In Double Down: Game Change 2012, Mark Halperin and John Heilemann report that Neil Newhouse, Mitt Romney’s pollster, “was certain that no convention could have appreciable impact on voters’ perceptions of whether the country was on the right or the wrong track,” (p. 391). After the Democratic convention in Charlotte, however, as public opinion dramatically shifted in favor of Barack Obama, Newhouse was forced to re-evaluate his certainty. Conventions, it seems, can matter quite a lot.

The shift toward Obama in September of 2012 would not have been a surprise to Newhouse and his colleagues if they had read an advance copy of Bob Erikson and Chris Wlezien’s The Timeline of Presidential Campaigns (published in October of 2012). Amassing every publicly available trial-heat poll from election years between 1952 and 2008, Erikson and Wlezien search for patterns in the polls that suggest whether and how presidential campaigns affect election outcomes. In doing so, they bring the tools of time series analysis to bear on the fundamental tension that haunts every scholar, reporter, or consultant trying to understand the effects of campaigns: how much of the final outcome is determined by what the candidates do (or what happens to them) in the weeks leading up to the election and how much is driven by the things out of their control, like the state of the nation’s economy, the distribution of party identification in the country, or presidential approval? To describe this tension another way, are elections mainly determined by the choices candidates make or the chance circumstances in which they find themselves?

The answer, of course, is both—and Erikson and Wlezien add their voices to the small but growing set of scholars showing how the process of persuasion in campaigns is slow—and built on solid foundations set in place before the candidates are even known. Despite the media’s obsessions with game-changing campaign moments, very few “games” are changed by a single, transformative event happening in the 300 or so days before presidential elections. Nowhere is this revealed in starker terms than in Timeline, which leverages the weight of 15 contests, 30 major-party candidates, and generations of voters spanning 60 years. This design lends gravitas to the argument and is a welcome antidote to trendy work on campaign effects that relies on survey-experiments in isolation of competition and is increasingly disconnected from campaign realities.

The book makes three important points. The first is simply to illustrate for readers the impressive aggregate-level stability that defines most election-year poll results and how closely those polls are tied to eventual outcomes. Beginning in April of each year and going through to November, Erikson and Wlezien show that in each of the 15 years they investigate, the early polls (even several hundred days out) do a pretty good job of characterizing the eventual outcome, but more importantly, as the polls change over the months of the election year and Election Day approaches, the change in poll results is systematic and slow. You would be hard pressed to find a game-changer in any of these elections since 1952. Political pundits, take note.

This is not to say that things don’t move around; they do. And Erikson and Wlezien do us a great service by separating the types of changes found in aggregate poll results over the years into two types: bounces (in which poll results go up but eventually back down) and bumps (in which the shift in poll numbers is permanent). Bounces are largely ignorable, unless they happen in the final days of the campaign. But bumps—bumps can be important and interesting, but they don’t happen very often, and when they do, it’s typically because of something predictable like a convention.

In an unexpected but delightful chapter, the authors abandon their aggregate data for individual-level panel data from the American National Election Studies (and Gallup, too). Using these data, they offer even more evidence that change happens, but not to the degree that pundits and political reporters suppose. Using these data, Erikson and Wlezien put an upper bound on the number of voters who change their minds. Just how many voters switch their party vote at least once between election years? The largest shift, given the years for which they have data, comes between 1972 and 1976, as 21 percent of Nixon voters in 1972 report voting for Carter in 1976; and 3 percent of McGovern voters switch to Ford. If roughly a fifth of the electorate is switching between elections, the authors argue, it seems unlikely that any more than that are switching within a single election year. Similar numbers are found when they investigate party defection within a campaign year—about a fifth of the electorate, on average, abandons their party candidate early in the election year (April), but slowly and steadily, those wandering voters come home—and by November, more than half of them have returned to their original party choice.

Finally, Timeline delivers on the promise in its title—it reveals the systematic, predictable ways in which voters are affected by campaigns in the 300 days leading up to election day—and how three important moments in the timeline are crucial for understanding election outcomes. The first is the early part of the election year, about 300 days before the election, when the nominating process is in full swing and the candidates are just being introduced to most voters. Erikson and Wlezien find a lot of shifts during this period, as people’s vote choice initially takes shape and the process through which information is translated into decisions may be a little noisier for most people (they don’t yet know what considerations to give the most weight to because the campaigns are not yet in full swing).

After this initial period, the next important time is the conventions. During this period a massive amount of information is revealed to voters and the fact that the election is right around the corner becomes difficult to ignore. It is during and after the conventions that the final outcome starts to solidify in the trial heat polls—partisans come home to their party candidates and undecided voters who eventually will drop out begin to do so. Finally, the last few days before the election are a campaign’s final chance to nudge the outcome their way because bounces can be consequential if the election is very close.

If you study presidential politics or time series analyses, there is a lot to like in Timeline. The connection between the method and the substance is close and tight, which makes this book a great example of how the right method can help illustrate important nuances in the substance of a problem. For example, the thoughtful discussion of whether the polling time series is stationary or integrated helps illustrate the important differences between bounces and bumps. But by far, the most important contribution the book makes is to illustrate that presidential campaigns matter in predictable ways, and voters’ intentions evolve incrementally over the course of the election year.

Election outcomes don’t always reflect where the polls start in April, but they always begin to reflect the outcome early in the year. There are no outcome-changing gaffes, only underlying fundamentals and campaigns that help voters make sense of the state of the world around them. To be clear: for Erikson and Wlezien, presidential campaigns matter—despite the regularity of attitude change in elections, the campaigns are an (if not the) important catalyst in the process. Without them, the noisiness of decision-making might grow instead of shrink. The regularity comes from something—and it’s unlikely that thing is just the passage of time or the proximity of Election Day.