Do you think 1,000 is sufficient as a number when polling?

Question

Considering we have over 40million voters now should the minimum poll size be 4,000 (0.01% of the electorate)

In the UK I have today learned that you can sign up to be consulted, this ensures you get contacted lots of times by the pollsters, surely that scews the polls to the activists opinion rahter than the wider electorate?

OKay I get the statistical theory in simple terms, but as opinions fragment aren't views that are not staple likely to be overlooked?

Surely if 1,000 works on 10 Million there is some logic that it would require a bigger sample to produce the same accuracy from 40Million?

I have heard that the +/- 3% relates to a poll of polls where the range of answers was limited to three or four well specified options.
Not an individual poll where the margin for error was more typically +/- 7% depending on the scope of the questions and the sample methods used anyone else heard that?

Answer 1

For a poll result to be reflective of the public as voters it would need to me adjusted to account for the quantity of participating voters (35%?). Polls are misused continually. The results of a poll of a thousand people is that of thousand, x respond '+', y respond '-', and the proverbial margin of error. What this means is, if the results are made available, we can learn what the inherent biases of the poll author were, and how up to a thousand people responded, between answering the cell phone and answering the call of nature.

Answer 2

Poling is a science, so this is a math/statistics question. The sample size is important and so is the sample distribution. But with a proper distribution, a poll of 1000 will give a pretty true indicator.

When polls go wrong (which is not all that often) it can be a sampling error, or it could be poorly worded questions. In 1948, the prediction of Dewey defeating Truman was based largely on telephone polls. Well it happened that in 1948 a large portion of the voting public still did not have telephones and they were not part of the sample. The same thing may be happening again. People who have no land line (only a cell phone) will not be included in a telephone pool. If that number gets big enough and there is a big enough difference in the opinions of people with only cell phones versus land lines, then results can get skewed.

The pollsters know this. They are working out the impacts.

Answer 3

The sample size will determine the margin of error. This is usually provided by any half decent poll. 1000 will generally produce a fairly low margin of error -3-4%. It is the size of the sample, not the % of the population that determines this. For instance in a population of 2, a poll of 1 will give a 25% margin of error even though it is 50% of the population.

Answer 4

Its been a while since I took the class, but you can get the statistical likelihood of your poll of being correct. This is why every poll shows a margin of error (+/-3%).

At 1,000 polled, you can be 95% sure that your results are within 3% of the actual figures. From here, increasing accuracy is an exponential deal, and it takes 10,000 new polled people to increase the certainty by 1 additional percent. To reach 100% absolute certainty, you would have to ask everyone.

Answer 5

1,000 is statistically accurate enough for the purposes of most polls: it's a balance between accuracy and cost given the number of polls conducted regularly in countries such as the UK/US. As said above, the number of persons polled expands exponentially for any increase in statistical accruacy, so 1000 is a good compromise.

However, I would add that better cross sections and more fairly worded questions needs to be worked towards by all pollsters.

Answer 6

1000 can be okay the main thing is to have a proper cross section of the particular population you are trying to poll. If you want an opinion from a cross section of the whole population then taking the first 1000 people coming out of Watford station one evening is no good. You would need to take opinions from 10 people picked at random from 100 towns spread all over the UK.

Answer 7

Mathematically,statistically, 1,000 is enough providing that it is a genuinely random sample.

Unfortunately, the sample would typically not be random or representative of the population because it would mainly comprise of sad people such as us with nothing better to do.

Therefore the sample size as you say needs to be significantly larger

Answer 8

You are correct, we have to many voters and a sample of 1,000 voters is not an accurate depiction of the true American voting concensus. Besides most os the polls are run by the media and they favor thier ideals and beliefs.

Answer 9

I think as long as that 1,000 is a fair cross-section of society, then it seems a reasonable amount. If you make the pool too large, it makes the data very difficult to work and collect.

I think 1,000 is fair enough.

Answer 10

1000 is plenty if the people that are included come from a good sample from across the country.