail outliers from distorting causal inference.
"Amount sent: total in Philippine pesos (PhP), summed across all transactions, winsorized at 99th percentile of the full distribution of non-zero weekly amount sent in the data."
Implement dynamic recall window prompts based on individual prior submission timestamps in mobile survey applications to prevent reporting gaps.
"Because participants could respond to the survey on different days, the recall period asked about was always the period since the last date on which the participant recorded a survey response (which accommodated potential non-response to the weekly survey)."
Conduct brief, non-survey introductory welcome calls with remote panel study participants to build rapport months before administering lengthier endline surveys.
"During the welcome call, the household respondent was simply given an overview of the research study and asked if they would consent to being surveyed at a later date. No survey questions were asked in this welcome call. The purpose of this call was to establish a positive connection with the respondent to encourage high response rates during the endline survey..."
Incorporate individual participant fixed effects in panel RCT regressions to absorb time-invariant individual heterogeneity and boost statistical precision.
"Individual fixed effects account for individual-specific time-invariant factors affecting remittances. Because the treatment is randomly assigned, individual fixed effects are not necessary for identification, but they improve precision by reducing residual variation, and also deal with chance imbalances..."
Elicit and compare sender and recipient spending priorities prior to financial product deployment to identify structural alignment gaps in cross-border transfers.
"Migrants' preferences on how remittances should be spent often differ from the beneficiaries' preferences. Migrants often have a stronger preference that remittances be used for purposes that may have general development benefits in the long run, such as investment (in physical or human capital) or savings."
Test empirical treatment interactions across recipient kinship categories to verify whether generational hierarchies alter financial preference signaling.
"In Appendix Tables A1 through A5 we examine treatment effects in other subsamples defined by different baseline characteristics and survey responses, but do not find heterogeneity in treatment effects along other dimensions."
Construct composite Kling z-score summary indexes when evaluating multi-variable treatment impacts to prevent false positive risks from multiple hypothesis testing.
"The last column of the table examines a 'labeling index', a Kling et al (2007) index of the four other outcomes in the table. This labeling index is also positive and statistically significantly different from zero..."
Construct a Kling standard deviation normalized composite index when evaluating interventions with multiple correlated outcome variables to control false-positive rates.
"The last column of the table examines a 'labeling index', a Kling et al (2007) index of the four other outcomes in the table."
Integrate customizable spending category tags into remittance mobile applications to enable senders to transmit specific spending labels alongside transaction notifications.
"The ability to label remittances with the migrant's intended uses leads migrants with low levels of baseline (pre-treatment) remittances to increase their remittance levels."
Track participant financial behavior across multi-year follow-up windows to determine whether novel remittance control features suffer from learning-driven decay.
"This raises the possibility that migrants may later learn that labeling is not having an effect on household expenditure patterns and reduce their remittances as a result. This would mean that labeling has a short-term positive impact on remittance sending (for low-baseline-remittance migrants), but this impact declines or disappears in the long term."
Establish linked survey protocols for both financial senders and primary household spending managers to measure bilateral transaction dynamics.
"We successfully interviewed 1,987 migrant respondents and 2,075 household respondents at endline, among whom 1,377 migrant and household respondents are paired."
Enforce a multi-week active user screening period prior to feature rollout in digital RCTs to prevent differential attrition from skewing experimental results.
"By week 4, those who had complied with the participation requirements were considered retained while the others were dropped from the study. ... Those assigned to the treatment group were randomly selected to receive the intervention starting week 6/7/8/9."
Incorporate receipt photo upload requirements paired with convertible point rewards in data collection applications to verify self-reported financial behavior.
"Participants were asked to upload a photo of the transaction receipt showing all the above items so that the data can be verified. Records that included the receipt photo were awarded points... Upon reaching 100 points, the participant was awarded with a gift certificate of their choice... with a value of AED25 (UAE dirham) (US$6.81)."
Implement direct vendor payment solutions rather than soft informational labels when aiming to guarantee remittance allocation toward recipient education or housing expenditures.
"The labeling treatment does not lead to higher expenditures on uses that migrants report as priority items (in the full sample or in subsamples split by baseline remittances)."
Publish participant photo endorsements alongside institutional co-branding to mitigate privacy mistrust in migrant financial field studies.
"We regularly posted pictures of respondents holding their certificates of participation on the Facebook page and announced the current number of participants. Seeing that a large number of fellow Filipinos are participating gave people more confidence in the study."
Stagger experimental feature activation relative to individual participant onboarding dates to establish uniform baseline observation periods.
"Those assigned to the treatment group were randomly selected to receive the intervention starting week 6/7/8/9. All respondents, regardless of treatment assignment, completed the study after week 30..."
Use conditional prize nomination protocols during baseline migrant surveys to objectively pinpoint primary recipient households for downstream evaluation.
"To identify which corresponding household in the Philippines we should interview, the migrant participants were asked to name (and provide contact information for) an individual in the Philippines who would be the recipient of a US$500 lottery prize... The study participant's choice identifies an individual (referred to as the 'target beneficiary') and household... whose well-being is important to the study participant."
Direct soft remittance labeling towards health and medical spending categories when working with low-baseline remitting migrant populations.
"In the low-baseline-remittance subsample, labeling does lead to higher medical expenses, but has no effect on any of several other household expenditure items."
Monitor participant behavioral data across extended multi-year horizons to determine whether initial financial responses to soft commitment tools persist or decay over time.
"It is possible that migrants in the low-baseline-remittances subsample believe that labeling will be effective at shifting household expenditure patterns, and therefore send more remittances in response to labeling being available. They may not realize that households are not actually substantially shifting their expenditures in response to labeling."
Pause early-stage field intake to secure formal commercial partner integration when facing participant trust or logistical bottlenecks.
"A baseline survey was first launched in September 2018 (wave 1), which was suspended in November 2018 after reaching half of the sample size while we worked with UAE Exchange to formalize our research partnership. The second half of the baseline (wave 2) resumed in February 2019..."