Year: 2014
Type: dissertation
Abstract: Review Factor ModelsFor the past few decades, factor models have experienced increasing popularity in economic and …nance related studies.This is primarily attributable to a growth in the availability of large datasets as well as advances in technology.The growing mass of information broadens the horizons for in-depth …nancial analysis, and modern technology helps to revolutionise the data processing techniques, which makes the analysis of large datasets both feasible and worth while.Growing application of large datasets presents new challenges for the traditional modelling practices, which experience the "curse of dimensionality" issues during the process of modelling large dimensional panels.Factor models o¤er an elegant mathematical solution to the problem by introducing a methodology, which reduces the dimensionality by searching for common patterns between variables (see, Forni et al.(2000)).Many argue that pattern recognition, a bi-product of this technique, is a core reason for the popularity of factor models among practitioners, and in particular in the social sciences.Some indicators such as business con…dence, are commonly treated as an easily quanti…able variable, despite them being qualitatively di¢ cult to measure.The method by which to quantify such an hypothetical variable is to summarize the information from a large number of observable variables, by employing pattern recognition techniques in factor models.The latent factors of the models become a "measure index" of qualitative variables.More rigorous …elds, such as …nance, employ pattern recognition to identify trends in the large timeseries panels.In some cases it is possible to determine the latent variables with observable time-series.The model then bares a striking resemblance to the standard multiple regressions, for example traditional asset prising models, such as Arbitrage Pricing Model by Ross (1976), CAPM by Sharp (1964), or Gordon's triangle model for the in ‡ation rate forecast (see Gordon (1988)).It is more intuitive to regress a panel of observable variables, such as multiple regressions.However, it is near impossible to …nd new observable variables (that have a strong correlation with the data) without a preliminary examination of main trends in the dataset.Factor models provide an ideal preliminary analysis of large datasets, which can subsequently lead to an interpretation of established trends.Even without direct interpretation, latent factors can demonstrate leading patterns of the data.Pattern recognition is a crucial part of factor analysis, and it is the feature that is most commonly 3 First-Di¤erenced Inference for Panel Factor Series
Author: Ekaterina Ipatova
Language: English
Subfield: Finance
Domain: Social Sciences
Open Access status: green
OpenAlex ID: W170489029