(Photo: AP/Mel Evans)
Although news reports and commentators during this year's presidential election have focused on twists in the race and shifts in polls, the real story about campaigns since the mid-1990s is how little movement there is during a general election. American voters are much less open to persuasion by the other side than they used to be. With increased partisan polarization among voters, how states vote from one election to the next also changes much less than it previously did. That's the reason we can talk about "red states" and "blue states" and focus on only a few battlegrounds. Election outcomes aren't foreordained, but all the movement has been taking place within a relatively narrow range.
The volatility of public opinion during an election year can be measured using the standard deviation, a number that describes how far sentiment has departed from the average. Like the stock market, presidential races may be volatile (high standard deviation) or placid (low standard deviation). As the figure below shows, low-variability campaigns suddenly became the norm in the mid-1990s.
Six of the seven least volatile races in the past 64 years have occurred since Bill Clinton's re-election in 1996. During the general election in 2008, for example, the Obama-McCain margin spent about two-thirds of its time within 2 percentage points of its average. Compare that remarkably low level of variation to 1980, when the Reagan-Carter margin had a standard deviation of 12 percentage points-six times as volatile. Shown in red is an even more precise measure of volatility that I developed in 2004-the state poll Meta-Margin-which confirms the stasis of recent campaigns.
What happened between 1992 and 1996 to make presidential campaigns more stable? For one thing, Americans started voting more predictably on the basis of national party affiliations and doing so for both the presidency and Congress. Starting from the mid-1990s, the presidential popular vote and the national congressional vote have come into close alignment, differing by an average of only 2.9 percent. Earlier, there was some truth to the dictum that "all politics is local," as Tip O'Neill, the Democratic speaker of the House from 1977 to 1987, famously put it. But today all politics is national, and it has been that way ever since 1994, when Newt Gingrich and the House Republicans swept into power with a national platform, the Contract with America.
Of course, these developments were long in the making. The role of the two parties in the civil-rights revolution in the 1960s brought about a massive regional realignment, which had its final chapter with the Republican takeover of the House in 1994. But other factors in the 1990s, particularly innovations in media and technology, may have contributed to the nationalization of politics, especially among Republicans. Conservative talk radio and cable news-above all, Rush Limbaugh and Fox-began providing their audience with a strong, one-sided, partisan message on a daily basis. Local, person-to-person communication has also become more ideologically consistent. As documented by Bill Bishop in The Big Sort and confirmed in detail by later researchers, the last few decades have seen a clustering of like-minded voters with one another in the same communities.
The geographic sorting of voters facilitates the creation of noncompetitive districts. The number of swing districts that can go either way has decreased, a trend that is driven both by population clustering and partisan gerrymandering. In most districts, general elections for state legislatures or the House of Representatives carry little suspense. Primary elections, when turnout is low and dominated by the most motivated partisans, determine who gets elected.
The hardening of partisan geography explains why the terms "red states" and "blue states" now make sense to us.
Since 2000, only two states, Virginia and Colorado, have shifted from being more Republican than average to being more Democratic than average; only Missouri has shifted in the opposite direction. The rest have stayed the same.
To determine when geographic voting patterns change in consecutive elections, I use the state-by-state correlation coefficient, a measure that is based on whether individual states are above or below average in their partisan strength. Correlation coefficients are not affected by national swings in opinion; in other words, if one party wins an election by gaining an additional 10 percent of voters in every state, there is no change in this measure.
The maximum possible value for the correlation coefficient is +1.00, which indicates perfect proportionality with only an offset for the average national swing. If there is no relationship between state-by-state voting in consecutive elections, the correlation will be zero. Consider two elections won by Democrats. In 1992, the Democrats won 32 states under Bill Clinton after winning just ten states under Michael Dukakis in 1988. That change was principally driven by a nationwide shift in Democratic-Republican vote margins by an average of nearly 13 percentage points. The state-to-state correlation coefficient was +0.90.
In contrast, when Democrats won a landslide victory under Lyndon Johnson in 1964, they didn't just benefit from a nationwide swing from 1960, when John F. Kennedy won by a narrow margin. The 1964 election shook up the entire map. The state-by-state correlation between 1960 and 1964 was -0.03, close to zero and indicating no relationship.
In recent years, consecutive elections show very little change in state-by-state voting patterns. In the last 15 elections, the five largest election-to-election correlations occurred between 1992 and 2012-and 2016 looks to be similar. The color-coded map below left shows polling margins, with darker colors indicating larger leads.
As of early September, the pre-election polling margins between Donald Trump and Hillary Clinton were a near-replica of the Romney-Obama race, with a correlation of +0.93. Despite the radical nature of Trump's candidacy, the distribution of his support reflects the same pattern as the support for George W. Bush, John McCain, and Mitt Romney.
No matter who wins the presidency, these trends suggest that polarization will remain after November. That's not to say the current patterns are fixed forever. Although primaries favor each party's diehards, the fissure that opened this year between GOP party leaders and voters may open the way for surprises-including even future realignments. The 2018 campaign may show what kind of fruit arises from seeds planted in 2016.