This paper studies short-run cyclical behaviour of public (government and social) and private health expenditure and GDP using both time series and panel data techniques. First, national time series data have been used within a multivariate Beveridge–Nelson decomposition framework to construct the permanent and cyclical components. The correlation analysis results for the cyclical components suggest that current public health expenditure is pro-cyclical while there is no clear evidence of a correlation between cycles in private health expenditure and in GDP growth. Next, using an instrumental variable method and the generalised method of moments estimator, provincial-level panel data analyses confirm pro-cyclical impacts of government spending on health. The provincial analysis also suggests that private health expenditure in urban China has a pro-cyclical association with GDP growth, but a lack of good instruments makes it difficult to identify a clear causal link between cycles in income growth and private health expenditure. The results suggest two policy recommendations relevant to public health expenditure, in line with China's current health reforms. Copyright © 2013 John Wiley & Sons, Ltd.