A parametric estimation of the policy stance from the central bank minutes

I estimate the policy stance on the monetary policy rate decision, which stands for the intensity of the committee’s consensus to adjust the policy rate, with the minutes of the Bank of Korea from 1999 to May 2018. The policy stance is estimated more directly without relying on external dictionary information by using a parametric model that combines two observations-the text of the minutes and the actual policy rate decision, both of which are closely related to the policy stance. I also propose an algorithm matching non­consecutive collocations to improve the quality of the Korean text data. The estimated policy stance is applied into a vector-autoregression model, and shows that it improves the identification of the policy rate shock without losing many degrees of freedom. The results show one example where the text data work as an alternative to the traditional economic data, possibly even more parsimoniously and efficiently