Market Efficiency With High-Dimensional Input Data

his Thesis is concerned with market efficiency with high-dimensional input data at disposal of economic agents. The relevance of this topic is justified by the fact that volumes of information in financial markets are increasing quickly, and data science methods are used for analysis of this data. Empirical research of this question demands research techniques which allow for evaluation of dynamic changes in market efficiency. However, the methodological apparatus, which is used for assessment of the efficient market hypothesis, cannot be used for this purpose. As a solution, the author proposes to consider the rapidly developing branch of literature which is concerned with empirical evaluation of the adaptive market hypothesis. This hypothesis can be considered an extension of the efficient market hypothesis, allowing for existence of time dynamics of market efficiency. Using the methodology of mentioned works (in particular, the rolling window approach), the author constructs metrics for assessment of changes in market efficiency for the past 30 years. Data in daily resolution is gathered on the following stock market indices: 1) S\&P 500; 2) NASDAQ Composite; 3) NIKKEI 225; 4) FTSE 100; 5) XETRA DAX; 6) TSX Composite. In addition, analysis of structural breaks is implemented with the help of a numeric algorithm. A conclusion is made that there exist cyclical shifts in stock market efficiency over the last 30 years, and some of the sharpest shifts are connected with major economic shocks. However, some of the dynamics cannot be assessed with these methods exclusively. Hence, an additional investigation into the question is required.