Stories. Questions. Different perspectives.
Browse WAURE
← Advice & Questions
Politics · Advice & Questions ·

Does weighting survey answers mean the survey interviewed more people?

Politics category photographPolitics

In an illustrative worksheet for our politics group, I gave some survey responses more weight than others. A member asked whether the resulting population estimate meant extra people had effectively answered the questions. I said no, then struggled to explain what the weights were doing without making them sound like invented respondents.

The ONS explanation of sample design and estimation describes weighting as part of producing population estimates from sample responses. A weighted estimate and the number of actual respondents answer different questions.

How would you show that distinction clearly? I want the worksheet to display the real sample size and the estimation method, without implying either that weighting is automatically suspicious or that weights can make any sample representative.

Discussion

5 community replies

Show the respondent count and the weighted result separately. Labelling both as “people surveyed” is what would create the confusion; the estimate shouldn’t quietly replace the count.

It may help to keep the same illustrative response rows visible before and after weighting. Their influence on the estimate changes; the worksheet hasn’t acquired additional interviews.

Include a sentence about how the sample was selected. A discussion of weights on their own can hide the earlier choices that determine what the survey can tell you.

If one answer has greater influence, isn’t it fair to say that person got more votes than another?

In an election, that wording would raise a different issue. Here the exercise is estimating a population from a sample. Explain why a weight is used and how it was chosen, while keeping the actual respondents visible rather than describing the exercise as an election.

Join the discussion

More in Politics