Source: UNIV OF MINNESOTA submitted to
IMPROVING LIVELIHOODS FOR AGRICULTURAL WORKERS IN MALAWI THROUGH INCREASED ACCESS TO SAVINGS
Sponsoring Institution
National Institute of Food and Agriculture
Project Status
TERMINATED
Funding Source
Reporting Frequency
Annual
Accession No.
1010016
Grant No.
(N/A)
Project No.
MIN-14-164
Proposal No.
(N/A)
Multistate No.
(N/A)
Program Code
(N/A)
Project Start Date
Oct 1, 2016
Project End Date
Sep 30, 2021
Grant Year
(N/A)
Project Director
Kerwin, JA.
Recipient Organization
UNIV OF MINNESOTA
(N/A)
ST PAUL,MN 55108
Performing Department
Applied Economics
Non Technical Summary
In this project, I will evaluate two savings products for agricultural workers in rural Africa. The first product is simply the option to receive one's pay through a mobile money service that allows people to transfer money electronically using their phones. The second product enhances the first one, by allowing workers to defer the income they put into the account until the end of the growing season. This "deferred wages" product is popular among workers but rarely offered by employers, and allows people to save money in a context where traditional bank accounts are rare or unavailable, and where informal saving is costly or risky. We will study the demand for these two products, by measuring their uptake and utilization, and their benefits, by examining how access to them shapes workers' savings, expenditure, and labor supply.
Animal Health Component
0%
Research Effort Categories
Basic
0%
Applied
100%
Developmental
0%
Classification

Knowledge Area (KA)Subject of Investigation (SOI)Field of Science (FOS)Percent
60661203010100%
Goals / Objectives
For this research project, I propose to use data from an RCT that I will conduct at the Lujeri Tea Estate to examine the effect of improved access to savings on workers' livelihoods and on the investments they choose to make. The improved savings options arethe ability to defer part of their wages to the end of the growing season (deferred wages, or DW)the ability to receive part of their pay through mobile money accounts (mobile money, or MM)As discussed above, DW is an extremely popular but underexploited savings tool in this context. An initial pilot at Lujeri found that nearly 70% of people who were offered DW took it up, and that they saved an average of 15% of their income in the accounts.[1] The goal of the study is to estimate the effect of DW on savings, expenditure, and labor supply, and to understand the mechanisms behind its impacts. MM offers potentially-useful insights into the latter: if the impediment to savings is a very negative effective interest rate, or having my assets and income be observable to my social network, then MM should have similar takeup and impacts to DW. If the problem is that people are tempted into spending money on things they don't really want, or tempted into sharing with relatives, then DW should have much larger effects - sophisticated agents can use it as a commitment device to avoid spending money until they really want to.I will examine the impact of the two savings technologies on four major outcomes:1.. Uptake and utilization. Using data from the mobile money provider and the tea estate, we can see what percentage of workers utilize each of the savings options and how much money is saved in each. This will help us understand what uptake of such a savings product would look like in the future.2. Expenditures and saving. We will use survey data to measure the effect of access to each savings option on expenditures over time and savings amounts.[2] Savings will be measured using a standard instrument that captures all the savings vehicles normally used at the tea estate, in addition to the MM and DW options introduced by the project.[3] We will also measure the degree of trust workers have for the various savings options they have access to, including ROSCAs, formal saving, and deferred wages through their employer. Our trust measure will follow Chuang and Schechter (2015) and will also include the elicitation of subjective expectations about losing money, following Delavande, Gine, and McKenzie (2011). We will use trust in the main savings option being employed as one of our outcome variables of interest.3. Lumpy investments. Our survey will also ask about purchases of key lumpy investments in the local area such as bicycles, roof sheeting, school fees, and subsidized fertilizer.4. Labor supply. Lujeri collects data on the daily attendance and output of each worker in order to compute their wage payments. We will study the effect of the savings options on these administrative measures of labor supply, which is a first in the literature. The DW treatment could increase labor supply through two distinct channels. First, the lumpy nature of certain purchases means that having $100 on a single day is more valuable than having $10 on ten different days. By making it easier to save up wages into a useful sum, DW should make working more attractive. Second, workers can utilize the DW option as a commitment device to get themselves to work more: if they commit a share of their wages to be deferred until the end of the season, then they have to work harder now to achieve their desired current consumption level.[4] On the other hand, DW could also reduce labor supply through a couple of different mechanisms.My project would thus evaluate the potential of these savings technologies, both by measuring their effects on livelihoods and by measuring the demand for them.[1] This is much higher than the takeup and utilization of other products in The Dupas and Robinson (2013) study of bank accounts in Kenya found uptake of 40% and an average savings of about 6% of revenue.[2] All workers in the study are paid the same piece rate, irrespective of study arm (it is fixed across the whole firm).[3] All three study arms will be asked about each savings option, even if it was not offered to them (e.g. control workers will be asked about DW and MM).[4] DW could also affect labor supply directly, through a motivation effect. To address this and other such Hawthorne effects (where just being in the treatment arm causes effects, rather than the actual intervention), we will explore extensions where we offer DW again in future seasons. If the effects persist then Hawthorne effects are not a likely explanation.
Project Methods
The data will come from an RCT that will take place during the 2016-2017 main growing season at the Lujeri estate. The RCT will randomly assign workers to receive one of the two savings options or to be in a control group. In order to equalize access to the cell phone and mobile money systems, all workers will be provided with a SIM card that is linked to a mobile money account.[1] The RCT will enroll all (consenting) tea pluckers employed by Lujeri, yielding a total sample of about 3000 workers, or 1000 per study arm. We will conduct several rounds of surveys on our sample: a baseline, a short-term followup just after the end of the season, and a long-term followup several months later. The short-term followup will take place after the deferred lump-sum payments have been made, to measure changes in the composition of expenditure. The long-term followup will allow us to examine any lasting effects. Each survey will cover formal and informal savings, asset ownership, expenditure levels and composition, labor supply, and purchases of key lumpy investments. The baseline will also include demographic information.Statistical ApproachWe will supplement the surveys with three forms of administrative data. First, the project's administrative records will show how much of their paycheck each worker has saved in the scheme. Second, we will collect usage data from the mobile money operator, showing us balances and withdrawals over time. Third, we will utilize Lujeri's administrative data on workers' attendance and output to measure labor supply.Our analyses will compare the three study arms in a regression framework. We will regress the outcome on indicators for each of the two treatment variables and a vector of controls X:Outcome = a + b*OfferedMobileMoney + c*OfferedDeferredMobileMoney + X ′d + uThis will yield consistent estimates of the effect of each savings option on the outcome of interest. The only exception is in studying uptake (outcome 1), which is defined only within each study arm. Since the study arms are randomly assigned, the uptake and utilization of each savings option within the relevant arm of the study is a consistent estimate of the company-wide demand for the product. Thus for this outcome we will examine simple means. For MM, we will also estimate a hazard function of withdrawals over time, to understand the determinants of how workers deplete their savings.For labor supply, we will analyze the data at the worker-day level, exploiting our access to a unique administrative dataset on attendance and output by day. These regressions will allow us to control for determinants of output that vary at the daily level, and also to control for worker fixed effects by exploiting data that is collected prior to the treatment. This substantially enhances our precision for studying labor supply.Power CalculationsWe conducted a small pilot of the planned intervention during the current (2015-2016) harvest season, to work out logistical details and confirm its feasibility. This pilot, which included 50 treated workers and 150 untreated ones, did not use random assignment, but the exact workers offered the pilot deferred wages product were selected idiosyncratically. We can therefore run a difference-in-differences estimator to get a sense of the treatment effects and precision we can expect in the full study. For labor supply, we estimate a treatment effect of 4.4 KGs of tea picked per day (against a mean of 53 KGs), with a standard error of 2.6 KGs. Expanding our sample to 2000 workers (1000 for each study arm) will reduce our standard errors by a factor of 3.2. This will give us 80% power[2] to detect an effect of 2.3 KGs/day, which is about 4% of output. For treatment effects as large as those detected in the pilot, we have nearly 100% power. Our other outcomes are even more significant in the pilot, and so our power will be even better for those outcomes.Potential CollaborationsI am working on this study with co-PIs Lasse Brune of Yale University and Eric Chyn of Brown University. Our principal collaborator on the project is the Lujeri Tea Estates, a large agricultural firm in southern Malawi. We are also partnered with Innovations for Poverty Action (IPA) on the research implementation and data collection aspects of the study; IPA funded the data collection for the study through their Financial Inclusion Program fund. For the mobile money aspect of the study, we are considering a partnership with TNM, a local mobile operator.There are several possibilities for future collaborations on this project. The most logical group to reach out to are the other members of the Tea Association of Malawi, an trade group made up of all the tea companies in Malawi's Southern Region. Their employees work in similar conditions and face similar financial challenges. More broadly, our findings might be of interest to organizations that study and support agricultural development in the developing world, such as IFPRI or the FAO. Our context - wage workers at an agricultural firm - will be increasingly relevant across the developing world as economic growth and development leads to consolidation in the agricultural sector.[1] Mobile money accounts can be accessed without a phone, by using one's SIM card with a friend's phone or one that belongs to one of the mobile money "agents" that allow withdrawals from the system.[2] Power is the probability that a statistically-significant effect can be detected, assuming the treatment effect is actually a given size. It depends on how precisely treatment effects are estimated, for which sample size is of primary importance.

Progress 10/01/16 to 09/30/21

Outputs
Target Audience:We have two target audiences for this work: policymakers who are designing interventions to help people in rural Africa, and social scientists two are interested in improving livelihoods financial inclusion among the rural poor. We have had very substantial success reaching both audiences this year. For the first audience, our work was featured on VoxDev, a website for disemminating research about development policy. Innovations for Poverty Action is also hosting a seminar about our results and potential applications to cash transfers, workfare programs, and payroll systems for companies. Both of these policymakers focused on improving livelihoods learn about our approach, as a possible way to raise the savings rates for their beneficiaries. We are also working on the design of a project with GiveDirectly to scale up our intervention with their cash recipients, and exploring the possibility of doing something similar with the Government of Indonesia. In terms of reaching the second audience, our paper has been published at the American Economic Review, the premier journal in the field of economics. The final version - which acknowledges NIFA's support - was published online and in print in July. Changes/Problems:As noted in previous progress reports, we were unable to study the impacts of the second savings option, mobile money (MM). What opportunities for training and professional development has the project provided?We are working with a graduate student on follow-on work that uses the data from this project. It focuses on using the data on worker's daily output to measure agricultural labor supply elasticities via bunching estimators, and is the subject of my new MAES project that started this year. How have the results been disseminated to communities of interest?As noted above, our work was featured on VoxDev and is also the subject of an online seminar run by Innovations for Poverty Action. Both of these channels target policymakers who are interested in improving the livelihoods of people in developing countries, especially those in rural areas. What do you plan to do during the next reporting period to accomplish the goals? Nothing Reported

Impacts
What was accomplished under these goals? We achieved all the major goals set out for the project, with one exception: we were unable to implement the planned MM arm due to logistical constraints at the partner firm. We completed the planned fieldwork, collecting data on nearly 900 workers who were interested in the savings product and tracking them for over three years. Half of them were enrolled in the savings scheme. The savings scheme was very popular and led to a substantial increase in short-term savings, which ended up leading to the accumulation of additional assets and workers making improvements to their homes. We presented the results from the study twenty times, including at leading conferences such as the AEA meetings in 2019 and the NEUDC meetings in 2017. We also presented our work at institutions that include Yale, the Philadephia Federal Reserve, and the University of Washington. In 2021 our paper was accepted by the American Economic Review, the leading economics journal in the world. It was published in the July issue of the journal. We also circulated the paper as an NBER working paper. The paper has already been cited ten times. We have also worked to publicize our findings in other ways. Our results were featured on VoxDev and shared by the World Bank's Development Impact blog. Innovations for Poverty Action is hosting a panel this December to discuss our findings and how they might be useful in other settings. To achieve the long-run goals of improving policy based on our findings, we are in talks with GiveDirectly and Inclusive Financial Innovation Initiative about ways to incorporate the "Pay Me Later" approach into other payment systems.

Publications

  • Type: Journal Articles Status: Published Year Published: 2021 Citation: Brune, L., Chyn, E., & Kerwin, J. (2021). Pay Me Later: Savings Constraints and the Demand for Deferred Payments. American Economic Review, 111(7), 21792212. https://doi.org/10.1257/aer.20191657


Progress 10/01/19 to 09/30/20

Outputs
Target Audience:We have two target audiences for this work: policymakers who are designing interventions to help people in rural Africa, and social scientists two are interested in improving livelihoods financial inclusion among the rural poor. We have had very substantial success reaching both audiences this year. For the first audience, our work has been included in a publication by the Abdul Latif Jameel Poverty Action Lab (J-PAL) about commitment savings products for people in developing countries. This will help people involved in designing workfare or cash transfer programs in rural areas learn about our approach, as a possible way to improve the savings rates and well-being of their beneficiaries. In terms of reaching the second audience, we submitted our paper for publication last year, and now it is revise-and-resubmit at the American Economic Review, the premier journal in the field of economics. We recently submitted a revised version of the paper to the journal; if accepted, this would pay major dividends in terms of the paper's impact. Changes/Problems:As noted in previous progress reports, we were unable to study the impacts of the second savings option, mobile money (MM). What opportunities for training and professional development has the project provided?I have not yet involved graduate students on this project, but we have ideas for follow-on work to improve the measurement of assets in rural communities in the developing world, that would exploit the same data collected for this project. I think that a follow-on study would be a good opportunity for involving a graduate student, who could gain hands-on experience doing research. How have the results been disseminated to communities of interest?In previous years, we have presented this work at academic conferences. It has also been picked up by prominent development economics newsletters and blogs, such as the World Bank's Development Impact blog and the J-PAL newsletter. This year, our results are being featured in a publication on commitment savings products being put out by J-PAL. What do you plan to do during the next reporting period to accomplish the goals?Our plans for next year are focused on completing the publication process for the first paper. This will likely involve another round of revision at the American Economic Review; if the paper is not accepted there, we will send it to other journals, starting with the American Economic Journal: Applied Economics. Another goal for next year is to begin the aforementioned follow-up study on improving asset measurements. Rural households hold the majority of their wealth in assets and livestock, and measurements of this wealth are key outcomes in many studies of efforts to improve livelihoods, such as the famed "graduation" programs (Banerjee et al. 2015, Science). Asset measurements are also used as proxies for consumption, food security, and economic growth. Our dataset contains unique data where self-reported assets (collected away from the home) are matched with in-person validations by our survey enumerators. This provides us with the opportunity to assess the quality of various methods of measuring physical assets and livestocks in rural Africa, and investigating their implications for previous evidence on anti-poverty programs.

Impacts
What was accomplished under these goals? Last year we completed the data collection for the project and wrote up the paper. This year, it received a revise-and-resubmit from the American Economic Review, the leading economics journal in the world. We received the decision on July 12th, and significantly overhauled the paper based on the comments from the editor and the two referees. The paper is greatly improved as a result. We turned in the revision on December 8th.

Publications

  • Type: Other Status: Under Review Year Published: 2020 Citation: Brune, Lasse, Eric Chyn, and Jason Kerwin. "Pay Me Later:Savings Constraints and the Demand for Deferred Payments." Working Paper


Progress 10/01/18 to 09/30/19

Outputs
Target Audience:We have been very successful at reaching our target audience of policymakers and social scientists who are interested in promoting livelihoods and financial inclusion among the poor. This year we presented the project at three major academic conferences: the American Economic Association conference in January, the Centre for the Study of African Economies conference in March, and the Northeast Universities Development Consortium conference in October. I also presented the results in an academic seminar at the University of Washington in Seattle. Beyond those presentations, the results have also been disseminated in fora that are more targeted toward policymakers. They were picked up by the World Bank's Development Impact blog in October and also included in the december edition of J-PAL's newsletter, which has 30,000 subscribers. Changes/Problems:As noted in previous progress reports, we were unable to study the impacts of the second savings option, mobile money (MM). What opportunities for training and professional development has the project provided? Nothing Reported How have the results been disseminated to communities of interest?As summarized above, we have disseminated the results to policymakers and researchers in several forms. We also previously presented these results to the firm where we conducted the experiment. What do you plan to do during the next reporting period to accomplish the goals?Next year we plan to continue the publication process for the paper. This will depend on how things go at initial journals; if rejected at the American Economic Review, we would next try the American Economic Journal: Applied Economics, then the Review of Economic Studies, and then the Journal of Development Economics. I will also work to continue to disseminate our findings. I plan to promote them on my own personal blog as well as Twitter, and to submit them to several other public-facing blogs such as VoxDev. A key goal is to get these ideas out to people in control of payment systems, such as employers, but also governments who run social support and cash transfer programs. Finally, I aim to work toward finding other firms who would be interested in trying out thistechnology. Our previous attempt to partner with Beyonic, a payroll company, did not pan out because we targeted the wrong grant. I would like to try again with them and also look for potential partners in the US or other developed countries.

Impacts
What was accomplished under these goals? This year we completed data collection for this project and found a set of very exciting long-term results. Repeated exposure to the intervention leads to permanent increases in asset holdings, in the form of improved roofing materials (metal roofs) on houses. This is a major savings goal for the workers in our sample and carries multiple benefits - metal roofs last longer than grass roofs and have much lower maintenance costs. They keep people warmer and drier during the rainy season. This result also serves as a proof of concept: deferred wage payments can help people not just save in the short run but grow wealthier in the long run. An important open question is whether we can replicate this finding in other populations. Based on these results, we have finished a heavily-revised draft of the paper which we have begun submitting to academic journals. The paper has been under review at the American Economic Review since October. (Note that, as described in a previous progress report, we are now only studying the deferred wages savings product; the mobile money accounts were not a viable option for our partner firm.)

Publications

  • Type: Conference Papers and Presentations Status: Other Year Published: 2019 Citation: Pay Me Later: Savings Constraints and the Demand for Deferred Payments (with Lasse Brune and Eric Chyn). Presented at: - American Economic Association 2019 Annual Meeting. Atlanta, GA. - Centre for the Study of African Economies 2019 Conference. Oxford, UK. - North East Universities Development Consortium Conference. Evanston, IL. - University of Washington Department of Economics. Seattle, WA.
  • Type: Other Status: Other Year Published: 2019 Citation: Work featured in the following blogs and newsletters: - Development Impact Blog (October 2019) - J-PAL Newsletter (December 2019)


Progress 10/01/17 to 09/30/18

Outputs
Target Audience:The target audience reached during the current reporting period has included social scientists who are studying livelihoods in rural Africa and other parts of the developing world, as well as those who study financial inclusion in the US and other developed countries. In the past year we have presented the work at one conference and in seven invited research seminars. I gave seven of those nine presentations myself. The audiences have included development economists, African studies experts, researchers studying the Chinese economy, and business school professors. At a previous conference (the Financial Inclusion Program conference), we were connected with Beyonic, a payment services organization that manages payroll for hundreds of firms and over 100,000 employees in the developing world, and are in talks with Beyonic to roll out the deferred wages savings plan to their clients. The interest in our research from this wide range of audiences suggests that our savings product could be a low-cost way to improve livelihoods for agricultural workers and other low-wage employees around the world. Changes/Problems:As described in our previous progress report, there were logistical barriers to rolling out the alternate savings product (MM). We completed a small pilot study in which we tested the feasibility of using the product, but the transaction fees were too high to permit a broader rollout and so a randomized evaluation was not possible. Moreover, the capacity of the mobile money company to actually manage the payroll process was tenuous at best. Fortunately, MM was the less-important of the two products we wanted to study, as it has already been studied in other contexts. We have conducted no additional testing of MM in the current reporting period. What opportunities for training and professional development has the project provided? Nothing Reported How have the results been disseminated to communities of interest?As noted above we have presented our results at a wide range of academic conferences and invited seminars. We have also shown our findings to the management of the firm. What do you plan to do during the next reporting period to accomplish the goals?Our plan is to revise our paper to include the results of our supplementary experiments and also the second wave of the randomized trial, and then submit the revised paper to academic journals.

Impacts
What was accomplished under these goals? We previously completed an initial randomized trial of DW (which we are now calling "Pay Me Later" as described above), which included several waves of data collection. We measured the uptake and utilization of the product and its impact on expenditures, lumpy investments, and labor supply. During the current reporting period, we ran a set of supplementary experiments in which we varied different components of the savings product to test which factors are the main drivers of demand for the product. We also conducted a second wave of the randomized trial, in which we re-treated the original treatment group two additional times. We will begin collecting outcome data for this second wave in December 2018; this will let us measure the longer-term impacts of the product. Based on the data we have collected so far, we have completed a draft paper, which we have presented at numerous academic conferences and invited seminars during the current reporting period.

Publications

  • Type: Conference Papers and Presentations Status: Other Year Published: 2018 Citation: - Hong Kong University. Hong Kong, China. - Hong Kong University of Science and Technology. Hong Kong, China. - Peking University. Beijing, China. - China Agricultural University. Beijing, China. - Midwest International Economic Development Conference. Minneapolis, MN. - Federal Reserve Bank of Philadephia. Philadelphia, PA. - University of Minnesota African Studies Initiative. Minneapolis, MN. - University of Utah. Salt Lake City, UT.


Progress 10/01/16 to 09/30/17

Outputs
Target Audience:The target audience reached by this research so far has been both policymakers and social scientists who are studying livelihoods in rural Africa and other parts of the developing world. We have presented the work at several conferences already. At one of those conferences (the Financial Inclusion Program conference), our work was seen by Beyonic, a payment services organization that manages payroll for hundreds of firms and over 100,000 employees in the developing world, and they are very interested in the savings product we developed. We are in talks with Beyonic to roll out the deferred wages savings plan to their clients. If successful, this could become a standard feature of payroll systems in low-income, financially-excluded locales. Our hope is that the Beyonic rollout will help us reach fair trade activists as well, since our savings product could be a low-cost way to improve livelihoods in the developing world, especially for agricultural workers. Changes/Problems:As noted above, we were unable to study the impacts of the second savings option, mobile money (MM). What opportunities for training and professional development has the project provided? Nothing Reported How have the results been disseminated to communities of interest?Yes, we have presented preliminary results at four conferences so far and also shown them to the management of the firm. As noted above, a payroll-processing company, Beyonic, is planning to roll out Pay Me Later to their clients, including 100+ firms and 100,000+ workers. We are in talks with them about evaluating that rollout. What do you plan to do during the next reporting period to accomplish the goals?We plan to finish our research paper and submit it to academic journals, and also to partner with Beyonic to study the potential of Pay Me Later in other settings.

Impacts
What was accomplished under these goals? We successfully completed a randomized trial of DW (which we are now calling "Pay Me Later" as described above), and several waves of data collection. We have measured uptake and utilization of the product and its impact on expenditures, lumpy investments, and labor supply. We have completed a draft of a paper on our results which we are presenting at academic conferences. There were logistical barriers to rolling out the alternate savings product (MM). We completed a small pilot study in which we tested the feasibility of using the product, but the transaction fees were too high to permit a broader rollout and so a randomized evaluation was not possible. Moreover, the capacity of the mobile money company to actually manage the payroll process was tenuous at best. Fortunately, MM was the less-important of the two products we wanted to study, as it has already been studied in other contexts.

Publications

  • Type: Conference Papers and Presentations Status: Other Year Published: 2017 Citation: Pay Me Later: A Simple Employer-based Saving Scheme (with Lasse Brune and Eric Chyn). Presented at: - North East Universities Development Consortium Conference. Medford, MA. - Innovations for Poverty Action, Researcher Gathering on Financial Inclusion. New Haven, CT. - Consumer Financial Protection Bureau. Washington, DC. - Yale University Department of Economics. New Haven, CT.