An excellent proxy of the stance of the monetary authority is the yield on the two year note. As the Fed has hiked at a pace unseen since 1994, the two year yield has responded dramatically. It has risen from essentially 0.0% to 4.5% as of Friday’s closing auction. After so many years of rock-bottom rates, that sure sounds like a lot. How much is it exactly? Just how tight are monetary conditions?
While the two year yield is a good proxy, it doesn’t tell us whether the present level is appropriate given prevailing macro conditions. The same 100 basis points (bp) of hikes may be too little in an environment where inflation is high and accelerating, and too much under more benign conditions. In order to get a handle on the question we need a model of “appropriate” rates conditional on prevailing macro conditions.
The usual approach to this problem is to use some version of the Taylor rule. These alternatives yield widely varying estimates of the appropriate rate. We follow a simpler, less theoretical and more empirical approach. Instead of asking what rates ought to be in theory, we ask what rates would look like, conditional on a small set of important macro variables, if the Fed’s response function has not changed since the Volcker coup.
In other words, we give up entirely on theory. Instead, we want to pin down the answer to a simpler question: has the Fed’s response function changed? That is, has the Fed become more hawkish or dovish in the recent past relative to the historical norm since the 1980s? Ultimately, we want to know whether the Fed is responding too much or not enough. We cannot answer that question in general. What we can answer is whether the Fed is responding too much or not enough, relative to its historical response function.
We pull data on two year yield, unemployment rate, headline CPI, core CPI, and the oil price (WTI) from FRED. We compute year-on-year inflation rates. We resample quarterly to allow the Fed to respond to innovations in the macro variables. And we detrend by first-differencing. Our response is the two year yield. We estimates single-factor models using OLS, including a lagged term for the response (necessary to get rid of autocorrelation). The next table displays our estimates. DW is the Durbin-Watson test statistic that tests for autocorrelation (DW near 2 means we’re good). The elasticities are all significant and bear the expected signs.
Next we use these four models to obtain predicted values for the two year yield. We then take their average to get ensemble predictions. In order to visualize the results, we compute six-quarter rolling sum of the changes in two year yields and the predicted values of the same. Since monetary policy works with a lag, we’re interested in this cumulated series more than the underlying shocks (where we have to do the estimation).
The gap between the actual yield and the predicted yield contains information on the relative tightness of monetary policy. We call it the Tight Money Index, which is displayed next. The graph reveals six significant tight money episodes: 1988Q3, 1994Q2-1995Q2 (when Greenspan went berserk), 2000Q3, 2005Q1-2006Q2, 2018Q2 and 2022Q3. Also clear are four cycles of extraordinary monetary accommodation: 1992Q2-1993Q3, 2001Q2-2004Q1, 2008Q1-2009Q3, and 2020Q2-2021Q3.
The accommodative episodes make perfect sense — they all happened during or in the aftermath of US recessions (next graph). We don’t find much evidence that tightening episodes plunge the economy into recession. The only case where a tight money episode is clearly followed by a recession is the 2000Q3 episode. At any rate, there are already too many false positives in our small sample. Unlike the slope of the yield curve, then, the Tight Money Index does not predict recessions.
The Yellen Fed made a policy error when it began to hike in Dec 2015 — after bamboozling Lael Brainard into signing up. It followed up the initial quarter-point hike (oh, the good old days of the New Normal) with guidance for more hikes. We called that error in real time. Yellen was soon forced into a course correction.
In the aftermath of the Covid recession, the Powell Fed made a policy error in the other direction. They expected inflation to be transitory. But the damn thing underwent a metamorphosis. What had seemed like transitory, pandemic-related price instability morphed into something altogether more dangerous — entrenched, broad-based inflation that is threatening to destabilize the inflation expectations of price setters.
Powell seems to have realized the error around the same time as we did, in November 2021. Monetary conditions did not really tighten until the summer. Of course, monetary policy works cumulatively (which is why we’re looking at six-quarter rolling sum of changes). Because the Fed was so slow off the mark, it now has to hike faster and further than it would otherwise have. This problem was pointed out to the Fed through the summer for 2021 by people who were more informed than your analyst.
Of course, the present struggle against inflation is very far from over. So far, there are no signs that underlying inflation has slowed. To the contrary, underlying inflation is still accelerating. And our expectations signal — the lowest volatility quintile — has climbed to 5.9%, as of the last print. Recall that the signal-to-noise ratio of this measure is higher than any other measure of underlying inflation.
Markets are now pricing a terminal rate of about 5.0% by the end of 2023. That is almost surely too rosy. In order to get this sucker down, the Fed may have to go considerably further. That means that last week’s enthusiasm in risk assets is likely to be short-lived; a dead cat bounce if there was ever one.
As of writing, US monetary policy is no tighter than it was in 2018, all considered; and well short of the mid-2000s, not to mention Greenspan’s fuck-up that caused the Bond Market Massacre of ‘94. But we’ll keep an eye on our Tight Money Index, and try to alert you, dear reader, when conditions are finally ripe for a Fed pivot.
It might be a minute.