Economics Letters 9 (1982) 153-160 North-Holland Publishing Company 153 zyxwvutsrqponml EFFICIENT SPECIFICATION TESTS FOR LIMITED DEPENDENT VARIABLE MODELS Carlos M. JARQUE and Anil K. BERA Austrulia~~ Natroml Universi<v, Cmberru ACT 2600, Au.~aho Received 23 September 1981 The procedure of Jarque and Bera (1980a, b), consisting of the application of the Lagrange Multiplier (LM) test to the Pearson Family of distributions, is used to derive efficient normality and/or homoscedasticity tests for limited dependent variable (LDV) models, 1. Introduction LDV models arise when the dependent variable is restricted in some way (e.g., to non-negative values - like supply of and demand for given goods). These models are now being increasingly used in economic modelling, and methods for their efficient estimation - given a set of assumptions - are presently available (e.g., see the Fall 1976 issue of the Annals of Economic and Social Measurement). In comparison, few results exist regarding inferential problems. In previous papers we have sug- gested the use of the LM test on a general Family of distributions - the Pearson Family ’ - to derive simple diagnostic tests which have maxi- mum asymptotic local power and are therefore efficient [see Jarque and Bera (1980a, b) and Bera and Jarque (1981)]. In this paper, using the same approach, we derive tests for residual specification in LDV models. Here the presentation is made for the Tobit model (section 2) and the Truncated model (section 3) but our approach has more general applica- bility. ’ We have also used other families such as Gram-Charlier (type A). and obtained the same test statistics as with the Pearson Family. This makes our tests more appealing. 0165-1765/82/0000-0000/$02.75 0 1982 North-Holland