Testing Linear and Log-Linear Regressions for Functional Form

Godfrey, L. G.; Wickens, Michael

doi:10.2307/2297160

Cited by 119 publications

(49 citation statements)

References 15 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…(1) PE is a test proposed by MacKinnon, White, and Davidson (1983) for testing linear versus log-linear models (and vice versa); similarly, BM is a test due to Bera and McAleer (1989); GW is a test suggested by Godfrey and Wickens (1981). (2) In the case of the PE and BM test, the figures reported are the t-statistics of the coefficient of the relevant variable in the comprehensive model on which the non-nested tests are based.…”

Section: Discussionmentioning

confidence: 99%

“…Bera and McAleer (1989) propose a test which is based upon the following two regression (Davidson and MacKinnon, 1985;Godfrey, McAleer, and McKenzie, 1988). One such test is proposed by Godfrey and Wickens (1981), whose suggested strategy leads to H 1 and H 2 being tested, respectively, against:…”

Section: Discussionmentioning

confidence: 99%

“…While this issue is postponed until the next section, here we tackle the statistical discrimination between linear and log-linear models, a topic which has traditionally attracted the interested of both econometricians and applied economists, particularly from the viewpoint of non-nested hypothesis testing (see, among others, Godfrey and Wickens, 1981;Davidson and MacKinnon, 1985;Godfrey, McAleer, and McKenzie, 1988). Not only are the two specifications separate from one another, but unlike standard non-nested models, they also entail a known transformation of the dependent variable.…”

Section: Model Selection and Specification Testingmentioning

confidence: 99%

“…and McAleer (1989), and the GW test of Godfrey and Wickens (1981). All these tests are asymptotically valid, the first are based on the principle of artificial regression, while the third is based on the well-known Box-Cox transformation.…”

Section: Model Selection and Specification Testingmentioning

confidence: 99%

See 3 more Smart Citations

Desperately Seeking (Environmental) Kuznets

Galeotti

Lanza

1999

SSRN Journal

View full text Add to dashboard Cite

Abstract. The number of studies seeking to empirically characterize the reduced-form relationship between a country economic growth and the quantity of various pollutants produced has recently increased significantly. In several cases researchers have found evidence in favor of an inverted-U "environmental Kuznets" curve. In the case of a major greenhouse gas, CO 2 , however, the evidence is at best mixed. This paper attempts to shed further light on this issue by using a newly developed data set covering over one hundred countries around the world for the last twenty five years and by considering alternative functional forms together with an effort to rigorously discriminate among competing alternatives. DESPERATELY SEEKING (ENVIRONMENTAL) KUZNETS Non-Technical SummaryThe relation between economic development and environmental quality is a very complex issue. For this reason, in the last few years several studies have tried to characterize this problem as an empirical reduced-form relationship. In particular, a few studies have identified a bell shaped curve for the pollution intensity of GDP. This behavior implies that, starting from low (per capita) income levels, (per capita) emissions or concentrations tend to increase but at a slower pace. After a certain level of income, emissions or concentrations start to decline as income further increases. In the case of a major greenhouse gas, CO 2 , however, the evidence of an inverted-U "Kuznets curve" is at best mixed.While all the studies have focused upon the empirical emergence of the environmental Kuznets curve and have typically discussed its implications with special reference to the value of the income turning point, the analysis concerning the robustness of the basic findings has not been, somewhat surprisingly, a major concern. The issue of the functional form for the reduced-form relationship between CO 2 emissions and GDP appears to be critical for the emergence of a bell shaped curve and for the crucial policy implications that could be drawn from such an empirical finding.We argue in this paper that the empirical relationship between a country GDP and CO 2 emissions ought to undergo a more careful scrutiny from at least two standpoints. In this paper we start from the observations that nearly all papers in this literature use the same source of data for CO 2 emissions and almost invariably fit to the data either linear-in-variables or log-linear functional relationships. Departing from this practice, we estimate the "standard" relationship on a newly developed data set for emissions. We also apply the theory of non-nested hypothesis testing in the attempt to choose between linear and log-linear models. After having noted a few theoretical and empirical drawbacks of these functional specifications, we propose two alternative functional forms, Gamma and Weibull, which are subsequently implemented. The estimated results are satisfactory and the features of the EKC relationship reasonable. Finally, interestingly enough non-nested tests produce the unambi...

show abstract

Section: Discussionmentioning

confidence: 99%

Section: Discussionmentioning

confidence: 99%

Section: Model Selection and Specification Testingmentioning

confidence: 99%

Section: Model Selection and Specification Testingmentioning

confidence: 99%

See 2 more Smart Citations

Desperately Seeking (Environmental) Kuznets

Galeotti

Lanza

1999

SSRN Journal

View full text Add to dashboard Cite

show abstract

“…In certain situations it may be difficult to take expectations, for example, when testing linear and log-linear models using the Godfrey and Wickens (1981a) Table III indicates the characteristics of the regression disturbances. Recall that pi and p2 denote "weak" and "strong" first order serial correlation, ctl and a2 denote "weak" and "strong" heteroscedasticity and, "t" and "log" denote that the errors e. were obtained from student t and log-normal distributions respectively.…”

Section: Introductionmentioning

confidence: 99%