This commentary was written yesterday by Dr. Daniel Thornton of D.L. Thornton Economics on 1/4/21. Thornton spent over three decades at the St. Louis Fed as vice president and economic advisor.
In my previous essay, Requiem for Fed Independence, I noted that in the early 1950s the Fed began conducting its daily open market operations in very short-term Treasuries. They did this with the intent of having little or no effect on interest rates, especially long-term rates.
The Fed believed correctly that interest rates should be determined by the financial markets, not the Fed. Policymakers understood that the interest rate is a price—the price of credit. They also understood that arbitrarily fixing any price is bad. Price fixing distorts prices and, hence, the allocation of economic resources from their best and most productive use. This monetary-policy philosophy ended after Lehman Bros. bankruptcy in September 2008 when the Bernanke Fed began a massive long-term bond buying program.
The intent of the program was to drive long-term rates lower than the market would have. They understood that short-term rates—which the Fed had been influencing since the late 1980s—had little if any effect on spending. They needed to reduce long-term rates for their interest rate policy to have any appreciable effect on spending.
I argued that the Fed’s massive bond-purchasing program, known as quantitative easing (QE), would have at best a tiny and transient effect on long-term interest rates. This essay presents evidence that I was correct.
The figure below shows the 10-year sovereign rates for the United States (US), Germany (GR), the United Kingdom (UK), France (FR) and Australia (AUS) for the period January 1990 to October 2020.
The correlation between any two rates ranges from a low of 94% to a high of 99%. The extremely high correlations among these rates suggest that the five rates are being driven by the same force. Indeed, a statistical method called principal components shows that first principal component accounts for 97.4% of the total variation of all five rates.
The second explains about 1.5% of the variation. The remaining three components account for tiny amounts of the remaining “generalized variance” among the five rates. The correlations between the first principal component and the five sovereign rates range from 97.4% to 99.4%, confirming that these rates are almost entirely being driven by forces that are common to each of the five rates.
The common downward trend in these rates is likely due to the global and country specific downward trends in inflation and the rate of economic growth over the period. The common higher frequency (month-to-month) variation in the rates is a consequence of the fact that financial markets are “efficient.” A large body of evidence shows that financial markets are efficient in the sense that yesterday’s interest rate reflects all of the information known yesterday.
Hence, the change in the interest rate between yesterday and today reflects the response of these rates to today’s new information, i.e., “news.” Note that news is plural. On any given day interest rates may respond to a variety of news.
In the figure the monthly changes in the rates reflects their accumulated response to all of the economically relevant news during the month. The magnitude of the response varies across rates, but the direction of the response is nearly always the same.
The fact that all of these rates are driven by the same factors rules out the possibility that QE reduced the 10-year Treasury rate since late 2008. If it did, it would have had to have caused other sovereign rates to decline too. While the Fed’s long-term bond buying program was large, its purchases are small relative to the size of the global bond market. Specifically, the Fed increased the supply of credit to the financial markets by about $3 trillion between September 2008 and October 2014. While the increase is unprecedented in Fed history, it is small, perhaps even tiny, relative to the size of the global financial market. Such a small increase in the supply of credit over a six year period could have, at best, a tiny and, most likely transient, effect on long-term rates. Consequently, claims that QE reduced long-term rates is absurd. [For additional evidence that the Fed’s purchases did not reduce the 10-year Treasury rate, see Has QE Been Effective?]
Perhaps because he was aware of this dilemma, Bernanke has suggested the Fed’s purchases of Treasuries most likely affected long-term Treasury rates by reducing the term-premium on long-term Treasuries.
The term premium is the additional return an investor requires to hold long-term bonds relative to otherwise equivalent short-term bonds. This is required because long-term bonds have more interest rate risk than short-term securities, i.e., their price changes more than the price of short-term bonds for a given change in the interest rate. An example of the term premium is the interest rate differential between a 10-year and a 2-year Treasury bond The term premium accounts for the fact that this spread in most often positive..
Bernanke and others suggested that QE reduced Treasury term premiums because the Fed’s purchase of a large quantity of long-term Treasuries removed a large amount of securities with high interest rate risk from the market. Bernanke hypothesized that this would cause the term premiums on the remaining Treasury to decline because either: 1) the remaining securities would be held by investors more tolerant of risk, or 2) investors may not require as large of a premium because they would have less risk in their portfolios.
I have shown (Requiem for QE) that this cannot happen. The reason is easy to understand. Specifically, the term premium that an investor requires on a given long-term security depends on two things. The first is the investor’s tolerance for risk, which is unique to the investor and most likely embedded in an investor’s DNA and, hence, unlikely to change. The second is the fact that the interest-rate sensitivity of any long-term security depends solely on the specific characteristics of the security—e.g., its maturity, its default risk, call provisions, collateral, etc. which does not change for the remaining Treasuries.
Neither of those things that determine the risk premium can be affected by the quantity of securities available in the market. Consequently, removing a quantity of 10-year Treasuries from the market cannot affect the interest rate differential between the remaining 10-year Treasuries and say 2-year Treasuries.
Removing a large quantity of long-term securities from a market can only affect the term premiums on the remaining securities if the most risk-averse investors leave the low risk Treasury market, while the most risk-tolerant investors remain. It is extremely unlikely that most risk-averse investors to flee the security of the default-risk-free Treasury market, while the least risk-averse investors remain.
The Bernanke Fed also attempted to reduce long-term rates using what is called “forward guidance.” This policy is the brainchild of Columbia University Professor, Michael Woodford. Woodford’s idea is based on the belief widely held by many economists and financial market analysts that long-term interest rates are determined by investor’s expectation of the short-term rate over the maturity of the long-term security, the so-called “expectations hypothesis,” (EH). Woodford used the EH to suggest that central banks could reduce long-term rates by committing to keep their short-term policy rate low for a longer period than the market would otherwise expect. For example, if the Fed typically began increasing its policy rate—the overnight federal funds rate—shortly after a recession ended, it could reduce long-term rates by committing to keep the policy rate lower for a longer period after the recession ended.
There are reasons to expect that this policy would be ineffective. For one thing, the commitment must be credible. Investors would have to believe that the central bank will not renege on its commitment no matter what happens. But, of course, this is unlikely. Indeed, the Fed did renege. At the December 12, 2012 policy meeting the Fed committed to keep the federal funds rate “exceptionally low” as long as the unemployment rate remains above 6.5%. In an essay I wrote in early 2013 (here) I suggested this was a bad idea because the unemployment rate is strongly affected by the labor force participation rate which had been pushing the unemployment rate lower and over which the Fed has no control. The Fed effectively removed all forward guidance at its March 2014 meeting because further declines in the labor force participation rate had pushed the unemployment rate uncomfortably close to 6.5.
There is a more compelling reason to doubt the EH. Specifically, the EH requires investors to behave irrationally. The reason is simple: financial markets are efficient. This means that investors can predict tomorrow’s interest rate, if and only if, they can predict tomorrow’s news.
But, of course, no one can do this consistently. In order to predict the future interest rate at any horizon, one would have to correctly predict not only the news over that horizon, which is absurd, but the markets’ reaction to it.
No one can do this! Investors know that interest rates are essentially unpredictable. Consequently, the EH requires investors to price long-term rates on their expectation of the future short-term rate, which they know they can’t do: The EH requires investors to behave irrationally! What is the likelihood the interest rate on today’s 10-year bond is determined by investors’ expectation for the federal funds rate over the next ten years? Correct answer: nil!
However, this does not mean that investor’s expectation for the direction of interest rates generally does not affect both long-term and short-term rates. It clearly does. But it does mean that the price of long-term securities is not solely or even principally driven by investors’ expectation of future short-term rates as the EH hypothesis requires. Consequently, it is not surprising that EH has been tested and massively rejected using a variety of long-term and short-term interest rates, sample periods and monetary policy regimes. It is also not surprising that there is no compelling empirical evidence that forward guidance works e.g., see Kool and Thornton (2015).
For all of the reasons presented here, it is extremely unlikely that QE or forward guidance had any persistent effect on long-term rates. My research (here) showed that long-term rates didn’t even respond significantly to QE announcements as Bernanke and some other economists claim. Specifically, the March 18, 2009, announcement that the Fed would purchase up to $ 1.75 trillion in long-term securities is the only QE announcement that had statistically significant “announcement effect” on the 10-year Treasury rate, but not other long-term rates.
This announcement effect was large but offset in little more than a month. This quick reversal caused the Fed Governor, Janet Yellen to rethink the effectiveness of the Fed’s QE policy on Treasury rates, but it did not alter her recommendation for its use (see Requiem for QE for the details).
EDITOR'S NOTE
This is a Hedgeye Guest Contributor piece written by Dr. Daniel Thornton. During his 33-year career at the St. Louis Fed, Thornton served as vice president and economic advisor. He currently runs D.L. Thornton Economics, an economic research consultancy. This piece does not necessarily reflect the opinion of Hedgeye.