[drxzcl]

If you plug in the numbers into a sample size calculator you will find that a sample size of 2000 enables you to get an estimate within 5% of the true value 99.999% of the time for an infinite population.

So, not exactly small, no.

https://www.calculator.net/sample-size-calculator.html

[midasuni]

As long as you have a decent sample set, which the reputable polling companies do.

[drxzcl]

And as long as the thing you are polling is about 50:50. It’s hard to poll more skewed things, and you need larger sample sizes.

[zh3]

As long as the sample is representative, of course.

2000 does seem small enough that it would be hard to ensure no bias creeps in.

[drxzcl]

If only there were a whole field of study dedicated to this, and you could hire experts to run studies for you…

[darajava]

It doesn’t say where the sample was taken. London? Yorkshire? It almost certainly wasn’t taken at random.