New nonparametric regression tests improve accuracy of economic forecasting models.
The article introduces tests for checking if a nonparametric instrumental regression model fits well. These tests use series estimators to compare the general model with specific ones and to check if the regressors are exogenous. The tests' distributions are derived under correct specification and shown to be consistent against any alternative model. The tests also perform well under local alternative hypotheses and are uniformly consistent as the sample size grows. A study using simulations examines how well the tests work in real-world situations.
