Academic journal article Brookings Papers on Economic Activity

Comment by John C. Williams

Academic journal article Brookings Papers on Economic Activity

Comment by John C. Williams

Article excerpt

This paper by Marco Del Negro, Domenico Giannone, Marc Giannoni, and Andrea Tambalotti sets the ambitious goal of providing a coherent, theoretically founded explanation for why the natural rate of interest, or r*, has declined during the past two decades in the United States. The authors present evidence that a sustained rise in the convenience yield--the spread between the return on assets related to the real economy, like corporate bonds, and returns on ultrasafe, highly liquid assets like Treasury securities--combined with a pronounced slowdown in trend productivity growth, have driven down r*. By incorporating these medium-term real and financial factors in a dynamic stochastic general equilibrium (DSGE) model, this paper makes an important step in advancing DSGE models' usefulness for studying issues like r* (Williams 2017a). In so doing, it builds a bridge between the existing literature on r*, which has primarily used reduced-form models, and the DSGE models used at central banks.

The starting point for this paper is the emerging consensus that r* has declined over the past few decades in the United States (Williams 2017b). The shaded region of my figure 1 shows a range of estimates of r* at each point in time from seven models taken from the literature (Laubach and Willams 2003; Kiley 2015; Lubik and Matthes 2015; Johanssen and Mertens 2016; Holston, Laubach, and Williams 2016; Crump, Eusepi, and Moench 2017; Christensen and Rudebusch 2017). To put these on a consistent basis, in cases where the model uses inflation measured by the consumer price index (CPI), the estimates are increased by 0.23 percentage point to reflect the trend difference in the inflation rates between the CPI and the personal consumption expenditures (PCE) price index (Christensen and Rudebusch 2017). The white line shows the mean of the seven estimates at each point in time.


Although these estimates are based on models that differ in terms of specification, methodology, and data, all the estimates reached historically low levels in recent years. As seen in the figure, the mean estimate of r* fluctuated between 2 and 3 percent in the late 1980s and 1990s, fell to about 2 percent in the early 2000s, and subsequently declined to 0.5 percent in 2016. A striking aspect of these estimates is that they show no signs of moving back to previously normal levels, despite the fact that the U.S. economy has now fully recovered from the Great Recession.

Del Negro and his colleagues add to this literature by estimating a number of vector autoregressive (VAR) models with time-varying intercepts. They use a combination of macroeconomic, survey, and financial market data to estimate these models using Bayesian techniques. This approach builds on the insights, originally due to Sharon Kozicki and P. A. Tinsley (2001), that (i) time variation in intercepts effectively captures most of the instability evident in VAR models, and (ii) survey and financial market data can help pin down these endpoints. In the present paper, the estimates of r* are heavily influenced by the survey and market-based measures.


Despite the intuitive appeal of augmenting macroeconomic data with survey and market-based data, it is worth noting that these measures carry with them their own set of issues. First, because they represent the perceptions of market participants, they are indicators of what people think r* is, rather than evidence of structural change inferred directly from economic data. In a world of rational expectations, this is an advantage, because economic agents are efficiently processing all available information to come up with their estimates of r*. However, in practice, this search for r* has aspects of a "hall of mirrors," where market participants are trying to discern what the Federal Reserve thinks r* is, while the Federal Reserve's economists are using the views of market participants to estimate r*. …

Search by... Author
Show... All Results Primary Sources Peer-reviewed


An unknown error has occurred. Please click the button below to reload the page. If the problem persists, please try again in a little while.