Summary: | This thesis investigates the evolution of the relationships between the idiosyncratic components of inflation subindices over time. Utilizing machine learning techniques, specifically Principal Components Analysis (PCA) and Graphical Lasso, the study identifies key components of price changes and maps the network of idiosyncratic elements. The analysis period is divided into two intervals, 1997-2010 and 2010-2023. Findings indicate a strengthening of the network, evidenced by an increase in nonzero edges and enhanced connectedness in the latter period. Contributing factors to this change include shifts in the competitive dynamics of product markets and economic cyclical variations. Notable changes were observed in categories such as small home electronics and housing-related services, reflecting broader economic trends. Key words: Inflation, Consumer Price Index, subindices, Iceland, machine learning, principal components, PCA, graphical lasso, dependence structure.
|