Every U.S. president elected in my lifetime was either a Democrat or a Republican. My parents could have said the same thing. Also my grandparents and great-grandparents. These facts will come as no surprise to anyone who has the slightest acquaintance with American politics and history. After all, we have a “two party system.” Nevertheless, the longevity of the two dominant parties seems remarkable to me. Issues come and go, voters shift their allegiances, the candidates and platforms vary wildly from one electoral cycle to the next, but every four years the same two teams come back for another big matchup. The Democrat/Republican rivalry has lasted longer than the Yankees and the Red Sox, or the Harvard-Yale football game.

But the parties’ persistence is a bit of a fraud. The Republican party of Donald Trump is not the same outfit that brought us Abraham Lincoln or Calvin Coolidge. The Democratic voters who elected Barack Obama differ in many ways from the coalition of interests that supported Franklin Roosevelt. Only the labels truly endure; the parties themselves glide across the ideological landscape in an elaborate dance, adapting their principles and policies to the shifting whims of the voters.

Again, it’s hardly shocking news that parties change their stripes to match the political mood. But I had not realized how thoroughly they’ve remade themselves. Over the past 50 years the Democrats and the Republicans have just about swapped places. Two pivot points mark this curious pas de deux: the presidential elections of 1964 and 1992.

The maps below summarize the metamorphosis. The first map shows each state’s party choices averaged over the 16 presidential elections from 1900 to 1960. The second map portrays a transitional state, giving the averaged results for the seven elections between 1964 and 1988. And the third map covers the seven most recent elections, from 1992 through 2016. Hover over a state to see the tally of Democratic and Republican victories. (Data and base map from Wikipedia. Third-party votes are ignored.)

The changes are pretty dramatic. In the early years of the 20th century, a great swath of Southern states, from the Carolinas west to Texas, were a monolithic Democratic bloc, painted bright blue. Most of the rest of the country was reddish purple, voting Republican more often than Democrat but capable of going either way. Then, starting in 1964, the South flipped from Democratic to Republican, and a dozen other states in the western half of the country also took on a redder tinge. The years since 1992 have brought a rebalancing. The Northeast and parts of the Midwest have turned blue, and so have the states along the Pacific coast.

The net effect is to turn the geography of American politics upside down. In particular, the South and the Northeast have traded party loyalties. The state of Georgia was won by the Democrats in every presidential election from 1868 until 1960, while Vermont was Republican from the party’s founding in 1856 until 1960. Since 1992, Georgia has been in the Republican column six times out of seven, and Vermont has been steadfastly Democratic.

In looking at the maps, it’s also notable that the colors have lately become more intense: red states are redder, blue states bluer, and purple has all but disappeared. One reason may be that the 1900–1960 map includes twice as many elections, thereby offering twice as many opportunities to muddy the record. But the rising polarization of the electorate is also surely a factor.

Here’s another way to look at the same data, sacrificing geography to make the temporal patterns clearer. In the matrix below each row represents a presidential election year and each column corresponds to one of the 50 states or the District of Columbia. The box at the intersection of a row and a column is colored to indicate which party won that state in that election. I have grouped the states into rough geographic regions, but in forming the groups I’ve also kept an eye on political affinities; for example, I put Hawaii with the West Coast states but Alaska with the Mountain states. The horizontal divisions correspond to the time spans of the three maps above.

Rectangular table of colored squares, with one row of squares for each presidential election year from 1880 to 2016 and one column for each state.

I have been staring at this tableau for about a week, trying to make up my mind what the patterns mean. In the South the transition from blue to red stands out clearly enough. Most of the Border states follow the same trend, although the contrast between before and after is less extreme. The region labeled Mountains/Plains has become even redder than the Deep South—it’s very nearly monochromatic—but that’s less remarkable because those states have leaned Republican throughout their history.

The recent blueing of the West Coast and the Northeast are also visible here. The change is abrupt: In the 1970s and 80s both areas were strongly Republican; in the seven election cycles since 1992, with 98 presidential elections across 14 states, all but two contests have swung Democratic.

But it’s not just these conspicuous, localized blocks of solid color that catch the eye. I believe there are other patterns, spanning large areas of the matrix, that may be more important. Based on eyeball analysis alone, I feel I can partition the matrix into rectangles that differ in texture. In particular, I see horizontal stripes in the lower part of the diagram (before 1992) and vertical stripes in the more recent years. The presence of horizontal stripes would indicate that many states tend to agree on the same candidate, regardless of party, whereas vertical stripes suggest that states tend to vote for the same party year after year, regardless of candidate.

Do these stripes really exist in the data, or are they a perceptual figment? That’s debatable. In an attempt to settle the question in my own mind, I have tried tidying up the matrix with a procedure I call combing: I brush across the pattern either vertically or horizontally, depending on which way each region “wants” to go. The algorithm is applied separately to each block of the matrix, where the blocks are defined by the seven regions and the three time periods. Here’s the recipe. First, make two copies of each block. On one copy, examine each column of cells and recolor the entire column to match whichever color is in the majority; if the red and blue counts are equal, choose the color that has an overall majority in the block. Now turn to the other copy of the block and apply the same procedure to the rows rather than the columns, letting the entire row take on the majority color and breaking ties where necessary. Finally, having “combed” the matrix both vertically by columns and horizontally by rows, choose the recoloring that alters fewer cells and thus causes less disturbance.

Behind this procedure is the supposition that regularities in the voting pattern are partially obscured by random noise, or statistical fluctuations. The combing procedure, by applying a winner-take-all rule to each row or column, suppresses the noise and allows the pattern to stand out more clearly. Or maybe it creates a pattern where none existed. I have qualms about the reliability of the method, in part because the result depends on how the matrix is carved up into blocks. I hope to come up with an algorithm that avoids all subjective judgments. In the meantime, for what it’s worth, the combed matrix does support my intuition that horizontal stripes have recently been replaced by vertical ones.

A 'combed' election table, emphasizing areas with horizontal stripes where many states tend to agree on the same candidate, regardless of party, and vertical stripes, where states tend to vote for the same party, regardless of candidate.

If this analysis can be taken seriously, it appears that American presidential politics underwent a significant transformation in 1992. Until then, the horizontal stripes across the matrix suggest a tendency for consensus across broad sections of the country (excluding the South). Blue horizontal stripes are associated with Democrats Woodrow Wilson (1912–16), Franklin Roosevelt (1932–44), and Lyndon Johnson (1964); red stripes mark Republican dominance in the earliest years of the 20th century, in the 1920s, in the Eisenhower years of the 1950s, and during the Nixon and Reagan-Bush regimes of the 1970s and 80s.

Since 1992, the dynamic is different. No presidential candidates in the past seven elections have put together a strong cross-regional coalition. Instead, we have states that always vote for the same party, year after year. If this combed view of modern politics were literally true, there would be no need to hold elections. We could just add up the electoral college votes of the red and blue states and know the outcome in advance. (The specific allocation shown here favors the Democrats, 285 to 253.) Only the small variations that I have dismissed as random noise leave the contest in doubt. This is a new thing, quite unlike the political order of the past century, but people have quickly gotten used to the idea that the nation’s fate will turn on the decisions of voters in Ohio and Florida.

Maps of averaged votes and tables of red and blue boxes—it’s all so abstract! There are no issues or candidates in these elections, no personalities or strategies. An intensely social process that engages the attention and often the passions of millions of people is reduced to mere numerical data—which are then simplified further by averaging or “combing.” This reductionist approach to politics may seem brutally mechanistic, but I would defend it precisely because it wrings all the humanity and emotion out of events. In the resulting dessicated view, with all the brush and foliage cleared away, we can see some long-term and long-range forces at work under the surface. It’s like the statistical approach to understanding automobile accidents or wars or epidemics: The numbers won’t tell you who’s to blame in each incident, but they may well offer hints about how to avert future disasters.

On the other hand, at some point we must look up from the orderly world of maps and matrices and re-enter the land of individual choices, stories, controversies, human lives. Let us look into the election of 1964, when Democrats in the Deep South first bolted to the other party. Like so much of American history, the story is about race.

Even before the Civil War, the South was largely Democratic; afterwards, the Republican party was anathema to white Southerners. It was the party of abolition, of emancipation, of Reconstruction. The Democrats, shut out of the presidency from 1860 through 1880 and greatly outnumbered in Congress, actively courted white Southern voters as a means of regaining power. Generations of yellow dog Democrats in the South voted the party line even when they disliked the candidate, just to spite the Republicans.

Symmetry suggests that black Southerners should have supported the Republican party. Initially, they did. During Reconstruction, while the Union army occupied the conquered Confederate states, blacks not only voted Republican but were also elected to state and national office as Republicans. But then the carpetbaggers went home and left the black population to its fate. For decades, blacks in much of the South had no opportunity to vote either Democrat or Republican. Nationwide, neither party showed much enthusiasm for becoming the voice of black America.

Then came the 1960s. John F. Kennedy proposed a civil rights bill, but Southern Democrats blocked it in a House committee. After Kennedy’s assassination, Lyndon Johnson made passage of the act his first priority. After six months of maneuvering and stalling, including a 60-day filibuster, Congress passed a weakened version of the bill, and Johnson signed it into law. A few weeks later, at the Democratic national convention, Johnson had to deal with competing delegations from Mississippi—one all white, the other mostly black—both demanding to be seated as the state’s representatives. He imposed a compromise that infuriated both factions. Many white Southerners felt betrayed, and they expressed their resentment at the polls in November.

Map of 1964 electoral college votes. Red for Republican in SC, GA, AL, MS, LA, AZ and blue for Democratic everywhere else.I didn’t vote in 1964 (I was 14 at the time) but I followed the conventions and the campaigns. Johnson, an unelected incumbent, was not a popular candidate, but he got a lot of help from his opponent, Barry Goldwater, the right-wing senator from Arizona. Goldwater was given to off-the-cuff remarks about abolishing Social Security, lobbing nukes into the Kremlin, and sawing off the Eastern seaboard of the U.S. and letting it float out to sea. His campaign slogan declared, “In your heart you know he’s right,” but people were swayed by the Democratic riposte, “In your guts you know he’s nuts.” Johnson won 61 percent of the popular vote, close to the all-time record. In the electoral college Goldwater carried his own state by a sliver, and the only other votes he received were from five states of the Deep South. White Southerners had been offered a choice—sharing the lunch counter with their black neighbors vs. nuclear annihilation—and they had chosen the latter.

(I can’t resist noting that our most recent presidential election looks to me a lot like 1964, but reflected in a funhouse mirror. The Republicans nominated somebody even goofier and scarier than Goldwater. If only the outcome had been the same.)

The lopsided vote in 1964 was interpreted as a repudiation of Goldwater’s “extremism.” The morning after, James Reston wrote in the New York Times:

Barry Goldwater not only lost the Presidential election yesterday but the conservative cause as well. He has wrecked his party for a long time to come and is not even likely to control the wreckage.

That’s not how it turned out. The Republican party was not wrecked for a long time to come; they won back the White House four years later, and they’ve won eight of the thirteen presidential elections since 1964. The radical right is rampant within the party; it’s the liberal wing that’s struggling to survive.

The South’s rebellion was treated as a minor footnote to the 1964 election—a protest rather than a serious attempt to elect a candidate. In retrospect, the South’s conversion to Republicanism looks much more important than it seemed at the time, because it was not a mere fluke; it became the new normal, though not immediately. In 1968 five Southern states supported George Wallace’s third-party bid, and in 1976 the whole of the South got behind Jimmy Carter, the liberal Democratic Georgia peanut farmer and nuclear engineer. But since then the region has been reliably red in presidential elections, and Republicans have also taken over governorships and state legislatures.

The shift in Southern voting patterns after 1964 left the national political system out of equilibrium. Of the six presidential elections from 1968 through 1988, the Republicans won five. In the absence of any counterbalancing movement, the Democrats would have had a hard time ever again winning the presidency. But it seems the electoral machinery shifted gears again in 1992.

Looking back on the 1992 election, I see nothing about it that signals a major turning point. George H. W. Bush was a mainstream Republican incumbent who had won handily in 1988 (after serving eight years as Ronald Reagan’s VP). He was popular enough to deter some prominent Democrats from risking a run against him. Their absence doubtless helped Bill Clinton, then little known outside of Arkansas, to get the nomination. I’m sure there are many factors that might explain Clinton’s success that year—a lagging economy, the third-party candidacy of Ross Perot, Clinton’s energetic campaigning, Bush’s inept oratory (“born with a silver foot in his mouth”). It’s a little harder to fathom why Clinton’s victory was such a clean sweep in certain regions. He won every state in the Northeast and on the West Coast, regions that had leaned Republican for more than a century. He won in nine states that had voted Republican in all six elections since Johnson-Goldwater. He won 15 states that Jimmy Carter had lost in 1976.

A still broader question continues to puzzle me. Why have all the Northeastern and Western states continued to vote for Democrats in every subsequent election, even as the candidates and the issues change? In New England, in the Midatlantic region, and on the West Coast, 15 states plus the District of Columbia have voted in the electoral college 16 × 7 = 112 times since 1992; only two of those votes went to Republicans (New Hamshire in 2000 and Pennsylvania in 2016).

In trying to make sense of these observations, I can come up with several categories of possible causes:

  1. Noise. There's actually no pattern to be understood; it's just a statistical fluctuation. Keep watching for a few more election cycles and the maps will look different.
  2. Demographics. The Vermonters voting Democratic today are not the same Vermonters who voted Republican in the 1970s and 80s. People move, people die. Millions of voters in 2016 were not yet born in 1992.
  3. Reconsideration. As the years pass, voters develop new beliefs, convictions, sensibilities. A change of heart or a change of mind leads to a change of party preference.
  4. Realignment. The inverse of Item 2. The people stay the same but the party moves on to a new platform.
  5. Strategy. In a smoky back room political operatives decide to shift campaign funds and other resources from one region to another.
  6. Shenanigans. As our president tells us, elections are rigged and millions of illegal immigrants are voting in precincts all across the country.

In the South in the 1960s, option 3 seems to have the most explanatory power. Southern white voters perceived that their interests were in better alignment with the Republican platform than the Democratic one on the issue they cared about most. Strategy has also played a role in consolidating the Republican hold on the South, with backing from donors such as the Koch brothers. And recently shenanigans have also become a worrisome factor, as states enact measures of voter suppression and intimidation, and refine the art and science of partisan gerrymandering.

What about the red-to-blue flip in the Northeast and West? Realignment is also plausible here; the Republican party has certainly shifted to the right, perhaps enough to alienate more centrist voters in the Northeast. But I don’t see any single compelling issue that would unite and motivate voters from across the region in the way that race relations did in the South. Also, the timing is hard to understand. If New Englanders were looking for a voice of moderation, they would not have been so enthusiastic about Ronald Reagan. Also puzzling is the swiftness and depth of the shift in allegiance. In 1988 New England was still so Republican that four of the six states voted against a local Democratic favorite, Governor Michael Dukakis of Massachusetts. By 2012 the region was so Democratic that all six states opposed another Massachusetts governor, the Republican Mitt Romney. I don’t have a secure grasp of what’s going on here.

My in-house political consultant points out that important detail is lost when you look only at lists of states won or lost. At the very least, it would be helpful to know the margins of victory. Furthermore, within each state there are urban/suburban/rural conflicts, gradients of wealth, subpopulations distingusihed by race, age, ethnicity, linguistic community, employment status, and so on. The voting is done by millions of individual citizens, who are not colored red or blue but have multiple interests that may pull them toward different candidates. All of these objections are well founded. A handful of voters can make the difference between 49.999 percent and 50.001. And Vermont’s conversion experience may be quite different from California’s. Any serious political analyst would want to dive into the finer-scale data. Nevertheless, when more than a dozen states change sides simultaneously and remain in lockstep for seven election cycles, it’s reasonable to look for a single widespread cause.

The American system for choosing presidents is designed to amplify small advantages in the popular vote in order to create a clear majority in the electoral college. Within each state (except Nebraska and Maine) the contest is winner-take-all; the ticket with the largest number of popular votes—even if it falls short of a majority—gets all the electors. This is a recipe for lopsided outcomes, for landslides. Between 1900 and 1988 there were 10 elections where one party won at least 80 percent of the states. But there have been no such runaway contests since then. Somehow the system has found a configuration where either party has a chance to win every election, but not by much. And thus whoever wins must govern a nation where nearly half the citizens are hostile. More than half in the case of Mr. Trump.

A final note. If the South had stuck with the Democratic party in 1964 and after, American history would look somewhat different. The whole Nixon presidency might have been avoided. But I can’t quite imagine how the alliance with the Dixiecrats would have been held together. If the party had held onto those votes by continuing to deny the rights of black people, who would want to be a Democrat?