Term spread and real economic activity in Korea: was the crisis predictable?

Abstract

This paper has three objectives. First, it examines the link between the term spread (difference between long-term and short-term rate of interest) and GDP growth in the Korean economy for the period 1980–1999. Second, it tests for the independent information content of the term spread by including current and expected monetary policy indicators. Third, it explores the usefulness of the spread as a leading indicator of recessions and poses the question, was the crisis of 1997–1998 predictable?

Acknowledgements

We are grateful without implication to David Peel for helpful comments and advice.

Notes

1 Unit root tests, augmented Dickey–Fuller (ADF) and Phillips–Perron (PP), have been conducted for the one-to-one quarter change in GDP growth and real money growth. In all cases, we were able to reject the null hypothesis of unit root at the 1% significance level. In the case of the spread, the ADF test was –2.37 and the PP test –2.96, rejecting the null of unit root at the 5% and 1% level respectively. All variables are then considered to be stationary, I(0) and no cointegration analysis is needed for estimation purposes.

² The standard errors of the OLS estimations are corrected for autocorrelation created by overlapping forecasting horizons and for heteroskedasticity using Newey and West (1987) variance–covariance matrix.

3 For a theoretical and empirical discussion, see Estrella and Hardouvelis (1991), Plosser and Rouwenhorst (1994), and Estrella and Mishkin (1997).

4 The correlation coefficient between k-year real GDP growth and k-year inflation rate and k-year money growth for the period 1980–1999 are the following:

Table

The variables have included stock prices, interest rates, monetary aggregates and macroeconomic indicators such as growth in real GDP, trade-weighted dollar and housing permits.

In-sample estimations by Stock and Watson (1989) using a logit specification and Estrella and Hardouvelis (1991) using a probit model on US data have shown similar performance by the yield spread.

UK, Japan, France, Italy, Germany, Belgium, The Netherlands and Canada.

One exception was identified by Bernard and Gerlach (1996) in the case of Japan, where the predictive power was lowest.

Specifically, we have taken the value of 2% as the threshold, that is, we have dated X _t as 1 in the quarters where the quarterly real GDP growth was lower than 2% two quarters in a row.

The literature suggests that the spread of an influential economy, such as the US, could have explanatory power over the output of other economies (Plosser and Rouwenhorst, 1994). Further, the US and German spreads have been proved successful when trying to predict recessions in other economies such as the UK, Japan or Canada, using probit estimations (Bernard and Gerlach, 1996). We have then included the spread of other economies that could influence the Korean economy in the probit estimations. In particular, we used the US spread and Japanese spread in our calculations but the results were unsatisfactory; those foreign spreads were insignificant.