Term Structure as a Leading Indicator for Output

Abstract

The term structure defined as the spread between a long-term (10 years) and a short- term (3 months to 2 years) government yield rate accounts for a non negligible predictive power for both output and ination under most circumstances. Despite the fundamental drawback that no standard theory model explains this relationship Estrella (2005) de- rived a rational expectations model which can be analytically solved. Estrella and Trubin (2006) proposed a probit model which is extended in this thesis to the Bayesian Regression Framework and the Gibbs Sampling technique to a collection of thirty countries.

The general topic of interest covers the econometric measures to assess the interconnection be- tween Government Yield Rates and the Gross Domestic Product output. Estrella and Hardou- velis (1991) formulate a probabilistic model using the US Government Yield Rates to forecast possible recessions in the US Economy. There exists an associated relationship between a positive slope in the yield curve with a future increase in the (real) economic activity level. According to the authors, this measure (yield curve slope) historically gives a predictive power that is independent of monetary policy and outperforms indicators such as surveys and lagged growth in economic activity. The attening phenomenon predicts a drop in the future interest rates which is associated with a lower level of real GDP. The next step that I will introduce is to extend this model to include a diversity of countries to study the connections between GDP output and government yield rates among different nations. Historic Example: In the case of the US, the flattening of the curve predicted the imminent recessions of the early 90s, early 2000s and the 2007-2008 crisis. Moreover, a considerable set of countries suffered the same yield curve flattening prior to the latter crisis.

One of the most recurred explanations about the reason why yield spreads have a predictive power over the future output is given by the notion that yield curves tend to `flatten' whenever there is a monetary policy tightening, which is itself related to a slowdown in the future levels of economic activity and ination. Recall the IS model as the downward slopping curve which considers the Gross Domestic Prod- uct as the dependent variable and the interest rate as an independent term. Remind as well of the Phillips curve as an inverse relation between inflation rate and unemployment.

There seems to be, therefore, a predictive role contained in the spread between the long-term and the short-term yields. The effect of the yield, however, is affected by all three parameters of monetary policy: gy, g_pi and gr. Let us take a closer look to the last equations derived in the model above. In the extreme positive scenario, where gr = 1 and g_pi = 0, the yield spread would thereby become the solely remnant predictor1. In the other sense, a very negative scenario would develop if both g_pi and gy are positive and very large, yet gp/gy tends to a constant value. In such case the coeffcient on the yield spread would become diluted. Apart from both extreme scenarios, the model suggests a predictive role of the yield spread in the expectations of output and ination. As a final remark, the model denotes that the role of the yield spread is not structural and may vary along time in case of changes in the monetary policy. We will see in the next section how the coeffcient for the yield spread in the probit models has structural changes after economic recessions occur and new monetary policies are enforced.