ESSAY 1

This paper examines the Sharpe-Lintner CAPM using time-varying-parameter models in addition to the static market model. Prior evidence does not support the CAPM in that it suggests that market risk is not priced or that the price of the beta risk is significantly negative for a thin European stock market, eg the Finnish stock market. This paper shows explicitly that this phenomenom is due to static ordinary least squares beta estimates which are spurious. We reduce the errors-in-variables problem by estimating firm-specific betas using Kalman filter techniques and employ the betas forecasted on the basis of these estimated betas in a cross-sectional analysis. Analysis of pooled data shows that the price of conditional risk is positive and that the mean-variance efficiency of the market index cannot be rejected. These results support the CAPM. The data covers all Finnish common stocks listed on the Helsinki Stock Exchange throughout the period 1972-1989.

ESSAY 2

This paper examines the sensitivity of tests of the Sharpe-Lintner CAPM to different beta estimation methods and asset-return samples in a thin European asset market, ie the Finnish asset market. A time-varying-parameter model is introduced as an altemative to the static market model. We also employ pooled data in the analysis in addition to second-pass Fama-MacBeth regressions. The tests are, furthermore, carried out with four asset-specific samples. It is shown that in every case, the analysis on OLS betas leads to rejection of the mean-variance efficiency of the market index and the price of market risk is statistically significant, but negative. The corresponding tests on the time-varying betas indicate just the opposite. We are not able to reject the mean-variance efficiency of the market index in any of the samples. The price of market risk tums out positive and statistically significant, especially for the stock-return data set that most closely resembles the normal distribution.

ESSAY 3

This paper examines cointegration and Granger causality among the stock markets in the United States, the United Kingdom, Germany, Sweden and Finland. The first three nations are the biggest trading partners of the two small open Nordic economies, Finland and Sweden. We apply standard univariate VAR models and a system of VAR models under the assumption of multivariate cointegration, first introduced in Johansen (1988). The cointegration analysis suggests that the stock markets are cointegrated, having one cointegrating vector when prices are measured in local currencies or in Finnish markkas and two cointegrating vectors when prices are measured in US dollars. It is also found that the Finnish and Swedish markets may deviate from the equilibrium path without having a significant impact on the three other markets, which indicates that the causality is from other stock markets to Finland and Sweden. The Finnish stock market is always found to be led by the German market, instead of the Swedish market as previously suggested, and also by the UK market when returns are measured in local currencies or in Finnish markkas. The Swedish stock market is Granger caused by the UK market instead of the US market, as previously suggested. The data covers the period 1974-1989.

ESSAY 4

This paper studies the driving forces of predictable variation in Finnish stock returns. The dynamics of Ferson and Harvey's (1991) methodology are extended and applied within the Sharpe-Lintner CAPM. Wc find that market risk is conditionally priced in the thin Finnish stock market. Most of the predictable variation of stock returns is attributed to the time-varying risk premium, which supports the hypothesis of rational behavior by Finnish investors in setting stock prices. However, the conditional residual term accounted for a larger part of the predictable variation of the stock returns than is found in the US market.