Grasping Reality on TypePad, by Brad DeLong

Taylor (1999) presents two versions of his rule. In the first, Rule #1, each 1% of GDP gap vis-a-vis economic potential induces an 0.5%-point shift in the target interest rate. In the second, Rule #2, each 1% of GDP gap vis-a-vis economic potential induces a 1.0% shift in the target interest rate: the policy rule reacts more strongly to both booms and busts. In both versions, each 1% of difference between actual and target inflation induces an offsetting 0.5%-point shift in the real, and thus a 1.5%-point shift in the nominal, short term safe interest rate dictated by the policy rule.

It is fair to say that this approach took the macroeconomic policymaking community, at least here in the United States, by storm. Within three years after the publication of Taylor (1993), some of the most senior policymakers in the Federal Reserve system were noting how the Taylor Rule was extraordinarily helpful and was providing essential intellectual discipline to the FOMC’s discussions.

Taylor (1999) used this framework to describe and assess monetary policy since the early 1960s. In his view then, it made little difference whether one took the rule in its more aggressive or less aggressive variant in its responsiveness to the output gap. With the exception of the unemployment rate peak at the nadir of the Volcker disinflation itself, output gaps had not been large enough and certainly not persistent enough to cause the two versions of the rule to produce significantly different policy recommendations.

By this yardstick Taylor found that monetary policy in the early 1960s had been too tight; in the late 1960s and through the 1970s too loose; “on track” over 1979-81 and 1985-1997; and too tight over 1982-84. The magnitudes of these deviations of actual policy from what the Taylor rule recommended were, however, of wildly varying sizes. The “too tight” policies of the early 1960s generated 12%-point-years of cumulative excess monetary tightness. The “too loose” policies from 1965-1978 generated 65%-point-years of cumulative excess monetary looseness. (And, indeed, over that era expectations of future inflation went from roughly 3% to roughly 10%.) The “too tight” policies of 1982-4 generated 10%-point-years of cumulative excess monetary tightness; and deviations of actual interest rates from those projected by the Taylor Rule over 1985-1997 were, in Taylor’s (1999) estimation, too small to matter.

One might therefore presume that assessing how different standard interest rate-based monetary policy has been in this most recent downturn and recovery would be straightforward: Take the Taylor Rule. Assess by how many cumulative percentage-point-years short-term safe nominal interest rates have deviated from the Taylor Rule value as a result of the zero lower bound. And you are done.

The difficulties are more than threefold:

• First, the rise in the unemployment rate in 2008-9 and its high value thereafter mean that, for the first time, the gearing between the output gap and unemployment rate on the one hand the the short-term safe interest rate on the other really matters: one’s view of what was following the policy rule varies significantly depending on whether one’s preferred policy rule as the gearing of Rule #1 or Rule #2 in Taylor (1999).

• Second, the early 2000s see the appearance of what Bernanke (2005) characterized as the “global savings glut”, and what appears to be a consequent secular fall in the Wicksellian neutral safe real interest rate. Did the policy rule we were following back before 2008 implicitly include a one-for-one adjustment for shifts in the Wicksellian neutral rate as estimated along the lines of Laubach and Williams (2003)? Tastes and views differ.

• Third, there is no reason to believe that the coefficients of the true policy rule as it stood before 2008 were linear—that there were no cross-terms by which the appropriate response to an unemployment rate rise in terms of percentage-point reductions in interest rates varied depending on whether the Federal Reserve was in inflation-reduction or full-employment mode. There is not enough data to estimate such cross terms, even though the scatter-plot differences between “murder” and “financial crisis” episodes suggest that there might be. And whether or not policymakers act as if there are or are not such terms, they do not think that way.

• Fourth, we have not yet noted the knotty question of whether the unemployment rate remains a sufficient statistic for the state of the labor market, let alone for the output gap, after 2000.

These difficulties matter. Depending on whether one takes Taylor (1999) Rule #1 or Taylor (1999) Rule #2 adjusted for Laubach-Williams (2005), one sees either 5%-point-years or 2%-point-years of excess looseness over 2002-2006. And Higgins (2016) calculates that, depending on whether one follows Taylor Rule #1 or Taylor Rule #2 adjusted for Laubach-Williams (2003), either one sees interest rate policy from 2008-12 roughly following Taylor Rule recommendations and then 4%-point-years of excess looseness since 2013; or one sees 40%-point-years of excess tightness induced by the zero lower bound between 2008 and the present—excess tightness that is only now dissipating. How one comes down on whether interest rate policy has or has not been sharply constrained since 2008 and thus made monetary policy in this recovery very different from previous patterns.