Band 14, Heft 3

Misspecifications of econometric models can lead to biased coefficients and incorrect interpretations of error terms, which in turn can lead to incorrectly estimated models and incorrect inference. There are specific techniques such as instrumental variables, which are used in the economics literature to deal with some individual forms of model misspecification, only addressing one problem at a time. The joint and separate solutions to the problems of unknown functional forms, omitted variables and measurement errors, discussed in this paper, prove that instrumental variables do not exist. Therefore, the specific techniques used in the literature are not feasible. This paper proposes a general method for estimating underlying parameters in the presence of a range of model misspecifications. It is argued that this method can consistently estimate the direct effect of an independent variable on a dependent variable with all of its other determinants held constant even in the presence of an unknown functional form, measurement error and omitted variables.

With their article on Maximum Entropy (ME) densities for time-varying moments, Rockinger and Jondeau (2002) set a milestone for the application of information theoretic principles to the analysis of financial market data. In this note we briefly discuss the application of their approach to financial data, point out some shortcomings that it encounters and show how these can be overcome. Applying our model to different market indices, we find evidence for time-variability of skewness and kurtosis.

The specification of Smooth Transition Regression models consists of a sequence of tests, which are typically based on the assumption of i.i.d. errors. In this paper we examine the impact of conditional heteroskedasticity and investigate the performance of several heteroskedasticity robust versions. Simulation evidence indicates that conventional tests can frequently result in finding spurious nonlinearity. Conversely, when the true process is nonlinear in mean, the tests appear to have low size adjusted power and can lead to the selection of misspecified models. The above deficiencies also hold for tests based on Heteroskedasticity Consistent Covariance Matrix Estimators but not for the Fixed Design Wild Bootstrap. We highlight the importance of robust inference through empirical applications.

This paper derives, extending the work of Rilstone, Srivastava and Ullah (1996), an analytical expression that takes account of first and second order asymptotic bias of nonlinear estimators in a non-parametric setting. By using moment expansions, we obtain a first and a second order bias removal mechanism. We specialize our results on the smoothed maximum score estimator of the coefficient vector of a binary response model in the dynamic setting of De Jong and Woutersen (2009). First order asymptotic theory has already been provided, although very large samples are needed to reach the asymptotic outcome of normality in this model. We provide a second order asymptotic expansion and, with the appropriate estimated quantities, we design a new bias-corrected estimator. Finally, a simulation study shows the advantages of our proposed bias-correction procedure.

Recently, there has been much interest in modeling time-varying higher-order conditional moments in the density estimation context. These studies employ a moment-based methodology to test their specification of higher-order conditional moments. We compare the size and power of these moment-based tests with the size and power of a recently developed set of nonparametric tests. The results show that the moment-based tests have good size only for conditional skewness and have a large size distortion when all the moment conditions are tested jointly and used as an overall specification test. The nonparametric tests have a slight size distortion for conditional higher moments, which can be eliminated. The power of the nonparametric tests for detecting overall misspecification and lack of dynamics in the conditional higher moments is better than the power of the moment-based tests.