Tuesday, October 21, 2008

Population and CO2

I read an interesting piece over at Prometheus about how European countries have - per capita - not generally performed better than the United States in the period of 1997-2005. There are a few troubling things about this post for me (and it's not because I'm European, a Euro-phile, or a US-hater). The central graph seemed wrong to me. I think it was because the CO2 per capita graph is analogous to the one I wrote about a while ago: it seemed to be comparing things that could not be equally compared across set items.

If per capita CO2 emissions was something that could be standardized across countries, then its base values (CO2 emissions and population) should share some characteristics. Minimally, they should have similar slope values in a direct comparison with each other. If they don't then the value of "per capita CO2" is not a standard unit. Looking at the graph at Prometheus, I chose to compare Denmark (the best actor), Spain (the worst actor) and Luxembourg (one that ends up with near-zero change) against the United States.

For data on CO2, I used data found here. For data on population, I used this data. The values appear to be slightly different from that used on Prometheus, but do not vary so much as to make comparison invalid for these purposes.

As you can see in this first graph, the untransformed values of CO2 emission and population do not provide linear regressions of similar value across all four groups.

 Even after a log-log transformation, the values for slope are not very similar, regardless of looking at the trend from 1950-2005 (the available data) or 1990-2005 (since the Kyoto Protocol requires a 10% decrease below 1990 levels).

From this comparison, it seems to me that there is little statistical justification to compare trends of per capita CO2 emissions across countries, even after you do a log-log transformation. So... that would mean there seems to be a problem with the overall argument at Prometheus, since it seems (to me) to presuppose that addressing the issue by looking at per capita CO2 emissions would be a good enough control for population growth.

One interesting thing (to me at least) is looking at what is happening at different time-frames. I will use here Prometheus' own methods of comparing change in per capita CO2 emissions against a base level. (I know that I cast doubt on it, but bear with me...) I will compare these per capita graphs against both the population growth and CO2 emissions of the same period. However, instead of only showing 1997-2005, I will also show 1990-2005 and 1950-2005.
 What is apparent is that population growth is having an effect on the values of CO2 emissions in Luxembourg, especially when one considers changes since 1990 or 1997 (since Luxembourg has the highest population growth rate during this period of time, and has the smallest population to begin with). Due to this issue of scale, Luxembourg (and likely other small-population countries) should be dropped from the initial analysis. (Luxembourg was also a country that had little similarity between other countries from above).

No comments: