Fear of Hell Might Fire Up the Economy

Read article

Letter Writer:

Garry Shilson-Josling, economist, Australian Associated Press, Sydney

Date Posted:

March 23, 2005


As a professional economist, I am dismayed that you have not properly explained why the entire original analysis was excised from this article. Your readers, and the readers of numerous web sites referring to the article, deserve better.

The original article simply misconstrued the rank correlations between corruption and belief in hell. In the corruption data published with the article, the highest rank of 1 was given to the most "corrupt" nation (Nigeria), while the highest hell-believing rank of 1 was given to the LEAST hell-believing nation (Latvia). The negative correlation between these two indicators actually implied a positive correlation between corruption and belief in hell (and therefore a negative correlation between religious belief and per-capita GDP). This is the exact opposite of the original claim made by your researchers, and STILL implied by the introduction to the revised article as well as its title. I have checked the original data, as well as larger data-sets across different time periods from the same sources.

The results are clear: ceteris paribus, fervent belief in hell and corruption go hand-in-hand. The USA and Ireland are outliers rather than typical hell-believing, wealthy nations as claimed in the article as first published.

Perhaps believers should consider how to stop wealth eroding spirituality rather than claiming spirituality brings material rewards. In any case, please do the right thing and set the record straight.

Editor's Note:

A professional economist should be aware of the difference between simple correlations and regression results. As stated in the editor's note accompanying the online version of the article, the original version of the article used simple correlations to illustrate serious regression results found in work by Barro and McCleary.

It turns out, however, that the simple correlations are not consistent with the regression results. This has absolutely no bearing on the validity of the regression results, however, which are the results to be taken seriously.