Fractional Integration and Its Influence on Unit Root and Co- Integration Analysis

Resumen

This study assesses the power of traditional unit root and co-integration tests when they are applied to fractionally integrated stochastic processes in the 0 ≤ d ≤ 1 range. Monte Carlo simulations were conducted to evaluate the sensitivity of the unit root tests in distinguishing the I(1)−I(0) conditions of the fractional conditions. Our results showed that unit root tests have individually low power when applied to small sample series with long-memory. However, we found that under specific conditions the unit root tests can produce results that can help avoid the over-differentiation problem. In the co-integration analysis for fractional alternatives on the interval 0 ≤ d ≤ 0.6, we found some conditions that can lead to satisfactory results