Tesis doctoral de Olga Klinkowska
This doctoral thesis has arisen as a result of my interest in two research fields: asset pricing theory and financial econometrics. It consists of three self-contained essays. Although united under the same title, they differ with respect to the topics considered, research question asked and econometric methodology. Each of the essays accounts for a separate piece of work and as such can be approached individually and read separately. the topic of the first essay is related to empirical tests of the classical capital asset pricing model. The second essay studies the cross-section of government bond returns within the framework of consumption capital asset pricing model with long run risk. The third essay investigates the stochastic properties of aggregate consumption growth. All the essays are empirical studies and as a result an important part of each of them is the econometric framework used and the empirical implementation. Despite the apparent differences there is a common element in all the three essays: each of them at some stage makes use of the common factors which, extracted from a large panel of economic and financial data, summarize the information in this data. the theory of factor analysis has quite a long tradition in psychology and social sciences. It was pioneered more than 100 years ago by a psychologist charles spearman (spearman, 1907), who hypothesized that the great variety of tests of mental ability e.G. Measures of mathematical skill, verbal abilities etc. Could be explained by one underlying «factor» of general intelligence, that he called the g factor and that this factor was the only one common to all the measures of mental ability. Though an interesting idea, it turned out to be not fully correct. Nevertheless, charles spearman played an important role in the development of the factor theory. currently factor analysis is used to study the patterns of relationship among many variables representing different phenomena, not necessarily related to psychology or social sciences, with the goal of discovering something about the nature of the independent variables that affect them, even though those independent variables are not measured directly. In particular, it seeks to discover wether the information included in a set of observed variables can be effectively summarized and reflected in terms of a smaller number of unobserved independent variables called factors. This issue has been particulary relevant for a while since as the time passes and informational technology improves, the availability of the data greatly increases in terms of both, the number of the variables and the length of the data span. As a result the obvious possibility to explain economic phenomena better with more data may be limited by the informational overload if more data is not organized in some sensible way. Thus the factor analysis, which might be viewed as a dimension reduction technique, seems to be a tailor-made solution for that. in recent years the analysis of large dimensional data has received an extensive attention in economics. In early works sargent and sims (1977) and geweke (1977) applied the factor approach in the economic environment for macroeconomic analysis purposes. The last decade has given a rise to a huge number of research on factor analysis at both, theoretical and empirical grounds. Special attention has been given to the econometric theory related to factor estimation with an emphasis on statistical properties of the estimated factors and conditions under which they are consistent. Bai and ng (2008) give a comprehensive overview of the main findings on these issues. Stock and watson (2010) provide a complementary survey of the applications of dynamic factor approach in economic environment, especially in the context of macroeconomic forecasting. The three essays which account for this doctoral thesis contribute thus to literature on dynamic factor analysis with examples of novel application of the factors extracted from a large panel of economic and financial data. I now provide a brief overview of the each of the essays. the topic of the first essay is related to the classical capital asset pricing model (capm). One of the central pieces of this model important in empirical tests is the wealth portfolio. According to the theory, this portfolio should include all the risky assets in the economy. As it is practically impossible to find such a portfolio, a broadly defined market index proxy is usually used. I propose to account for two important elements of this portfolio, specifically human capital and housing capital. These two assets were shown to be important components of the aggregate wealth in the economy and to influence greatly investment decisions of the investors. I investigate if and how a better approximation of the wealth portfolio affects the empirical test of the capm model. Further, since the capm model is developed in a static framework, i also introduce dynamics into this model. It is important to account for that as the investment opportunities of the investors are changing over time. The novelty of this extension is related to the application of the estimated common factors. Selected factors are used as conditioning variables in the conditional capm model. Since, by definition, the factors summarize and reflect the information included in a large panel of data, they are more accurate representatives of the unobservable information set of the investors than other single and arbitrary chosen variables. my results imply that indeed a better proxy for a wealth portfolio significantly improves the empirical performance of the classical capm. An extended capm model is able to explain 52% of the cross-sectional variation in tested assets in comparison to only 9% using the canonical capm. Further, when the conditional information is introduced into the model, this explanatory power rises to 83%. The obtained results are of particular relevance for corporate finance practitioners who extensively use the capm model to determine the required rate of return when assessing whether to launch new projects or not. the second essay is a joint work with abhay abhyankar and soyeon lee. In this work we use a consumption-based asset pricing model with epstein–zin–weil recursive preferences to explain the cross-section of excess returns on nominal u.S. Treasury bond portfolios. Specifically we ask the following question: do investments in nominal government bonds pay off in good times and are they therefore risky assets that investors need an inducement to hold or do they pay off in bad times and thus might help investors hedge macroeconomic risk. we follow the theoretical framework of long run consumption risk model, which arises as a result of both the use of epstein–zin–weil recursive preferences and specific assumptions for the evolution of aggregate consumption. Our model has two pricing factors which influence the equilibrium asset returns: the innovations to current consumption growth and the innovations to expected future consumption growth. The novelty of our study is specifically related with the estimation of these innovations. We use a vector autoregressive (var) model where appropriate state variables are selected that are known to forecast consumption growth well. However, instead of choosing specific predictor variables we use a set of dynamic factors obtained from a large panel of macroeconomic and financial time series. We then estimate a factor-augmented var and extract shocks to current and expected future consumption growth. This approach has some advantages. First, we can be agnostic in our choice of state variables, second, there is evidence that common factors have good forecasting properties even in the presence of structural breaks, third, the pre-estimation of the factors does not influence the consistency of ordinary least squares (ols) estimates in the var model which is relevant in our application. we find that, over the period 1975–2006, nominal u.S. Government bonds are risky assets that pay off in good times which are characterized by good prospects for future consumption growth. The risk premium related to news in expected future consumption growth is positive and significant. Further, we find that the current consumption growth does not play a significant role in pricing the cross-section of average excess nominal government bond returns. We also find that bond portfolios with long maturity bonds are riskier in terms of higher and positive consumption betas as compared to the portfolios with short term bonds. The model explains well the cross-sectional variation in average excess returns on portfolios of u.S. Treasury bonds with differing maturities, provides plausible estimates of the structural parameter and compares favorably with other competing models. Our results are robust to using alternate test assets and sample periods, different definitions of consumption and estimation methods. in the third essay i investigate whether aggregate consumption growth is predictable. The stochastic properties of consumption growth and those of the state variables that possess valuable information about the future path of consumption growth are central to many theoretical models in finance and macroeconomics especially to consumption-based asset pricing models. Thus the importance of time series evolution of consumption and its implications for asset pricing especially in the context of long-run risk models cannot be overstated. My study contributes to an extremely sparse literature on consumption growth predictability. The novelty, however, is in the forecasting variables used — i use as predictors the common factors estimated from a large panel of macroeconomic and financial data. I compare their forecasting abilities with those of other specific variables that were used in previous research on the predictability of consumption growth. i find, using quarterly data over the period 1960-2007, that a selected group of common factors has an extraordinary forecasting power for consumption growth. The selected factors are: the first dynamic factor, which represents real economy, the second factor, which summarizes interest rates and spreads, the fifth factor, which incorporates mainly housing market measures and the eighth factor representing aggregate equity market. A combination of the four dynamic factors explains as much as 36% of the variation in future consumption growth. Interestingly, i find as well that the second common factor is highly informative about the future consumption growth — it explains individually around 23% of the consumption growth variation. None of all the other specific variables considered in this study possesses that amount of information about future consumption growth. I also conduct a battery of robustness check that further support my main conclusions. references: bai, j., Ng, s., 2008. Large dimensional factor analysis. Foundations and trends in econometrics 3(2), 89-163. geweke, j., 1977. The dynamic factor analysis of economic time series. In latent variables in socio-economic models, ed. By d.J. Aigner, a.S. Goldberger, amsterdam, north-holland. sargent, t.J., Sims, c.A., 1997. Business cycle modelling without pretending to have too much apriori economic theory. In new methods in business cycle research, ed. By c.A. Sims, minneapolis, federal reserve bank of minneapolis. spearman, c., 1904. General intelligence objectively determined and measured. American journal of psychology 15, 201-293. stock, j.H., Watson, m.W., 2010. Dynamic factor models. Prepared for the oxford handbook of economic forecasting, ed. By m. P. Clements and d.F. Hendry, oxford university press.
Datos académicos de la tesis doctoral «Essays on empirical finance«
- Título de la tesis: Essays on empirical finance
- Autor: Olga Klinkowska
- Universidad: Autónoma de barcelona
- Fecha de lectura de la tesis: 22/07/2011
Dirección y tribunal
- Director de la tesis
- Abhay Abhyankar
- Tribunal
- Presidente del tribunal: jordi Caballe vilella
- laurence Copeland (vocal)
- (vocal)
- (vocal)