Hacking the Electorate: How Campaigns Perceive Voters

It's as if there are two books here, and I'm writing two reviews. First, there's the book the author meant to write, which is academic and only moderately interesting. But there's a second book, meaningful only to a very narrow audience, which discovers and explains very important and interesting aspects of American politics, which were previously unknown.Hersh has a thesis, which he develops in the first few pages of the book, and then belabors to death. It goes like this: campaigns that engage in direct voter contact perceive the individual voters through the prism of the available data. Because the data that's readily available varies from state to state, the campaigns have different understandings of the voters in different places.To provide a concrete example: in a handful of states, the voter's race is recorded and is part of the data available as public data. In those states, voters are treated differently according to their race, since race is so strongly correlated with politically important attributes, such as partisan affiliation, turnout, and persuadability. In other states, which don't record race, campaigns use other data, such as precinct-level voting statistics, to tap into the same predictive power. In those states, a campaign may break the electorate into categories like "lives in a precinct whose typical vote is 95% Democratic", or "lives in a 98% white Census tract". As a result, a white voter who happens to live in a heavily black precinct is treated quite differently, depending on whether his state happens to record race. In a state which records race, he may be targeted by Republicans, who try to pick him out like a raisin in oatmeal. In another state, where race is unavailable, he might be completely written off by the Republicans, but treated by the Democrats as one of their own. And of course, this divergent treatment is likely to lead over time to divergent political behavior.That's Hersh's thesis, stated in 200 words. With additional examples, evidence, footnotes, and discussion, it might make a NY Times Magazine article. But why does it deserve a book? I didn't count the actual number of re-statements, but I guess he makes exactly the same argument fifteen times in his 220 pages. For that book, I would have awarded three stars. His writing is perfectly competent, and he seems to know his way around the academic literature. I'm a specialist in the field, but a week after reading his book, it had mostly grown dim in my memory. It just isn't much of a step forward; it isn't surprising; and it doesn't provide much illumination.But then, there's the OTHER book. The one to which I would have awarded five stars. This is the book Hersh accidentally addressed to the five hundred people in the U.S. who live and breath the rarefied air of voter databases. In the course of gathering and presenting data to explain his intended topic, Hersh goes a long way toward resolving a battle which has been quietly raging among the database nerds. That question, which is unknown to the general public is whether complex modeling based on non-public data is a significant part of political campaigning. Hersh's answer: "No."Every state provides some sort of access to the data it collects as it registers voters. (Technically, we might exclude North Dakota, which doesn't actually register its voters, but still keeps track of them.) That data, which varies widely from state to state, is certainly important to voter contact programs run by local, state, and national campaigns. It always includes name, address and voting precinct, but may also include age, gender, race, voting history, veteran status, political party preference, previous use of absentee ballots, length of residence, and - depending on the state - other surprising tidbits. Everybody uses this freely available public data; the controversy is over whether OTHER data is much used.The popular press carries frequent reference to more complex modeling of voter behavior, based on "psychodemographics" or similar approaches, which supposedly depend on privately maintained data such as survey results, data collected by commercial marketers pertaining to consumer purchases, automobile registration data, and so on. It isn't far from parody to say that campaigns supposedly try to figure out which voters prefer cats to dogs, use that information to unravel their individual psychologies, and attempt to design different pitches that will be uniquely effective to the persona that emerge. Applying the technique to cat ownership may seem to trivialize it. Exactly the same sort of analysis is said to have been applied to possession of a valid passport, subscription to hunting magazines, ownership of muscle cars, being divorced, undergoing bankruptcy, and a thousand other latent categories.Supposedly, the practitioners of this esoteric art have learned to manipulate the electorate in subtle an powerful ways. They charge high fees, because they need to purchase or assemble large and expensive datasets, and they possess deep insights into statistics, psychology, political science, and communication strategy. The humble practitioners (of whom I am one) who merely apply the data available from the local election office are left far behind - at least in marketing their services.But has Hersh rumaged through hundreds of campaigns - including the Obama campaign, which is universally acknowledged the most sophisticated America has ever seen - he found little evidence of such magic. He found lots of evidence such data has been collected, and applied to political campaigning, but he found very little evidence that it made any major difference. And everywhere, he found a sense that those who had explored the more complex modeling had either given it up as impractical, or continued to pursue it mainly as smoke-and-mirror self-promotion.The reasons, as Hersh explains, involve the practicalities of cost and benefit. Privately available data is expensive, it can't be obtained for much of the electorate, and its insights don't really have much political bite. Whether you own a cat simply doesn't tells us anything useful about your political preferences or predilections. Hersh presents his evidence on these points over several dozen pages, complete with tables, footnotes, and details.Hersh argues - correctly, I think - that publicly available voter data has been implicitly designed by the political class to satisfy their data needs. Of course, the data is supposedly collected for the purpose of administration of the voting process, but Hersh (repeatedly) points out that much of the data collected has no genuine use to the people hired to collect and count ballots on election day. So it's no accident that the data available from the Michigan Bureau of Elections happens to be very useful when applied to the practical needs of political campaigns. If the data weren't useful, the Legislature would seize some excuse for changing the process of voter registration, in order to mandate the collection of whatever additional data might be desired. Where a particular data isn't collected, it may be that the party in power prefers the status quo because they fear a change would disproportionately help the opposition.Anyway, Hacking the Electorate is a very important book - but only to a handful of people. For people whose interest in politics has nothing to do with SQL or chi squares, it may be a dud.

One of the more important takeaways for me was how inadequately consumer data and social networks are in predicting voter behavior. I’m also more cynical about micro targeting practices. This is a great book - very compelling narrative.

I enjoy the irony that I got to read this book right before the data breach scandal involving the Bernie Sanders campaign and the DNC. I'll leave that debate aside, and the book was written long before that scandal, but since it focused primarily on the campaign databases used by Democratic candidates (as Republican campaigns wouldn't give the authors access) it was great at giving context on why that scandal was such a big deal. The authors do a great job at showing just what campaigns think of voters and how access to voter information informs campaign strategy. As technology marches on and candidates take advantage of tech to get more information (such as the Ted Cruz campaign and their app), this book is a great way to get an understanding on just why campaigns want this information so badly. And for those of you who are particularly concerned about privacy, it might also be a little unsettling just how much information campaigns have access to (mostly in the form of census information and voter registration/records).

An extremely important book for anyone who wants to understand the much-touted role of "big data" in political campaigns.

Quite informative overview about the current state of campaign data. It would be a good companion to The Victory Lab.

This book profoundly changed how I understand elections. A great read!

Great book if you are interested in understanding how campaign see the electorate and the voters see the candidate.

"Most of what campaigns know about voters comes from a core set of public records" - that's one of the key messages from Eitan Hersh's pathbreaking study of what data American political campaigns really use, how good it is and where it comes from.Or in other words, the big data, grassroots-fired war of microtargeting isn't so much about super accurate data on voters linking cereal eating preferences to views on immigration as about careful computer programmers pulling in a limited number of rather prosaic publicly available datasets. That which generates the media excitement is but a small adornment on top, adding little to the accuracy of the public data or purposes to which campaigns put it, Hersh argues.This also means that telling the story about political campaigns and data should start with looking at what public data is available - such as names and addresses on the electorate register - yet political journalism on this topic tends to be about the glitzy technology and eager party workers, not about summarising the sources of public data.For the US, the one country he studies, Eitan Hersh moreover concludes that commercial data doesn't really matter much as it has "comparatively limited predictive power".That is in the context of a county which is rather different from the UK when it comes to free public data about voters because not only do many states have voters record their political affiliation in public when registering (thereby giving parties and candidates free canvass data for the electorate) but also in nearly every state the age of each voter is publicly available. In the UK the former doesn't happen and the latter only happens for voters who are going to turn 18 during the year.The availability of the age data means, as Hersh points out, that campaigns therefore are particularly keen to target messages and to segment voters by age because they have that data. Were the state to provide, say, the income of every voter but not their age, then campaigning would flip around to using that far more.The rules over data determine how campaigns classify and address the electorate and at heart even how campaigns see voters: the data categories which are widely available become the prisms through which voters are classified, understood and targeted.In the US, Hersh goes on to argue, politicians have in fact often shaped the laws over what data is available in order to benefit their campaigning. That may be different in the UK, but what is the same is how often parties are reliant on small volumes of gathered data supplemented by much larger volumes of publicly available data.What Hersh writes about the US applies to the UK too: "the fallacy is that modern campaigns are assumed to have accurate, detailed information about the preferences and behaviors of voters. When I dig into advanced campaign databases and show that much of what campaigns perceive about voters is a result of a limited set of public records, and even the most sophisticated campaigns often misperceive voter preferences, I hope to push back against the hype of the information fallacy".So there is much that is original myth-busting in Hacking the electorate. Alas there is also rather a lot too of that ponderous academic style of saying what you are going to say, saying it and then saying what you've said. But the quality of the myth-busting and the illuminating details of the evidence makes it well worth battling through reading things in triplicate.One to read, one to learn from and one whose conclusions can be enjoyed - just don't expect the road to them to be too lively.