Triangulating health expenditure estimates from different data sources in developing countries: the case of...

This article deals with the accuracy of statistical records used for political decision making and international comparative analysis. In developing countries, even major macroeconomic indicators can include data inadequacies and methodological differences in data generation between statistical agencies. Existing data show that total health expenditure as a percentage of GDP is about 50% lower in Pakistan than in other low-income countries (LIC). To determine whether these results reflect the actual situation in Pakistan or whether they are due to statistical error, Pakistan produced National Health Accounts (NHA) for the first time in 2009 to assess health spending in 2005-6. Improved NHA estimates are also being made for 2007-8, which will be based on the following: public expenditure data published with time lags; survey results for 2007-8; and multivariate analyses of data from 2010 and 2011 surveys on health-specific out-of-pocket (OOP) expenditure, healthcare providers, non-profit institutions and census data on autonomous bodies and large hospitals. Since these data are not yet available, a best estimate of health expenditure has to be made to support policy decision making and to provide a point of comparison for future NHA results. Health expenditure data are available from different data sources and estimates have been made by applying different methods, leading to a range of health spending estimates. As a result of this diversity of estimates and data, each with its own inaccuracies or gaps, there was a clear need to triangulate the available information and to identify a best possible estimate. This article compares estimates of household health expenditure from different sources, such as the Household Integrated Economic Survey, the Family Budget Survey and National Accounts (NA). The analysis shows that health expenditure figures for Pakistan have been underestimated by both WHO and the NHA. An adjusted estimate shows OOP spending to be twice as high as previously thought. Previous per capita total health expenditure estimates ranged from $US16 to $US19. The revised estimate showed per capita total health expenditure to be $US33, based on NA data. This puts Pakistan in a different position in international comparisons, with health expenditure exceeding the level of India ($US32.5) and the average of all LIC ($US24.5). Methodological differences in estimating expenditure and the multiple and conflicting estimates might cause stakeholders to make potentially adverse or even erroneous policy decisions on the allocation of resources. Because policy makers make decisions based on the estimates provided, the provision of a best estimate, made following a review of the advantages and limitations of existing sources and methods, is key.