Michael Beenstock and Yaniv Reingewertz have a fairly new working paper that applies the method of polynomial cointegration to testing the global warming hypothesis. This paper has apparently been discussed quite a bit in the blogosphere.
The authors find that they cannot reject the null hypothesis that the forcing variables that they consider do not polynomially cointegrate with global temperature. They do find some short run effects of greenhouse gases on temperature. From this they argue that greenhouse gases do not permanently alter the climate and that, therefore, climate change is a minor problem. The problem is that not being able to reject the null is not the same as accepting it. Non-cointegration can mean two things - either the variables are unrelated to each other or that important non-stationary variables that are neccessary to producing a statistically adequate model have been omitted. If these variables were added to the model then cointegration would be achieved. But we can't know a priori which of these is the case.
I wrote two papers on multicointegration and climate change - one was published in Computational Statistics and Data Analysis and one I never published. Both these papers include ocean heat content as well as atmospheric temperature. The former variable is omitted in all other econometric studies of climate change including the new Beenstock and Reingewertz paper. But I think now that it is rather critical.
I've sent Beenstock my papers. Will be interested in his reaction.
Hat-tip to VS who pointed this paper out to me in the comments on my recent post.
No comments:
Post a Comment