Have public health spending and access in South Africa become more equitable since the end of apartheid?

Abstract

This study investigates whether health spending and access to services in South Africa have become more or less pro-poor over time. We find that over the post-apartheid period health spending has become significantly more pro-poor. In addition to the rising share of the health budget allocated to public clinics, there has been an increase in the share of public clinic and hospital spending going to the poor and a rising share of the health budget allocated to public clinics. In addition, between 1993 and 2008 there were improvements in both financial access to public health services – as measured by the incidence of catastrophic costs – and physical access to public health facilities – as measured by reduced travel time. Given that substantial progress has been made with fiscal equity and access to health, problems that users complain about – rude staff, long queues and lack of medicine – have moved higher on the policy agenda.

Keywords:

• fiscal incidence
• equity
• benefit incidence
• financial protection
• health care utilisation
• access
• health expenditure

JEL codes:

• H51
• I18

Acknowledgements

The findings, interpretations and conclusions expressed in the paper are entirely those of the authors, and do not represent the views of the World Bank, its Executive Directors, or the countries they represent.

Notes

¹Between 2000 and 2008 real public health spending per capita increased by 5% per year. Prior to 2000 it is more difficult to obtain comparable spending estimates, but the work of McIntyre et al. (1998) indicates that there was a slight decrease in real public health spending per capita between 1995 and 1999. However, the relative magnitudes are such that the growth in per capita real spending in the second period dominates the slight decrease in the first period.

²For example, the constant unit cost assumption (where each patient visit is assumed to cost the same) or the constant unit subsidy assumption (where each visit is assumed to be subsidised by the same amount, regardless of patient or whether it is the first, second or last visit).

³It is encouraging that fairly similar patterns are obtained when using alternative welfare indicators to examine spending incidence in 2003. The results are thus not an artefact of the modelling process used.

⁴In 1993 the recall period (i.e. the length of time that the survey asked respondents to recall) was two weeks, whereas it was a month in 1995 and 2003, so the 1993 survey estimate was adjusted to be more comparable with that of 1995 and 2003. Estimates for 2008 are not shown, as questions were not comparable to those asked in 1993, 1995 and 2003. The 1995 and 2003 questions were reasonably comparable, but in 1993 a sentence was added (‘This includes people who have some form of permanent injury, disability or ailment’) to make it clear that the question also referred to chronic conditions.

⁵Note that comparability between the top quintile in 1993 and 1995 may be compromised because the analysis assumes that medical scheme members always paid for their services and the 1993 survey does not capture medical schemes membership. This assumption regarding payment by medical scheme members was introduced because of inconsistencies in how the question on payment for services was perceived and answered by medical schemes members across survey years. The assumption does not affect our analysis much because utilisation of public facilities is rare amongst medical scheme members and the focus of the research is on the uninsured and those at the bottom of the distribution.

⁶Medical scheme members are excluded from both these estimates because of the variation in how comprehensively the expenditures of medical scheme members were captured, and because the focus of this analysis is on the poor.

⁷On the basis of their 2006 SACBIA survey, McIntyre & Ataguba (2009) report that average out-of-pocket health expenditure represented less than 2% of total expenditure.

⁸This table is based on a pooled version of the GHS surveys. The wealth quintiles are based on a wealth index that was compiled using multiple correspondence analysis (MCA) to weight a series of characteristics and household assets. MCA is similar to principal component analysis, but considered more appropriate when the variable list includes discrete variables with no cardinal interpretation. Respondents were asked whether they had complaints about any of seven potential problems – with ‘other’ as an additional category. We report the prevalence of ‘yes’ responses to the five most frequently cited complaints. The other two complaints not reported here were ‘opening times not convenient’ and ‘facilities not clean’, both rarely cited as complaints by respondents.