Presentation on theme: "Willingness to accept conditional economic incentives to reduce HIV risks: Experimental evidence from men who have sex with men in Mexico City Omar Galárraga,"— Presentation transcript:
Willingness to accept conditional economic incentives to reduce HIV risks: Experimental evidence from men who have sex with men in Mexico City Omar Galárraga, Brown University & INSP Sandra G. Sosa-Rubí, INSP César Infante, INSP Paul J. Gertler, UC Berkeley Stefano Bertozzi, Gates Foundation Vienna, Austria July 2010 Research Funding: NIH K01-TW (Fogarty International Center) & Mexican Ministry of Health / National Center for HIV/AIDS Control and Prevention (CENSIDA)
Motivation l Mexico (most of Latin America) has a concentrated HIV epidemic: General population HIV prevalence < 1% HIV prevalence estimates in Mexico for: men who have sex with men (MSM) > 13.5% male sex workers (MSW) > 18% l Traditional approaches (based on education and mass media prevention campaigns) have not sufficiently reduced the high incidence rates. l New and innovative programs are needed.
Background l Mexico PROGRESA/Oportunidades: poverty reduction using conditional cash transfers (CCT) for preventive check-ups, keeping school-age children in school, health education workshops. l HIV-related experimental international work: Small incentives can improve return rates for HIV testing in rural Malawi (Thornton, AER 2008) l NEW: incentives to improve take up of known prevention methods, linked to sexual health outcomes in MSM including MSW; first work in Latin America / concentrated epidemic.
Objectives l To measure willingness to accept (WTA) conditional economic incentives (CEIs) for staying free of sexually transmitted infections among men who have sex with men (MSM) of low socio-economic status in Mexico City. l To identify determinants associated with willingness to accept / willingness to participate in CCT to prevent HIV & other sexually transmitted infections (STI). l To provide an incentive elasticity of the demand and potential participation rate in a prevention program.
Model l Interaction between participants in the prevention program, and health planner with fixed budget. l Social planner maximizes number of participants (minimizes sexual risk behaviors, SRB). l Individuals participate if disutility of reducing SRB is at least compensated (“reservation level”) with an increase in utility derived from the incentive. l Planner will offer optimal incentive which is high enough to spur participation but staying within a fixed budget constraint.
Methods l We interviewed 1,820 MSM in Mexico City, sampled through time & place, as well as respondent-driven methods. l Questionnaire described hypothetical prevention program where participants would receive economic incentives monthly to attend prevention talks; and every three months as long as they remained free of STIs (verified by periodic STI testing). l Experiment with contingency evaluation methods: random starting- point and iterative bidding (bargaining game) to estimate WTA.
Methods (2) l Each subject was asked if he would enroll in the prevention program given a set level of incentives. l Regression models (linear, probit, and quantile regression) to estimate incentive elasticity of demand and potential participation in the program. l Covariates controlled for: socio-economic status (SES) and sexual risk levels (measured through condom use, number of sexual partners, and commercial sex).
WTA questions (experiment) l WTA 1. WOULD YOU BE INTERESTED IN PARTICIPATING IN THE TALKS AND THE TESTS IF YOU RECEIVED $_"x1"__ PESOS PER MONTH? Yes =1 or No =0 COMPUTER WILL RANDOMLY PICK NUMBER "X1" FROM LIST: $0 TO $200 PESOS IN $25-PESO INTERVALS IF ANSWER IS YES, THEN COMPUTER PROPOSES A LOWER INCENTIVE: X2 = X1 * 0.7 IF ANSWER IS NO, THEN COMPUTER PROPOSES A HIGHER INCENTIVE: x2= ($200 + X1) /2 l WTA 2. WOULD YOU PARTICIPATE IN THE TESTS AND ATTEND THE TALK FOR $ "x2" PESOS PER MONTH? Yes =1 No =0
WTA questions (2) l WTA 3. IF YOU WERE GIVEN A PRIZE FOR REMAINING STI-FREE FOR A PERIOD OF 3 MONTHS, WOULD YOU REMAIN FREE OF STIs IF YOU RECEIVED A PRIZE OF $”Y1" PESOS EVERY THREE MONTHS? Yes =1 No =0 COMPUTER WILL PROVIDE RANDOM NUMBER "Y1"as FOLLOWS: Y1 = X1*3 IF YES, THEN PROPOSE A LOWER INCENTIVE: Y2=Y1*0.7 IF NO, THEN PROPOSE A HIGHER INCENTIVE AS FOLLOWS:Y2 = Y1*2 l WTA 4. WOULD YOU REMAIN FREE OF STIs IF YOU RECEIVED A PRIZE OF $"Y2" PESOS EVERY THREE MONTHS? Yes =1 No =0 l WTA 5. WOULD YOU BE WILLING TO REMAIN FREE OF STIs WITHOUT RECEIVING A PAYMENT TO ATTEND THE MONTHLY TALKS, BUT RECEIVING A PRIZE OF $ ”Y3" EVERY THREE MONTHS? Yes =1 No =0 COMPUTER WILL CALCULATE "Y3" AS FOLLOWS: Y3 = Y1+MAX(X1,X2)*3
Model specification l WTA i = person i is willing to accept CCT to reduce risk (yes=1 ; no=0) l incentive i = random incentive level for person i (in natural logs) l i subscripts in X vector: individual respondent characteristics that might influence choice: education, wealth index, risk level (sex work, condom use, number of partners) l Random error: e l Estimate with linear, non-linear (probit models), & quantile regression l Calculate elasticity:
Table I. Descriptive Statistics
Table I. Descriptive Statistics (2)
Table II. Checking for balance across observable characteristics
(1)(2)(3)(4)(5) VARIABLES Monthly incentive X1 Monthly incentive X2 Quarterly incentive Y1 Quarterly incentive Y2 Quarterly incentive Y3 Age (in years) [0.582][0.721]+[1.747][2.813][3.469] Has HIV or other STI (=1) [4.280][5.298]**[12.840][20.672][25.492] Any condom use (=1) ** [4.066][5.034][12.199][19.640][24.220] Sex partners (# per month) [0.266][0.330][0.799][1.287][1.587] Last sex with someone just met (=1) [2.687][3.326]+[8.060][12.976][16.002] Has a stable partner (=1) [2.704][3.347][8.112][13.060][16.105] Sex worker (=1) [6.132][7.591][18.397][29.618][36.525] Occupation: student (=1) [2.685][3.323][8.054][12.966][15.990] Highest level of education * [2.270][2.810]**[6.810][10.963]+[13.520]* Has correct knowledge (=1) [3.263][4.040][9.790][15.761][19.437] Wealth index ° [1.789][2.215][5.368][8.642]*[10.657] Constant [12.956]**[16.039][38.869]**[62.580]*[77.172]** Observations900 R-squared F-test that all coefficients are zero p-value of F-test
Table III. Economic incentives to achieve different participation levels Note: Figures are expressed in $ Mexican pesos of December 2008; 1 USD~11 MX pesos
Unadjusted density for willingness to accept monthly prevention talks (wta1)
Willingness to accept monthly prevention talks & quarterly STI testing (wta1*wta3)
Table III. Non-parametric adjusted demand elasticity for HIV prevention Simultaneous quantile regressions controlling for all covariables in Table I. Bootstrapped standard errors in shown below coefficients (with 200 repetitions).
Quantile regression and probit marginal effects, and confidence intervals
Table IV. Switching point for accepting HIV prevention Positive SwitchersNegative Switchers (1)(2)(3)(4) Switched to yes from wta1 to wta2 Switched to yes from wta3 to wta4 Switched to no from wta1 to wta2 Switched to no from wta3 to wta4 Conditional incentive (log) [0.019]**[0.009]**[0.011]**[0.009]** Observations Switchers (persons) Observed switching probability Computed elasticity Mean conditional incentive ($ pesos)120.99/mo /qtr.43.28/mo /qtr. Mean total conditional yearly incentive ($pesos) /yr /yr. Probit regression, marginal effects controlling for all covariates presented in Table 1. Robust standard errors in brackets: ** p<0.01, * p<0.05, + p<0.1
Table V. Benefit Cost Estimations Yearly testing111 Program duration (years)511 Time Horizon (years)25 Incidence per 10,000 MSM180 Effectiveness in reducing incidence16% ART costs per person per year $USD 5,585 8,064 Benefit Cost Ratio
Discussion l High willingness to participate in prevention program with CEIs; acceptance rates: 60-80% depending on modality. l Median expected incentive conditional on staying free of STIs (with quarterly testing) was $1760 pesos per person per year; for switchers it was $2,641 (~ $200 USD). l Models show positive coefficient between incentive and WTA. Elasticity: if we increase incentive by 10% then participation (in talks & STI testing) increases by about 20%. l Models robust over the quantile distribution: stable coefficients.
Conclusions l MSM & MSW generally well disposed to participate in a CCT program for HIV and STI prevention in Mexico l Results corroborate previous qualitative work: Infante et al 2009; program could be generally accepted by the target population. l The average willingness to accept estimate is within the range of feasible allocations for prevention in the local context. Expected impact can be highly cost beneficial (cost saving). l Mexico, a global leader in CCT for human development, could extend that successful model to prevent HIV and other STIs.