The effective number of parties implied by early-August voting-intention polls is between 5.2 and 5.6. At the 2024 general election, the postwar record, it was 4.70; in 1951 it was 2.14. Polls measure stated intention, not votes cast.
British elections are usually described by who is ahead. A different question — how many parties the vote is genuinely divided among — has a standard answer in political science. The effective number of parties, set out by Markku Laakso and Rein Taagepera in 1979, is one divided by the sum of every party's squared vote share.4 Two parties on 50% each score exactly 2.00. Two on 45% with a third on 10% score 2.44. Because shares are squared, the index weights parties by size: a long tail of very small parties barely moves it, while a genuine fourth or fifth contender moves it a lot.
Computed from House of Commons Library vote shares,1 the postwar series is flat and low for a quarter of a century. It bottoms out at 2.14 in 1951, when Labour and the Conservatives took 96.8% of the UK vote between them, and stays near 2.2 through the mid-1950s. It steps up to 3.12 at the February 1974 election, on the back of a Liberal share of 19.3%, then drifts upward from 1997 to reach 3.91 in 2015. It does not rise monotonically: in 2017 it fell back to 2.87, the sharpest single-election reversion in the series, as Labour and the Conservatives recovered to a combined 82.3%. In 2024 it reached 4.70,2 well clear of anything else in the postwar period.
YouGov's voting intention poll for The Times and Sky News, fielded on 2 and 3 August 2026, put Reform UK on 23%, Labour on 22%, the Conservatives on 19%, the Greens on 13% and the Liberal Democrats on 12%, with Restore Britain on 4%, the SNP on 3% and Plaid Cymru on 1%.3 That distribution gives an index of 5.60. The PolitPro poll trend of 4 August, which aggregates published polls and reports a full party breakdown, gives 5.41.6 The PollCheck seven-poll moving average of 6 August publishes only the leading five parties; holding the remaining 8.9% as a single bloc gives 5.22.7
The mechanism is visible in the columns. Labour and the Conservatives took a combined 57.4% of the UK vote in 2024, already the lowest postwar figure for the two of them; in the current trend they take 43.3% between them. Reform is up by about ten percentage points on its 2024 result and the Greens by about five, while the Liberal Democrats are close to flat. The index rises not because more parties exist, but because the share held by the largest two has moved down into the range where three or four others are competitive.
The effective number of parties depends on how finely the sources resolve small parties. Published polls fold minor parties into a single "Others" line, and historic election tables do the same for earlier decades. Treating that residual as one party understates the index, because it counts scattered votes as if they were concentrated. We therefore computed two figures at every point: a lower bound with the residual held as one bloc, and an upper bound with its contribution removed entirely. Every figure quoted in this piece is the lower bound.
That matters for the headline claim, and it survives the test. The largest upper bound for any postwar election before 2024 is 3.93, at the 2015 election — still well below 2024's lower bound of 4.70. A second check runs the same way. Poll shares cover Great Britain, while the election series is UK-wide and so includes several Northern Ireland parties that push the index up. Recomputing 2024 on a Great Britain basis gives 4.48, not 4.70, which widens the gap between the polls and the record rather than closing it.
Fragmentation of the vote is not fragmentation of the House of Commons. Applying the same formula to the 2024 seat distribution rather than the vote gives 2.24 effective parliamentary parties2 — a figure consistent with an independent published calculation of about 2.2 including Northern Ireland.5 The vote-based index in 2024 was 2.1 times the seat-based one. Under first past the post, a vote splitting more ways does not produce a Commons splitting more ways; it produces a Commons in which smaller pluralities win seats. Reform took 14.3% of the vote in 2024 and 0.8% of the seats; the Green parties took 6.7% and 0.6%.2 Nothing in the current polling changes that mechanism, and this piece makes no claim about how the poll shares would convert into seats.
Voting intention is not a vote. These polls were taken with no election called and roughly three years since the last one, and the index is computed on stated preference, not on ballots cast under a real turnout. We do not know how much of the current spread would survive a campaign. The 2017 reading is the standing warning: the index fell from 3.91 to 2.87 in a single cycle. Published polls also differ on who leads — one of the two aggregators we used has Labour first and the other Reform — and gaps of one or two points between the leading parties are within normal sampling error and should not be read as movement. We could not separately source UKIP's 2017 share, so it remains inside that year's residual; the effect on the plotted value is under 0.02. No causal claim is made here about why the vote has spread.