Introduction: This paper applies panel vector autoregression (VAR) to data on health spending by US counties and county life-expectancy. Methods: Life expectancy data for 3133 counties for 1985 to 2010 were included in a panel VAR together with reports of logged per capita non-hospital-related health spending, hospital health spending, and total county spending by each county government. Each variable depends on its own lags and the lags of all the other variables in all the counties in the state. Granger causality was tested using a Wald statistic on the joint significance of all of the lags of non-hospital health spending on life expectancy in each US state. Results: Granger causality between county level non-hospital health spending and life expectancy was found in 14 states, more than would be expected by chance. States with gains, gained a cumulative 9 days of life expectancy as the result of an impulse of $100 per capita. This is equivalent to $3868 per life year gained in the 14 states where the association holds. Conclusion: County government health spending does not cause measurable improvement in population health in every state. The health spending strategies used in successful states need to be identified and spread.