Estimating Monetary Reaction Functions at Near Zero Interest Rates: An Example Using Japanese Data

The importance of truncated distributions for bias in estimation of regression coefficients has been well understood by econometricians, but the relevance of truncation when estimating policy reaction functions has not been fully appreciated. Due to the emergence of low interest rates and the proximity of a zero lower bound (ZLB) on interest rates, coefficient estimates can be biased upwards. This paper illustrates the importance of measuring and correcting estimates for this bias using Japan’s unique experience of prolonged low inflation/deflation. While we would expect the monetary policy reaction function in Japan to differ from other countries in the G4, we show the bias from truncation of the interest rate distribution is significant and needs to be taken into account.