Thursday, May 03, 2007

Component Analysis Snark

A competent analyst looking at a complex problem often is able to break things down into more manageable chunks. Some of these chunks are constants which can be skimmed over, while other components are highly variable and deserve an in-depth inquiry to figure out what is actually going on and what probable future scenarios could look like. Good analysts are then able to pull together a coherent story to tell their boss, to tell their clients, or to inform their own actions. This story incorporates the known realities and does not exclude unfavorable data.

The NewsBusters is reputedly a conservative analytical site. One of their posters took a look at the State Department's annual terrorism report and thinks he has a story that the media is overly trumpeting the worldwide increase in incidents and deaths from terrorism. The story he is trying to tell is that if you exclude a couple of components the rest of the report is not that bad.

State Department released its annual Country Reports on Terrorism showing a number of interesting findings, including steep declines in terrorist attacks and murders in many regions of the globe......

Instead of saying that terrorism has increased markedly in Iraq (the truth), the media are extrapolating beyond that to claim that, as Reuters puts it, "U.S. sees sharp rise in global terrorism deaths."

Once you get past the lede of these Reuters and Associated Press pieces, you'll discover the small detail that the increase in terrorism was almost entirely due to Iraq. Nowhere in either piece do you learn the fact that aside from the Middle East (which does not include Afghanistan according to State), the number of terrorist attacks worldwide is down from a year ago by over 300 incidents. The number of deaths from terrorism was only up 14 percent.


So if you exclude Iraq, he argues that Bush's strategy is successful as we only saw a global increase in fatalities by 14%. This 14% increase in fatalities is the good news that the guy is trying to sell.

On the other hand, a good analysis of the economy by Caroline Baum [h/t Barry at the Big Picture]uses component analysis to look at the impact of housing on the broader economy. The picture is not pretty but I just want to pull one great quote on the limitations of component analysis from her.

"Excluding housing, the U.S. economy is doing just fine.

That's the latest rationalization of a select group of operators who think that the Bush administration's 4.6 percentage point cut in the top marginal tax rate and 5-point reduction in the top capital gains rate can protect the economy from any and all ills.

To say that ex-housing the economy is doing just fine is tantamount to claiming that, ex-Iraq, Bush's Middle-East policy is a rousing success."


Being able to exclude data to illustrate points, to find strengths and weaknesses and look for underlying trends is a good thing. However the best analysts remember that they made these exclusions and include this information when they actually think about their recommendations and descriptions of reality. Permanently excluding Iraq from any broad discussion of terrorism, national security or foreign policy instantly marks the analyst as incompetent.