Woodrow Wilson suffered a catastrophic stroke inside the White House and never returned to the presidency as he had practiced it before. The office remained occupied. Access to the president, information about his condition, and much of the work placed before him passed through a small circle around the sickroom.
For seventeen months the constitutional form held while the practical presidency narrowed. Cabinet officers struggled to see him. Senators sent visitors to judge his capacity for themselves. Edith Wilson described her role as a stewardship; whatever name one gives it, she and a few unelected attendants decided what reached the president and what the public could know.
That is the first regency in this story. The second is not medically equivalent, and this essay does not pretend otherwise. The recurrence lies in the institution: a protective court managing exposure until the governing party had almost no time left to arrange a credible succession.
The obvious analogy for 2024 is 1968: a damaged Democratic presidency, a vice president inheriting the nomination, and a Republican comeback. That comparison explains the electoral handoff. It does not explain the defining procedural failure.
Lyndon Johnson withdrew in March, publicly and under his own authority. The country saw the decision as it happened. Hubert Humphrey received a difficult inheritance, but not a succession improvised after months of official reassurance had collapsed in a single evening.
The closer precedent is 1920 because the political problem began before the heir appeared. It began when loyalty to the principal became a reason to postpone disclosure, adaptation, and an orderly transfer of responsibility.
Edith Wilson controlled papers and visitors. Physician Cary Grayson disclosed as little as he believed the office could bear. Secretary Joseph Tumulty kept the machinery moving around a president who could no longer sustain its center. Their motives were protective; the consequences were governmental.
A court optimizes for the person at its center. It reduces shocks, filters demands, and treats access as a resource to be rationed. An election requires almost the opposite: exposure to contradiction, rapid adaptation, a willingness to hear that the current arrangement is failing, and a clear answer about who comes next.
Those tasks can coexist for a time. Near the end of a presidency, they begin to compete. The court can protect the principal so successfully that it leaves the coalition unprotected.
Wilson wanted the election to become a “great and solemn referendum” on the League of Nations. But the hidden injury did not merely create a public-relations problem. It removed the politician who would have had to bargain, persuade, discipline the message, repair estranged constituencies, and decide when a cause required compromise.
The treaty failed in the Senate in March 1920. By then the country was judging more than the League: war, repression, inflation, labor conflict, and a presidency that had become difficult to see. The regency could preserve Wilson’s claim to leadership. It could not perform the political work that leadership required.
James Cox had not built the regency and had never occupied the sickroom. The Ohio governor emerged after forty-four convention ballots, with a thirty-eight-year-old Franklin Roosevelt as his running mate, and inherited the full verdict on a presidency he had not run.
Warren Harding won 60.3 percent to Cox’s 34.1, a 26.2-point margin and 404 electoral votes to 127. Cox carried the South and almost nothing else. Even the national popular margin understated the electoral collapse: Rhode Island, the tipping-point state, was thirty-one points Republican.
The result was larger than any single cause. Postwar prices, the influenza years, civil-liberties abuses, labor conflict, the League fight, and defections among Irish- and German-American voters all mattered. The regency’s contribution was subtraction: months in which the party could neither use Wilson as an active coalition manager nor openly organize a future beyond him.
For a century the pattern remained exceptional. Presidents aged, became ill, lost elections, and left office. The Twenty-Fifth Amendment supplied a constitutional mechanism Wilson’s era lacked. Modern campaigns developed polling, opposition research, professional communications, and an endless appetite for exposure.
Yet institutions are not self-executing. A mechanism works only when the people closest to power are willing to use it—or at least willing to tell the political system what problem it must solve.
The meaningful similarity between Wilson and Biden is not personality, ideology, or diagnosis. It is the temptation created by an incumbent president whose allies believe the governing project is too important to expose to a destabilizing succession fight.
That belief can be sincere. It can also become self-sealing. Evidence of weakness is reclassified as unfairness; requests for unscripted proof become hostile demands; the absence of exposure is offered as evidence that nothing is wrong. The court protects not only the principal but its own prior assurances.
The problem is democratic before it is partisan. Voters cannot judge a succession question that the governing party insists does not exist. When the question finally becomes unavoidable, time itself has become an opponent.
The public record before June 2024 showed an unusually managed presidency: limited unscripted interviews, carefully bounded appearances, repeated concern about age, and repeated official assurances that those concerns were misleading or politically motivated. None of that, by itself, establishes incapacity. It does establish a system designed to minimize uncontrolled exposure.
Inside such a system, personal loyalty and institutional risk become difficult to separate. Family, longtime aides, senior officials, and campaign professionals all have reasons to avoid a rupture. Each can believe someone else will deliver the hard judgment. A court does not require conspiracy; aligned incentives are enough.
On June 27, ninety unscripted minutes supplied the public test that the managed presidency had postponed. The debate did not create the succession crisis. It made the crisis impossible for the party to keep treating as private.
Biden remained the nominee for twenty-four days after the debate. He withdrew on July 21 and endorsed Kamala Harris that afternoon. The decision resolved the immediate question of who would run and revealed the cost of resolving it so late.
Harris had 107 days to become the nominee, define a program, build distance from the administration where distance was necessary, introduce herself to low-information voters, answer for an unpopular record, and prosecute a case against a former president who had been campaigning for years.
A succession made in open daylight two years earlier would not necessarily have produced a victory. It would have allowed the party to test candidates, arguments, and coalition weaknesses before they hardened into the general election. The regency’s measurable cost was not a guaranteed outcome. It was lost optionality.
Harris inherited the administration’s achievements, its liabilities, and the compressed calendar at once. A sitting vice president cannot credibly run as a complete outsider; nor could she easily explain why a new direction was needed without reopening the question of why that direction had been delayed.
Her answer that nothing came to mind when asked what she would have done differently was damaging because it exposed the structural trap. Separation risked indicting the administration she served. Continuity left her carrying a record the electorate was already judging harshly.
Trump won all seven battleground states and the national popular vote by 1.5 points. This was not 1920: the national race was close, and millions of voters distinguished among candidates below the presidential line. But the heir again received the bill for a succession the court had delayed.
Of the 3,143 counties with comparable results, 2,804 moved toward the Republican between 2020 and 2024. The typical county shifted about 3.2 points right. The movement crossed cities, suburbs, and rural regions.
That breadth matters because it limits easy explanations. Candidate errors in one battleground cannot produce a nearly national directional shift. The result reads as a judgment on the incumbent order, even though the size and reasons for that judgment varied from place to place.
Down-ballot results show that the electorate did not reject every Democrat equally. In 2024, Democratic Senate candidates won Arizona, Nevada, Wisconsin, and Michigan while Harris lost those states. In Arizona, Ruben Gallego drew more votes than Harris while Kari Lake drew fewer than Trump.
The same signature appears, less dramatically useful, in 1920. Democratic Senate and House candidates often ran ahead of Cox even as the Harding wave defeated them. In a landslide, outperforming the top of the ticket merely narrows the loss. In a close election, it can save a seat.
The distinction is central to the regency argument. The penalty was not a generalized banishment of the party. It was concentrated at the point where the concealed succession question touched the ballot most directly: the presidential ticket.
Several old immigrant and machine strongholds moved sharply toward the Republican in both periods: the Bronx, Hudson County, Suffolk County, Cook County, and Wayne County. The constituencies, party systems, and grievances were not identical. The recurrence is geographic and institutional, not genealogical.
In 1920, war, repression, Versailles, and ethnic grievances broke parts of the urban Democratic coalition. In 2024, prices, immigration, foreign policy, public order, and dissatisfaction with the administration formed a different bundle. In both cases, communities the party treated as inherited proved available to a challenger who made a more visible claim on them.
Both elections followed a pandemic and a disruptive rise in the cost of living. Both governing parties faced protest on their left. Both Republican nominees had been underestimated by opponents who considered them too compromised, too slight, or too exhausted to become the vehicle for a national verdict.
Those forces are ordinary engines of anti-incumbent elections. They can explain defeat without any regency at all. The comparison becomes useful only after that admission: concealment did not create the weather. It reduced the governing party’s ability to navigate it.
A fair causal reading is therefore conditional. The court did not single-handedly produce either result. It delayed information and succession in environments where time, adaptation, and coalition repair were already unusually valuable.
The two elections are separated by scale. The movement from 1916 to 1920 was 29.3 points toward the Republican. From 2020 to 2024 it was 5.9. One was a historic annihilation; the other was a close national result with a broad directional shift.
That difference forbids a claim of repetition. It does not erase the procedural rhyme. In both cases a protective circle limited disclosure, adaptation came late, the succession was poorly prepared, and a nominee other than the protected president carried the administration into the election.
History is most useful here not as prophecy but as a controlled comparison: hold the institutional mechanism in view, then measure how much everything around it differed.
Harding’s urban gains did not become a permanent Republican ownership of the immigrant city. Al Smith recovered much of the coalition in 1928; Franklin Roosevelt transformed it after 1932. Votes obtained during an opponent’s failure can be retained, but only if the winning party builds institutions capable of serving them.
The same uncertainty follows 2024. Trump won the popular vote but remained below fifty percent, at 49.8. A broad shift can reveal a new coalition without completing one. The next elections will show whether these voters transferred allegiance or merely issued a demand for better delivery.
A regency begins as an act of protection. Its political danger appears when protection becomes a substitute for succession: access narrows, contrary evidence is treated as disloyalty, and the future is postponed until the calendar can no longer absorb the truth.
Cox and Harris were not responsible for constructing the courts they inherited. Each became the point at which the public could finally charge the governing party for what had been delayed. The protected principal did not stand for election. The heir did.
The lesson is neither that illness determines elections nor that every aging presidency becomes a regency. It is simpler and more durable: a party that cannot speak candidly about succession eventually loses control of the timing, the terms, and often the verdict.