An article forthcoming in the Federal Reserve Bank of New York’s Economic Policy Review—Signal or Noise? Implications of the Term Premium for Recession Forecasting—sheds new light on the sources of the yield curve’s success in predicting U.S. recessions.
As authors Joshua Rosenberg and Samuel Maurer explain, studies have shown that when the yield curve inverts—that is, when short-term interest rates rise above long-term interest rates—a recession has followed in twelve months. One view holds that the ability of the yield curve’s slope to predict recessions stems from interest rate expectations: the markets anticipate an easing of monetary policy in response to an upcoming deterioration in the economic outlook, and the decline in expected future short-term rates drives down current long-term rates.
While the yield curve reflects a term premium in addition to interest rate expectations, term spread models used to forecast recessions typically combine the effects of these two distinct components. The term premium arises because investors require higher yields on long-term Treasury securities to compensate for greater risk.
To understand how these two components affect recession predictions, Rosenberg and Maurer construct a forecasting model that excludes the term premium. They then compare the performance of their model in predicting past recessions with that of the standard term spread model.
The authors find some evidence that the model using only the expectations component predicts recessions more accurately than the standard model. While the authors caution that the historical data are insufficient to produce a definitive conclusion, their findings do support the view that interest rate expectations are the source of the yield curve’s predictive power. Significantly, a model using only the term premium component performs poorly in predicting recessions.
Taking their analysis one step further, the authors compare recent recession predictions from several models. While signals from the standard term spread model indicate an imminent recession, the model with only the expectations component does not signal a recession. To date, the authors note, it is unclear which model’s prediction is correct.
Joshua Rosenberg is a research officer and Samuel Maurer an assistant economist in the Capital Markets Function of the Research and Statistics Group.
Signal or Noise? Implications of the Term Premium for Recession Forecasting ››