Election letter: There is more to the margin of error than voters are being told

The public are being misled by polls and their interpretation, particularly in respect of political parties which are not contesting every electoral area.

These parties consequently attract negligible support in the uncontested areas, which is reflected in their low average, nationwide, support. Polling data should not be published for such parties, for it is knowingly doing them and the public a disservice.

Additionally, pollsters and commentators are wont to talk of a ‘margin of error’, often stated to be plus or minus 3 per cent when the nationwide sample of voters is about 1,000.

They rarely explain it and when they do, they often get it wrong. It is invariably incorrectly interpreted for parties with small levels of nationwide support. Intuition tells us that the voting intentions expressed by a sample of 1,000 electors will most likely be almost the same as the expressed voting intentions of the whole population.

The margin of error is a statistical concept. It is the difference in the expressed voting intentions of the sample and that of the whole population of voters which may be exceeded in one poll in 20 on average, if the sampling were repeated continuously over the whole electorate at one time.

However, contrary to pollsters and commentators usage of the ‘margin of error’, it is not constant. The margin of error where the national support is 2 per cent (typified by the Greens and Renua), is plus or minus 0.9 per cent; for 15 per cent (eg Labour on a really good day) it is 2.2 per cent; for 30 per cent (eg Fine Gael) it is 2.7 per cent; and for 40 percent, it is plus or minus 3.1 per cent.

Commentary that is put on the data without regard for these facts is little better than ‘pub talk’ and harmful talk at that.

John Colgan

The Toll House,

Dublin Road Street,

Leixlip,

Co Kildare