precautionary savings to manage common health risks among the poor
DESCRIPTION
Precautionary Savings to Manage Common Health Risks among the Poor. Jessica Pickett University of Pennsylvania. Research Problem. Poor households struggle with unpredictable out-of-pocket health expenses Recent trend towards government, NGO microinsurance Primarily hospitalization - PowerPoint PPT PresentationTRANSCRIPT
2012 Research Conference on Microinsurance
Precautionary Savings to Manage Common Health Risks among the Poor
Jessica PickettUniversity of Pennsylvania
Insurance in combination with other financial services 1
Jessica Pickett
• Poor households struggle with unpredictable out-of-pocket health expenses
• Recent trend towards government, NGO microinsurance• Primarily hospitalization• Pharmaceutical & outpatient expenses rarely covered
• Even minor illness can reduce consumption in inefficient capital markets• Productivity losses from under-treatment• Future losses from high-interest loans or asset sales
• Precautionary savings as an alternative• Constrained by biased beliefs, inattention & inconsistent preferences
Research Problem
12 April 2012 Insurance in Combination with Other Financial Services 2
Jessica Pickett
Insurance
• Catastrophic coverage• Adverse selection &
moral hazard• Regulatory constraints• Oversight, monitoring
& fraud prevention• Most clients will not
benefit
Savings
• Longer time horizon• Contribute while
healthy / productive• Individual
accountability• Must trade current
period consumption for uncertain future benefits
• No protection against worst shocks
Loans
• No uncertainty• High interest rates• Illness may increase
need but decrease ability to repay
Options for Risk Management
12 April 2012 Insurance in Combination with Other Financial Services 3
Jessica Pickett
• Financial impact of illness is lower in households with access to rural deposit accounts in Indonesia (Gertler, Levine & Moretti, 2009)
• Market women in Kenya with access to individual savings accounts spend less working capital on medical costs & more likely to receive treatment for malaria episodes (Dupas & Robinson, 2010)
• But recent research on health-specific savings technologies with Kenyan ROSCAs (Dupas & Robinson, 2011) and the impact of saving reminders more generally (Karlan, McConnell, Mullainathan, & Zinman, 2011) suggests households systematically underestimate future treatment costs• Financial access is not enough to ensure savings – must also address behavioral
constraints
• Supported by observed demand for health savings & loan products among Indian SHGs and MFI clients in Burkina Faso (Reinsch & Ramirez, 2011)
Savings & Health
12 April 2012 Insurance in Combination with Other Financial Services 4
Jessica Pickett
• Does outpatient spending affect household consumption?
• Is there demand for insurance and/or precautionary savings for health?
• Why don’t the poor currently save enough for that purpose?
Hypothesis: Insurance should be supplemented with better savings vehicles to overcome biased beliefs, inattention, inconsistent preference & other behavioral constraints to manage common illness
Research Questions
12 April 2012 Insurance in Combination with Other Financial Services 5
Jessica Pickett
• Permanent income hypothesis predicts households should always opt for remedial treatment for health shocks & thus always prefer full insurance, if actuarially fair
• Predicts positive savings rate when insurance markets are not feasible or loading exceeds household’s willingness to pay• Assuming reasonable levels of risk aversion & interest rates
• But precautionary savings for health still entails significant uncertainty• Exact timing even of frequent illnesses is unpredictable
• Savings goals require accurate information on probability & costs of illness, as well as treatment efficacy
• Households must diligently deposit savings & protect those funds from expropriation• Limited attention, narrow framing & commitment
• Model of healthcare utilization under these forecasting constraints predicts negative savings rate & decreased future consumption due to asset sales, loans or decreased productivity
Theoretical Context
12 April 2012 Insurance in Combination with Other Financial Services 6
Jessica Pickett
• Wage rate Yi, productive assets At, debt Dt & non-medical consumption Ct
• Insurance at given loading α
• Disability-adjusted labor Ht determined by exogenous probability of illness & depreciation • Superscripts denote realized incidence of illness: 1 indicates a state of the world if sick & 0 if healthy
• Remedial treatment at price pm
Expected Utility Model
Insurance premiumCurrent period earnings Loan paymentsMedical expenses
12 April 2012 Insurance in Combination with Other Financial Services 7
Jessica Pickett
• Forecast illness with probability) due to:• Inattention or poor cognition• Imperfect information, innumeracy
or biased beliefs• Disutility from considering negative
prospects
Behavioral Model (Savings)
Current period earnings Loan paymentsMedical expenses
Current period utility from illness
Expected forecast of future utility
Current utility of full health
12 April 2012 Insurance in Combination with Other Financial Services 8
Jessica Pickett
• Model suggests demand for savings products that would allow households to save to smooth consumption against health shocks• Ongoing research proposes the use of reminders, conditions & restrictions targeting individual
accounts• Alternatively, existing community-based savings groups (CBSGs, aka VSLAs) “nudge” members
to save more than simply increasing access to financial services
• CBSG meeting structure and corresponding social dynamics simultaneously serve to maximize trust while mitigating many crucial behavioral constraints: • Mandated savings implicitly represents target savings goal & explicitly reminds members to
save (compounded by peer effects)• Salience of medical costs & other financial risks is amplified by observing other members apply
for loans or receive social fund grants• Reduces “negative” expropriation of funds by others by introducing formal loan structure
where loans are likely to be repaid & member has access to liquid capital during interim• Loan requests subject to approval by the group, serving as a semi-liquid commitment device
protecting against temptation goods or time-inconsistent preferences
• Better way to manage outpatient expenses than ROSCAs or MFIs, where existing health savings pilots observed demand for health-specific accounts• But still subject to information problems
Community-Based Savings Groups
12 April 2012 Insurance in Combination with Other Financial Services 9
Jessica Pickett
Case Study: Tajikistan
12 April 2012 Insurance in Combination with Other Financial Services 10
Jessica Pickett
• Poor, land-locked country in post-Soviet Central Asia; rural & mountainous, with a population of 7 million• Few natural resources; economy is heavily dependent on agriculture & remittances• Per capita income $1,860 (PPP); 47% poverty rate; GDP growth of 3.4% per year• High education levels & 100% literacy rate in Tajik (Persian) or Russian• Communicable disease ~ 72% of the overall disease burden, especially TB (MDR)
• Crumbling Soviet health system: ostensibly free network of government-run hospitals (61 beds per 10,000) & salaried doctors (20 per 10,000)• Unsustainable excess institutional capacity
• Vast majority of health costs paid out-of-pocket, mostly for pharmaceuticals, transportation & informal payments to government providers• 64% of households rate it “difficult” or “impossible” to find money for needed treatment
• Aggravated by inefficient capital markets (17% interest rate spread)• Rural households lack access commercial banks, especially for relatively small transactions• MFIs increased access to credit but prohibited from accepting deposits or insurance
products
• Recent, widespread expansion of CBSGs has potential to mitigate these challenges
Case Study: Tajikistan
12 April 2012 Insurance in Combination with Other Financial Services 11
Jessica Pickett
Tajikis
tan IndiaLao
s
Vietnam
Mexico
Nicarag
uaKen
yaGhan
a
Indonesia
Thaila
nd
United St
ates
Netherl
ands
0%
10%
20%
30%
40%
50%
60%
70%
80%
0.0%
0.5%
1.0%
1.5%
2.0%
2.5%
3.0%
3.5%
4.0%
Out-of-Pocket Health Spending (Sample Countries)
Private out-of-pocket expenditures as % of total health expendituresOut-of-pocket health spending as % of income per capita (PPP)
% T
otal
Hea
lth E
xpen
ditu
res
% P
er C
apita
Inco
me
Case Study: Tajikistan
12 April 2012 Insurance in Combination with Other Financial Services 12
Jessica Pickett
• Existing: Tajik Living Standards Survey (TLSS) contains data on household health expenditures, consumption & financial access• Nationally representative sample of 4,860 households in 2007 (stratified
by region & population density)• WB Living Standards & Measurement Survey research program
• Subset of panel data from 1503 households (10,069 individuals) in 2009• 3rd wave of data collection began in November 2011
• Proposed: Detailed health & expenditure logs• Sub-sample of TLSS participants via SMS or phone interviews
• “Large T” analysis can dramatically increase power with more frequent observations of a smaller overall sample for health outcomes, household expenditures & other noisy outcomes with low autocorrelation (McKenzie, 2011)
• Also important given recall bias for frequent health spending (Das, Hammer & Sanchez, 2012)
• CBSG expansion as a natural experiment
Data
12 April 2012 Insurance in Combination with Other Financial Services 13
Jessica Pickett
(1) (2) (3) (4)
∆ Medical Expenditures -0.480*** -1.604*** -0.127 -0.823* (0.080) (0.449) (0.068) (0.373)
Constant -111.0*** -95.94*** -88.35*** -71.53*** (12.24) (1.077) (11.10) (0.895) Observations 1,451 245 1,451 245R-squared 0.401 0.039 0.262 0.017Number of PSUID 167 118 167 118Community FE Yes Yes Yes YesWealth Decile FE Yes Yes Poorest Quintile Yes YesFood Only Yes Yes
Empirical Test of Full InsuranceEffect of Health Expenditures on Per Capita Non-Medical Consumption
Robust standard errors in parentheses, clustered by strata*** p<0.01, ** p<0.05, * p<0.1
12 April 2012 Insurance in Combination with Other Financial Services 14
Jessica Pickett
• Households should be willing pay a “risk premium” in excess of expected expenses to insure against illness• Based on variance of losses, income & risk tolerance:
• Use TLSS to estimate maximum % of expected benefits that households would pay to fully insure ambulatory medical care• Typically lower for outpatient care than for hospitalization due to lower
variance (Pauly, Blavin, & Meghan, 2010)
• Compare to plausible levels of administrative loading• ~30% in other developing countries
• Higher regulatory & monitoring costs for informal payments to public providers
• Higher loading costs for outpatient care given greater number of transactions & providers • Also aggravates selection problem (especially with relatively low risk aversion)
Theory of Insurance
12 April 2012 Insurance in Combination with Other Financial Services 15
Jessica Pickett
Monthly Per Capita Non-Medical Consumption
Mean 185.87
Std. Dev. 83.46
Monthly Per Capita Outpatient Expenditures
Mean 8.68
Std. Dev. 15.68
Coefficient of variation 1.81
Risk Premium 1.32
% of mean 15.2%
Estimated Demand for InsuranceSurvey-Adjusted Risk Premium
(2007/2009 Panel)
* CRRA
12 April 2012 Insurance in Combination with Other Financial Services 16
Jessica Pickett
• Do outpatient medical expenditures affect household consumption?• TLSS data suggests health expenditures are only partially offset through informal
coping mechanisms & negatively affects current consumption• More pronounced among the poorest households
• Is there demand for insurance and/or precautionary savings for health?• Absence of insurance for ambulatory care appears consistent with low demand
relative to plausible loading• Variance of OOP spending still high enough to affect behavior & policy →
suggests high value of savings & other mechanisms to smooth consumption more cheaply
• Consistent with observed demand for health savings in India, Kenya & Burkina Faso
• Potential for bundling with hospitalization insurance
• Why don’t the poor currently save enough for that purpose?• Inefficient capital markets combined with biased beliefs, inattention, inconsistent
preference & other behavioral constraints
Policy Implications
12 April 2012 Insurance in Combination with Other Financial Services 17
Jessica Pickett
• Would savings increase utilization and/or out-of-pocket health spending?• Depends on demand elasticity & price discrimination• Preliminary evidence from recent pilots & objective of future
research
• Do savings actually improve health outcomes?• Unknown → mitigated by poor treatment-seeking behaviors on the
part of patients combined with low quality & corruption on the part of providers• More ambiguous demand function (price elasticity & non-price factors),
combined with less obvious outpatient health benefits• Need for patient education plus incentives for high-quality care &
evidence-based treatment
Future Research
12 April 2012 Insurance in Combination with Other Financial Services 18