The Sovereign Mind

Free thought on politics and real life

How to Mislead with Charts: Who’s Responsible for the Great Divergence?

with 3 comments

I’ve been following with interest Timothy Noah’s series on income inequality in the United States. In the latest installment, he cites Larry Bartels’ book in which Bartels supposedly proves that Republicans are responsible for the “great divergence”:

Bartels came to this conclusion by looking at average annual pre-tax income growth (corrected for inflation) for the years 1948 to 2005, a period encompassing much of the egalitarian Great Compression and all of the inegalitarian Great Divergence (up until the time he did his research). Bartels broke down the data according to income percentile and whether the president was a Democrat or a Republican. Figuring the effects of White House policies were best measured on a one-year lag, Bartels eliminated each president’s first year in office and substituted the year following departure. Here is what he found:

That looks pretty impressive. According to the chart, not only have Democratic presidents created more equitable income growth, but they’ve created larger growth for every income category! This seems to be a slam dunk case against Republicans. But there are some problems. Firstly, is it really right to consider only a one-year lag before a president is fully responsible for the economy? President Obama might object to that! Secondly, of course the president is not the only one who affects economic growth. We don’t live in a dictatorship. What about the congress? Again, you can ask President Obama how easy it is for a president to get exactly the policy he wants, even when his own party controls congress, much less when it doesn’t. I decided to take a look at these two questions.

Firstly, I wanted to reproduce Bartels’ data. Unfortunately the Census Bureau’s historical tables that I found only go back to 1967, so I had to start from there. In any case, that’s about when the “great divergence” started, so that should be the most interesting data set anyway. I get similar results as Bartels:

But what happens when I tweak the parameters to have a two-year lag instead of a one-year lag?

Now we see a slightly different picture. Republican presidents still help the rich more, at the expense of the middle class, but the over-all economic growth picture is more fuzzy. Is a two-year lag better than a one-year lag? I don’t know. The point is that Bartels’ decision to use a one-year lag is arbitrary, and I’ve demonstrated that we get a very different result by just tweaking one arbitrary parameter. That’s not a sign of solid scientific evidence. What if I were to tell you that a climate model could be tweaked to predict global cooling instead of global warming just by tweaking one little parameter that was chosen arbitrarily to begin with?

Ok, but still even my tweaked graph doesn’t look good for Republicans: it still supports the argument that Republican presidents help the rich at the expense of the middle class. But what about congress? What if we looked at which party held the majority and ran the same analysis? I did that, dropping the years were there was a split legislature with one party controlling the senate and the other controlling the house. I’m actually left with only a handful of years with Republicans in control of both chambers, but that illustrates yet another problem with Bartels’ methodology: we’re talking about precious few data points to begin with, not to mention we’re not controlling for any other variables. In any case, here’s the result with a one-year lag:

Hmm… this graph looks very different from the first one we saw. Never fear, Democrats, using a two-year lag makes things look a little better for you:

So, which party’s policies are contributing more to income inequality? Which party is better at producing economic growth? My point is not to answer those questions. My point is to show that Bartels’ guess is no better than yours or mine. His methodology is interesting, but unfortunately fatally over-simplified.

Advertisements

Written by Mike

September 11, 2010 at 10:30 pm

3 Responses

Subscribe to comments with RSS.

  1. […] Bartel’s methodology is interesting but fatally over-simplified. See here for my critique: https://sovereignmind.wordpress.com/2010/09/11/how-to-mislead-with-charts-whos-responsible-for-the-gr… I won’t go through Mike at SoveriegnMind’s logic, but he calls the bar graph into […]

  2. […] a comment » Since my last post on chart-truth was such a hit, I thought I’d come back for more and do another spin-off of Timothy […]

  3. […] top of that, Mike at A Sovereign Mind tweaks the charts from Slate in order to demonstrate how they are misleading.  His conclusion is […]


Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

%d bloggers like this: