In mid-October 2024, Călin Georgescu was polling in the low single digits for the Romanian presidency, inside the margin of error of almost every published poll. Six weeks later, on November 24, he finished first, with 22.94 percent of the vote and a nearly four-point lead over the second-place finisher, Elena Lasconi. Two weeks after that, on December 6, the Constitutional Court annulled the first round. The rerun, held in May 2025, elected Nicușor Dan to the presidency and closed the file.
What happened between October and November is the clearest case I know of a saturation event driven by Frequency rather than Volume. The two are easy to confuse. Volume is how much material there is. Frequency is how often the same audience sees it. The Brasília case last week was a Volume case: a curve that stayed high for weeks, across many channels, until it converged on a physical location. The Romanian case is a Frequency case: a curve compressed into a short window and concentrated on one platform.
Volume asks how much. Frequency asks how often. The two variables answer different questions, and they leave different fingerprints.
The Court’s December 6 decision states the grounds plainly. It cites three findings from intelligence assessments presented to the Supreme Defence Council on November 28 and declassified by President Iohannis: preferential algorithmic treatment of one candidate’s content on a single platform; undisclosed paid promotion that was not labeled as political advertising; and coordinated inauthentic behavior in the weeks before the vote. The single platform was TikTok. The window was the last weeks of the campaign.
None of those three findings, taken alone, describes a Volume problem. A single-platform surge lasting a few weeks covers a small share of Romania’s total online public, but for a specific slice of that public it is inescapable. That is the Frequency variable at work: the number of times the median sympathetic user encountered adjacent, on-message content per day, held above the threshold at which repetition begins to substitute for evidence. Repetition produces familiarity, and in a compressed window before a vote familiarity can pass for credibility.
Two features of the case are worth pulling out for the rubric.
The first is the choice of surface. TikTok’s recommender system is unusually well suited to Frequency plays. It does not require a follower graph to deliver a video to the same user repeatedly; the For You feed will do so on its own if engagement signals are strong enough, and engagement signals can be manufactured. A campaign that would need weeks to build reach on a follower-graph platform can, on the right recommender surface, build cadence in days. The Commission’s December 17 opening of Digital Services Act proceedings against TikTok cites this dynamic directly, and it is why the case is not replicable, in the same form, on every platform.
The second is the absence of a paid-advertising trail. Paid political content is supposed to be labeled. Among the Court’s grounds is paid promotion on TikTok that was never declared as political advertising. That is the second distinguishing mark of a Frequency play at scale: the cadence is bought, but the receipts are hidden. Visible receipts would have let journalists and rivals see the price of that cadence in time to answer it.
The lesson for the rubric is again narrow. When a candidate’s curve goes from flat to first inside a month, the question to ask is how often the same audience saw it, on which surface, and whether the money is on the public record. If the answer is daily, one platform, and no, you are looking at a Frequency case, and the annulment is downstream. The Romanian court drew that line at the level of the election result itself.
Next week: E is for Emotion, and the 2018 Vosoughi, Roy, and Aral paper, revisited.
— J.W.B.
Note 08 of 24