Hi,

Granger-Causality is mainly used in economics and finance research and it can be used in other disciplines as well. Therefore, most of the literature are in those two fields. Since you have only two time series, you can go gor pair-wise granger causality tests to assess the nature of causality, i.e whether the causality is uni directional or bi directional if there is causality exists between the two variables.

Please see below for some articles in your field.

https://www.tandfonline.com/doi/abs/10.1080/00036840601019083

https://www.tandfonline.com/doi/abs/10.1080/13504850010017357

https://www.sciencedirect.com/science/article/pii/S0304407602001458

Article Is health capital formation good for long-term economic grow...

http://www.emeraldinsight.com/doi/pdfplus/10.1108/JPBAFM-20-04-2008-B002

https://www.sciencedirect.com/science/article/pii/S0165176512001814

Good luck with your research.

Kind regards

Thushara

Well, predictive causality is quite different from treatment evaluation. The statistical models behind each of those two tests are not the same at all. Please provide the specific question you are adressing and how the time series you have are related to your goal of evaluation health policies.

In Granger causality tests a key point is the election of lags to be included. results may be sensible to the number of lags considered. I would suggest you to perform a robustness check running multiple Granger causality tests with varying lags and evaluate if results change. Also, look if the sign of each coefficient maintains (it should in policy evaluations).

Finally, why do you think that Granger causality is the best approach to apply in your problem?