Our manifesto outlines “22 ways to build resilience and aspiration in people and communities” across five key areas. Download your copy.

Dismiss close

Poll-itics: the art and science of predicting who’ll win

A badge is overlaid on the Catch22 green gradient background with the text "Catch22 Election Watch" and a cross inside a box.

In the run up to the general election taking place on the 4th July, Catch22 will publish its “election series” of blogs. In this week’s blog, Magid El-Amin, Director of Evidence and Insight, provides a run-down of the different polling methods that are employed across the UK to gauge public opinion and predict the outcomes of elections.

Last year, we published our manifesto, which outlined the ways in which we want to see all political parties, in the run up to the 2024 general election and beyond, commit to policies that ensure everyone in society has good people around them, a safe place to live and a purpose in life: we call these our 3Ps.

The upcoming general election will affect every member of our society so we believe that it is important for everyone to understand the information that they are receiving about it. Below, we have outlined some of the most popular methods for polling voters, with some detail on how they work, and the key positives and negatives of each technique.

Multi-level regression and post-stratification (MRP)

Multi-level regression and post-stratification (MRP) is a sophisticated and powerful polling technique. It helps predict winners in individual seat – which is why it’s so important to political campaign strategists. It helps them decide where in the country their candidate should visit, where to push campaign ads, and crucially, which topics to amplify. So, when you see Sunak pouring pints in Erewash, East Midlands or Starmer in a high-vis in Derbyshire, MRP will have played a part in how they got there.

Here’s how it works:

  • MRP takes a large sample of poll respondents (e.g. over 10,000) and using a statistical model to predict how each individual respondent will vote, based on a combination of individual-level data (e.g. race, gender, age, education, past voting record).
  • It combines the above with local area info, commonly the parliamentary constituency: this is the ‘multi-level’ part.
  • The model calculates a percentage probability of how each group of voters is likely to vote: this is the ‘post-stratification’ part. For example, the model might predict that a white, 30-year-old woman who went to university has a 35% chance of voting Labour. The probabilities are then aggregated to predict a constituency seat winner.

Pros and cons of using MRP

  • MRP provides granular estimates at local levels, which can be useful for understanding regional variations.
  • MRP accounts for demographic factors and adjusts for potential biases.
  • MRP requires extensive data collection and complex statistical modelling.
  • The accuracy of MRP depends on the quality of the underlying data and assumptions made in the models.

Opinion and exit polling

An opinion poll is where people are asked about their voting intention, whereas in an exit poll, voters are interviewed as they leave polling stations after casting their votes. Pollsters ask a representative sample of voters about their choices and demographic information.

They both provide a snapshot of the electorate’s preferences and can help predict election results before the official count is complete. But, they can go spectacularly wrong… as they did in 1992 and 2015 where polls predicted either a hung parliament or a slim Labour majority (1992), when in reality Conservatives won with a small majority in both cases and Labour were wiped out in Scotland by the SNP (2015).

Pros and cons of opinion and exit polling

  • Opinion and exit polling provide early insights into voting patterns and potential outcomes.
  • Opinion and exit polling allow for analysis of voting behaviour based on demographics.
  • Opinion and exit polling can be inaccurate if the sample is not truly representative – as happened in the examples above.
  • Voters may not disclose their true choices or may change their minds after being polled. For example, the ‘Shy Tories’ phenomenon in 1992, where polled voters cast for the Conservatives but told pollsters otherwise.

Tracker polls

A tracker poll is a type of survey which is conducted repeatedly over a period of time to track changes in public opinion or attitudes. These polls are typically conducted at regular intervals, monitor trends, and identify shifts in public sentiment.

Pros and cons of tracker polls

  • Regular updates from tracker polls offer timely information, enabling quick responses to changing public sentiments or emerging issues.
  • The consistent methodology used in tracker polls makes it easier to compare data across different time periods, ensuring that changes observed are due to real shifts rather than variations in survey techniques.
  • Conducting surveys regularly can be expensive and resource-intensive, especially if large sample sizes are required.
  • Respondents may experience survey fatigue if they are repeatedly asked to participate, potentially leading to lower response rates and less reliable data over time.

Some of these techniques aren’t restricted to elections: a social organisation like ours can use MRP to understand how our interventions can be tailored to fit with different group of service users, or to better understand risk factors affecting different populations and how we might design programmes to fit.

While no technique is perfect, the key is in having accurate data, a rigorous methodology and being cognisant of potential biases. Yes, keep an eye on the numbers but always remember, someone could just as easily throw a milkshake in the mix.