Methodological challenges in evaluating health care financing equity in data-poor contexts: Lessons from Ghana, South Africa and Tanzania

Objective – Measurement of the incidence of health financing contributions across socio-economic groups has proven valuable in informing health care financing reforms. However, there is little evidence as to how to carry out financing incidence analysis (FIA) in lower income settings. We outline some of the challenges faced when carrying out a FIA in Ghana, Tanzania and South Africa and illustrate how innovative techniques were used to overcome data weaknesses in these settings. Methodology – FIA was carried out for tax, insurance and out-of-pocket (OOP) payments. The primary data sources were Living Standards Measurement Surveys (LSMS) and household surveys conducted in each of the countries; tax authorities and insurance funds also provided information. Consumption expenditure and a composite index of socio-economic status (SES) were used to assess financing equity. Where possible conventional methods of FIA were applied. Numerous challenges were documented and solution strategies devised. Results – LSMS are likely to underestimate financial contributions to health care by individuals. For tax incidence analysis, reported income tax payments from secondary sources were severely under-reported. Income tax payers and shareholders could not be reliably identified. The use of income or consumption expenditure to estimate income tax contributions was found to be a more reliable method of estimating income tax incidence. Assumptions regarding corporate tax incidence had a huge effect on the progressivity of corporate tax and on overall tax progressivity. LSMS consumption categories did not always coincide with tax categories for goods subject to excise tax (e.g. wine and spirits were combined, despite differing tax rates). Tobacco companies, alcohol distributors and advertising agencies were used to provide more detailed information on consumption patterns for goods subject to excise tax by income category. There was little guidance on how to allocate fuel levies associated with ‘public transport’ use. Hence, calculations of fuel tax on public transport were based on individual expenditure on public transport, the average cost per kilometre and average rates of fuel consumption for each form of transport. For insurance contributions, employees will not report on employer contributions unless specifically requested to and are frequently unsure of their contributions. Therefore, we collected information on total health insurance contributions from individual schemes and regulatory authorities. OOP payments are likely to be under-reported due to long recall periods; linking OOP expenditure and illness incidence questions – omitting preventive care; and focusing on the last service used when people may have used multiple services during an illness episode. To derive more robust estimates of financing incidence, we collected additional primary data on OOP expenditures together with insurance enrolment rates and associated payments. To link primary data to the LSMS, a composite index of SES was used in Ghana and Tanzania and non-durable expenditure was used in South Africa. Policy implications – We show how data constraints can be overcome for FIA in lower income countries and provide recommendations for future studies.