The Corner

Re: National Review Jukes The Stats

In a guest post at Timothy Noah’s generally very good blog at The New Republic, Nathan Pippenger accuses me of “juking the stats” in my recent post comparing inequality data for Europe and America

He says of my post:

Here’s an even simpler demonstration of Manzi’s folly: The CIA already constructed a ranking of inequality by its most common measure, the Gini index—and it included the European Union, allowing for easy comparison with the United States. The result: Not even close.

I don’t think that Pippenger understands his own data source.

The CIA Factbook page that he cites lists the Gini Index for the European Union as 30.4 in 2010. This is the reported 2010 value for the EU-27 in the official Eurostat database. You can find the data page here. This number is defined as the weighted-average of the individual country-level Gini Indexes (This issue is described and discussed at length in the papers that I linked to in my post, and you can confirm the calculation for yourself by downloading Eurostat’s 2010 population numbers by country here, and doing the weight-averaging with the available country-level Gini Indexes — I did this before I published my post.)

But the central analytical point of my post was that this is different than calculating a Gini Index for Europe as a whole. If different countries have materially different average incomes, this difference will be severe.

By illustrative example, Finland and Slovakia each have around five and a half million people. Each country has a very similar Gini Index: According to the base Eurostat data for 2010, Finland’s is 25.4, and Slovakia’s is 25.9. If I weight-average these two Gini Indexes to create a Gini Index for the hypothetical reporting district of Finland plus Slovakia, I will get a Gini Index right around 25.7. But the actual Gini Index for the roughly 11 million people who live in Finland and Slovakia is enormously greater than this, because per capita income in Finland is more than 50 percent higher than it is in Slovakia.

Estimating such an actual income distribution for the whole European population was the purpose of the analysis in my post. This is why I spent all that time describing this difference at a conceptual level with the Laketown-Hillsville analogy, and why I described the exact methodology and data sources I used in such detail.

The balance of Pippenger’s post seems either to ignore qualifications that I was careful to make at length, or to impute to me positions that I have not stated and do not hold.

Jim Manzi is CEO of Applied Predictive Technologies (APT), an applied artificial intelligence software company.


The Latest