At nine in the morning on Thursday, Chris Kennedy was the Green Party's candidate for the Makerfield by-election. By six in the evening, he was not. According to the Independent, the party announced his selection in the morning and confirmed his withdrawal roughly nine hours later, citing "personal and family reasons," after a social media post about an attack on a Jewish ambulance resurfaced. Kennedy has since apologised.
It is a small story in the grand scheme. A by-election candidate in a single Greater Manchester seat, vetted poorly, gone before lunch the next day. But it is the kind of story that should interest anyone who thinks about how political contracts are priced, because it captures something the polling industry handles badly and prediction markets handle, at best, unevenly: the candidate-level shock that arrives without warning and reprices a contest in hours.
The vetting failure as a market event
Political parties are, among other things, risk-management operations. Candidate vetting is the unglamorous compliance layer that stops a local activist's old Facebook posts from becoming a national headline. When the layer fails, as it clearly did here, the cost is not just embarrassment. It is a forced reset of the contest itself, with a new candidate selected under time pressure, weaker local roots, and a press cycle that has already framed the seat as a problem for the Greens.
For a major-party contest with deep liquidity, an event like this would show up as a clean intraday move. Polymarket's US election contracts routinely repriced double-digit points on candidate-level news through 2024, and the order flow was visible in near real time. UK by-election markets are thinner, quieter, and often only exist on niche venues, but the mechanics of how prediction market odds work are the same. New information lands, traders adjust, the new midpoint reflects the updated probability.
The interesting question is whether nine hours is fast enough for a market to price it before the headline cools. Often it is not. The story breaks, the candidate withdraws, the replacement is announced, and the market quietly settles at a new level without any single dramatic candle. That is what makes candidate-vetting failures genuinely hard to trade, even when they are easy to spot in hindsight.
Why polls miss this and markets sometimes do too
Polls are calendar instruments. They run on fieldwork windows of several days, weight against demographic frames set weeks in advance, and report with a lag that makes them structurally blind to one-day candidate stories. A poll fielded Wednesday and reported Friday cannot tell you that the Green candidate was replaced on Thursday afternoon. The number it produces is already historical by the time it lands.
Markets, in theory, fix this. They update continuously, they price the candidate as much as the party, and they let a trader with better information move the line before the next poll catches up. The full case for that advantage is laid out in our piece on prediction markets versus polls, and the Makerfield episode is a textbook example of the kind of event where the theory holds.
In practice, though, the advantage only materialises if the market exists and has enough liquidity to absorb a real position. A single UK by-election rarely clears that bar. What does clear it, increasingly, is the meta-contest: which parties are gaining ground, which are bleeding candidates, what the running tally of selection disasters implies about a party's national readiness. That is the level at which the Makerfield story actually moves money.
The bigger contest the small one feeds into
Look at the House majority market on Polymarket, or the equivalent UK-flavoured contracts on the legislative balance, and you can see how individual selection stories aggregate. Each candidate-vetting failure is a small piece of information about a party's internal operation. One is noise. A pattern is a signal. Traders pricing the broader political map are doing exactly this kind of aggregation, weighting each small story by what it implies about the parties' wider competence.
The Greens have spent the last two years presenting themselves as a serious electoral force, not a protest vehicle, and the Makerfield episode cuts against that framing. Whether it moves any contract meaningfully depends on whether traders treat it as a one-off or as part of a trend. Reasonable people will argue both sides. The market, eventually, will settle on one.
For anyone tracking UK political risk, the practical question is which venues actually list contracts granular enough to catch these moments. The honest answer is: very few, and the ones that exist are mostly offshore. The UK regulatory position on these contracts is restrictive enough that retail access is patchy, which is covered in more detail in our explainer on whether prediction markets are legal in the UK.
The editorial take
A nine-hour candidacy is a comic data point on its own. As a signal, it is more interesting than it looks. It tells you something about the Green Party's vetting bandwidth, something about how fast modern news cycles can shred a selection, and something about the kinds of events that prediction markets are theoretically built to price but in UK contests rarely have the depth to actually catch.
The story to watch is not Makerfield specifically. It is whether episodes like this start to compound, and whether the broader political contracts begin reflecting a discount on parties that look operationally fragile. iPredicta tracks UK political markets across the venues that list them, and candidate-level shocks like this one are exactly the kind of event we flag when the order flow starts moving before the polls do.
Frequently asked questions
Why don't UK by-elections have liquid prediction markets the way US races do?
Two reasons, mostly. UK regulation treats political event contracts more restrictively than the US, which limits retail access to the venues that list them. And by-elections are local, fast-moving contests with small audiences, so even where contracts exist they tend to attract thin volume and wide spreads.
Would a candidate withdrawal like this actually move a prediction market price?
On a single-seat by-election contract, yes, if one existed with real liquidity. On a national or aggregate contract, the impact is usually small unless the withdrawal is part of a visible pattern. One vetting failure is noise. A sequence of them across multiple seats is what traders pricing party-level contracts actually react to.
