Free Distribution or Cost Sharing?

Transcription

1 Free Distribution or Cost Sharing? Evidence from a Randomized Malaria Prevention Experiment Jessica Cohen & Pascaline Dupas Kaylyn Fraser & Justine Robinett - February 4, 2010

2 Outline of the Paper 1. Introduction 2. Simple Model of Pigouvian Subsidies 3. Background & Experimental Set-Up 4. Data & Results 5. Cost-Effectiveness Analysis 6. Conclusion & Discussion

3 1. Introduction Standard microeconomic theory: goods with positive externalities should be subsidized For what kinds of goods is this not the case? Cost-sharing : charging a positive price What is the argument behind this?

4 1. Introduction: Effects of Cost- Sharing Possible positive effects on usage: Selection effect Psychological effect Interpretation of price as signal of quality Other counter effects to consider: Lower demand Cash and credit constraints

5 1. Introduction: Key Factors 1. Elasticity of usage with respect to price 2. Elasticity of demand with respect to price 3. Impact of price variation on consumer vulnerability 4. Presence of non-linearities in the health production function This paper s goal is to estimate parameters 1, 2, and 3 using a field experiment with ITN s in Western Kenyan pre-natal clinics

6 Outline of the Paper 1. Introduction 2. Simple Model of Pigouvian Subsidies 3. Background & Experimental Set-Up 4. Data & Results 5. Cost-Effectiveness Analysis 6. Conclusion & Discussion

7 2. Model C = marginal cost of an ITN T = subsidy h = # nets used for health purposes H = average # of nets used for health purposes/household n = # nets not used for health purposes N = average # of nets used for non-health purposes/household k = positive health externality Household Utility: U = u(h) + v(n) - (C-T)(h+n) + kh both u & v 0 & both u & v 0

8 2. Model (continued) Household Utility: U = u(h) + v(n) - (C-T)(h+n) + kh from the F.O.C.: u (h) = v (h) = C - T marginal cost of increasing the health externality: T*[d(H+N)/dT] marginal benefit: k*[dh/dt] The efficient subsidy sets these equal: T*[d(H+N)/dT] = k*[dh/dt] Ť= k*[dh/dt] / [d(h+n)/dt]

9 2. Model: Implications Optimal subsidy level, Ť: depends on responsiveness of non-health purchases to changes in subsidies, dn/dt dn/dt = 0: optimal subsidy is simply the level of the positive externality (Ť= k) dn/dt > 0: optimal subsidy is lower than the positive externality (Ť < k) Ť - k =?

10 Outline of the Paper 1. Introduction 2. Simple Model of Pigouvian Subsidies 3. Background & Experimental Set-Up 4. Data & Results 5. Cost-Effectiveness Analysis 6. Conclusion & Discussion

11 3. Background: ITNs Why ITNs? Very effective against infected bites Not widely used, cost is high Failure to achieve higher coverage Positive public health benefits Some facts specific to Western Kenya: Malaria endemic to the region, with 2 peak seasons Pregnant women may receive as many as 230 infected bites during their pregnancy Up to 1/3 of all infants either premature or abnormally small

12 3. Experimental Set-Up 20 rural public health centers across 4 districts in Western Kenya 4 clinics: control group 5 clinics: free (100% subsidy) 5 clinics: 10 Ksh (97.5% subsidy) 3 clinics: 20 Ksh (95% subsidy) 3 clinics: 40 Ksh (90% subsidy) Financial incentive for the clinics to adhere to assignment Programs introduced within 3 months of each other and kept going through the peak malaria season

13 3. Experimental Set-Up: (cont.) Two-Stage Price Randomization: why? Only in 11 clinics charging positive price Subsample offered lottery for possible discount, after agreeing to purchase an ITN Implemented at least a month after program started, varied day of week

14 Outline of the Paper 1. Introduction 2. Simple Model of Pigouvian Subsidies 3. Background & Experimental Set-Up 4. Data & Results 5. Cost-Effectiveness Analysis 6. Conclusion & Discussion

15 4. Data Data were collected using 3 different methods: 1. Administrative records 2. Surveys during clinic visits 3. Surveys during household visits During household visits, surveyors asked women: 1. To show their ITN 2. Whether they used their net 3. Who slept under the net Outcomes of interest: ITN up-take, usage rates and health

16 4. Data (continued) Indicators for outcomes of interest: Up-take (demand) Average weekly sales of ITNs (as recorded by clinics) Share of surveyed pregnant women who acquire an ITN Usage rates (effective coverage) Share of surveyed pregnant women who not only acquire an ITN but also report using it during household visits

17 4. Results Up-take of ITNs: Drops by 60% with an increase from 0 Ksh to 40 Ksh Falls by 43% with an increase from 20 Ksh to 40 Ksh Decreases at even highly-subsidized cost-sharing prices For each increase in price by 10 Ksh, 8 fewer ITNs are sold per week (20% decrease in weekly ITN sales) Clinics distributing fully subsidized nets hand out an average of 41 ITNs per week

18 4. Results (continued)

19 4. Results (continued) ITN usage rates: 25/100 of women receiving the nets for free would purchase one at the average cost-sharing price No relationship between the price paid and the probability of usage; although usage is highest with the net is fully (100%) subsidized Ratio of hanging nets to acquired nets is close to 1 During household visits, 62% of women who acquired a net through the study claimed to be using it, while 95% of this group had the net physically hung during the visit Usage rates are lower among pregnant women (90% for overall Kenyan population)

20 4. Results (continued) ITN-related health outcomes: When ITNs are fully subsidized, the greatest number of children s lives are saved On average, women who benefit from the 100% subsidy are healthier than the average prenatal woman (although, those who pay a higher price are not sicker than the average prenatal woman) Why? Compared to women in the control group, those receiving the full subsidy: Were 12% more likely to return for a follow-up visit Were 12% less likely to have arrived at the clinic by foot Spent 3.5 Ksh more on travel expenses to get to the clinic

21 4. Results (continued) ITN-related health outcomes: (continued) Women acquiring the ITNs for free are significantly less likely to be anemic While in absolute terms, free ITN distribution reaches more anemic women than through cost-sharing When compared to free distribution, cost-sharing results in 60% less effective coverage for the anemic population

22 Outline of the Paper 1. Introduction 2. Simple Model of Pigouvian Subsidies 3. Background & Experimental Set-Up 4. Data & Results 5. Cost-Effectiveness Analysis 6. Conclusion & Discussion

23 5. Cost-Effectiveness Analysis Presented in terms of number of children s lives saved and cost per life saved Authors propose 2 indices: Protection index for non-users - a logistic function of the share of users of the total population Protection index for users - a weighted sum of a physical barrier effect of the ITN and the externality effect, the weights depending on the share of users 3 possible values for each index, leading to 9 different scenarios for each of the 4 pricing strategies

24 5. Cost-Effectiveness Analysis (cont.) Assumptions: The difference in the subsidy is the sole explanatory value for the variation in cost per ITN between free distribution and cost-sharing 65% of households will experience a pregnancy within five years, therefore rendering themselves eligible for the program

25 5. Cost-Effectiveness Analysis (cont.) Key findings: Free distribution leads to a greater reduction in child mortality than any cost-sharing strategy In terms of cost per life saved, a cost-sharing scheme of 40 Ksh is more cost-effective than free distribution when there exists a strong physical barrier effect When this physical barrier effect is weak, free distribution is at least as cost-effective, if not more, than any cost-saving initiative Cost-sharing is, even if marginally, more cost-effective than free distribution, ALTHOUGH free distribution saves more lives : trade-off!

26 5. Cost-Effectiveness Analysis (cont.)

27 Outline of the Paper 1. Introduction 2. Simple Model of Pigouvian Subsidies 3. Background & Experimental Set-Up 4. Data & Results 5. Cost-Effectiveness Analysis 6. Conclusion & Discussion

28 6. Conclusion Cost-sharing in this context: had no effect on the reduction of wastage had no effect on usage intensity decreased demand had no effect of anemia likelihood

29 6. Conclusion (continued) Limitations to the analysis: Relatively small sample size Small sample imprecision Missing data Leakages and mismanagement observed in 4/11 clinics Wealth effect within 40 Ksh group No information on the behaviour of the distributive agents Potential correlation between clinic attendance and ITN prices Services offered may not be homogenous across clinics