U.S. zero-coupon bond yield data, 1991-2004

Buckley, George (2016). U.S. zero-coupon bond yield data, 1991-2004. [Data Collection]. Colchester, Essex: UK Data Archive. 10.5255/UKDA-SN-852298

Data description (abstract)

This data collection consists of 2 data files (ZEROYLD3.txt and ZEROERR3.txt), giving the zero-coupon bond yields implicit in coupon-paying bonds and the associated standard errors, respectively. Data are given for 166 months for the period 03/1991-12/2004, and for 56 maturities which are monthly from 0 to 18 months, then quarterly to 2 years, then semi-annually to 3 years, then annually to 35 years, and finally for 40 years.
This is an extension on data from McCulloch and Kwon(1993), (ZEROERR1.txt, ZEROERR2.txt, ZEROYLD1.txt, ZEROYLD2.txt).

The primary objective is to test whether behavioural models can explain the overwhelming evidence that yields on long bonds do not move in the way predicted by economic models. Theory implies that the yield on a long bond is determined by the expectation of the short yield over the life of the long bond, henceforth the REH. This gives rise to a number of testable implications for the movement of bond yields and these are widely rejected. The core idea that characterises Behavioural Finance is that investors may be subject to the same behavioural biases when they trade in financial markets that have been widely demonstrated by psychologists in laboratory experiments. Models that build on these two biases have been successful in explaining well-established anomalies in equity markets, especially short-term momentum and longer-term reversals in equity returns. If investors display this type of bias in the equity market when forecasting company earnings, we would expect them to display the same bias in the bond market when forecasting the short rate.

We also investigate whether the apparently irrational movements in long rates can be explained if we assume investors are rational but are uncertain about the model of the short rate. In this approach the rejections of the REH are not due to a failure of rationality but a failure of information. The conventional definition of the REH assumes expectations are generated “as if“agents know the true model. The key idea in the learning literature is that this information assumption is too strong. Movements in stock prices and returns may therefore reflect the process of learning. Learning may result in systematic patterns in stock returns that look remarkably like those that result from behavioural biases.

Downloads

Altmetric