The New Data

Last week, the third largest pornography website in the United States [Pornhub] released data on its annual pageviews per capita by state. A chart of pageviews by state is shown at right. [see image]

The chart as presented by the website is limited in its applicability, because relevant demographic variables are left uncontrolled for, but the conclusion nevertheless appears favorable for Mormons. Utah’s pageviews per capita in 2013 were 40th in the US. Idaho and Wyoming, the other states with large Mormon populations, are even lower on the list, at 49th and 46th respectively.

In order to find a more meaningful interpretation of the data that would adjust for possible confounding variables, I went to the trouble of gathering the most recent demographic data I could find for each state, so I could perform a controlled regression. I included variables for GDP per capita, internet penetration per capita, male/female ratio, age distribution, race and each state’s marriage rate.

Using ordinary linear regression methods, I generated a difference between a state’s actual views per capita and the views that would be predicted based on demographic variables. In this analysis, Utah’s deviation from the views predicted by demographics was 45th in the United States, while Wyoming was 46th and Idaho came 50th.

In other words, when controlling for other variables, there is an even stronger suggestion than before that Mormon populations do not have abnormally high rates of porn use (at least as represented by the website). We might even suggest that their rates of use are especially low.

I also decided to directly analyze the relationship between Mormonism and porn use (again, as measured by this particular metric). This is something the author of the 2009 study did not do. I included a variable for the percentage of a state’s population that is LDS, as measured by official LDS membership statistics and the most recent population projections based on census data.

The regression finds, roughly speaking, when controlling for the variables already mentioned, that a 10 percentage point increase in a state’s LDS population is associated with an approximate 16% decrease in the amount of porn consumption.

This result is highly significant, even at the 0.001 level. In fact, “percentage of Latter-day Saints in population” had a higher statistical significance than any other single variable I included in the regression (the next most significant variable was internet penetration). The proportion of overall explained variation in the regression is 66%, and a test for overall significance is highly conclusive, suggesting that the model as estimated is meaningful and significant.

Why do the results appear so different for these two sets of data? It’s almost impossible to know. The author of the 2009 study did not reveal the identity of the “top ten” porn vendor who gave him credit card data, and he admitted that there was no way to evaluate whether the users of that vendor were representative of the porn industry in general. His data, which was gathered from 2006 to 2008, also did not measure consumption, but rather paid subscriptions. A possible explanation of the discrepancy is that Utah’s porn use is skewed toward paid pornography.