Today, we are pleased to present a guest contribution by Saiah Lee, Ph.D. candidate in economics at UW Madison.
Central banks make monetary policy decisions based on noisy and incomplete real-time data, and reliance on inaccurate data may yield suboptimal policy decisions. In my new paper, Monetary Policy under Data Uncertainty: Interest-Rate Smoothing from a Cross-Country Perspective, I examine if the persistence in policy interest rates are caused by central banks’ sluggish adjustment in the face of data uncertainty, as conventional belief predicts.
“As a general rule, the Federal Reserve tends to adjust interest rates incrementally, in a series of small or moderate steps in the same direction. … Relatively gradual policy adjustment produces better results in an uncertain economic environment.” – Ben S. Bernanke, May 20, 2004.
This conventional belief can be found in the cross-country estimates of Taylor rules in Figure 1, where the data uncertainty is measured by the standard deviation of differences between real-time data and their revisions after a year. However, this data pattern is not representative of the degree of data uncertainty but rather is an artifact of what we as the econometrician observe from ex-post data.
The study builds a simplified New Keynesian model and allows central banks to make inferences about the true data from noisy observations using Kalman filter. I find that central banks adjust interest rates less gradually in the face of more data uncertainty, but it creates an illusion of timid and sluggish adjustment of interest rates because central banks’ inference about the true data is not observable to an econometrician in the ex-post data and the central banks’ inferences are highly correlated with lagged interest rates.
Figure 2 (a) reports the actual central banks’ weight on lagged interest rates, and Figure 2 (b) reports the reduced form estimate of the coefficient on lagged interest rates based on ex-post data. Even if we force the central banks not to place any weight on the lagged interest rates, we have the illusion of timid and sluggish adjustment of interest rates in the face of more data uncertainty, as reported in Figure 2 (c). Central banks’ beliefs about the true data are highly correlated with the lagged interest rates, and they are not observable in the ex-post data.
There are two reasons why central banks adjust interest rates less gradually in the face of more data uncertainty. First, central banks’ learning process effectively, but not completely, filters out data noise. Second, central banks’ learning process helps to tease out additional information about the true data from the persistence of the noise.
The paper is available here.
This post written by Saiah Lee.