Estimation of Exchange Rate Exposure in the Presence of Cross-sectional Dependence

Abstract

This article investigates exchange rate exposure (at the firm level and overall) of 211 non-financial firms to four major currencies across two overlapping sample periods employing both static (OLS) and dynamic (ARDL, CS-ARDL) estimation techniques. A novel model specification (model 2) replaces market return with cross-sectional average return in model 1, the conventional specification, to find relatively more efficient estimates, mitigating cross-sectional dependence (CSD) in most cases. OLS finds a highest of 14.7% firms to Rupee, whereas ARDL finds a highest of 13.2.% firms to Rupee and Yuan contemporaneously exposed in the full-sample and pre-covid period, respectively. ARDL detects lagged exposure in all cases with a smaller number of firms exposed in subsequent lags. Positive exposure to Dollar (USD) in most cases, and a larger number of firms exposed to Yuan and Rupee in the full-sample period (which includes the COVID-19 and Russia–Ukraine war) are detected. Overall, contemporaneous exposure detected by pooled OLS finds the least, followed by CS-ARDL long-run estimates, with short-run CS-ARDL detecting the highest number of cases (in both periods, model 2 detects no exposure in the full-sample period like OLS estimates). While lagged exposure is detected under model 1 to all currencies, model 2 detects no lagged exposure in any case.