Examining the influence of temporal aggregation on the power of the ADF residual-based stock-Watson cointegration test

Abstract

This paper examines the effects of standard and random sampling temporal aggregation on the power properties of the Dynamic Ordinary Least Squares (DOLS) residual ADF-based test of the null hypothesis of cointegration. The results indicate that the conducted Monte Carlo simulation experiments reveal that the use of temporally aggregated data significantly impacts the power of the DOLS cointegration test. An empirical application is then provided that investigates the nature of the equilibrium relationship between UK nondurable consumers’ expenditure and disposable income. This confirms the results obtained from the Monte Carlo analysis that the estimation of the equilibrium parameters of this relationship using temporally aggregated data significantly impacts the power of the DOLS cointegration test. Comparative analysis shows that a higher level of temporal aggregation reduces test power; however, the increase in sample size can easily compensate for this effect. Thus, researchers are advised to use the largest feasible sample sizes to ensure robust cointegration testing, especially when dealing with high aggregation levels. The practical application results further confirm that increased aggregation lowers test reliability, with anomalies arising at higher aggregation levels.