Dr Jonathan Birch (Letters, 16 May) confidently asserts that ComRes has stopped asking respondents how likely they are to vote and instead weights responses “by demographic variables such as age and class”. ComRes has always weighted by those and other demographics to ensure that samples are representative of British society, and we have not stopped asking respondents how likely they are to vote.
Having reflected on many years of modelling voting intention, and aware that most final prediction polls across the industry have overstated Labour support for the past 34 years, we have changed our treatment of different demographic groups in order that we can take account of what we know to be their actual, as opposed to self-declared, voting behaviour. For example, if an affluent, university-educated 65-year-old tells us they are absolutely certain to vote, then their propensity to vote tends to be greater than that of a low-income, secondary-educated 25-year-old making the same claim. That is not a value judgment but a statement of reality.
We have a choice of either taking that reality into account when modelling voting intention or deliberately choosing to be less accurate in order to avoid offence to those who would wish to see our polls flatter one demographic group at the expense of another. We make no apologies for striving to be as accurate as possible.
• Join the debate – email email@example.com
• Read more Guardian letters – click here to visit gu.com/letters