The UK Prime Minister, Rishi Sunak, has called a General Election for July 4th, firing the starting gun on 6 weeks of frenzied campaigning.
It’s also going to be a busy time for pollster and research companies as they reach out to the electorate to try to predict the result and understand the issues that are driving their votes.
How Things Stand
On the face of it, this maybe looks like an easy election to predict. As of May 2024, the Labour Party, led by Keir Starmer, holds a historically huge lead over the incumbent Conservative Party.
Recent polls give Labour on 48% and the Conservatives on 27%, which give Labour a landslide win with a majority of around 150 – approaching a level of dominance not seen since Tony Blair.
With the Conservatives having been in power for 14 years, Brexit not having gone as planned, and the economy in a parlous state, the consensus is that it should be a shoo-in for Labour.
However, there are some lessons from recent elections that pollsters and commentators are no doubt heeding to avoid complacency:
2010 Election: Polls accurately predicted a hung parliament but overestimated support for the Liberal Democrats by about 4% following the party’s surge after Nick Clegg’s performance in the first televised debate. This highlighted the limitations of polling methodologies in capturing the true extent of voter intentions, and potentially was a case of pollsters creating and believing their own hype (Wikipedia).
2015 Election: Polls predicted a close race between the Conservatives and Labour, but the Conservatives won a clear majority. It turned out that the polls had consistently underestimated the Conservative vote by around 4%, leading to significant scrutiny and subsequent inquiries into polling methodologies (Wikipedia) (Electoral Reform Society – ERS).
2017 Election: Polls anticipated a comfortable Conservative victory, but the result was a hung parliament, with Labour performing much better than expected. Polls underestimated Labour’s vote share by 4-5%, with most of them failing to capture the scale of the late surge in Labour support (Wikipedia).
While the 2019 election was more or less accurately called, it genuinely was a foregone conclusion once Prime Minister Boris Johnson turned it into a single issue campaign on “Getting Brexit Done”.
Potential Pitfalls This Election
As Election Day approaches, polling companies are going to be extremely wary of potential biases which could skew their results. Getting the most accurate polling data means uncovering the intentions of so-called shy voters, making an effort to reach out to demographics that are less likely to participate in opt-in online polls, and aiming for higher participation and completion rates for more complete data.
Some of the pitfalls pollster and commentators will be watching for include:
- Underestimating Conservative Support: Similar to past elections, there may be an underestimation of Conservative support if “shy” Conservative voters are reluctant to disclose their true preferences.
- Overestimating Labour Support: Polls might overestimate Labour’s lead if they fail to account for lower voter turnout among younger, more Labour-leaning demographics.
- Impact of Smaller Parties: The influence of smaller parties like the Liberal Democrats, Greens, and Reform UK could be underestimated, affecting the overall prediction of seat distribution.
- Voter Turnout Models: Predicting who will actually vote remains challenging. Inaccurate turnout models can lead to significant discrepancies between poll predictions and actual results.
- Shy Voter Effect: Social desirability bias may lead some voters, particularly those supporting controversial candidates, to conceal their true preferences. This was evident in past elections and remains a concern for pollsters.
- Nonresponse Bias: Declining response rates and the difficulty of reaching a representative cross-section of the electorate continue to challenge pollsters. Those who do respond may not accurately represent the broader population.
Methods to Improve Polling Accuracy
We’ve written a whole White Paper – which you can download here for free – on how to improve the accuracy and representativeness of political polling. It boils down to several key things:
- Do More Phone Polling: Live phone polling is the gold standard for getting accurate polling data. You get higher response rates, more representative samples as you’re not relying on opt-ins, and clearer, more detailed responses as interviewers can clarify information. The reason most pollster rely on online surveys is due to cost (however that needn’t be a barrier with offshore CATI polling).
- Focus on Sample Representativeness: Ensuring that the sample accurately reflects the population, including hard-to-reach demographics who don’t voluntary participate in polls, is crucial. This includes addressing nonresponse bias and utilising sophisticated weighting techniques to correct for uneven representation.
- Embrace Transparency: Openly disclosing methodologies, sample demographics, and potential biases can foster greater public trust and scrutiny, leading to more reliable and accepted polling results.
- Collaborate and Innovate: Partnering with specialised providers and adopting innovative methodologies can enhance the efficiency and accuracy of polling efforts. For examples, this could include integrating online or SMS surveys as initial screening steps to gather demographic data before detailed CATI (Computer Assisted Telephone Interviewing) surveys.
The Benefits of Live Phone Polling
Live phone polling has dramatically declined in use during the past two decades with just 10% of pollsters now relying on this method. This change has largely been driven by the higher cost compared to online polling – which we’ve solved now by doing it from offshore. With live phone polling you get:
- Higher Response Rates: Using CATI for polling offers higher response rates compared to online and automated methods, ensuring a more representative sample of the population and reducing bias.
- Human Interaction: Live interviewers build trust and verify identities, reducing the risk of scams and increasing the reliability of responses. They can ask follow-up questions and clarify answers, enhancing data quality.
- Controlled Environment: CATI provides a controlled environment for data collection, minimising errors and inconsistencies. This leads to more accurate and reliable data, crucial for making informed decisions.
- Representative Sampling: CATI allows for easier weighting of responses to match the demographics of the target population, enhancing the accuracy of the data and ensuring underrepresented groups are included.
We’ve put together a Case Study to show some of the results and cost reductions we’ve achieved when conducting political polling offshore – download it here.
For more information about TKW Research’s Live Phone Polling solutions – and to find out how we’ve made it cost-competitive again – visit this page.
