Q.A survey asks respondents: 'Don't you agree that this excellent new government policy has helped the poor?' 80 per cent of respondents said 'yes'.

The question is worded in a leading, loaded way ('excellent... don't you agree'), which pushes respondents toward answering 'yes' regardless of their true opinion. …

1. This question illustrates the problem of a leading or loaded question. By including the word 'excellent' and phrasing the question as 'Don't you agree...', the question itself signals the answer the researcher expects or wants to hear, making it far more likely that respondents will simply agree, out of social desirability or a wish to avoid seeming to disagree with an apparently settled, positive framing, rather than because they genuinely believe the policy helped the poor. The chapter notes exactly this general problem with survey methods: people sometimes give the response they think the researcher wants to hear rather than their own honest opinion, and a loaded question like this one actively invites that distortion.
2. A well-designed questionnaire can reduce this problem through neutral, balanced item-wording, for example, rephrasing the question as 'To what extent, if at all, has this policy helped the poor?' with response options ranging evenly from 'not at all' to 'a great deal', so that no particular answer is implied by the question's own wording. Using closed-ended response scales with genuinely balanced anchors, rather than a wording that presumes a positive answer, removes the pressure toward agreement that the original question created. …