Paul Haguenauer, Valerie Ross, Gyuzel Zaripova Master IEP 2012 Saving Constraints and Microenterprise Development Evidence from a Field Experiment in Kenya Pascaline Dupas, Johnathan Robinson (2009)
Structure Background and aim Context and Framework Design, Data and Results Critical Assessment Policy Implications
Background & Aim General debate on eased credit vs. eased saving (Dis)Advantages of credit easing: Credit can be available immediately but requires saving to repay. Saving as a time influenced decision - consume today less to have more tomorrow? Present bias (in Banerjee, Duflo (2007): The Economic Lives of the Poor: issue of self-control and self-awareness). For micro-entrepreneurs: running investment or costs rely on selling activity. Yet, investment "lumpy", thus cash necessary but often not available. Only alternative save until enough working capital. Saving depends on interest rate r (lower r => longer saving period). Main finding of Dupas and Robinson (2009): poor willing to pay premium for safe saving conditions. Suggests bad saving conditions at home (theft, neighbours and family members asking for money).
Rough results Estimate of the treatment control difference amounted to 108 KSH (1.6 $US), thus a 40% increase in average investment, 4-6 months after opening the account (yet only 60% of women used the account in these months After 6 months, daily expenditures of women: 37-44% higher, food expenditures: 14-29% higher Some evidence of better health risk coping (less financial health induced risk shocks) Case for inefficient traditional saving opportunities consistent with other research: Aportela (1999): Effects of Financial Access on Savings by Low-Income People ( expansion of savings institute - increased average savings rate) and Burgess, Pande (2005): Do Rural Banks Matter? Evidence from the Indian Social Banking Experiment (expansion of banking program - significant decrease in rural poverty)
The experiment: free savings accounts with high withdrawal fees (thus a negative interest rate). Only if utility of saving at home is lower than depository saving, would the agents use the savings account. Context and Framework Rotating Savings and Credit Associations (ROSCAs) are already in place. Functioning: group loans money every 18 months to 1 group member. Members take turns at specified intervals. Disadvantage: not always timely availability of credit, because of their rotational character. Members not always entrepreneurs but may as well spend the sum on consumption. Alternative: secure saving at depository institutions. Questions: impact of health shocks on financialsituation reduced? What welfare implications? Saving incurring costs onto other factors? Model of utility functions: 1) output (Y) requires financial capital (K) and labor (L) 2) lump capital required for production 3) agents cannot borrow Yt= f(kt, Lt) at a given time t. Moment t-1: Stock depleted, decision on how to allocate wealth: cash-on hand for consumption (ct) or cash-on hand for investment (kt), or savings (st). In other terms: investment depends on output (which depends on capital and labor), and on the interest rate on savings minus the consumption costs and the sum saved.
Design The randomized evaluation took place in Bumala Town, in Busia District, Kenya Focus on non farm micro entrepreneurs: 169 individuals were sampled to open saving accounts, who were randomly devised into treatment and control groups Small scale operations: mean number of articles traded is below 2, mean daily investment is 6$ Treatment group was offered the possibility to open no cost saving accounts with minimum balance corresponding to 1.43$ Sampling took part in 2 waves: 2006 and 2007, the control group of the first wave became treatment Group in the second wave Take-up determinants identification,
Data 4 sources of data used: background survey, administrative data from the village bank, data on time and risk preferences and daily self reported logbooks Trained enumerators helped filling in logbooks correctly twice a week. Daily visits were illiterate individuals. 0.71$ offered monthly if logbooks were properly filled Logbooks provided information on income, expenditures, health, investment, labor supply and adverse income shocks Poor data on revenues disabled further analysis on profits Non-differential attrition during the logbook exercise will not bias estimates Pre-treatment differences are managed by controlling for gender, years of education, marital status, occupation, age, literacy and ROSCA contributions in all of regression specifications
Take- ups: Results - From 122 individuals offered the opportunity to open saving account 11% refused and 34% never used the account. - Women tend to depose larger amounts of money - Accounts considered "active" if at least one deposit was made during 6 month following the opening of the account - Take-up determinant identification by regressing "active" on baseline characteristics: men are less likely to take up the account. Possible explanation is ROSCA participation, as members are 26 points more likely to have active account, and ROSCA participation is much higher among women. Impact estimation strategy: estimating the effects of saving account on average investment, expenditures, transfers and other outcomes Two levels of interest for each outcome: the average effect of being assigned to the treatment group (ITT) and the average effect for those who actively use the account, using instrumental variable approach (IV)
Results: impact on productive investment and labor supply No evidence on labor supply, but sizable effects on daily amount of investment for women Because of the small size of the sample, impossible to estimate this effect separately for married and single women. Present biased women are less likely to use a savings account Significant effect of the accounts on food and private expenditures, which increase for women
Results The effects of the account No effect of the account on labor supply A sizable effect of the account on the average daily amount invested in the business Female market vendors sampled for the account are more likely to give credit to customers Robustness checks Large fraction of the sample purchased shares, which might have affected savings and investment decisions. The same analysis as in Table 3 after restricting the sample to those who didn't invest in shares gave results suggesting that the effects on productive investment and labor supply are not driven by the anticipation of loans. Effects on investment and expenditures should be larger for those women who withdrew money from their accounts. => effects are larger for all expenditure categories, but not for investment Given the median size of deposit, the need for saving facilities becomes questionable. Main regression excluding those who made large deposits check that the results are not driven by those who make large deposits
Results on Risk Coping To identify pathways by which savings are accumulated, transfer and expenditures in a given week are regressed on the profits made this week. => Women in the control group spend larger share of their income on consumption and private expenditures Risk coping Estimation strategy: test if the treatment improved individuals ability to resist to health shocks. therefore are estimated: o impact of weekly variations in health level on outcomes o the effect of the "active" account with having been assigned to the treatment group Results: - women without saving account reduce their labor supply when they are sick themselves, they invest less in weeks when they get malaria and increase medical expenditure in case of illness and decrease food expenditures=> working capital affected by malaria could be the reason of the lack of growth of microenterprises - women in the treatment group maintain their labor supply and investment levels when they are sick, because they were better able to afford medical expenses
Critical Assessment Assessment of Data Noisiness of data on business investment Unreliability of data on revenues, sales levels (don't have time) with the consequence that there are no reliable profit figures 88 by men and 97 by women Standard errors on both ITT and IV estimates are large BUT food and private expenditures increased significantly for sampled women Data Reliability Exclusion of those who already had an account and those who declared not being interested Difficult reliability of logbooks: lying, impossibility to read or write (enumerators), a high number of participants did not follow the logbook for the whole 3 months Many individuals refused to keep the book but similarity between 2 Financial incentive might have biased the filling in of the notebook External Validity The noisiness of the results and the restricted sample for the data may compromise the external validity of the results
Critical Assessment o The impact of the experiment is lessened by the fact that savings account was significant for women and not for men. o No possibility of treating the heterogeneity effect because of the small size of the experiment (married/ unmarried)
Critical Assessment ROSCA FSA Crowding out effect? Not seen with the regression in the household. Treated women transfer more to their spouse, and investment and expenditure are increasing at some cost to the larger social network Take-up is higher with the population already using ROSCA Suggests that women face demands on their income from their extended family rather than just their husband ROSCA comprise credit, insurance against risk and social commitment, not with savings account, therefore imperfect substitutefor ROSCA Animal savings are protected from inflation when they are not at the bank Present-biased women tend to use credit to commit themselves to saving rather than using savings accounts.
Critical Assessment Anticipation of loans FSA mechanism, have to buy share to be able to loan. Only 3,2% got loans in the first 6 months but 16% bought shares. Did the women invest more because of the anticipation of the loan? Excluding them from the data does not change the results of the regression, so rather no. But this reduces the sample size and increases the size of standard errors. Withdrawals The effects are larger for the women who made withdrawals But the investment is similar between women who did and did not. Cost in time and efforts to go to the bank (deposits as well): women save up small amounts and withdraw big sums. Some unanswered questions Self control problem or social pressure? Intra-household or extra-household causes for savings constraints? Why did they not take up actively the accounts : lack of information or inadapted? Finally do small businesses have the potential to grow? Returns high enough?
Policy Implications Building on existing institutions that rely on peer pressure and social commitment to ensure trust in a financial service. Lowering the price and accessibility (opening hours, distance) of savings institutions would push towards saving which would create a larger output - or at least a higher consumption as seen in this experiment. Could be achieved, for example through mobile banking solutions. There is a need for information about the utility of these savings accounts for the people Savings account fosters credit both informal and formal In order to target efficiently the savings constraints we need to identify if the constraints are internal or external to the household. Could solve the classical critic of microcredit institutions accused of being too slow and too reliant on external capital.