Unbiased Analysis of Today's Healthcare Issues

Government Expenditures and Health Outcomes

Written By: Jason Shafrin - Apr• 25•07

Do increases in government spending affect health outcomes? While this seems like a simple question, proving whether or not spending impacts outcomes is difficult. There are questions of reverse causality: the governments of countries or regions with more serious health problems ceteris paribus may decide to increase their allocation of health spending; thus one may erroneously conclude that government spending worsens health outcomes. Also, whenever one examines different regions or countries, a researcher must take into account heterogeneity across these geographical units. For instance, observing that Florida has higher Medicare spending and a higher death rate may not imply that government spending increases mortality, but simply that Florida has a higher percentage of elderly patients. Further, raising spending levels may increase the amount of unnecessary procedures preformed, and thus worsen health outcome measures.

Two studies which analyze government spending and health outcomes are papers by Bokhari, Gai, and Gottret (2007) and Byrne et al. (2007), both published in the Health Economics journal. The first paper analyzes cross-country government spending variation and the second looks at regional disparities in Veterans Affairs (VA) spending across U.S. regions.

Bokhari, Gai and Gottret (2007)

To control for the problems above, the Bokhari paper controls for income level (GDP), the level of donor funding, the deviation in donor funding from its historic average, and some infrastructure variables such as literacy levels, miles of roads in the country and measures of access to improved water sources and sanitation. The authors also use an instrumental variables approach to control for endogeneity in income and government health expenditures. The instrument for income is the consumption-investment ratio since the authors claim that it is correlated with GDP per capita, but not with infant mortality (the dependent variable). The authors use the military expenditures of a nation’s neighboring countries as an instrument for the proportion of spending on health. This is a decent instrument–better than using a country’s own military expenditures–but if there was a war it would be likely that another country’s military expenditures would be correlated with infant mortality measures.

While cross-country regressions should always be viewed with some skepticism, the authors do find that increasing government health expenditures decreases infant mortality as well as maternal mortality. According to the authors, “[t]he elasticity of under-five mortality with respect to government expenditures ranges from -0.25 to -0.42 with a mean value of -0.33. For maternal mortality the elasticity ranges from -.042 to -0.52 with a mean value of -.050. For developing countries, [the] results imply that while economic growth is certainly an important contributor to health outcomes, government pending on health is just as important a factor.” In developing countries where many of the top health problems come from contagious diseases, one would expect public health efforts to be particularly effective in reducing mortality rates.

The authors do wisely qualify their claims by stating that increased government health spending in countries with corrupt government can reduce health outcomes. Also, one may worry that increased government health spending may decrease government spending in other areas important to health (e.g.: water works, utilities, network of roads, and education) . For instance, having poor roads may prevent the population from easily accessing care in a hospital or outpatient facility.

Byrne, Pietz, Woodard and Petersen (2007)

The Byrne et al. paper looks at health care funding and risk-adjusted mortality in 22 VA geographical networks over a six year period. The risk-adjustment is accomplished by controlling for Diagnostic Cost Groups (DCG).

The authors conclude the following: “in cross sectional regressions that VA Networks with higher funding have lower risk-adjusted mortality when all male veterans were analyzed. However, when we analyzed a multi-year data set consisting of six years, using a hierarchical linear regression with clustering on Network, funding levels are no longer significantly associated with mortality, but Network was highly significant. This indicates that some characteristics of the Networks themselves are driving this result.” The authors, however, found a positive correlation between spending and poor health outcomes for the sickest patients in the sample. One can not be sure if this is due to increased spending leading to unnecessary procedures or because unobserved sickness levels are correlated with mortality.

Overall, it seems that in developed countries, government health care spending is not strongly correlated with health outcomes. In developing countries, however, government spending has a positive association with health outcomes likely due to public health efforts to control infectious diseases.

You can follow any responses to this entry through the RSS 2.0 feed. Responses are currently closed, but you can trackback from your own site.