Hacker News Reader

Comment by frereubu | Hacker News Reader

Any correlation like this - as implicitly stated in the article - should be taken with a massive pinch of salt. They're only given any credence because they "make sense" - i.e. fit preconceived notions of how the world works, mixed with a magic PR-friendly sprinkling of supposedly counter-intuition - rather than anything meaningful in the stats.

For more, see http://www.tylervigen.com/spurious-correlations where you will find that, for example, the number of people who drowned by falling into a pool correlates with the number of films that Nicholas Cage appeared in that year.

View on HN

But the article does enumerate flaws wit the cellphone-violent crime hypothesis:

The University of Leeds criminologist Graham Farrell, who is closely associated with the hypothesis that better security technology is the primary cause of the crime decline, also took issue with some of the paper’s data analysis. “At first glance, it seems to be that antenna [density] increased mostly after homicide already declined,” he wrote to me in an email.

The data that the economists presented don’t match the chronology of the decline of homicides, especially considering that their proxy variable—how many antennas were up—would almost certainly precede cellphone usage by some period of time. The timing, he said, is “not even close.”

Correlation doesn't imply causation, but no one said it did.

> their proxy variable—how many antennas were up—would almost certainly precede cellphone usage by some period of time.

I find this assumption VERY suspect.

Cell tower density is going to increase in the most profitable areas first--and those are not necessarily going to be correlated with high-crime areas (in fact--probably anti-correlated).

Cell phones are useful even if just a small number of drug dealers have them (remember when drug dealers used to have pagers?). You would probably have to correlate cell phones with pagers in some way in order to make a valid correlation.

It's the title, combined with the fact that they've even given the idea air time, which made me write that comment. My argument is essentially that this article shouldn't have been written.

Yup, through the 90's you can likely find all sorts of factors that correlate pretty well just because they were on the rise too. The question really is how to pick out causation given spurious correlations, noise, and multiple causal factors.

Exactly. A motivated individual could find dozens of supposed causes of reduced crime in the 90s.

Exhibit A - Murders by sharp object correlates with visitors to SeaWorld - http://tylervigen.com/view_correlation?id=41291

Reply on news.ycombinator.com