Parameter stability in the market model: tests and time varying paramete r estimation with UK data

Abstract

This paper investigates parameter stability in the market model using a data set constructed from the companies that have remained in the London Stock Exchange FT-SE 100 index for the first decade of its existence. A battery of tests of structural stability are performed, finding that parameter instability is prevalent in the fitted market models. Two specific time varying parameter models are then investigated: a smooth transition regression and a random walk parameter model. Shifts in -parameters at certain non-firm-specific events, particularly the deregulatory Big Bang of October 1986, the crash a year later and the general election and later European exchange rate mechanism crisis of 1992, are found to be a feature of the data set.

Time Varying Beta Risk: An Analysis of Alternative Modelling Techniques

“…In addition, the random walk model took a shorter time on average to converge to a solution than any other model. Given that this model appeared to dominate all other models, we follow Coutts, Roberts and Mills (1997) in carrying out a test of parameter constancy for the random walk specification and unsurprisingly find that the null hypothesis of parameter stability is rejected at the 5% level for every industry in the sample.…”

European stock market integration: Fact or fiction?

“…What constitutes a sufficiently long period, however, seems to be disputed in financial literature. For example, Coutts et al (1997) argue for shorter periods based on the likelihood of structural breaks occurring in stock price data over a 5-year period. Brav and Lehavy (2003) are using a cointegration approach to study the long-term behavior of markets and prices based on a 500 trading day sample.…”

The time profile of risk in banking crises: evidence from Scandinavian banking sectors

“…7 First, smooth transition regression models (STR) are utilized, a recent contribution in nonlinear modelling of economic timeseries. The application is based on the framework recently proposed by Granger and TeraÈ svirta (1993), Lin and TeraÈ svirta (1994), TeraÈ svirta (1994) and applied by Coutts et al (1997). Important for this exercise, the STRmodel lends itself to both testing whether the parameters change continuously over time and estimating a modi®ed market model with time-varying parameters.…”

“…Hyytinen 2 Even though there exist many studies suggesting that the basic version of the market model is empirically mis-speci®ed, it still has remained as the most frequently invoked workhorse for the bulk of ®nancial practitioners to whom no established or appealing alternative exist. For recent evidence that the assumptions of the basic market model are violated with actual data, see Coutts et al (1997), Jong et al (1992 and Mills (1996), and the references therein. However, both Campbell et al (1997) and Elton and Gruber (1995) argue that despite its de®cits, the market model is still perhaps the most appropriate in, e.g., the event-type studies since in practice the gains from employing either multifactor models or economic models (such as the APT) are limited.…”