How to sell a condom? The Impact of Demand Creation Tools on Male and Female Condom Sales in Resource Limited Settings Fern Terris-Prestholt 1 & Frank Windmeijer 2 Fern.Terris-Prestholt@LSHTM.AC.UK 1 London School of Hygiene &Tropical Medicine 2 Bristol University HESA retreat Sept 14, 2015
HIV, risky sex and condoms End of HIV not in sight, despite recent advances in HIV prevention (VMMC and ARV-based prevention) Condoms highly effective at preventing HIV: 50 million infections averted (Stover 2015), and very cheap Annual condom gap: 8 million condom shortfall (UNAIDS 2014) 8 condoms per sexually active persons in Africa (Desperthes, 2014) Demand and supply constraints; Need for accessible and consistent supplies Demand creation to increase (and sustain) use
Social Marketing: Why can t we sell brotherhood like we sell soap? Sweat et al, WHO bulletin 2011
AIMS To estimate key drivers of aggregate demand for male and female condoms (static panel model) Identify differences between products Estimate short and long run effects of demand creation tools and obtain unbiased estimates of effect (dynamic panel estimator)
Aggregate Condom Demand Model Demand (country level male/ female condom sales) as a function of: Price (own) Price of substitutes (P male condom ; Presence of female condom in marketing mix) Income (GDP per capita) Promotion: Mass media advertising and IEC (interpersonal communications) Programme effort (Local Staffing levels) Product maturity Adult HIV prevalence Population Size All prices in 2011 PPP
Data 1: Sales over time Male condom Female condom n=11 n=52 n=3 n=28 n=430 n=155
Data 2: Descriptives Variables Mean (min-max) Sales 15,900,000 (249-197,000,000) Price (ppp$2011) $0.10 ($0 - $0.95 ) Advertising expend. (ppp$2011) Information & Education Campaign expend. (ppp$2011) $778,457 ($0-$8,263,314 ) $467,164 ($0-$12,400,000 ) Product maturity 8.89 (0.2-24 ) HIV adult prevalence 5.53% (0.10%-28.70% ) Mean (min-max) 147,802 (26-4,238,960) $0.37 ($0-$4.14 ) $64,170 ($0-$690,505 ) $28,930 ($0-$441,515 ) 4.21 (1-13 ) 8.16% (0.10%-28.70% )
Econometric approach 1: static panel- OLS and FE ln( q kit ) FC j j j j ln( X j kit ) FC *ln( X j kit ) v i w t kit k= 1, 2; i =1...N; t =1...T, Where : q= natural log of: Sales quantity X = natural log of: Price, Advertising, IEC, Programme Effort, Product Age, HIV, Income, Population k = product type (male condom, female condom) i = country t = time
Results 1: Differences in drivers of condom sales Static model FE, Year as trend, VARIABLES Main effects Constant -84.41 0.385 Ln Advertising 0.192*** -0.192* Ln IEC 0.0812** 0.176* Ln Price -0.123-0.179 Ln Programme Effort 0.131-0.0349 Ln GDP per capita -0.53 0.146 Ln Population Size -0.0364-0.317*** Ln Adult HIV prevalence 0.445* 0.336*** Ln product maturity 0.383** -0.0477 Year 0.0496 Observations 515 R-squared 0.93 Country effects Female condom dummy LR-test (FE versus OLS) 335*** Country effects not shown *** p<0.01, ** p<0.05, * p<0.1
Econometric approach 2: Dynamic panel- systems GMM j j ln( qit ) ln( X j it ) ln( qi, t 1) v i w t it Where : q= natural log of Country Sales quantity X = natural log of: Price, promotion, Programme Effort, country vars λ= sales persistence i = country t = time k= 1, 2; i =1...N; t =1...T, Long run effect: β_ (1- λ)
Estimation results: dynamic panel data model of the demand for condoms Male Condom Female Condom VARIABLES GMM Long-run GMM Long-run Lag Ln Condom Sales (λ) 0.503*** 0.378*** Ln Promotion 0.165** 0.332** 0.185** 0.297** Ln Price -0.085-0.17* -0.365*** -0.587*** Ln Programme effort 0.128 0.256* -0.077 Ln GDP per capita -0.004-0.022 Ln product age 0.102 0.314 Ln Population size 0.191** 0.385*** 0.166* LN Adult HIV prev. 0.0205 0.206** 0.331** ar1p 0.128 0.073 ar2p 0.514 0.141 hansenp 0.201 1.000 Observations 280 85 Number of Countryc 58 26 Robust standard errors in parentheses, *** p<0.01, ** p<0.05, * p<0.1 Parameters for year, country and dummys for: no promotion, free condoms and no staffing and their lags not shown
Limitations Data limitations Income distribution should have been included to account for aggregation bias but was not widely available for full panel Narrow definition of substitute market Potential measurement errors in routine program M&E data. Sample relatively small, though large for subject area
Robustness Estimates largely intact when estimated: Excluding: large countries; Eastern Europe & Central Asia; Without country context variables alternative specification of lags (2-4); excluding 0 value observations; using GMM-Differences did affect results considerably, but Sargan test could not reject GMM-SYS in favour of GMM-DIF
Conclusions Demand stimulation tools affect demand products differently: No one-size fits all marketting for new products Advertising only effective for male condoms IEC important for both products, but far greater impact on female condoms Price elasticity of demand is > female condoms Demand for condoms slow to adjust Sales of both products are persistent, though male condom more so.
Acknowledgments Support for the write up of this work was provided by Leverhulme Trust s International Academic Fellowship I thank University of Toronto for hosting my sabbatical June to August 2015. We thank Population Services International, Research & Metrics Department, Washington D.C. for providing the data. These were analysed independently and the authors take full responsibility for the conclusions drawn there from.