Covered interest rate parity in the presence of time-varying transaction costs: an application of the fractional integration approach

Purpose We demonstrate that fractional integration offers an improved framework for assessing the reliability of the covered interest rate parity condition compared with traditional unit root tests. Specifically, in the presence of transaction costs, the parity condition will tend towards one of the arbitrage bounds, making the process fractionally integrated at that bound which renders traditional unit root tests unsuitable. Given the widespread adoption of parity conditions in practice, in areas such as risk management and foreign exchange rate forecasting, an approach which permits an assessment of the reliability of these models, particularly for volatile markets such as emerging markets, is crucial. Design/methodology/approach We propose a threshold autoregression framework that more effectively identifies near-unit-root behaviour and improves market classification relative to conventional approaches. The analysis uses data from the industrialized markets of Australia, Hong Kong, Japan, New Zealand and Singapore, as well as the emerging markets of India, Indonesia, Korea, Malaysia, the Philippines and Thailand, with the USA serving as the reference market. Findings Our results show that shocks to covered interest rate parity are asymmetric and that the parity conditions tend towards one of the arbitrage bounds. Originality/value The fractional integration approach is well established in the econometric literature, though we are the first to apply this in this context. In relation to the international finance literature on tests of parity, recent attention has been almost exclusively on the causes of parity deviations than on assessing the reliability of the condition as an exchange rate forecasting instrument.