| dc.description.abstract | This thesis consists of four wp. In the first one we investigate the nature of the links between, on the one hand, stock market reactions to news in the information set, and on the other hand, the way firms from a panel revise their real investment plans in light of the same information. Using neSted covariance and variance a>mponent models we decompoee the uncertainty that tl 2 firms faced during the period 1976 to 1993 into three levels of shocks: micro, 86ctor, and macro. The main findings are that the link between stock market reactions to news and the firms revisions of inv&ment plans is weak, and that the dominant source of uncertainty for investment spending is found at the micro level. In the second essay we model real investment behaviour for an unbalanced panel of Swedish manufacturing firms during the period 19831993. An Euler equation approach ii9 uzxxl, and the specification is choeen to test for the impact of liquidity constraints. Various sample separation criteria are us6d to test for croes-sectional differences among the separated firms. The criteria are whether a plant is owned by a firm that, (1) is listed versus not listed at the Stockholm Stock Exchange, (2) has more versus less than 23% of its employees abroad, (3) is listed at the Swedish bourse and had a pre-sample dividend income ratio higher versus bwer than %. The final separation criterion delivers the most theory consistent estimab and there is weak evidence of liquidity constraints.In the third eeeay we model the returns for 14 large Swedish firms stocks with a conditional multifactor model with time-varying be& terms. The data is monthly and the sample period is June 1992 to August 1997. The beta term are modelIed as linm functions of lagged firm attributes, which are either taken from published acoounting data or from a-us fad data. The main findinga are that the stock exchange is not e&ient with respect to the co~nsus information and the la@ yield sprsed. We also find that the lag@ firm attributes are mainly &a&l with risk exposures. For mot& firms, the Ix& models are not rejected in out-of-sample forecast tests for the period September 1997 to December 1997. The final eesay contains modeb of an overnight ix&zest rate from the Swedish interbank market during the period January 1986 to May 1991. A continuous time specification is estimated, using a c&cre&ized venGon of the process. This estimator is evaluati with a Monte Carlo study. In out-of-sample foreca&, a simple AR model with ARCH errors is slightly better than the continuous time specification. Using spectral analysis, a filtered interest rate has a spectrum with a peak at an intermediate frequency implying a periodicity of one month in calendar time. | en |
