Financial Education vs. Access to Finance in Transnational Households

Individuals who are more able and self-motivated enough to enroll in financial literacy training are also more likely to make sound financial choices. There is an obvious selection bias if those with high levels of financial literacy are different than those with low levels of financial literacy. OLS estimates of the effect of financial literacy on financial decisions and outcomes would be biased as we might end up attributing differences in savings levels to unobserved differences rather than financial literacy. For instance, Meier & Sprenger (2008) show that individual time preference may explain who chooses to become financially literate. At the same time, discount rates are an important determinant of individual saving patterns as a person with a high discount rate would place lower value on future rewards. If we do not control for this unobserved heterogeneity in time preferences across individuals, we will get biased estimates.

In order to establish a causal effect, our study uses random assignment of transnational households to financial literacy training. We seek to evaluate the effect of a financial literacy program and access to financial services.

Filipino overseas workers and their families are our target population. The Philippines is the second largest migrant-sending country and the third largest remittance receiving country in the world. The concentration of Filipino women as international migrants is striking; 87% of international migrants in the services sector from the Philippines in 2010 were women. Among these, 70% were domestic workers.

2014-04-30

2015-12-31

Through financial education migrant households may be better able to allocate the remittances to productive investments. After all, besides their direct impact on current family incomes, remittances are used to fund future investments. In fact, for many households with limited access to credit, remittances may be their only source of finance. Thus, remittances are channeled into investment goods or for small businesses. Families also use them to pay school fees, thereby investing in human capital. A small component of remittances are also used to purchase land and property, as well as to repair and upgrade homes. We would establish the impact of the program by tracking these outcomes via surveys implemented among migrant families during the course of the research project. Further, through the administrative data provided by ASKI and BPI, we would be able to study the usage of financial products and microloans by both migrants and their households.

If we have enough funds to survey migrants located abroad, we will also measure the effect on migrant savings, remittance amounts and channels, frequency of remittances and communication between the migrant and households.

The study team will administer migrants and origin households a baseline survey. Usage of financial products at ASKI and BPI will be tracked using administrative data. A pre-analysis plan for analyses to be conducted will be lodged with The Abdul Latif Jameel Poverty Action Lab (JPAL) prior to the start of baseline fieldwork to forestall later concerns about selective reporting of results. Endline surveys conducted 12 months later will establish impacts on financial outcomes more broadly (beyond outcomes that can be observed in the administrative data).

Randomization ensures that observations are similar on average across treatment conditions, so simple comparison of means across treatments establishes impacts. The primary interest is in outcomes such as control over remittances, savings, remittance amounts and frequency, educational expenditures, housing investments etc.

Examining the impact of financial education and the provision of financial services on various financial decisions and outcomes suggests running the following regression:

Yi = f + b1R1i + b2C1i + b3R1iC1i + ei

Let Yi be the value of a dependent variable of interest, such as savings, control over remittances, amount and frequency of remittances, educational or housing expenditures of households in the Philippines, etc. R1 is an indicator variable indicating assignment to treatment groups in row 1 of the chart, and C1 is an indicator variable for assignment to treatments in column 1. The constant term f is the value of the dependent variable for households who were assigned to Treatment 0, while ei is a mean-zero error term. The coefficients b1 and b2 provide the impact of the financial education program and financial services access respectively while the coefficient b3 measures of the complementarity of providing both financial education and services to the migrant and their household.

The IPA survey team will administer a baseline survey to consenting migrant households upon locating them using the contact information provided by OWWA or through the door-to-door campaign. Migrants at this point will most likely be already abroad. We interview the household head among remaining household members.

The study sample will then be randomly subdivided into different treatment conditions on-site. First, households will be randomized into either a financial education treatment, or into a group that receives no financial education at all. Second, households will be independently cross-randomized into being offered access to formal credit, insurance, and savings products, or into a group that is not offered financial access. This will generate four treatment conditions with 450 households in each. The chart below maps the four groups:

Treatment 0 (control group): This is the control condition: No offer of financial education program and/or financial services. The sealed envelope will merely contain a “thank you” letter for participating in the baseline survey.

Treatment 1: Invitation to attend a financial education program: The household head will be invited to attend a short workshop on financial education in ASKI’s training center. The workshop will be free and will be scheduled on a Saturday. It will last 6-8 hours and will be completed in one day. To facilitate take-up, the household head will be allowed to bring a companion. A letter will provide details and will ask the household head to indicate interest in the program. OWWA will also provide a letter endorsing participation in the financial education program, which we expect to raise take-up.

Treatment 2: Access to financial services and products: This treatment group will only be provided access to financial services and products and will not be offered any form of financial literacy training. In particular, they will be invited to avail of 4 different products:

1. A BPI Savings Account – a regular savings account with low required maintaining balance, which can be used to receive remittances from migrant members abroad.

2. ASKI microloan – a small loan for enterprise development.

3. ASKI micro-insurance - accident or property insurance. Accident insurance covers losses caused by accidental death or bodily injury due to accident occurring in any place in the world. Property insurance covers property damage resulting from a peril.

Households in this treatment group will be provided with a letter that specifies requirements, and would indicate instructions from where and how they can avail of the products in addition to specifying a contact person to talk to. In addition, the surveyors will offer households immediate registration for a ASKI micro-insurance product, should they wish to register on the spot.

Treatment 3: Invitation to a financial education program and access to BPI Financial services and ASKI microloan facilities: This treatment group will be invited to attend both a financial education workshop and given access to financial services offered by BPI and ASKI, as in households in treatment groups 1 and 2.

The offer to attend a financial education program and to avail of financial products constitutes an encouragement design, since in practice we cannot prevent migrants and households from accessing these services independently. That said, past research (Ashraf, Aycinena, Martinez & Yang 2011) shows that mere offers and assistance with opening bank accounts lead to large differences in bank account usage in transnational households. Doi, McKenzie, and Zia (2012) find that attendance rates for financial literacy training in Indonesia have a high take-up rate when offered for free and done in coordination with accredited recruitment agencies, even when the training is done separately for migrant workers and family members. Their experiment involved assigning migrants to a migrant-only training group, a family member-only training group, and a joint training group. The take-up rates were 81.2 percent, 76.5 percent, and 65 percent respectively. We expect our offers to be similarly effective, as the government endorses it.

Randomization will be conducted on-site

Randomization will be at the household level.

No

1,800 households

1,800 households

450 households control
450 households financial services only
450 households financial education only
450 households financial education + financial services

Assumptions: Assume a minimum detectable effect size of .2 standard deviations when standardizing the mean values of the outcome variables of interest. Assume a take-up rate for the invitation to attend the financial education workshops of 90%. This is based on Gibson, McKenzie, and Zia (2012) and Doi, McKenzie, and Zia (2012) who also run financial literacy training workshops for migrants. Assume take-up rates for offered financial products to be 80%. In other words, 80% of study participants use some financial product, whether credit or savings, when offered to them. The assumption is based on a midrange of take-up rates for field experiments that study random provision of credit or savings products. Effect 1: The impact of the financial education workshop (b1) Power = 0.9684 Effect 2: The impact of financial services (b2) Power = 0.9242 Effect 3: The joint impact of financial education and financial services (b3) Power = 0.8508 Ultimately, even beyond these power calculations, we are confident that 1800 households is an adequate sample size to detect our main effects, even for our mini-experiment in Singapore and Hong Kong where the financial education treatment group involves training migrants abroad. The study by Doi, McKenzie, and Zia (2012), who look at similar outcomes, had enough power to detect statistically significant effects even with a sample of only 400 migrant households randomized into three treatment groups. According to administrative data by OWWA, 19.7% of migrants departing for abroad from Cabanatuan City went to Singapore and Hong Kong in 2013. Hence, we expect to obtain a similar subsample of around 355 for the mini experiment.

Pre-Analysis Plan (Financial Education vs. Access to Finance in Transnational Households)

We implemented a randomized controlled trial among transnational households in the Philippines estimating impacts on financial behaviors of a financial education treatment, a financial access treatment, and the combination of the two. We test whether there are complementarities between financial education and financial access interventions, and also provide insight into the nature of constraints operating in financial services markets. We find no evidence of complementarities between the financial education and financial access treatments. In addition, while we find no evidence of constraints in access to formal credit and savings products, our results do suggest that access constraints exist in the formal insurance market. Impacts on other financial behaviors are suggestive of the importance of information constraints in financial decision-making. These results provide guidance to designers of financial interventions in similar populations.

Abarcar, Paolo, Rashmi Barua, and Dean Yang. "Financial Education and Financial Access for Transnational Households: Field Experimental Evidence from the Philippines" Working Paper, April 2017

http://sites.lsa.umich.edu/deanyang/wp-content/uploads/sites/205/2016/09/abarcar-barua-yang-financial-education-financial-access.pdf