Agent Zero

Mark Buchanan doesn't like the dynamic stochastic general equilibrium (DSGE) model. He doesn't like these macroeconomic models because he believes that a handful of elegant equations can not accurately reflect a "messy reality." And while I share his enthusiasm over the development of agent-based models, and agree that current macroeconomic models have obvious limitations, I disagree with the overall sentiment of his article: that macroeconomic models are inaccurate because they consolidate too many agents into too few equations.

First, representing all households and all firms as one agent each is not as bad as it sounds. If we know that - in aggregate - households spend 70% of their income, there is not too much gained from modeling many different households with varying degrees of consumption, but whose average consumption level is still 70%. It may be interesting to see how different households react to different economic environments, but that's not the overarching goal of macroeconomic models. The overarching goal is to see how the overall economy reacts to different environments.

Second, it's unfair to blame the shortcoming of macroeconomic models in predicting the financial crisis and it's aftermath on the models themselves. Models use historical data. It's very hard to predict a once-in-a-generation economic crisis using a generation's worth of data. For example, look at historical housing prices:

Could your model guess what happened to housing prices after 1/1/2006? I'll give you a hint:

Please, don't hate on DSGE models, hate on DSGE models that use poor assumptions.

One Star, Two Parties

Micah Cohen of FiveThirtyEight writes about two of my favorite things: potential Democratic gains and Texas. The question he poses is how much would demographic change help Democrats make gains in Texas? Cohen reaches a pessimistic conclusion: that while Democrats close the gap by 6 points, Texas can still only be downgraded from super-duper Republican to super Republican by 2016. So just how optimistic do we have to be to see a blue Texas?

Well, a very optimistic assumption would be that Democrats could take what they learned in 2008 from every battleground state, and replicate their greatest hits in Texas. This would mean turning out African Americans like they did in Ohio, Latinos like they did in Nevada, and Asian American like they did nationwide. Using the same assumptions as Cohen but applying this new turnout model, Democrats would still only manage to garner 48.5% of the vote. This means that even the greatest hits of 2008 combined with much more favorable demographics still fail to deliver Texas Democrats to the promised land.

Now, one of the questionable aspects of this model is that it assumes that each ethnic group will vote the same way they did in 2016 as they did in 2008. But what if Democrats were able to not only increase turnout, but increase their vote share as well? If we assume that Democrats can win 70% of the Latino vote with a 52% turnout, then Texas finally turns blue as the Democratic nominee wins the state 50.3% to 49.1%.

So, in order for Democrats to win Texas in 2016, they can't just rely on a few big breaks. Every single thing has to go right. But it's not unprecedented for a state's Latino population to go from under 30% turnout to 50% and become much less conservative within one election cycle. Just look at the difference between Nevada in 2004 and Nevada in 2008.

The model for Texas in 2016 won't be Ohio or Virginia. It'll be Nevada.