Afghan opinion poll- Things improving?

1,500 adults is a pretty small sample, statistically speaking - and I'd like to see some more details on where the surveys were conducted and under what circumstances before I make my mind up.
smartascarrots said:
1,500 adults is a pretty small sample, statistically speaking - and I'd like to see some more details on where the surveys were conducted and under what circumstances before I make my mind up.
So small as to be statistically irrelevant: 45 people in each province! Wow! And that is without seeing the questions or the selection of respondants.
1500 is not unusually small for a sample size. A lot of political surveys for the newspapers by organisations like MORI use 1000.

The key point is where you interview, and that is down to the methodology of the surveyors. Naturally, if you interviewed 1500 Tajiks in the North, you would get a different result to 1500 Pathans in Helmand. A well distributed national survey would also be balanced for age and gender.

I haven't looked at the methodology used here, but I would guess it is fairly normal in structure to equivalent surveys done in the UK.
Yeah - the polling methodology isn't really the point here - 1500 is a decent size for a national multiple question survey. It would only be the story if last year the survey had asked 1500 people of one demographic in one area and this year completely changed it's sampling methodology to improve the results. That is because it is a comparative analysis.
I take the point about MORI, but still feel a bit of caution is necessary before breaking out the champagne. I'd like to know the answers to a few questions before I dare to hope:

1. What was the breakdown of people polled by province? People in the more peaceful areas might naturally have a different perspective to the areas where most fighting is going on.

2. What was the ethnic distribution of the sample. Tajiks and Uzbeks are heavily represented in the police and army and could have generally more favourable views on their kinsmen than the majority Pashtun.

3. What was the socio-economic distribution? Educated women would probably give a more favourable response to ISAF than an 80 year-old hill farmer. Does the sample really reflect the general population?

4. Were the pollsters accompanied by a heavy security presence? "Are you in favour of the ISAF presence?" becomes an easy question to answer for a variety of reasons if heavily-armed examples are swarming around the place but the answer is not necessarily going to be the same once they've gone. The Taliban might genuinely get a similar response from a war-sick peasantry with a conditioned reflex to tooled-up strangers if they conducted their own poll under those same conditions.
The survey document doesn't go into a lot of detail on this.

Survey methodology is fairly standard across the industry, and a lot of effort goes into designing a sample group. However, the best way to distort a survey isn't the sample group. It's in the questions you ask them.

This document shows what questions were asked, and most don't seem to be of the leading 'And when did you stop beating your wife' variety. Although asking 'In your opinion...' and then giving a fixed menu of responses does limit the possible answers, and could be seen as steering the response, it does allow for commonality on the larger scale. If, for instance, there was a pressing local problem in one village, it would only appear as 'none of the above' in a national poll.

The circumstances of the interview, whether it was done in private, in the street, by Westerners or local staff, would also have a bearing, but I suspect it would be a rare or insane Westerner who stood in an Afghan marketplace long enough to do these interviews. I would expect that most questioning was done by locally recruited staff, if only for the language.

Latest Threads