Joining up the dots

by Henry Farrell on March 29, 2005

Via Max, this gem from Doug Henwood’s Left Business Observer, deconstructing claims that the Heritage Foundation and Wall Street Journal’s “Index of Economic Freedom” tells us anything meaningful about economic growth. If you look at countries’ scores on the index in 1996, and graph it against their subsequent per capita GDP growth, there isn’t any meaningful correlation. If (as the Index’s authors do), you look at the relationship with aggregate GDP, you get a smallish correlation, which is very likely spurious. Short version: the Index is probably garbage – it doesn’t tell us about anything save for the ideological predilections of its authors. But take a look at the graph below and read Doug’s article, so you can judge for yourself.

Index of Freedom v. per capita GDP growth

{ 4 trackbacks }

Crooked Timber » » Joining up the dots II
03.29.05 at 4:15 pm » The Overwhelming Correlation Between Taxes and Growth
04.14.05 at 5:40 pm
Crooked Timber » » A mess of pottage
04.16.05 at 10:22 pm
Ragingpundits » Economic Freedom and Growth
08.21.05 at 3:08 pm



AdamSmithee 03.29.05 at 12:45 pm

The Index is pretty much bunkum (see here). But, of course, that still leaves open the possibility that better-measured indices of ‘good policies’ might be related to long-run growth…


Randolph Fritz 03.29.05 at 12:51 pm

This is the first time in a while that a graph has made me laugh. Thanks.


Oh Snap! 03.29.05 at 1:00 pm

Is that… Alan Greenspan’s liver?!

It all makes sense now!


Dylan 03.29.05 at 1:02 pm

While the chart is pretty funny, I have questions about its fairness. For instance, why use the rank of GDP per capita growth and “Freedom Index” rather than the numerical values?


Kieran Healy 03.29.05 at 1:22 pm

Heritage says the index is constructed from more than fify variables, which seems a little weird to me on the face of it. A bit of data reduction might go a long way there.


Matthew B. 03.29.05 at 1:24 pm

All you need to know about that “Index of Economic Freedom” is that Singapore routinely places near or at the top. Singapore, where you’re apt to be hit with a surprise conviction for tax evasion or defamatory libel if you’re foolish enough to, say, run for an opposition party. The Index folks still give it a perfect 1.0 score in property rights: “Despite reports of political influence, the legal system is sound and enforces contracts effectively.” Well, all right then.


Tim 03.29.05 at 1:45 pm

Henry, I agree the index is likely garbage, but I don’t read in your post much of a criticism beyond the charge of ideological bias and poor model fit (valid points, but obvious ones). It’s also interesting that the LBO article doesn’t critique the construction of the index much more than to point out that it’s long on negative freedoms and short on positive freedoms (again, an important point, but pretty obvious).

A more cogent criticism would be to point out that the overall index is a linear composite of other measures and that these indicators are assumed to measure a single, coherent underlying concept (i.e., “freedom”). What is more, it is assumed that these indicators all measure “freedom” equally well.

I’d like to see what could be done with factor analysis/structural equation modeling with these data; simple linear regression (i.e., kiddie stats) isn’t going to tell us very much. And lets gets some measures of positive freedom at the same time. The project could even be called the Correlates of Freedom (COF) project (pronounced “cough”).


Doug Henwood 03.29.05 at 3:32 pm

Just a little correction to the original entry: Heritage/WSJ’s preferred measure is the change in the Freedom Index over time and its relation to economic growth. I’m arguing that the change in the index can be contaminated by growth itself (not to mention scorer bias), so if the index means anything, the standing in a base year should have an effect on subsequent performance, because long-term decisions are made on the basis of structures that don’t change much over time.

Why rankings rather than growth itself? Because in economics (and the perception of welfare) it’s all relative.


Barry 03.29.05 at 3:44 pm

And use of ranks is a standard technique to minimize the influence of outliers.


DeadHorseBeater 03.29.05 at 4:04 pm

The LBO author correctly points out that 1st week Stat 1 students know that correlation isn’t causation. Good stuff that.
Apparently the LBO author did not get through the rest of the course, whereby they discussed things like CONDITIONAL probabilities.
Namely, we would like to know the correlation (and thus potential for causation) between our outcome (growth) and input (freedom score) variables.

That is: Shouldn’t we be attempting to control for other variables that might be relevant?
Like initial level of GDP? (If we think that economic laggards tend to converge to leaders.)
Or whether the country had a civil war during the study period?
Or about a zillion other things (besides those reflected in this Freedom Score) that might affect growth?

This is not to defend Heritage — think tanks in generally have hack and/or second-rate tendencies. They may have committed similar sins of analysis, either by ignorance or malign intent.
But it is truly surprising to see a Timberite make such a mistake of ignorance. Or partisan hackitude.


DeadHorseBeater 03.29.05 at 4:06 pm

Urk. That sentence after the cap-Conditional should point out that we want to know the conditional correlation. Essentially, we want to run a multivariate regression instead of a bivariate one.


Daniel 03.29.05 at 5:06 pm

That is: Shouldn’t we be attempting to control for other variables that might be relevant?

Give over. Maybe if we saw the inklings of a connection but wanted to know whether it was statistically significant. But look at that scatterplot. You can’t fit to white noise.


theCoach 03.29.05 at 6:00 pm

The ATOM feed is distributing a post titled ‘Joining up the dots II’ with a link to “” which results in a CrookedTimber 404 page.

Also, the nice quotes are sent down in an escaped CDATA section:


theCoach 03.29.05 at 6:02 pm

In the feed that shows up as


Decnavda 03.29.05 at 6:10 pm

Julian Sanchez bloged this at Hit & Run, flatly stating that the results were SO contrary to his beliefs that he refused to believe it. I lot of the commentors tried to bring up methodologic flaws. In my opinion the best of those were (1)to point out that this is ONE study – essentially, more data needed. Fine, I would like to see more data, I don’t think that will help them much, but let’s let the data do the talking. And (2) some people mentioned that the freedom axis showed rankings rather than absolute scores. Again, I would like to see the chart the way they suggest. Maybe slightly better for them, but not much.

Still, the comment that got my attention was this one, by “jc”:

If the chart can be believed, then it’s both a case against freedom AND a case against regulation. Economically speaking.

This really is a good point that I would like those here to respond to. What’s the point of arguing against ecconomic libertarianism, any more than for it, if the regulators are not likely to do any better. Sure, you can say that the TYPE of regulation matters, but this chart does not support that: Ecconomic regulation success stories appear to be as random as ecconomic libertarian success stories.


Nicholas Weininger 03.29.05 at 6:28 pm

The obvious conclusion, to this libertarian, is that the Index of Economic Freedom tells us very little about actual economic freedom.

Which shouldn’t be a surprise, given comments #5 and #6 above, and given that what Heritage and WSJ want to promote has less to do with actual free markets and more with crony capitalism on the Bush model.


Nicholas Weininger 03.29.05 at 6:30 pm

Which, now that I reread Henry’s original post, is pretty much what he was saying too. My bad for redundancy.


Randolph Fritz 03.29.05 at 8:53 pm

“You can’t fit to white noise.”

<tiredcrankyhungry>If you filter it, you can make music from it.</tiredcrankyhungry> :-)

Comments on this entry are closed.