Recent literature highlights the crucial role of understanding the mechanism between political uncertainty and financial market reactions. Along the lines of this topic, our study stresses a clear causal framework. Exploiting one unique natural experiment of the Taiwan Strait Crisis (1995-96), we provide a simple testing strategy which could precisely quantify the effects of political shocks on stock markets. This approach combines the features of one innovative panel estimator and new statistical learning methods for causal inference. Our results indicate, separating true signal from noise via the optimal benchmark, the political crisis had a substantial and significant negative impact on Taiwan's stock prices. This finding is consistent with the empirical evidence of risk premium in recent studies. Moreover, the optimal counterfactual could be an alternative option for the ceteris paribus assumption in non-lab controlled settings. Finally, this study shows predictor selection is needed for a convincing causal estimate in counterfactual studies.