Porn Statistics

Yesterday, Norm posted a link to post on boingboing, entitled, “Red states consume more porn?"  I have some things to say about the article's use of statistics, and rather than simply comment on Norm's blog, I thought I'd do so here, since I want to comment on it, and I want to make a blog post, but I don’t feel up to thinking of two things tonight. 

Please note that it is not my intention to dispute the validity of the article in question.  This is just a discussion on critical thinking. 

Church-goers bought less online porn on Sundays – a 1% increase in a postal code's religious attendance was associated with a 0.1% drop in subscriptions that day.

When I was a kid and was dragged to church, it lasted a total of 3 hours.  This is nearly 20% of the average person’s waking hours.  So a 1% increase in the number of people wasting 20% of their day, assuming that subscriptions-per-hour-spent-not-at-church (SPHSNAC) is a constant, should work out to a 0.2% decrease in subscriptions.  Since it only leads to a 0.1% decrease, the SPHSNAC must actually increase, not decrease as the statement would lead one to believe. 

Your SPHSNAC may vary. 

States where a majority of residents agreed with the statement "I have old-fashioned values about family and marriage," bought 3.6 more subscriptions per thousand people than states where a majority disagreed.

Did you notice they didn’t use a percentage this time?  Whenever I read any article that tries to convince me of something using a mixture of percentages and absolute numbers, I am immediately suspicious.  In this case, the number “3.6”, by itself, is almost meaningless.  It needs a reference.  Suppose in the other states that the number of subscriptions per thousand people was 2.  Then, the 3.6 increase is quite significant.  But if the number was 500, instead of 2, then an increase of 3.6 is not a big deal. 

In my experience, the only reason to switch between percentages and actual values is to choose the one that makes your point in the more compelling manner.

Remember, 50% of all statistics are just made up.