The Effects of Financial Inclusion on Children s Schooling, and Parental Aspirations and Expectations

The Effects of Financial Inclusion on Children s Schooling, and Parental Aspirations and Expectations Carlos Chiapa Silvia Prina Adam Parker El Colegio de México Case Western Reserve University Making Impact Evaluation Matter Better Evidence for Effective Policies and Programmes September 3, 2014

Introduction The poor have low access to the formal financial system They end up using imperfect substitutes (Collins et al. 2009; Rutherford 2000) Access to the formal financial system helps the poor to escape poverty (Aghion and Bolton 1997; Banerjee and Newman 1993; Banerjee 2004) Thanks to the financial benefits this access may bring Empirical literature on the impact of formal financial inclusion of the poor has mainly focused on the financial consequences of this inclusion on their lives (see Karlan and Morduch 2010)

Introduction Financial inclusion may bring changes in the lives of the poor that go beyond financial outcomes The more expensive and/or less efficient the set of financial resources, the harder the successful management of finances Lower sense of ability to control financial future (and life) Lower aspirations Lower expectations (Suboptimal) life choices

Introduction Access to the formal financial system may enable individuals to achieve a more stable and certain financial life Perceived ability to control financial life may increase Higher aspirations Higher expectations (Closer to optimal) life choices

What do we do? We exploit a unique field experiment in which Prina randomly offered a basic savings account to poor women in Pokhara, Nepal

What do we do? We exploit a unique field experiment in which Prina randomly offered a basic savings account to poor women in Pokhara, Nepal We show: 1 Most of the sample lacked access to the formal financial system 2 The accounts were accepted and used 3 Perceived ability to control financial life increased 4 Educational attainment of girls increased 5 Parental educational aspirations towards daughters increased 6 Parental educational expectations towards daughters increased

The field experiment Access to the formal financial system in Nepal is low (20%) Concentrated in urban areas and among the wealthy Prina, working with GONESA Bank, offered savings accounts to a random sample of poor households in 17 slums surrounding Pokhara, Nepal s second largest city In may 2010, before offering the accounts, a baseline survey was conducted with the women, aged 18-65, running each household 1,236 households were surveyed Separate public lotteries were held in each slum: 626 households were assigned to the treatment group and were offered the option to open a savings account at the local bank-branch office The remaining households were assigned to the control group and were not offered (could not open) the savings account

The savings accounts After completion of the baseline survey, GONESA Bank progressively began operating in the slums between the last two weeks of May and the first week of June 2010. The accounts have all the characteristics of any formal savings account: No opening, maintenance, or withdrawal fees No minimum balance requirement Nominal yearly interest rate of 6% (inflation = 10.5%) Local bank branches open twice a week for three hours; main office in downtown Pokhara open M-F during regular business hours

Data We use data from baseline and endline Baseline conducted right before the intervention in May, 2010 Household composition, education, income, income shocks, monetary and non-monetary asset ownership, borrowing, and expenditures on durables and non-durables Endline conducted one year after the beginning of the intervention Same modules as baseline Household perceptions of their ability to manage their financial daily lives, parental aspirations and expectations regarding each of their children s education

Sample and balance check Our sample consists of: 661 children aged 11-16 at endline 510 households where these children live We are interested in the children (and their households) in age to attend secondary school when many drop out Enrolment for children age 6-10: above 93% Enrolment for children age 11: 88% Enrolment for children age 16: 71% In general, our sample is well balanced

Balance check Table 1A: Descriptive Statistics at Baseline by Treatment Status Mean Obs. Sample Control Treatment T-stat 1 Characteristics of the woman running the household Age 510 37.35 37.45 37.25-0.29 (6.68) (6.86) (6.52) Years of education 416 2.26 2.21 2.31 0.56 (2.55) (2.45) (2.64) Characteristics of the spouse of the woman running the household Age 467 42.27 42.03 42.49 0.55 (8.22) (7.91) (8.50) Years of education 459 4.67 4.51 4.81 0.88 (3.40) (3.25) (3.52) Household demographic characteristics Household size 510 5.03 5.07 4.99-0.82 (1.63) (1.64) (1.63) Number of children 10-15 years old 493 1.36 1.37 1.36-0.21 (0.70) (0.67) (0.72) Number of female children 10-15 years old 493 0.70 0.73 0.67-0.93 (0.70) (0.72) (0.69) Number of male children 10-15 years old 493 0.66 0.64 0.68 0.76 (0.67) (0.65) (0.68) Characteristics of children 10-15 years old at baseline Average age of children 10-15 years old 672 12.50 12.53 12.48-0.39 (1.66) (1.64) (1.68) Average age of female children 10-15 years old 346 12.62 12.63 12.62-0.06 (1.70) (1.69) (1.71) Average age of male children 10-15 years old 326 12.38 12.41 12.35-0.36 (1.61) (1.57) (1.65) Average level of education of children 10-15 years old 667 6.26 6.30 6.22-0.44 (2.19) (2.09) (2.29) Average level of education of female children 10-15 years old 345 6.66 6.62 6.70 0.34 (2.04) (1.91) (2.16) Average level of education of male children 10-15 years old 322 5.83 5.93 5.74-0.70 (2.27) (2.22) (2.31)

Balance check Table 1B: Descriptive Statistics by Treatment Status Mean Obs. Sample Control Treatment T-stat 1 Household financial characteristics Total income last week 2 510 1,503.87 1,648.11 1,372.60-0.64 (5,011.10) (5,302.78) (4,736.28) Experienced a negative income shock 510 0.44 0.42 0.46 1.11 (0.50) (0.50) (0.50) Total net worth (assets - liabilities) 2 510 9,013.73 5,563.41 12,153.91 0.74 (136,656.40) (153,317.90) (119,703.30) Assets 2 510 48,698.86 42,762.08 54,102.00 1.76* (61,039.01) (45,743.50) (71,862.33) Proportion of households with money in a bank 510 0.15 0.15 0.15 0.19 (0.36) (0.36) (0.36) Proportion of households with money in a ROSCA 510 0.19 0.19 0.20 0.21 (0.39) (0.39) (0.40) Proportion of households with money in an MFI 510 0.56 0.58 0.54-0.71 (0.50) (0.50) (0.50) Proportion of households giving money to someone 510 0.03 0.02 0.04 1.01 trusted to safekeep (0.17) (0.14) (0.20) Proportion of households with cash at home 510 0.96 0.96 0.95-0.99 (0.21) (0.20) (0.22) Non-monetary assets from consumer durables 2 510 26,697.35 25,997.33 27,334.46 0.64 (25,758.86) (26,355.43) (25,236.36) Non-monetary assets from livestock 2 510 5,285.91 4,720.26 5,800.72 1.07 (12,419.06) (12,064.19) (14,546.52) Liabilities 510 39,685.13 37,198.67 41,948.09 0.50 (121,916.80) (140,471.20) (102,369.90) Proportion of households with outstanding loans 510 0.91 0.91 0.91 0.27 (0.29) (0.29) (0.28)

Empirical strategy We estimate the average treatment effect Y i of having been offered the savings account one year after the intervention Y is =α 0 + α 1 ITT is + X isα 2 + λ s + ε is Y is ITT is X is λ s ε is Outcome variable of household or child i in slum s Dummy equal to 1 if household was offered the account Vector of baseline characteristics (age and years of education of the account holder; number of children 16 years old or less; number of household members; household s assets; and most relevant source of household income dummies) Slum fixed effects Idiosyncratic error term Standard errors clustered at the slum level

Take-up and usage of the savings accounts offered At baseline, 15% of households had a bank account At endline, 84% of treated households accepted the account Of these, 93% used it actively Usage stats Average transactions: 46 (3 withdrawals; 43 deposits) Average deposit Rs. 137 (9% of weekly income at baseline)

Effects of offering the savings account on wealth Data very noisy: nothing is statistically significant Assets Liabilities Net worth Not clear whether households benefited financially from being offered (taking-up, and using) the accounts

Perceived ability to manage financial life Still, being offered the savings account might allow the households to feel more in control of their financial lives At endline, three questions were asked aimed at measuring the households perceived financial situation: 1 How would you describe your household s financial situation? 1 if live comfortably or meet basic expenses with little left for extras 0 if just meet basic expenses or don t even have enough to meet basic expenses 2 How financially stretched your household is, month to month? 1 if not very stretched or not at all stretched 0 if stretched to the absolute limit, very stretched or somewhat stretched 3 On the whole, I feel secure with the financial situation of my household 1 if strongly agree or agree 0 if feel neutral, disagree, or strongly disagree. What do we find?

Perceived ability to manage financial life Table 3: Effects on the Household's Self-Reported Financial Situation How would you describe your How financially stretched is your On the whole, I feel secure with the household's financial situation? household, month to month? financial situation of my household 1 if live comfortably, or meet basic 1 if not very stretched, or expenses with little left for extras. not at all stretched. 1 if strongly agree, or agree. 0 if just meet basic expenses, or don't even have enough to meet basic 0 if somewhat stretched, very stretched, 0 if feel neutral, expenses. or stretched to the absolute limit. disagree, or strongly disagree. (1) (2) (3) (4) (5) (6) ITT: Offered the 0.088* 0.072** 0.107** 0.082** 0.015-0.002 Savings Account (0.050) (0.033) (0.039) (0.035) (0.038) (0.032) Constant 0.267*** 0.659*** 0.230*** 0.172 0.206*** 0.169* (0.037) (0.185) (0.051) (0.122) (0.034) (0.092) Controls No Yes No Yes No Yes Obs. 510 510 510 510 510 510 R 2 (overall) 0.009 0.316 0.014 0.376 0.000 0.237 Note: Robust standard errors, clustered at the slum level, reported in parenthesis. Controls include age and years of education of the female head of the household, number of children 16 years old or younger, number of household members, assets at baseline, main source of income dummies, slum dummies. The dummies for the main source of household income are: sales of agricultural production, agricultural labor, sales of livestock and poultry, sand and stone collection labor, construction labor, driver, bus fare collector, helper, small shop, garnment and wool spinning, jewelry, government job, teacher, pension, rent, remittances, alcohol making, other full time job, other part-time job, other income source. Regressions with controls also include indicators for whether the years of education of the female head of the household, or the number of children, or the main source of household income were unavailable. These households are assigned the median values at the slum level of these variables. Statistically significant coefficients are indicated as follows: *10%; **5%; ***1%.

Children s school completion, and parental educational aspirations and expectations At endline we asked: School completion What is the education level that each child has completed? Aspirations What is the highest education level that you would like your child to complete? Expectations Given your child s ability and your household s economic condition, what is the highest education level that you think this child will actually complete? We estimate the impact of being offered the savings account on children s school completion, and on educational aspirations and expectations parents have for their children

Children s school completion Table 4a: Children's School Completion All Children Sons Daughters (1) (2) (4) (5) (7) (8) ITT: Offered the 0.133 0.102-0.355-0.298 0.569** 0.426* Savings Account (0.181) (0.161) (0.231) (0.236) (0.265) (0.230) Constant 6.830*** 2.852** 6.918*** 4.604*** 6.750*** 2.611* (0.136) (1.153) (0.166) (1.594) (0.207) (1.535) Controls No Yes No Yes No Yes Obs. 661 661 312 312 349 349 R 2 (overall) 0.001 0.156 0.008 0.199 0.016 0.239 Note: Robust standard errors, clustered at the household level, reported in parenthesis. Controls include age and years of education of the female head of the household, number of children 16 years old or younger, number of household members, assets at baseline, main source of income dummies, slum dummies. The dummies for the main source of household income are: sales of agricultural production, agricultural labor, sales of livestock and poultry, sand and stone collection labor, construction labor, driver, bus fare collector, helper, small shop, garnment and wool spinning, jewelry, government job, teacher, pension, rent, remittances, alcohol making, other full time job, other part-time job, other income source. Regressions with controls also include indicators for whether the years of education of the female head of the household, or the number of children, or the main source of household income were unavailable. These households are assigned the median values at the slum level of these variables. Statistically significant coefficients are indicated as follows: *10%; **5%; ***1%.

Parental educational aspirations Table 4b: Parental Educational Aspirations All Children Sons Daughters (1) (2) (4) (5) (7) (8) ITT: Offered the 0.404 0.468-0.361-0.176 1.089** 1.238*** Savings Account (0.339) (0.308) (0.428) (0.424) (0.448) (0.388) Constant 14.656*** 14.987*** 15.306*** 14.281*** 14.073*** 15.565*** (0.245) (1.890) (0.302) (2.822) (0.333) (2.129) Controls No Yes No Yes No Yes Obs. 661 661 312 312 349 349 R 2 (overall) 0.003 0.259 0.003 0.287 0.022 0.353 Note: Robust standard errors, clustered at the household level, reported in parenthesis. Controls include age and years of education of the female head of the household, number of children 16 years old or younger, number of household members, assets at baseline, main source of income dummies, slum dummies. The dummies for the main source of household income are: sales of agricultural production, agricultural labor, sales of livestock and poultry, sand and stone collection labor, construction labor, driver, bus fare collector, helper, small shop, garnment and wool spinning, jewelry, government job, teacher, pension, rent, remittances, alcohol making, other full time job, other part-time job, other income source. Regressions with controls also include indicators for whether the years of education of the female head of the household, or the number of children, or the main source of household income were unavailable. These households are assigned the median values at the slum level of these variables. Statistically significant coefficients are indicated as follows: *10%; **5%; ***1%.

Parental educational expectations Table 4c: Parental Educational Expectations All Children Sons Daughters (1) (2) (4) (5) (7) (8) ITT: Offered the 0.229 0.189-0.566-0.443 0.942** 0.779* Savings Account (0.324) (0.311) (0.436) (0.424) (0.410) (0.400) Constant 13.431*** 12.996*** 14.190*** 15.553*** 12.750*** 10.859*** (0.241) (2.071) (0.322) (2.920) (0.303) (2.062) Controls No Yes No Yes No Yes Obs. 661 661 312 312 349 349 R 2 (overall) 0.001 0.188 0.006 0.268 0.019 0.249 Note: Robust standard errors, clustered at the household level, reported in parenthesis. Controls include age and years of education of the female head of the household, number of children 16 years old or younger, number of household members, assets at baseline, main source of income dummies, slum dummies. The dummies for the main source of household income are: sales of agricultural production, agricultural labor, sales of livestock and poultry, sand and stone collection labor, construction labor, driver, bus fare collector, helper, small shop, garnment and wool spinning, jewelry, government job, teacher, pension, rent, remittances, alcohol making, other full time job, other part-time job, other income source. Regressions with controls also include indicators for whether the years of education of the female head of the household, or the number of children, or the main source of household income were unavailable. These households are assigned the median values at the slum level of these variables. Statistically significant coefficients are indicated as follows: *10%; **5%; ***1%.

Wealth effects The offer increased net worth (effect not statistically significant) Maybe this increase is behind our results, such that ITT NW PFS/SC/ASP/EXP We include net worth at endline in the regressions of: PFS Magnitude of coefficient of ITT decreases a bit, but remains statistically significant SC Coefficient of ITT decreases a bit, but remains statistically significant ASP Coefficient of ITT not affected EXP Coefficient of ITT not affected

Empowerment of the women account holders Maybe the savings accounts are empowering the account holders they are more in control of the household s finances This could probably explain the differential changes in SC, ASP, and EXP We check whether PFS, SC, ASP, and EXP are correlated with two proxies for the level of bargaining power women may have within the household: 1 Difference in schooling between wives and husbands (whole sample) 2 Money in savings accounts as a fraction of assets at baseline (treatment group only) None of these proxies seems to be correlated with PFS, SC, ASP, and EXP

Increase in the availability of cognitive resources Scarcity (Shah, Mullainathan, and Shafir 2012): Changes how people allocate attention Leads to engage more deeply on some problems (bring food to the table), than others (children schooling) Poverty impedes cognitive function (Mani, Mullainathan, Shafir, and Zhao 2013) Preoccupations with pressing budgetary concerns will leave fewer cognitive resources to guide choice and action At baseline, our population was facing scarcity in a number of key dimensions: 1 Scarcity of economic resources 2 Scarcity of formal financial alternatives to manage them 3 (Probably) scarcity of cognitive resources for general tasks in life Preoccupations caused by (1) and (2) may be behind (3)

Increase in the availability of cognitive resources Offering the savings accounts may have reduced scarcity of: Formal financial alternatives Economic resources Evidence suggests that treatment households may have less pressing needs, and fewer preoccupations with pressing budgetary concerns. In fact: Households perceive to be financially better Treatment households are eating better (more meat and fish) Probably more cognitive resources are becoming available for treatment households Unfortunately we do not have the data to test this (research in Mexico)

Discussion We find that offering a savings account: Increases the perceived ability to control one s financial situation Increases the the educational attainment of girls, and parental educational aspirations and expectations for these girls It is not clear what is driving our results 1 Evidence is not consistent with wealth or empowerment effects 2 Evidence may be consistent with an increase in the availability of cognitive resources Financial inclusion seems to have effects beyond financial outcomes