Do elections matter?
Oct 30, 2010
3 minute read

I am told that there will be an election next week. Actually nobody told me; it’s what I gathered from the yard signs. Elections, for me, are a spectator sport. I will probably be up all night watching the Prop 19 returns come in, but I haven’t participated since Bryan Caplan showed me the mathematics of voter decisiveness.

While I agree with Caplan and others that voting doesn’t matter, my own research (very much in progress, caveat emptor) is making me wonder whether it’s possible, even in theory, for an election to matter. We all know that an individual vote doesn’t change the outcome of an election. But does the outcome of an election causally change policy, or does it just correlate with policy change?

Increasingly, I think the latter. Here’s one thought experiment: what would policy in the US look like if it were an autocracy instead of a democracy and nothing else were different? That is, let’s impose a strong ceteris paribus condition and change the form of government from democracy to autocracy.

My tentative reaction is that everything would be roughly the same as it is now. Think about it in Coasian terms. Government policies impose negative externalities on some people. Those people have an incentive to bargain with other people in order to get that policy changed. That can be as simple as bribing the autocrat to change the policy, or as radical as a coup in which defectors from the regime are promised larger returns than they are currently getting.

As in any Coasian theory, transaction costs matter, and insofar as transaction costs prevent exchange, it is possible for the nominal form of government to make a difference. This is why the nasty autocrats are so nasty; they have mechanisms, based on ethnicity, ideology, or external support, of preventing the exchanges that would remove them from power. But in the US, transaction costs seem reasonable even if they are not negligible. Furthermore, transaction costs must be evaluated relative to the externalities to be addressed. Modest transaction costs mean only modest externalities remain. As the externalities increase in magnitude, holding transaction costs constant, the greater is the likelihood that exchanges will resolve them.

If the US were a ceteris paribus autocracy, with the modest transaction costs and wide distribution of power that now exist, we’d get basically the same outcomes we have today. I’m not saying it would be exactly the same, but it would be close. And if that statement is true, then it must also be the case that elections mostly just correlate with policy change, they don’t cause it. Changes in electoral outcomes reveal changes in constraints faced by the government, they don’t themselves drive the change in policy.

This is not just a semantic difference. It means that your opinion gets counted about the same whether you show up to vote or not. In the long run, it doesn’t matter if voter turnout is 90 percent or 10 percent, or if voter turnout is ideologically lopsided. The real constraint that the government faces is still the same. Elections don’t matter.

This is where I resist the urge to water down my conclusions with statements about how tentatively I hold these views. I hold these views. Tell me, is this my most absurd belief?