Modelling the Determinants of Naira/US Dollar Currency Exchange Rates Using Principal Component Analysis (PCA) and Singular Value Decomposition (SVD)

Abstract

In the literature, a large number of factors are mentioned as being responsible for the relative strength or weakness of a country’s currency with respect to the currencies of other nations. Many of these variables are usually highly correlated with each other leading to the twin problems of multicollinearity and redundancy in the explanatory variables.This study examines the main determinants of currency exchange rates, and applies Principal Components Analysis (PCA) and Singular Value Decomposition (SVD) to Naira/USD data regarding the factors responsible for the steady deterioration in the Naira/USD exchange rates over time. The aim is to rank the factors in order of importance and impact on the Naira/USD exchange rate in the Nigerian environment, as such ranking may not be of universal application in all countries. The study uses Machine Learning algorithms to achieve dimensionality reduction and thus address the problem of multicollinearity and high dimensionality. The study found that three principal components adequately explain more than seventy percent of the variance, thereby making it unnecessary to use more than three explanatory variables in similar studies predicting the evolution of currency exchange rates. In this study the number of explanatory variables was drastically reduced from twenty-one to just three, while solving the problem of multicollinearity.