The latest monthly ICM/Guardian poll , showing that Labour's lead over the Tories has shrunk from eight points to just five, has provoked much comment . How, when most surveys show the party around 11 points ahead, can Labour have sunk so low?
To begin with, it's important to note that ICM's polls have long shown a lower Labour lead than those from other groups such as YouGov and TNS-BRMB. Seven of the last 12 polls  by the company have had Labour ahead by five points or less. December's ICM poll, which showed a Labour lead of eight points, was also less flattering than the concurrent YouGov survey, which showed a Labour lead of 10.
Why the discrepancy? The main reason is the difference in how the two pollsters treat the "don't knows" . While YouGov simply discounts those voters who say they "don't know" which party they'd vote for, ICM reallocates 50 per cent of them based on the party they voted for at the previous general election. This approach was originally developed to respond to the phenomenon of "shy Tories " - those who intended to vote Conservative but were unwilling to say so. In 1992, this famously meant many polls showed Labour in front when in fact it was the Tories who were heading for victory. After studying the data, ICM found that a disproportionate number of those who said "don't know" to voting intention questions in 1992 had voted for the Conservatives in 1987. The reallocation method was born.
At present, since a significant number of 2010 Liberal Democrat voters are unsure how they'll vote at the next election, while most Labour voters are sure, ICM invariably shows a higher level of support for Clegg's party than YouGov and a lower level of support for Labour. The most recent poll  by the latter has Labour on 42 per cent and the Lib Dems on 11 per cent, while the most recent by the former has Labour on 38 per cent and the Lib Dems on a seemingly miraculous 15 per cent. There are good arguments for both YouGov and ICM's approaches but for those sceptical of ICM's method, it's worth pointing out that it was the second most accurate pollster  in 2010.
Finally, it's worth imagining how different the political narrative would be if it was ICM, rather than YouGov, that conducted a daily poll. Since the coalition came to power, ICM has never shown a Labour lead larger than 10 points (a figure the party has only attained once, in September 2012) or a Lib Dem vote share lower than 11 per cent (the party averaged 14 per cent in 2012). Conversely, YouGov has had Labour ahead by as much as 14 points and the Lib Dems as low as seven per cent. But fortunately for Ed Miliband and unfortunately for Nick Clegg, politicos pay more far attention to a daily pollster than they do to a monthly one.