How We’ve Changed (and not changed) Polling Methodology Since 2015
Survation’s methodology for UK-wide telephone polling has proved sound, and remains effectively unchanged.
- Our final telephone poll before the 2014 Scottish Independence referendum was the polling industry’s most accurate – a squeak ahead of Ipsos Mori’s similarly accurate work (by just one decimal point).
- Our embarrassingly-considered-rogue final 2015 General Election poll (and the only poll we conducted using our custom telephone methood) correctly picked up the correct Conservative lead over Labour – which was entirely missed by our online work and the industry (including all methods).
- Our published and private client polling for the EU Referendum accurately reflected the final result – the cause of the difference in our published week before 52% leave figure and our 49.5% leave figure 3 days prior to polling day was identified as a time of day-contact-by-age group issue.
- A final, published example was our 2016 Holyrood telephone polling for Daily Record, which correctly picked up the late Conservative surge which denied the SNP an overall majority in the parliament.
What is our telephone methodology?
Simply put, our telephone polling methodology involves targeting a pre-stratified sample of individuals, whom we believe to be representative of all persons in a chosen area – in the case of our UK polling – this includes Northern Ireland, rather than more usual “GB” region.
Calling most age groups using different combinations of mobile phones and landline contact points helps us reach all demographic categories – such as the “older old” and the 18-24 category. When trying to find hard to reach persons, we consider factors such as the best time of day and best days to call. For example, our Good Morning Britain series uses a “Friday through Saturday” method, to better reach different demographics.
Changes to report?
At present, the main difference between our final telephone poll of 2015 and our recent UK phone polling is in the voting intention prompt.
Our final voting intention poll in 2015 used a “ballot prompt” method whereby respondents are read out a list of all the candidates standing in their own constituency, including independent candidates and fringe parties. This method encourages the person interviewed to consider an alternative tactical choice than they might otherwise or a popular incumbent or challenger.
So far, in this election cycle, Survation have employed a standard prompt:
Conservative, Labour, Liberal Democrat and UKIP in Great Britain, including SNP in Scotland and Plaid Cymru in Wales.
The prompt in Northern Ireland contains Alliance, DUP, SDLP, Sinn Fein, and UUP. ‘Another Party’ is the final prompt across the UK.
Respondents who nominate Green or UKIP have been asked a secondary voting question, whereby they are asked who they would vote for if a candidate for their preferred party was not standing in their constituency. If it is found (via postcode lookup) that their preferred party is not standing, then their subsequent response is reallocated to another party, would not vote, undecided or refused, as appropriate.
Polling by online panel
Survation has taken steps to improve our sampling methods, after reviewing the issues that affected us and the industry in 2015, including:
- Targeting our online sample by household income and education (excl. Scotland polling – our Scotland online polling – e.g. our 2015 Westminster polling does not suffer from the UK online issues as we spend an extended 4-5 days quota balancing Scotland sample.
- We have introduced targeting sample by age using narrower age bands.
- We are running fieldwork at more appropriate times, based on the analysis of response rates of different demographic groups from pre-GE2015 surveys.
- We advise our clients better regarding times and days fieldwork should take place, to allow the time needed for all groups to be able to participate in surveys.
Our online polling is also being adjusted in a comparable manner to our telephone polling, where supporters of the Green Party or UKIP have their secondary voting intention reallocated to other parties where no Green or UKIP candidate is on their ballot paper or undecided refused or would not vote.
Finally, we are using data gathered in the run-up to this election to “fine tune” some elements of our methodology. We will publish any changes with full transparency.