Abstract:
This paper explores two very different models which might account for stock market crashes. A key innovative feature of our paper is that we use the models to show how their implications for stock market crashes may be tested using switching-regression econometrics. We are careful to show that our switching regressions reveal new patterns in the data which go well beyond known stylized facts (such as time-varying volatility, the leverage effect, and mean reversion.) We refer to the first model, which is based on historical accounts of "manias and panics," as a model of speculative behaviour; its key features are that "overvaluation" increases the probability and expected size of a stock market crash. We refer to the second model, in which there are regime switches in epxected dividend growth rates, as a model of switching fundamentals. Our results suggest that both speculative behaviour and news about fundamentals may be needed to explain stock market crashes.
JEL-codes:C1C2C3C4C5C8 (search for similar items in EconPapers) Date: 1995-02-07 Note: 45 pages of text & 8 pages graphs. Text and Graphs in separate Postscript files. Both files compressed in a single Info-zip archive, then uuencoded. View list of references