This presentation is part of the programme of the International Seminar "Social Protection, Entrepreneurship and Labour Market Activation: Evidence for Better Policies", organized by the International Policy Centre for Inclusive Growth (IPC-IG/UNDP) together with Canada’s International Development Research Centre (IDRC) and the Colombian Think Tank Fedesarrollo held on September 10-11 at the Ipea Auditorium in Brasilia.

1. AHORRO, CONTABILIDAD MENTAL, AHORRO POR DEFAULT Y TRANSFERENCIAS MONETARIAS CONDICIONADAS / DELIVERING CONDITIONAL CASH TRANSFERS VIA SAVINGS ACCOUNTS: DEFAULT AND MENTAL ACCOUNTING MECHANISMS*
Carlos Chiapa
El Colegio de México
cchiapa@colmex.mx
Silvia Prina
Case Western Reserve University
silvia.prina@case.edu
August 30, 2014
Project abstract
Our study aims to test whether and how mental accounting and default mechanisms improve the ability of poor households to save in the formal financial system, cope with negative shocks, and invest in health and education. We are also interested in testing whether making the financial lives of the poor easier has an effect on their cognitive system, and if this, in turn, affects their welfare. Designing savings tools that help the poor to save is a global challenge that could benefit many. The Mexican antipoverty program Oportunidades delivers conditional cash transfers to its beneficiaries via direct deposits into savings accounts. We plan to exploit this key feature of the program to evaluate the potential of default and mental accounting mechanisms for a vulnerable population. We will conduct a randomized controlled trial at the locality level with three treatment arms. Households living in control localities (C) will keep receiving their cash transfers via their savings accounts. Households living in our mental accounting localities (T1) will be offered access to an account for “emergencies.” They will be able to voluntarily deposit any amount into this account. Finally, households living in our mental accounting + default localities (T2) will also be offered access to an account for emergencies and will be given the chance to have 10% of their Oportunidades transfers automatically deposited into their account for emergencies. Comparison of C vs. T1 + T2 will identify the effect of mental accounting. Comparison of T1 vs. T2 will identify the effect of savings by default. Our results will help design better development policies associated with the delivery of conditional cash transfers programs or other government programs that make regular transfer payments to individuals.
* This project is part of, and has received generous financial support from, the Todas Cuentan (Everyone Counts) initiative, a project financed by the International Development Research Centre (IDRC) of Canada and implemented by the Economics and Business Department of the University of Chile, along with the Institute of Peruvian Studies (Instituto de Estudios Peruanos, IEP) and the Capital Foundation (Fundación Capital), and supported by the Ford Foundation and the IDRC of Canada. This project is also an Innovations for Poverty Action (IPA) project, funded by the Citi IPA Financial Capability Research Fund supported by the Citi Foundation. We also thank the Research Department of the Inter-American Development Bank (IDB) for financial support for this study.

2. 1
1 Context and target problem
Having access to the formal financial system helps the poor to escape poverty (Aghion and Bolton, 1997; Banerjee and Newman, 1993; Banerjee, 2004). Unfortunately, the poorest segments of the world’s population—beneficiaries of conditional cash transfer (CCT) programs included—lack access to the formal financial system. Research, however, suggests that the poor have some ability to save and use informal savings instruments, which have been shown to be more expensive and/or less efficient (Banerjee and Duflo, 2007; Collins, Morduch, Rutherford, and Ruthven, 2009; Karlan and Morduch, 2010). Recent studies show that there is untapped demand for formal saving devices and that access to a savings account can have a dramatic impact on the savings, investment and welfare of small entrepreneurs (Dupas and Robinson, 2013a) and households (Prina, 2014). This, in turn, may translate into an increased volume of domestic savings. These savings could be intermediated by the formal financial system and flow into productive investments. Hence, improving the financial access of the poor is a challenge that has yet to be achieved globally and that could benefit many.
Nowadays, in Latin America and the Caribbean alone, 22 million poor households are beneficiaries of a CCT program (Maldonado, Moreno-Sánchez, Giraldo, and Barrera, 2011). Some programs have started depositing their transfers directly into the individual beneficiaries’ bank accounts. Since January 2012, all six million beneficiaries of the Mexican CCT program Oportunidades receive their transfers directly into their own savings accounts at the Banco del Ahorro Nacional y Servicios Financieros (BANSEFI). Through this action, most of the poorest Mexican households are incorporated into the formal financial system. Still, beneficiaries make little use of their accounts, withdrawing most of their funds at once. Hence, saving a fraction of the transfer for rainy days could help them improve their management of their scarce resources and, eventually, get them out of poverty.1
Furthermore, mental accounting and default mechanisms may improve the ability of poor households in developing countries to save, cope with negative shocks, and invest in health and education. Such mechanisms have proven to be very powerful in developed countries. Research for the U.S. has shown that mental accounting can affect savings and financial accounting decisions (Feldman, 2010; Sahm, Shapiro, and Slemrod, 2010; Thaler, 1990, 1999). Defaults have also been shown to be very effective in increasing savings in developed countries (Thaler
1 Previous research has shown that it has been difficult for CCT recipients to get out of poverty (Campos-Vazquez, Chiapa and Santillán, 2013).

3. 2
and Bernatzi, 2004; Madrian and Shea, 2001). These mechanisms, unfortunately, have not been available for most of the poor, who do not receive regular payments in a bank account.2
2 Research questions
The intervention intends to test the effectiveness of mental accounting and default mechanisms for poor households in developing countries. Direct deposit of CCTs into a savings account provides a unique opportunity to do so. Working with the Banco del Ahorro Nacional y Servicios Financieros (BANSEFI)—the bank managing all the savings accounts of Oportunidades beneficiaries—we will implement a randomized controlled trial to address two questions. We aim to rigorously estimate the effects that:
(1) having an account denominated for emergencies in which individuals can voluntarily deposit some savings (mental accounting), and
(2) having an account denominated for emergencies where by default 10% of the CCT beneficiaries receive is regularly transferred (mental accounting + default)
have on savings at BANSEFI, on overall household savings, on the ability of households to cope with negative shocks, on investments, cognitive function, stress and mental health.
Our study is the first one to take advantage of direct deposits of CCTs into savings accounts to explore the potentially beneficial effects of these two behavioral mechanisms for poor beneficiary households. Recent research has shown that even a simple metal box labeled for health emergencies, that allows the poor to save for future health expenditures, can have a big impact on savings and health expenditures of households. Dupas and Robinson (2013b) explain that this is an effect of mental accounting: when money is deposited in the box, it becomes mentally allocated to cover health expenses only. In our study, labeling a formal savings account for emergencies could have a greater impact, as the risk of “losing” the money saved is less than when it is kept in a metal box inside the home. Our study will shed light onto this. Moreover, previous studies (Bernatzi and Thaler, 2004; Madrian and Shea, 2001) have shown that enrollment into a savings plan that, by default, deposits a small fraction of the wages of workers in a specific savings account, increases their savings. Thus, providing poor households with the ability to save by default could be a powerful tool to help them save. To our knowledge, our study will be the first to measure the effect of default saving mechanisms for the poor.
2 Previous research has shown the power of defaults in developing countries in other contexts, such as chlorine dispensers in Kenya (Kremer et al., 2009).

4. 3
Finally, while the poor face the same amount of risk—or even more in relative terms—as the rich, they do not have access to the many tools and mechanisms—among them, defaults, mental accounting, reminders, automatic deposits—the rich have access to. The use of such tools and mechanisms facilitate their decisions and improve their outcomes (Mullainathan and Shafir, 2009). Furthermore, poverty, by itself, creates additional psychological burdens (Mani, Mullainathan, Shafir, and Zhao 2013; Shah, Mullainathan, and Shafir, 2012). Mani et al. (2013) suggest that the human cognitive system has limited capacity. As such, the more preoccupations an individual has, e.g. due to pressing budgetary concerns, the fewer cognitive resources she will have to guide choice and action in important areas of her life such as investing in productive activities—be it in her business, her children’s education, etc. Indeed, recent evidence from lab experiments in the U.S. suggests that there may indeed be a link between poverty, cognitive resources, and economic returns.3
Hence, not only the poor lack access to these tools, but also, these tools could have a greater impact on their lives than the impact they have on the lives of the rich. Consequently, we hypothesize that savings and welfare of households receiving the mental accounting treatment should be higher than the ones of control households. In addition, the savings and welfare outcomes of households receiving the mental accounting + default treatment should be higher than the outcomes of the households solely receiving the mental accounting treatment.
3 Target population and partner organizations
Our target population is beneficiary households of Oportunidades that receive their cash transfers via a savings account and have been given a debit card to manage their funds.4 These beneficiaries live in urban, semi-urban, and rural areas neighboring urban areas. In order to reduce BANSEFI’s costs of opening and managing the accounts for emergencies, and to reduce the data collection costs, our focus is on localities in the central part of Mexico that have between
3 Evidence from lab experiments conducted in a U.S. college shows that people with low socio-economic status are highly vulnerable to short-term stress, which worsens their productivity and behavior, resulting in income losses as high as 30%. These effects seem to operate through a temporary cognitive impairment (Angelucci and Cordova 2013) and are consistent with a broader literature that documents how stress and poor mental health affects cognitive function and working memory (McEwen and Sapolsky 1995). Hence, it is relevant to test whether and to what extent these findings can be generalized to a real-life setting with poor households.
4 Only poor households are eligible to receive the program’s benefits. According to baseline data, monthly food consumption is MXN$1,037 per household, equivalent to USD$80. Average households size is on average 5.5. Hence, daily per capita consumption is less than USD$1 a day.

5. 4
120 and 622 Oportunidades beneficiaries.5
Data collected by Oportunidades in collaboration with the Instituto Nacional de Salud Pública (INSP) in 2012 allowed us to get relevant stylized facts about our target population at the time and to get relevant parameters for the design of our experiment. In particular, the 2012 data shows that, on average, our target population had been receiving the program for 4.3 years and had been given their debit cards 1.3 years ago. In 97% of the cases the recipient of the cash transfers—and owner of the debit card—was a woman. Female recipients had an average age of 40 and an average of 6.8 years of schooling. In 71% of the cases, they were the sole users of the card in their households. Prior to getting their transfers deposited in a savings account—and getting their debit card—the vast majority of the sample did not have any relationship with formal banks. Only 3.3% report having received job-related payments through a bank even though 97% report having a job. Furthermore, about three quarters of the women reported that they did not know basic characteristics of the savings accounts in which they received their CCTs, such as withdrawal fees (INSP, 2012).
Since April 2012, we have been working on this project jointly with two partner organizations: BANSEFI and Oportunidades. Both organizations understand the importance of the randomized nature of our study. BANSEFI is a Mexican Federal Government bank. It is responsible of distributing the monetary transfers of Oportunidades. Our second partner organization is Oportunidades. It helped us finalizing our sample localities and shared with us their lists of beneficiaries within each locality. Oportunidades has also committed to not making changes in the way our final sample of beneficiaries receives their payments for the duration of our study.
4 Study design and research methodology
We will use a randomized-controlled trial design to study the effects on savings of (1) having an account denominated for emergencies in which individuals can voluntarily deposit some savings, and (2) having an account denominated for emergencies where, by default, a certain amount is regularly transferred.
Our intervention has not started yet, but all Oportunidades beneficiaries in all our sample localities have already attended a first educational session (September/October 2013) on the
5 Our sample localities are located in the following states: Distrito Federal, Hidalgo, México, Michoacán, Morelos, Puebla and Veracruz.

6. 5
benefits of savings and a second educational session (November/December 2013) on why is it important to save for emergencies, what are the advantages of saving in a formal financial institution, the fact that they already have an account at a formal financial institution (BANSEFI), and some tips about the advantages of saving in a formal financial institution like BANSEFI and about the transactions they can carry out. Specifically, it should have been remarked that their money is protected by the Federal Government; that it is available 24/7 via ATMs; that the cardholder is the only person that can access the account, knows the balance of the account and can make withdrawals; that not having the money at hand may allow the beneficiaries to reach their savings goals faster; that they can make deposits into their accounts, for free; and that they can withdraw that deposited money the same way they withdraw their transfers from Oportunidades. We designed the content of the material for the two educational sessions and shared it with Oportunidades.6 Oportunidades was solely responsible for diffusing this material among all its beneficiaries.
From the end of April to the end of May 2014 we conducted a baseline survey on a subsample of beneficiaries in all sample localities. The survey was supposed to start in January, but was delayed due to delays in the process of signing the grant agreements from our two main funding institutions (Universidad de Chile – “Todas Cuentan” and Innovations for Poverty Action) and, consequently, in the reception of funds needed to cover the cost of the baseline survey and its monitoring.
We randomized our sample localities into one of three treatment arms so that all beneficiary households in the same locality are subject to the same treatment. Randomization at this level reduces concerns about contamination across treatments—i.e., the interaction between households in two different treatment arms in the same locality—affecting the nature of the experiment. The treatments are as follows:
Control C: Oportunidades beneficiaries keep receiving their transfers via their regular savings account and keep using their debit card to manage their funds. BANSEFI staff conducts a special educational workshop on how to use their regular account and provides each beneficiary with a booklet (see Figure 1) to help her track her savings.7
6 The material is based on Jaime Ramos Duffaut and Rita Carrillo Robles. 2012. “Una mujer que ahorra es una vida que cambia vidas: Serie de módulos para el desarrollo de capacidades financieras ‘Diva, la ahorrativa’,” Instituto de Estudios Peruanos - IEP.
7 In contrast to the educational sessions provided in September/October and November/December 2013, the emphasis of these special educational workshops is on how to make a deposit into their existing accounts, how to

7. 6
Treatment T1: Oportunidades beneficiaries keep receiving their transfers via their regular savings account and keep using their debit card to manage their funds. In addition, they get access to an account for emergencies. They can voluntarily make deposits of any amount in this account. Money in both accounts is fully liquid and can be withdrawn at any time. BANSEFI staff conducts a special educational workshop on how to use their regular account and their account for emergencies and provides each beneficiary with a two booklets (see Figures 1 and 2) to help her track her savings in both accounts.8
Treatment T2: Oportunidades beneficiaries keep receiving their transfers via their regular savings account and keep using their debit card to manage their funds. In addition, they get access to an account for emergencies. Moreover, they can agree to have 10% of their CCT deposited by default in the account for emergencies. The 10% transfer will happen automatically every time Oportunidades makes a transfer to the beneficiary’s account. Money in both accounts is fully liquid and can be withdrawn at any time. BANSEFI staff conducts a special educational workshop on how to use their regular account and their account for emergencies and provides each beneficiary with a two booklets (see Figures 1 and 2) to help her track her savings in both accounts.9
The comparison of C vs. T1 + T2 identifies the effect of mental accounting. The comparison of T1 vs. T2 identifies the effect of saving by default. Specifically, as the random assignment to the three treatment arms ensures that the treatments are independent of all other determinants of our outcomes of interest (e.g., savings),10 in order to estimate the effects of (1) mental accounting and (2) default, we will estimate two different sets of the following regression equation:
use their savings booklet and how to keep track of all their transactions using the savings booklet. We also suggest that saving a 10% of the Oportunidades transfer may be good in case an emergency happens, and show how to compute the 10% of a given amount.
8 The content of the special educational workshop for Treatment 1 is similar to the one for the Control group. The main difference is that we introduce the account for emergencies. We also suggest that saving a 10% of the Oportunidades transfer may be good in case an emergency happens. The two booklets the beneficiaries receive are identical in terms of their content—the only difference is the cover and the name of the booklet. We use the booklet associated with the account for emergencies when explaining how to use the savings booklets and how to keep track of all their transactions using their savings booklets.
9 The content of the special educational workshop for Treatment 2 is similar to the one of the other two groups. The main difference is that now we explain that saving has become easier: it is done automatically. Every time Oportunidades makes a transfer, 10% of the transferred amount gets deposited into their new account for emergencies. As in the case of Treatment 1, we use the booklet associated with the account for emergencies when explaining how to use the savings booklets and how to keep track of all their transactions using their savings booklets.
10 We will use our baseline data to check that our sample is balanced across treatment arms. In particular, we will check that average characteristics in treatment (T1 and T2) and control (C) localities are similar, conditional on locality fixed effects.

8. 7
Yim =α + βjTjm+ λXim + εim
where Yim is an outcome (e.g., overall savings in BANSEFI) for beneficiary i in locality m. Tjm, with j = 1, 2, is an indicator such that: T1m = 1 if beneficiary i belongs to treatment groups T1 and T2, and 0 if it belongs to the control group C; and T2m = 1 if beneficiary i belongs to treatment group T1, and 0 if it belongs to treatment group T2. Xim is a set of beneficiary household and locality characteristics. The variable εim represents unobserved determinants of Yim. β1 and β2 identify the effect of mental accounting and default, respectively, on the studied outcomes.
5 Data
5.1 Sources and information collected
The data will come from two sources. The first source consists of bank administrative data of all account transactions. This will enable us to compute and compare savings rates and account usage rates among C, T1 and T2 groups, as well as take-up rates of the accounts for emergencies offered to each treatment group. The second source consists of a baseline survey conducted from the end of April to the end of May 2014, before the intervention, and one endline survey to be conducted a year after the start of the intervention. The surveys were carried out on a subsample of Oportunidades beneficiaries in each locality. The surveys were designed to measure household socio-demographic characteristics, financial literacy, time and risk preferences, assets, access and amount saved in the formal and informal financial system, credit, vulnerability to shocks and ability and mechanisms to cope with them, consumption, investments in health and education, and bargaining power within the household. The surveys also measure the beneficiaries’ cognitive function (Raven’s matrices, digit recall), stress (self-reported), and broader indicators of mental health (life events, depression, locus of control). The institution we worked with to conduct the surveys is INSP—the same institution that has collaborated with Oportunidades to conduct a number of surveys.
5.2 Power calculations and baseline data collection results
For the power calculations, we made the following assumptions: a power of 0.8 and a

9. 8
significance level of 0.05; an intra-cluster correlation of 0.006;11 and an average take-up rate of the savings account denominated for emergencies and savings by default of 25% (based on take- up rates of other savings accounts in developing countries12). We considered as outcome variable self-reported savings in BANSEFI by Oportunidades beneficiaries who were given a debit card to manage their transfers and used the pre-existing data from INSP (2012). On average, beneficiaries reported having MX$12.48 (standard deviation MX$271.67) saved in their accounts. We assumed this to be the level of savings for the control group. Under the assumptions above, in order to detect a statistically significant difference of 0.2 standard deviations, considering 59 clusters (15 for the control group and 22 for each treatment group), we needed at least 20 complying households in each cluster.13
Considering the control group and assuming a 15% non-response rate at baseline, the number of households to be surveyed in each control locality was 24 (20/0.85). In addition, assuming a 15% attrition rate between baseline and endline, the number of households to be surveyed in each control locality was 29 (24/0.85 = 28.24 ≈ 29). Now, turning to the treatment groups, following the same reasoning and assuming a take-up rate of 25%, the number of households to be surveyed in each treatment locality was 113 (28.24/0.25 = 112.96 ≈ 113).
Thus, we planned to interview 435 beneficiaries in the control group (29*15) and 4,972 in the two treatment groups (113*22*2). Hence, our total to-be-interviewed sample was of 5,407 beneficiaries. The baseline data collection started on April 28 and ended on May 27, 2014. During fieldwork, an additional 162 replacements were added to the to-be-interviewed sample. With this, a total of 4,864 households were successfully interviewed at baseline. The average non-response rate was 12.7%, below the anticipated rate of 15%. And in each locality at the very least, the minimal sample size was reached.
6 Analysis of baseline data
Some stylized facts from the baseline survey are the following ones. Even though the accounts
11 This value was obtained using the loneway command in STATA. The data used are self-reported savings kept in BANSEFI by Oportunidades beneficiaries who were given a debit card to manage their funds. The data comes from INSP (2012).
12 Ashraf, Karlan and Yin (2006) find that for the Philippines, 28% of current bank’s clients opened an additional (commitment) account when offered one. Other studies offering a savings account to individuals that had no previous access find higher take up rates (e.g. Dupas and Robinson (2013a) find a 53% take-up rate in Kenya, Prina (2014) finds a 78% take-up rate in Nepal).
13 The number of 59 localities in the study was determined considering the maximum number of localities with a BANSEFI branch within 10 km and with a distance from other localities of at least 3 kilometers.

10. 9
they hold are savings accounts, 50% of the beneficiaries report not to save. Out of the 50% that does save, only 23% does it in their BANSEFI account. The rest of the beneficiaries (the 50% that does not save and the 77% that does save, but not in their BANSEFI account) prefer to withdraw their whole transfer as soon as the funds are available for withdrawal. That is, most of the beneficiaries choose not to use their BANSEFI accounts. In fact, 60% of the beneficiaries that do save, choose to do it informally: hiding money in a secret place at their homes or through ROSCAs.
A deeper analysis is being conducted to better understand why is it that the beneficiaries use their accounts so little to save. There are a number of hypothesis that could explain the lack of use of the accounts: high transaction costs, lack of trust in the financial system, lack of financial education, not finding the product offered attractive. For the time being, what we can say is that lack of financial education seems to be an issue despite the financial education sessions that we designed and Oportunidades supposedly imparted (apart from the own efforts of both Oportunidades and BANSEFI to financial educate this population).
Only 51% of the beneficiaries report to know that they can save in their BANSEFI accounts. Despite the fact that all beneficiaries regularly withdraw their transfers from their BANSEFI accounts, in general, they do not know how much it costs to withdraw their money. Further, 83% of the beneficiaries does not know how to deposit money into their accounts and 70% does not even know where they would have to go or what they would have to do in order to make a deposit in the event they wanted to make one. Also, 88% of the beneficiaries does not know that the money they save in BANSEFI is protected by the Federal Government. Nevertheless, 62% of the sample considers that keeping their money in a formal bank account does have an advantage over keeping it somewhere else. Thus, it does seem that the beneficiaries do not know or do not understand the product they have been given. Hence, if they could understand its characteristics better, the formal savings account at BANSEFI would be used more.
The statistics we have reported in the previous paragraph refer to self-reported knowledge of the way the accounts work. We also have information about more standard and objective ways to measure financial literacy. For example, we asked the beneficiaries to compute the 10% of MXN$1,250 and only 30% was able to do it. Then, we asked them to tell us how much they would owe, after three months, if they were to get a credit of MXN$100 at a monthly interest rate of 2%, and they had to pay the credit in full after those three months. In this case 43.2% correctly replied that they would have to pay back more that MXN$102. Finally, we asked them

11. 10
to consider a scenario in which prices are going up and they are keeping their money under the mattress. After that we asked them whether they considered that after a few months they could buy more, the same or less items with their money. In this case, 75% correctly replied that they would be able to buy fewer items.
The way the percentage of correct answers changes with these three questions suggests that as the financial concepts asked are more familiar to the beneficiaries, more are able to correctly understand them. Hence, it seems that more and better financial education could greatly benefit this population. In fact, it could stimulate the (educated) use of the savings account they already own at BANSEFI. On the other hand, offering more intuitive and easy to use products like the accounts we have designed could also help accomplish this objective.
7 Problems faced to date
As we have mentioned, our intervention was supposed to start on February 2014, immediately following the completion of the baseline data collection. The baseline data collection was supposed to start in January, but was delayed due to delays in the process of signing the grant agreements from our two main funding institutions (Universidad de Chile and Innovations for Poverty Action) and, consequently, in the reception of funds needed to cover the cost of the baseline survey and its monitoring. As a consequence, the intervention has also been delayed.
Both BANSEFI and Oportunidades have been kept updated about the change in the timeline of the project. BANSEFI has suffered changes in key positions and has been given direct others from the President of Mexico to design and implement an aggressive strategy to give credit to its current customers and to increase its customer base. Reportedly, due to the personnel changes and the presidential mandates, BANSEFI has not been able to start our intervention. We are hopeful the intervention will take place over the next few weeks.
An additional issue that we are currently facing is that, due to the depreciation of the Chilean Peso with respect to the US Dollar, the funds that El Colegio de México and the INSP have received so far—and must probably will receive in the future—are significantly lower than what we had budgeted in US Dollars. For example, the INSP was supposed to receive USD$47,000, while due to the depreciation, it received USD$40,795.66. This represents a 13% reduction.
8 Policy Impact
To date we know little about the potential benefits of delivering CCTs via savings accounts.

12. 11
Aside from increasing financial penetration for the poor, this new delivery method allows to introduce mental accounting and default mechanisms that have proven to work in increasing savings and welfare in the developed world and in other contexts. We will test whether this is the case for beneficiaries of Oportunidades, one of the largest CCT programs in the world.
Our results will help design better development policies associated with the delivery of CCTs programs or other government programs that make regular transfer payments to individuals. In addition, if our study demonstrates that our intervention has a significant impact, it may not be difficult to scale up the intervention and/or to replicate it in any country where the government makes transfer payments, on a regular basis, to poor individuals via a savings account.
References
Aghion, P., and P. Bolton. 1997. “A Theory of Trickle-Down Growth and Development.” Review of Economic Studies, 64(2), 151-172.
Angelucci, M. and Cordova, K. 2013. “Choice and Productivity Under Stress: Are Men and Women Different?” manuscript, University of Michigan.
Ashraf, N., D. Karlan, and W. Yin. 2006. “Tying Odysseus to the Mast: Evidence from a Commitment Savings Program in the Philippines.” Quarterly Journal of Economics, 121(2), 635-672.
Banerjee, A.V. 2004. “Contracting Constraints, Credit Markets and Economic Development.” In Advances in Economics and Econometrics: Theory and Applications, Eight World Congress, 3, eds. M. Dewatripont, L. P. Hansen, and S. Turnovsky, 1-46, Cambridge University Press.
Banerjee, A.V., and A.F. Newman. 1993. “Occupational Choice and the Process of Development.” Journal of Political Economy, 101(2), 274-298.
Banerjee, A.V., and E. Duflo. 2007. “The Economic Lives of the Poor.” Journal of Economic Perspectives, 21(1), 141–167.
Banerjee, A.V., and E. Duflo. 2011. Poor Economics: A Radical Rethinking of the Way to Fight Poverty. Public Affairs.
Campos-Vazquez, R., C. Chiapa, and A.S. Santillán. 2013. “Análisis de Trayectorias de los Hogares Beneficiarios del Programa Oportunidades.” El Trimestre Económico, 80(1), 77- 111.
Collins, D., J. Morduch, S. Rutherford, and O. Ruthven. 2009. Portfolios of the Poor: How the World’s Poor Live on 2 a Day. Princeton, NJ: Princeton University Press.

13. 12
Dupas, P., and J. Robinson. 2013a. “Savings Constraints and Microenterprise Development: Evidence from a Field Experiment in Kenya.” American Economic Journal: Applied Economics, 5(1): 163-92.
Dupas, P., and J. Robinson. 2013b. “Why Don’t the Poor Save More? Evidence from Health Savings Experiments,” American Economic Review, 103(4): 1138-71.
Feldman, N.E. 2010. “Mental Accounting Effects of Income Tax Shifting.” Review of Economics and Statistics, 92(1), 70-86.
Instituto Nacional de Salud Pública (INSP). 2012. “Cuestionarios de Medios de Pago para la Titular Beneficiaria de Oportunidades y Operadores.”
Karlan, D., and J. Morduch. 2010. “Access to Finance.” in Handbook of Development Economics, 5, eds. D. Rodrik y M. Rosenzweig, Amsterdam, Elsevier Science.
Kremer, M., E. Miguel, S. Mullainathan, C. Null, and A. Peterson Zwane. 2009. “Making Water Safe: Price, Persuasion, Peers, Promoters, or Product Design?” Unpublished.
Madrian, B.C. and D.F. Shea. 2001. “The Power of Suggestion: Inertia in 401(k) Participation and Savings Behavior.” Quarterly Journal of Economics, 116(4), 1149-1187.
Maldonado, J.H., R.P. Moreno-Sánchez, I. Giraldo, and C.A. Barrera. 2011. “Programas de Transferencias Condicionadas e Inclusión Financiera: Oportunidades y Desafíos en América Latina.” EnBreve, 22, Proyecto Capital.
Mani, A., S. Mullainathan, E. Shafir, and J. Zhao. 2013. “Poverty Impedes Cognitive Function.” Science, 341, 976-980.
McEwen, B., and Sapolsky, R. 1995. “Stress and cognitive function,” Current Opinion in Neurobiology, 5(2), 205-216.
Mullainathan, S., and E. Shafir. 2009. “Savings Policy and Decisionmaking in Low-Income Households.” In Insufficient Funds: Savings, Assets, Credit and Banking Among Low-Income Households, eds. M. Barr and R. Blank, 121–145. New York: Russell Sage Foundation Press.
Prina, S. 2014. “Banking the Poor via Savings Accounts: Evidence from a Field Experiment.” Unpublished.
Sahm, C.R., M.D. Shapiro, and J. Slemrod. 2010. “Check in the mail or more in the paycheck: Does the effectiveness of fiscal stimulus depend on how it is delivered?” NBER Working Paper No. 16246.
Shah, A.K., S. Mullainathan, and E. Shafir. 2012. “Some Consequences of Having Too Little.” Science, 338, 682-685.

14. 13
Thaler, R.H., and S. Bernatzi, 2004. "Save More Tomorrow: Using Behavioral Economics to Increase Employee Saving.” Journal of Political Economy, 112(1), S164-S187.
Thaler, R.H. 1990. “Anomalies: Saving, Fungibility, and Mental Accounts.” Journal of Economic Perspectives, 4(1), 193-205.
Thaler, R.H. 1999. “Mental Accounting Matters.” Journal of Behavioral Decision Making, 12, 183-206.

15. 14
FIGURE 1: COVER OF THE BOOKLET “EL LIBRITO DE MIS AHORROS”

16. 15
FIGURE 2: COVER OF THE BOOKLET “EL LIBRITO DE AHORRO PARA MIS EMERGENCIAS”