Networks, Violence, Externalities, and Denominators: an Analytical Puzzle

Something recently occurred to me as I continue to slouch towards a finished dissertation. The modern, tech-savvy economist's eyes will twinkle as she tells you tales of network externalities, of how being embedded in ever vaster webs of personal connections and voluntary associations yields gloriously cheap transmission. The world is edging ever closer to the caricature of the Coase Theorem, where near-zero transaction costs imply a wildflowering of voluntary, personal conflict resolution. This could even be one way to (incorrectly, since it misapprehends his claim) stand athwart Tyler Cowen's Great Stagnation thesis. More choice today includes more choice over whom to associate with. As a quick example, I am Facebook friends with Bill Leeb, Martin Atkins, cEvin Key, mc chris, and a smattering of other minor celebrities.

Hell, I blog here with Michael f'-omg-ing Munger! I have scores of friends I've never met in person, and thanks to knock-on effects of ever-denser social networks, I can solicit communication freely from the leading lights of my larger community. If you're old enough to remember typewritten post and wooden cabinets filled with 3x5 cards at the library, you'll recognize what a boon this is.

The unsung downshot of this is that even as GDP figures from the 1970s are probably drastically understated when piled next to today's figures, so are the per-capita violent crime statistics. The dis-utility of a violent crime event spreads wider and faster than it did 40 years ago.

And that's a feature, not a bug! It's good that murders, robberies, and assaults incidentally harm more folks per incident. Why? Because it gives folks a greater personal incentive to re-arrange the parameters of the crime and punishment game, particularly detection. By now, I'm sure you've heard tales of intrepid Internet sleuths hunting down bullies, thugs, hit-and-run drivers, and the like and hauling them to justice. The SPNE of this is... less crime.

So fleshy, broad networks are good not just because of the direct upside benefits, but perhaps also because they help mitigate downside risks. This makes them super-euvoluntary. This probably also suggests that to really reap the benefits of downside risk mitigation, legislators should avoid hobbling Web community-based police work.

At any rate, for those of us who work frequently with panel data on violent crime, I'd love to hear some ideas on how to attempt to control for this effect. It's something more than just time fixed effects, but I'm worried that it'll inevitably get captured as such. Maybe a natural experiment discontinuity regression around an ISP outage or something? Hm. It might be worth thinking about a little more.