“…While our bias-correction approach assumes that the regressor x it is strictly exogenous, we also analyse its robustness when the regressor is instead predetermined. In the empirical practice, the hypothesis of strictly exogenous regressors can be tested by adopting the tests considered in Su et al (2016) and Mayer (2020). To generate a predetermined regressor, we replace ǫ it in the data-generating process by ρu i,t−1 + 1 − ρ 2 ǫ it , and we set ρ = 0.4/σ ǫ .…”