How much can you trust a survey with such a small sample?

The detailed data on YouGov’s London mayoral survey is out today and there are two things to highlight – the survey took place AFTER the controversial Channel 4 Dispatches programme last Monday but the size of the sample was just about the smallest that I can recall from the firm.

In their normal national voting intention surveys YouGov generally involve about 2000 people. The figure for Saturday’s Daily Telegraph poll was 1992. In previous surveys for the London Mayoral race the figure has been about 1000. The December survey was based on 995 responses and a poll in November had a sample of 1036.

Yet only 339 people took part last week’s poll poll and as Anthony Wells on UK PollingReport points out if you exclude the “don’t knows” and “won’t votes” your are left with just 240 responders.

As can be seen above YouGov had to make some fairly major adjustments to deal with under-represented and over-represented groups. So just 25 people in the 45-54 year old age group took part and their responses had to be scaled up to fit the 50 target for that segment.

The smaller the sample, of course, the greater the margin of error.

It is estimated that about five million Londoners will be able to take part in the election on May 1st.

Let’s hope that this will not be the norm for the campaign and that future surveys will have at least 1000 people taking part. It costs more but we will be able to attach more importance to the findings.

London mayoral betting is here.

Mike Smithson