Dimensionality Reduction Techniques in Macroeconomic Analysis

Author

Abigail Rogot, CUNY Bernard M Baruch CollegeFollow

Date of Award

Spring 5-2023

Document Type

Thesis

Degree Name

Bachelor of Arts (BA)

Honors Designation

yes

Program of Study

Economics

Language

English

First Advisor

Sebastiano Manzan

Second Advisor

Anna D’Souza

Abstract

Over the past several decades, rapid innovation in data collection methods and technology has led to the development of dimensionality reduction techniques when dealing with a large number of predictors and time series observations. Especially relevant to the field of economics, many macroeconomic indicators rely on processing vast sets of data, often dealing with variables of different frequencies. Broadly, monetary policy is influenced by real-time evaluations of current and future economic conditions, meaning that lags in re- leases produce incomplete datasets. This paper closely examines the development and applications of two popular dimensionality reduction techniques: Principal Component Analysis (PCA) and Dynamic Factor Analysis (DFA). We give PCA a rigorous mathe- matical treatment to prove that the set of principal directions of a centered dataset is equivalent to the set of orthonormal eigenbases of the covariance matrix. The paper con- cludes with the construction of a Nowcasting model for United States Gross Domestic Product using data from Haver Analytics.

Recommended Citation

Rogot, Abigail, "Dimensionality Reduction Techniques in Macroeconomic Analysis" (2023). CUNY Academic Works.
https://academicworks.cuny.edu/bb_etds/166