VIEWS: 1 PAGES: 35 POSTED ON: 9/6/2013
Is it all about Money? A Randomized Evaluation of the Impact of Insurance Literacy and Marketing Treatments on the Demand for Health Microinsurance in Senegal Jacopo Bonan (Milan-Bicocca), Olivier Dagnelie (Namur), Philippe LeMay-Boucher (Heriot-Watt) and Michel Tenikue (CEPS) Thanks to: ILO/Bill and Melinda Gates Foundation, Carnegie Trust of Scotland, Fonds National de la Recherche Luxembourg and the GRAIM, Thies. 1 Introduction Health providers in Senegal: • Organized according to a tiered system: 1) health huts (staffed by community workers) 2) health posts (nurses and certified midwives) 3) health centres (with medical doctors, etc.) Survey: Thiès (2nd largest city in Senegal) Thies: 1 public hospital and 1 private hospital 2 Introduction • High health costs -Cost in health huts and posts are small -Hospitals: Xray (6000FCFA); Blood analysis (2000); Consultation (10-25000); Night (>25000) (sample median income 112500FCAF/month) → Often too Expensive Poor people face difficulties to access ‘modern-medical’ health care 3 Introduction 1) health shocks → direct expenditures (drugs & treatment) : out-of-pocket payments (OOP) 2) indirect costs → reduction in productivity • OOP in Senegal : 70% of total expenditure in health (WHO) (5% in GBR, 11% in France) • at low levels of income : strong link between health and income → Scope for more health insurance 4 Health Microinsurance Available • no universal coverage in Senegal No state insurance (ill-functioning: CESAME) • informal insurance (bilateral transfers; networks :De Weerdt et al.(2006); ROSCAS: Dagnelie et al. (2012)) 5 Health Microinsurance Available • IPM: treasury made up of employees’ compulsory contributions. (for large employers) • Private health insurers (PAMECAS, etc) • Mutual Health Organizations (MHO) 6 MHOs expansion • Self or informally employed: no IPM (>60% in our sample): ill-suited supply of health insurance • Senegal 1990 in Fandene: first ‘mutuelle de santé’ or MHO • Grass-root movement: managed by locals • MHOs in Senegal: 13 (1993) → 140+ (2007) 7 MHOs structure and rules • Open to everybody • Membership fees (1000-3000FCFA/hh) • Monthly premium 100-500FCFA/month/person • Period of observation (3 months) • MHO have contracts with all health providers • cover 25 to 75% of consultation fees cover 50% to 100% of medical exams and various inpatients cares fees 8 Microinsurance take-up in Thies • Out of our sample of 360 heads of hh: 32% of have health insurance, (for 73% of all household members) Of which: → 19% with IPM (public or private) (several private IPM are not working…) → 3% with private health insurance → 10% with MHOs 9 Low MHO take-up • Overall MHO take-up rate in Thies region = 5% Smith et al. (2008) & Lépine et al. (2010). → Our Research Question: Why is MHO take-up rate so low despite being well established and having the potential to reach poor people? 10 Why low MHO take-up? • From our sample, people who justified the lack of membership to MHOs: 1) lack of information about the product offered and/or MHOs existence (70% of hh) 2) lack of means (16%) 3) lack of interest (5%) 4) lack of trust (2%). 11 Lack of Information • Cai et al. (2009): farmers in China refuse to purchase a heavily subsidized insurance for sows → lack of awareness of the program. • Giné et al. (2007); Cole et al. (2009) and Gaurav et al. (2009): → Limited understanding of rainfall insurance mechanisms in rural India. • Jutting (2003): concept of insurance is alien to a large proportion of people in Senegal. → information campaign could have some impact 12 Lack of Means • Jutting (2003) poorer hh in Senegal less member of MHOs. • Chankova et al. (2008) similar results in Ghana and Mali • Giné et al. (2008) : take-up rate of rainfall insurance increases with household wealth in rural Andhra Pradesh. • Cole et al. (2009) low take-up rates of rainfall insurance : insurance is expensive. • In our sample: only mentioned by 16% Why? → willingness to pay (WTP) = actual premiums → a banana is 100FCFA (!?!) Bonan (2011) uses this data and the contingency valuation method to get WTP for MHOs premiums. 13 Lack of Trust • Cai et al. (2009) low take-up by Chinese farmers: lack of trust toward governmental institutions. • Cole et al. (2009) endorsement of rainfall insurance in India from a third party ↑ purchase by 40%. • In Thies: We asked the sample of non-members aware of the existence of MHOs: trust towards MHOs 1 to 10 Rescaled median score w.r.t. Trust in mother and family: 8/10 → Large positive a priori from locals towards MHOs 14 What if we inform and ↓ fees? • Factors at play: Information & Means impact on membership of: -more info -↓ financial barriers → Randomized Control Trials • Design and implement two treatments: 1)Insurance Literacy module 2)Marketing with 3 vouchers 15 Our 2 treatments on 360 hh 1) Literacy: 180 hh invited to: 3hour module on health microinsurance, MHOs and the concepts of risk and insurance. (given by GRAIM) After 2) Marketing: redeemable for 3 months Voucher 1: invitation to GRAIM (120 hh) Voucher 2: membership fees (120 hh) Voucher 3: memb. fees + 3000FCFA obs period (120hh) 16 17 Randomized Evaluation • Old technique: The first published RCT appeared in the 1948 paper entitled: "Streptomycin treatment of pulmonary tuberculosis“ which described a Medical Research Council investigation. -Austin Bradford Hill is credited as having conceived the modern RCT 18 Randomized Evaluation examples (I) • Very active field in recent years: impulse from MIT and Poverty Action Lab • Ex.: PROGRESA Mexico (1998) 506 communities (half randomly selected) Treatment: cash grants to women conditional on school attendance and preventive health measures. Results: Gertler et al. (2001) ↓ illness, ↑ height, ↑enrollment 19 Randomized Evaluation examples (II) • Ex: Kenya, free breakfast program effect on school participation. Results: Vermeersch (2002) ↑ school participation • Ex: Kenya, program providing uniforms and textbooks (to 7 randomly selected schools) Results: Kremer et al. (2002) ↓ Dropout rates 20 Randomized Evaluation Ex: avg test scores (Duflo et al. 2007) Schools with textbook (T) – Schools without (C) D = treatment effect + selection bias → with successful randomization : We can pretend selection bias = 0 21 Our data: Thies June 2010 • Second city of importance in Senegal population of 240,000 (2002 census) -its MHOs are the oldest in Senegal: well established supply of MHOs. • 360 randomly selected households • Urban area: 20 km2 • Baseline survey : housing information, household composition, info on head hh. • Uptake decision: head of hh 22 Survey timeline: Our dependent Invitation to Module is variable: subscribe made followed by or not to MHO Marketing treatment between May-Sept May-June 2010: September 2010: Survey of 360hh deadline for vouchers Greedy enumerators constantly asking for an increase in salary… 23 24 • ‘Strongly risk averse’ based on Voors et al. (2010) takes value 1 if individuals always opted for the certain outcome ‘A’ when presented with : 25 • ‘Patient’ We elicit discount factors (Voors et al. (2010)) → discount factors at one month of: 5%, 10%, 25%, 50%, 75%, 100%, 150%, 200%. dummy if head is more patient half of our sample. •Aside: experience with real money gave very different results… 26 •No significant difference if we look at the assignment across vouchers 1-2-3 (not shown) 27 Not good news Selection bias is clearly not equal to zero. Treatment group: -less insured (↑ intake) -less rich (↓ intake) -less knowledge about insurance (minor) -less public servant (↑ intake) Bias can be + or – difficult to guess 28 29 Empirical Strategy • Our model: -M takes value 1 if hh subscribes to a MHO following our treatments -E takes value 1 if hh was invited to module -Voucher takes value 1 if hh was given either voucher 2 or 3 • Results similar with probit (shown) and OLS 30 Our estimates • E measures the ‘invitation effect’ - does not measure the actual participation effect - α :‘intend to treat’ effect • Low compliance rate (58%) we compute also ‘treatment on treated’ or ‘average treatment effect’ (Imbens-Angrist 1994) à IV (instrument attendance by invitation) • Both techniques give similar results 31 32 Randomized Evaluation related to insurance • Rainfall insurance intake 1)Gaurav et al (2009): Gujarat, India (600hh) Treatments: insurance educ module & Marketing → module ↑ intake; little impact from marketing 1)Cole et al. (2009): India (2000hh) Treatment: insurance educ module → no impact from module 33 34 Conclusion • Literacy module has no significant impact Why? - Representative present (not being the head) - Health insurance is a simple product (relative to rainfall insurance): no need - Quality of our module delivery? • Vouchers 2-3 have strong and positive impact -efficient to only distribute voucher 2 35
"Is it all about Money_ A Randomized Evaluation of the Impact of "