Economists have long used macroeconomic variables to predict the Fed Funds rate — the short end of the yield curve for U.S. Treasury bonds. According to the Taylor rule, the Fed Funds rate should depend on inflation and real economic activity (GDP). But the rest of the yield curve has traditionally been the domain of bond traders, who use financial models based on characteristics of the yield curve — its level, slope and curvature — to predict yields. Although traders understand that yields are related to macroeconomic fundamentals, their models do not typically include macro variables.
“Practitioners are trading the whole yield curve, so they want to figure out how different prices are related to each other,” says Professor Mikhail Chernov. The problem with the standard finance approach is that it relies on endogenous variables, essentially describing yields with yields. “It’s basically a mechanical procedure that says, ‘It’s hard to track 10 yields; let’s track three factors, and we will be able to characterize the 10 yields,’” Chernov says. “But there is no explanation of what actually drives those three factors.”
By combining insights from finance and macroeconomics, Chernov and doctoral student Ruslan Bibkov developed a method for predicting Treasury yields using inflation and real economic activity. They found that a simple model based on these two variables predicts the Fed Funds rate with 50 percent accuracy. But since the yield curve reacts to the variables with a delay, the researchers devised a way to bring historical values into the model, raising its accuracy to 80 percent. Their model obeys the crucial no-arbitrage rule, which ensures that all of the yields are mutually consistent — that the long bond is not overpriced relative to the short bond, for example.
So how do you explain the 20 percent of the yield curve that diverges from the model? It turns out that these gaps correspond to monetary policy shocks — periods when the Fed is taking unusually aggressive measures to combat inflation. “There was a famous inflation scare in 1984, which shows up in the graph as a big deviation from our rule,” says Chernov. “There’s another famous episode in 1994, when Greenspan raised the interest rate by 75 basis points in one shot. Again, it was a situation where the Fed wanted to send a very clear signal to the market that it was fighting inflation.”
These deviations from the model are highly correlated with the spread on AAA corporate bonds. Since AAA bonds are virtually risk free, the spread between AAA yields and Treasury yields reflects the liquidity premium on Treasury bonds. “This premium of course varies over time, depending on economic conditions,” Chernov says. “So when the Fed is moving the interest rate one way or the other — effectively throwing cash into the market or withdrawing cash from the market — that liquidity effect is reflected in this totally separate market of quality corporate securities.”
Although the model explains 80 percent of the yield curve’s level, it explains only 55 percent of the curve’s slope. The gaps between the actual slope and the predicted slope appear to be linked to the federal budget deficit: a higher deficit corresponds to a higher slope. Although the correlation is not perfect, it suggests that a large movement in the budget deficit will affect the long end of the yield curve. Some economists have argued that the budget deficit has a major impact on bond yields, while others claim that it has virtually no effect, so this finding brings new evidence to the debate.
The research, which offers bond traders a relatively simple method of predicting bond yields, has generated considerable interest in the finance community. “This approach is very attractive relative to the standard approach,” says Chernov, “because with the standard approach, if you’re explaining yields with yields, you still have to predict something. Here you can build scenarios using published data and surveys of different professional forecasters. There could be surprises, of course, but still the band of possible values is pretty narrow, so people have a pretty good idea of what can happen.”
One constituency that has shown particular interest in the new approach is the Federal Reserve, which has invited Chernov and Bibkov to present their findings in Washington, D.C., New York and Atlanta. “Currently the Fed is extremely concerned with how the markets perceive what they do,” Chernov says. “This is very important for them, because one of the purposes of monetary policy is to establish credibility. Our approach gives them exactly the tools to extract information about the markets’ beliefs, because all we are doing is using market data on Treasuries and a little bit of macro data. Effectively, we are asking, how do the markets interpret these data?”
Mikhail Chernov is associate professor of finance and economics at Columbia Business School.