The ABCs of Financial Literacy

Transcription

1 The ABCs of Financial Literacy Experimental Evidence on Attitudes, Behavior, and Cognitive Biases Fenella Carpena (UC Berkeley) Shawn Cole (Harvard Business School) Jeremy Shapiro (Give Directly) Bilal Zia (World Bank) IPA Conference, Thailand, 2012

2 Motivation A combination of technology, financial liberalization and government/private sector initiatives have made financial products widely available An estimated 2 billion people worldwide will enter the formal financial system in the next 20 years (World Bank, 2008)

3 Motivation A combination of technology, financial liberalization and government/private sector initiatives have made financial products widely available An estimated 2 billion people worldwide will enter the formal financial system in the next 20 years (World Bank, 2008) Yet, not clear that these people will be equipped to make optimal financial decisions

4 Why Sub-Optimal Financial Choices?

5 Why Sub-Optimal Financial Choices? Not a user-friendly system financial products often complex and difficult to evaluate Calculating interest rates (flat or declining; compounding) Anticipating liquidity needs, inflation, etc. Comparing multi-dimensional products

6 Why Sub-Optimal Financial Choices? Not a user-friendly system financial products often complex and difficult to evaluate Calculating interest rates (flat or declining; compounding) Anticipating liquidity needs, inflation, etc. Comparing multi-dimensional products Inter-temporal trade-offs sometimes difficult Self control problems (Ashraf, Karlan and Yin, 2006) Difficulty understanding compound rates (Stango and Zinman, 2009) Other behavioral concerns: procrastination, channel factors

7 Can Financial Literacy Help? Compelling survey evidence from the developed world shows strong positive correlation. HHs with low financial education: Tend not to plan for retirement (Lusardi and Mitchell, 2007a) Borrow at higher interest rates (Lusardi and Tufano, 2008; Stango and Zinman, 2006) Acquire fewer assets (Lusardi and Mitchell, 2007b) Participate less in the formal financial system (Alessie, Lusardi and van Rooij, 2007; Hogarth and O.Donnell, 1999). Evidence from the developing world shows similar correlations (Cole, Sampson and Zia, 2011; Klapper and Lusardi, 2012)

8 Global Interest in Financial Literacy US: President s Advisory Council on Financial Literacy Indonesia: 2008 was Year of Financial Education India: RBI has established Financial Literacy and Credit Counseling Centers International and Private organizations are also pushing heavily for financial literacy programs: World Bank: $15 Million Russia Financial Literacy Trust Fund Citi Foundation: 10-year $200 Million global program on financial education, operating in 65 countries

9 But is there a Causal Relationship? Survey-based inference is difficult unobserved factors Observational evidence mixed Bernheim, et. al. (2003) find some effect, Cole and Shastri (2010) find no effect Experimental evidence shows only modest effects Cole, Sampson and Zia (2011)

10 Why No/Small Effects? Financial education may not be effective Behaviors difficult to change Generic courses may not be relevant, interesting or informative to individuals Skilled and engaging educators difficult to find Difficult to teach adults anything

11 Why No/Small Effects? Financial education may not be effective Behaviors difficult to change Generic courses may not be relevant, interesting or informative to individuals Skilled and engaging educators difficult to find Difficult to teach adults anything Measuring change may be difficult Basic surveys may not pick-up changes Administrative data typically not available Lack of statistical power to detect small changes

12 This Paper (1) Research design to overcome both these problems: Video-based delivery high quality, well produced, engaging content with local celebrities and ordinary people as actors Large sample size 1200 observations, randomized at the individual level. This design enables us to detect even small effects

13 This Paper (2) Research design to identify entire causal chain of impact: Financial literacy intervention changes in financial knowledge changes in financial behavior But how to measure changes in financial knowledge? We identify and test three distinct dimensions of financial knowledge: (1) Numeracy Skills, (2) Basic Financial Awareness, and (3) Financial Attitudes and Perceptions

14 This Paper (3) Research design to identify precise channels through which financial literacy may be effective: Pay for performance on exit test Angrist and Lavy (2010) find strong effects on test performance in Israeli schools Financial counseling individualized financial advice Financial goal setting Mullainathan and Shafir (2010) argue opening up of simple channels can facilitate behavior change

15 Experiment Design Study over several waves To accommodate large sample size To accommodate improvements in survey instrument Details of main intervention: One-third (treatment): Financial literacy videos One-third (treatment): Financial literacy videos + Counseling One-third (control): Health literacy videos Additional interventions: ½ of treatment: Financial goal setting ½ of treatment and control: Pay for performance

16 Sample Financial Literacy 1/3 Financial Literacy + Counseling 1/3 Health 1/3 Goal Setting 1/2 No Goal Setting 1/2 Pay for Performance on Video-related Questions 1/2 Pay for Performance on non-video-related Questions 1/2

17

18

19 What Does Our Sample Look Like? HH Characteristics Ave HH size: 6, Ave Monthly Income: Rs ($150) 84% has a phone, 26% have a non-farm enterprise, 83% has water connection Respondent Characteristics Ave Age: 39 57% female, 98% married Limited schooling: 49% completed elementary school, 4% completed secondary school 48% Saath MFI Clients

20 Analysis Step 1: Impact on Financial Knowledge Financial education has very limited role in equipping individuals to evaluate complex financial tradeoffs that require high numeracy skills Even individuals provided with financial incentives for their performance did not perform any better on financial numeracy questions

21 Analysis Step 1: Impact on Financial Knowledge Financial education has very limited role in equipping individuals to evaluate complex financial tradeoffs that require high numeracy skills Even individuals provided with financial incentives for their performance did not perform any better on financial numeracy questions Financial education leads to significant improvements in individuals' awareness of financial products and services

22 Analysis Step 1: Impact on Financial Knowledge Financial education has very limited role in equipping individuals to evaluate complex financial tradeoffs that require high numeracy skills Even individuals provided with financial incentives for their performance did not perform any better on financial numeracy questions Financial education leads to significant improvements in individuals' awareness of financial products and services Financial education significantly changes respondents' attitudes towards purchasing and recommending financial services and financial planning tools

23 Analysis Step 2: Impact on Financial Behavior We have positive treatment effects on real behavior (what a relief!) All treatments combined have the most significant effect (Financial Literacy + Counseling + Goal Setting)

24 Analysis Step 2: Impact on Financial Behavior We have positive treatment effects on real behavior (what a relief!) All treatments combined have the most significant effect (Financial Literacy + Counseling + Goal Setting) Strong effects on: Budgeting (and thinking budgeting is important) Bank accounts or other forms of savings (post office, chitti fund, etc) Real saving amounts Less likely to borrow to pay for unforeseen expenses

25 Analysis Step 2: Impact on Financial Behavior We have positive treatment effects on real behavior (what a relief!) All treatments combined have the most significant effect (Financial Literacy + Counseling + Goal Setting) Strong effects on: Budgeting (and thinking budgeting is important) Bank accounts or other forms of savings (post office, chitti fund, etc) Real saving amounts Less likely to borrow to pay for unforeseen expenses No effect on consumption

26 Conclusion Comprehensive analysis of the entire causal chain of financial literacy Evidence on the types of knowledge that can be effectively conveyed Impact on actual behavior (yay!) Stitching together causal chain is important for policy so we understand mechanism, not just end outcome

27 Transnational Household Finance: A Field Experiment on the Cross-Border Impacts of Financial Education for Migrant Workers Ganesh Seshan Georgetown University Dean Yang University of Michigan, BREAD, & NBER 1

28 Remittances vs. ODA, FDI (Billions of US$, ) Source: Data up to 2008 are from World Development Indicators Data are in billions of current US$, in total across developing countries (low & middle income as classified by World Bank). Variables displayed are: Net official development assistance and official aid received (current US$), Foreign direct investment, net inflows (BoP, current US$), and Workers' remittances and compensation of employees, received (current US$) data compiled from World Bank Migration and Remittances Factbook, OECD, and UNCTAD. 2

29 Motivating ideas and questions The microeconomics of financial decision-making in transnational households is poorly understood Insights into transnational household decision-making can help suggest innovative development policies Much interest in harnessing migrants (and their remittances) for development goals But decentralized nature of remittances poses major challenges Interventions need to be focused on individual knowledge and capabilities 3

30 Transnational household finance Household financial management that faces complexities associated with Extended separation from important income earners Management of international remittances Oftentimes, substantial increases in household income Concern: in combination with low levels of education of many migrants, may lead to sub-optimal financial decision-making 4

31 Some basic questions Do transnational households make high-quality financial decisions? And are there ways in which their decision-making can be improved? Do physically separated members of transnational households attempt to make joint decisions on financial matters? Or do such households operate as independent economic units? If transnational households do seek to make coordinated financial decisions, how constrained is their ability to do so? In particular, is intra-household asymmetric information an important constraint? 5

32 What we do We randomly assigned invitations to a savings-focused financial literacy workshop for migrant workers in Qatar who are from Kerala, India Via surveys of migrants as well as their wives remaining behind in India, we provide a unique window into financial decision-making in transnational households We exploit the intervention to examine impacts on Financial decision-making of the migrants Migrants attempts to influence the financial decision-making of their wives in the home country Migrant beliefs about their wives behaviors Wives actual behaviors 6

33 Summary of results Treatment led to substantial changes in migrant financial practices and savings goals Migrants with below-median baseline savings are most responsive to the treatment Migrants own savings and the remittances sent to wives rise substantially Migrants and wives commonly engage in joint financial decision-making, and the treatment increases this propensity Evidence of asymmetric information in transnational households In the low-baseline-savings subsample, positive impact of treatment on migrant report of wife s savings in India, but no corresponding impact on wife s report of same Likely due to treatment-induced upward updating by migrants of wives savings 7

34 The experiment After baseline, migrants randomly assigned to treatment with 2/3 probability Treatment: invitation to financial literacy workshop at local hotel in November 2010 Invitation by phone Three-hour workshop, plus two-hour dinner 41% of invitees attended the workshop Financial literacy workshop led by K. V. Shamsudheen Head of Pravasi Welfare Trust, NGO serving migrants from Kerala in Gulf Sample: married male migrant workers from Kerala, India whose wives have remained behind in India N=200 migrants, 200 wives 8

35 KVS in action 9

36 KVS s opening slide 10

37 KVS motivating question 11

38 Financial literacy workshop contents Unifying theme: importance of setting financial goals to ensure long-term financial security once the migrant returns home permanently Specific financial literacy topics covered: Creating and following a budget, for both migrant and the household in India Consulting entire family in financial planning Setting aside money from remittances to save Different types of investment options Secondary topics covered: time management; maintaining a positive attitude; good work ethics; healthy living 12

39 Analysis of treatment impacts Because those attending workshop in treatment group may be different from those not attending, simply compare entire treatment group with entire control group Intent-to-treat (ITT) effects Make comparisons separately for subsamples defined by: Low (at or below median) baseline savings High (above median) baseline savings

40 Impact on KVS workshop attendance 45% 40% 41.3% 35% 30% 25% 20% Treatment Control 15% 10% 5% 2.7% 0% Low savings subsample 14

41 Key results For migrants with low baseline savings Treatment has substantial impact on migrants financial outcomes Treatment increases joint decision-making with wives Migrant reports substantially overstate impact of treatment on wives savings 15

42 Key results For migrants with low baseline savings Treatment has substantial impact on migrants financial outcomes Treatment increases joint decision-making with wives Migrant reports substantially overstate impact of treatment on wives savings 16

43 Indian rupees (INR) Indian rupees (INR) Impact on migrant savings, remittances Savings Remittances 60, ,000 50,000 52, , ,333 40,000 35, ,000 30, ,000 20, , ,309 10, ,000 0 Low savings subsample 130,000 Low savings subsample Treatment Control Treatment Control 17

44 Key results For migrants with low baseline savings Treatment has substantial impact on migrants financial outcomes Treatment increases joint decision-making with wives Migrant reports substantially overstate impact of treatment on wives savings 18

45 Impact on joint decision-making Dependent variable: Migrant discusses household financial goals with wife (indicator) 60% 50% 52.4% 40% 40.5% 30% Treatment Control 20% 10% 0% Low savings subsample 19

46 Key results For migrants with low baseline savings Treatment has substantial impact on migrants financial outcomes Treatment increases joint decision-making with wives Migrant reports substantially overstate impact of treatment on wives savings 20

47 Indian rupees (INR) Impact on wives savings Dependent variable: Migrant s report of wife s savings 350, , , , , , , ,000 50,000 0 Low savings subsample Treatment Control 21

48 Indian rupees (INR) Indian rupees (INR) Impact on wives savings Dependent variable: Migrant s report of wife s savings Dependent variable: Wife s report of her own savings 350, , , , , , , , , , , , , , , ,000 50,000 50,000 0 Low savings subsample 0 Low savings subsample Treatment Control Treatment Control (difference not statistically significantly different from zero) 22

49 Indian rupees (INR) Indian rupees (INR) Impact on wives savings Dependent variable: Migrant s report of wife s savings Dependent variable: Wife s report of her own savings 350, , , , , , , , , , , , , , , ,000 50,000 50,000 0 Low savings subsample 0 Low savings subsample Treatment Control Treatment Control (difference not statistically significantly different from zero) Treatment effect on wives savings substantial if reported by migrants, but zero if reported by wives Entire difference accounted for by differences in reporting of wives gold holdings How to interpret this difference? 23

50 Explaining reporting treatment effect differences on wives savings No statistically significant difference in treatment impact on migrant and wife reports of other financial outcomes His own savings Remittances sent to wife So it is not the case that treatment induces wives to systematically understate financial outcomes, or husbands to overstate 24

51 Interpretation of migrant-wife discrepancy Two possibilities: 1. Non-cooperative behavior by wives In response to treatment, migrants ask wives to save more Wives tell migrants they are saving more, but are not saving in practice 2. Treatment-induced updating of migrants previous underestimates In response to treatment, migrants ask wives how much they have saved This can lead to systematic upward updating by migrants, if migrants normally (in control group) systematically underestimate wives savings Implications: Important intra-hh information asymmetries exist Future work should seek to distinguish between these two possible explanations 25

52 In sum Main conclusions (1) Financial literacy interventions can have substantial impacts on migrant financial decision-making and outcomes That such a short, simple intervention had large impacts suggests that migrants did not have strong beliefs that their previous decisions were optimal Migrants commonly seek to make joint decisions with families back home Financial literacy interventions can increase migrant interest in such joint decision-making, and change reported financial practices and goals of wives back home But limited evidence of effects on actual financial decisions of families back home 26

53 In sum Main conclusions (2) Important information asymmetries exist in transnational households Future work should explore whether these information asymmetries facilitate non-cooperative behavior For example: couple financial literacy with mechanisms that facilitate migrant monitoring and control of originhousehold financial behaviors Ashraf, Aycinena, Martinez and Yang (2012) show that providing migrants with better ability to monitor and control savings in home country increases savings in home country Impact could be magnified if savings intervention was combined with financial literacy training 27

54 28

55 Departure from existing research Our work is at the intersection of two literatures: intra-household decision-making financial literacy and its relationship with financial outcomes We use a randomized financial literacy intervention to look inside the black box of financial decision-making within the household Key innovations: Surveys of both migrants and their wives back home Results from incentivized budget allocation exercise bolster results from self-reports Complements previous work: Gibson, McKenzie and Zia (2012), Doi, McKenzie and Zia (2012) experiments on financial literacy Ashraf, Aycinena, Martinez and Yang (2012) migrant savings experiment 29

56 Sampling and surveys Convenience sample of migrants from Kerala residing in labor camps in Doha s Industrial Area Baseline survey: Aug Nov 2010 Follow-up survey: Dec 2011 Apr 2012 Surveys conducted in Malayalam by survey staff who were also migrants from Kerala Face-to-face interviews of migrants in Qatar (phone allowed at follow-up if face-to-face attempt unsuccessful) Phone interviews of wives in Kerala N=232 at baseline, N=200 at follow-up Attrition not correlated with treatment in full sample or in either subsample 30

57 The sample Married male migrant workers from Kerala, India whose wives have remained behind in India Occupations mostly manual, low-skill 79% services (of which, 2/3 are drivers ) 14% construction Migrant medians at baseline: 10 years of education, 40 years of age 8.4 years abroad Savings: INR 50,601 (~US$1,000) Annual remittances: INR 117,727 (~US$2,300) Annual income: INR 272,397 (~US$5,400) Wife medians at baseline: 12 years of education, 32 years of age Savings: INR 153,496 (~US$3,000) Annual income in India: none 31

58 Context Migrants from Kerala, India Major source of Indian migration to Gulf states In Kerala, 17.1 percent of households received remittances, and remittances are 31 percent of state GDP (Rajan and Zachariah 2011) Migration to Qatar Oil-rich Gulf state Country with most immigrants as share of population: 90% of population age 15+ were foreign-born US$8.43 billion in outgoing remittances in

59 Treatment impact on migrant own savings Notes: Distributions shown are for migrant report of own savings in follow-up survey, for low-baseline-savings subsample. 33

60 Treatment impact on remittances Notes: Distributions shown are for migrant report of annual remitances sent home to wife in follow-up survey, for low-baseline-savings subsample. 34

61 A migrant household s dwelling, Kerala 35 Note: This is not a household from financial literacy study, but another related study by Seshan. Migrant s household s dwelling is concrete house on left.

62 Migrant s wife and sons, Kerala 36 Note: This is not a wife from financial literacy study, but another related study by Seshan. Wife is in yellow and white, seated. Her sons are standing behind her.

63 Impact on KVS radio show listening 70% 60% 58.7% 50% 49.3% 48.3% 40% 30% 35.1% Treatment Control 20% 10% 0% Low savings subsample High savings subsample (difference not statistically significantly different from zero) 37

64 Incentivized budget and savings exercise Concern: self-reported savings and other outcomes might be biased by experimenter demand effects (desire to please the experimenter ) Such misreporting comes at no cost to respondent Approach: See if treatment impacts on decisions in incentivized choices correspond to impacts in self-reports What we do: enter respondents into a lottery Winner s household in Kerala gets INR 20,000 (~$400) Winnings must be pre-allocated among 18 categories of expenditures plus savings; cash is not an option Savings option is committed savings (must specify an eventual use) Migrant specifies allocation of money for wives Wife specifies how money will be allocated for herself Separate lottery for migrants and wives Respondents told that project staff will enforce allocations for winners Since actual money is potentially at stake, responses to this exercise are less likely to reflect reporting bias

65 Indian rupees (INR) Impacts on incentivized savings choices (migrants) 12,000 10,000 10,159 8,000 6,000 5,676 Treatment Control 4,000 2,000 0 Low savings subsample 39

66 In sum secondary conclusions Financial literacy training has very heterogeneous effects Average effects mask impacts in low-savings group Suggests possibility that tailored trainings could enhance benefits; should be explored in future work Examining treatment impacts on decision-making in incentivized choices bolsters confidence in interpretation of treatment effects on self-reported outcomes Important to survey both sides when assessing impact of interventions in transnational households Relying solely on one side s report (in particular, migrant s) leads to misleading conclusions 40