Thursday, November 17, 2011

From Correlation to Granger Causality Added to New Crawford School Research Papers Series

My paper From Correlation to Granger Causality has been added to the Crawford School Research Papers series on RePEc. I'm now working on revising the paper and doing new tests in collaboration with Kerstin Enflo. This version represents the conference paper I gave in Michigan at the "Annual Institute on Joint Outcomes for Sustainability" (which was not an annual institute but a once off workshop) organized by Arun Agrawal. Both Shuang and myself were invited to the workshop. Shuang's work was more directly relevant looking at decision-making for dealing with invasive species. I was asked to talk about "from correlation to causal inference" in a session of that topic where different econometric methods were discussed. Most relevant in that session to the topic at hand was Paul Ferraro's presentation on the effects of protected areas in Costa Rica on poverty in surrounding districts.


  1. Great paper, I look fowrad to final print version.

    However I think this from the Abstract of the Draft could be better put (if not already done): "Energy prices have a
    significant causal impact on both energy use and output while there is no strong evidence that
    energy use causes carbon and sulfur emissions despite the obvious physical relationship" as it looks like a big non sequitur, especially as not all energy use (eg Hydro, nuclear) has that "obvious physical relationship".

  2. Yes, maybe disaggregating energy into fossil and non-fossil would make a difference. Anyway, we don't plan on trying to formally publish this paper as it stands. It's just a record of the conference paper.

  3. But what about the non sequitur? The carbon tax is wholly dependent on the physical relationship between on outputs from combustion of hydrocarbons and those CO2 and SO2 emissions, but your Abstract seems to deny that!

  4. There's nothing about carbon taxes or climate change etc. in the paper. One of the points was that you have to be very careful about how you define variables if you want to test for causality. You've come up with a good potential reason why there is no causality here between energy use and carbon emissions which would demonstrate this point very well if it is true, which I haven't tested.