Review: American Nations: A History of the Eleven Rival Regional Cultures of North America

In machine learning, there is a class of algorithms known as "unsupervised learning" algorithms. They will take in a set of data and automatically figure out how to label these data points to, hopefully, tell a compelling and actionable journey. They have a few failure modes. One type of failure mode is picking too few or too many clusters for the algorithm to find.

Let's say you put in the color, alcohol content, bitterness and sourness of different drinks into an algorithm- well, if you say "only find me two clusters" and you give it wine, beers, and hard liquor it might start labeling vodka as a beer, or cider as a wine! It is liable to tell you the wrong thing. What if you don't limit the number of clusters? Well, it might keep going and going and going. Soon, you've given it fifty different drinks and it says there are 50 different clusters.

The information, in both cases, is useless.

American Nations seems to have this problem, or at least, a highly related problem: how do you cut up the United States geographically to understand what the hell is going on? How do you cluster the cities and the plains and the deserts and the mountains into separate and cohesive wholes?

There are literally thousands of maps that attempt to do this on the Internet. They do it by commuting zones or population size or voting or the way you say "soda" or "pop" or the regions top porn search on Google. Do any of these make more sense than the others? Who is to say?

Woodard's attempt is to do it based on colonial history. He reminds us that American history did not start in 1776- that it was building for 200 years before colonialists even began to question whether remaining with Britain made sense. He reminds us that the American nation had many, many secession movements from all quarters before 1860.

He posits the existence of 11 nations, each connecting with different colonial waves and geographies. The first two come from the French and Spanish on the continental US's geographical edges. The Puritans, the Dutch, and the Quakers make up the northern nation. To the south, there is the Tidewater gentry of Virginia and the Slavelords of South Carolina (who literally came from Caribean plantations). Greater Appalachia contains the remnants of war refugees from Britain's internal wars of conflict.

American history, in Woodard's version, is the history of alliances and rivalries between these different nations.

But by the end of his history, everything seems like a just-so story. Maybe that's the point. Maybe Woodard is being hyperbolic about how an "American nation" isn't real and that he actually just means some trends in the voting habits by region are useful to name and think about.

There is no model-fighting in this book, and model-fighting is what you'd expect to have in a book that is trying to convince you of something. A lot of Woodard's statistical reasoning comes from predicting who will vote for who in Presidential elections. This is great and all, but how does his model beat "naive models" that guess that regions that vote for one party will just vote for that party again? How does it beat more complicated models that cluster the United States into two regions- the rural and the urban? The haves and the have-nots?

Should San Francisco and Portland and Seattle really be lumped in with another? Is the NOVA region even a little Appalachian?

How did the post-World War 2 migrations change these dynamics? Surely the fact that millions of black Americans fled north and millions of freezing cold Yankees moved south should have changed the way these dynamics work?

On the whole, this book is entertaining, and some of the characterizations of each region certainly ring true, but it is only really interesting as a play model. It isn't trying to be a real one.