Opinion poll findings - October 18, 2007

In view of the fact that over the last five years I have carried out a number of studies for Allied Newspapers, mostly published in The Sunday Times, I have been approached by a number of persons asking for further information on the opinion poll on Maltese politics published in The Times last Saturday (October 13).

The study has not been conducted by me.

In reality, I have serious doubts about the validity and reliability of the said study because of the following:

The response rate in the report is said to have been a mere 21.3 per cent. This is extremely low and renders the validity of the study seriously in doubt.

The study generalises for each of the electoral districts. The number of respondents in the individual districts varies from 36 in district 1 to 58 in district 11. At the 95 per cent confidence level, and assuming a 50/50 split (as in the case of whom to vote for, given that there are two main parties), a sample size of 100 carries a sampling error of ±10 per cent. The sampling error for a sample of 50 is naturally much higher and, as such, the data is normally considered to be useless.

There are a number of other technical issues in this study, the results of which are all weighted, but the above are enough to put in doubt its usefulness.

I believe that, irrespective of who undertakes such studies, the reliability and validity of these studies should not be doubtful.

Editorial note: A response rate of 21.3 per cent is related to a total sample frame of 3,000 randomly selected households - equating to 638 complete interviews as reported in the analysis. This brings the sample error to an estimated ± 4.86 per cent at 95 per cent confidence interval. In his comments, Mr Vassallo is possibly inferring that the 21.3 per cent response rate was in relation to the 638 interviews, which is incorrect. What needs to be pointed out is that there must not be a specific pattern of non-responses and that the respondents are a reflection of the general population. It is true that telephone surveys produce higher response rates but the sensitivity of the information given the Maltese context significantly brings down such a response rate.

Surely Mr Vassallo has a better picture of the study as he seems to be in possession of details that did not appear in the report carried on The Times last Saturday.