Blog

One vote, one party - but how much do we agree with the party we vote for?

Posted in  by Matt Chocqueel-Mangan & Cyrus New
August 14, 2020

Data analysis: We crunched the numbers from the 1.1 million responses to our pre-election survey in 2019. The ‘average’ respondent bases their political decisions on around 6 issues, and selects policies from three to four different political parties. Just 1.9% of users choose policies from a single party. This article explains our approach to the analysis, and what the results mean for our democracy. 


Our democracy works best when education and information are freely available to us all. That’s why being able to make an informed decision when we vote is fundamental to our goal of increasing participation - it’s not just about voting, it’s about having the confidence that the party we vote for is the one that best represents the changes we want to see. 

What that means for our campaign to increase turnout, is that we simultaneously need to reduce ‘non-representation’ (not voting at all) while also reducing the risk of ‘mis-representation’ by voting with imperfect information - whether that’s media bias, our own bias and our past voting habits, or deliberate misinformation. 

The risk of mis-representation is clear. From polling some of our users during the 2019 election, 59% of the 1,941 respondents were surprised by the parties they chose policies from, and 31% said they were more likely to vote for a different party as a result. So without the right information, mis-representation when we vote is a common occurrence.  

The flip side to this, is that having the right information can also raise new challenges. If you have taken our pre-election survey you will most likely have chosen policies from more than one party. When this happens, you will have had to consider other factors - which issues you care most about, trust, party leadership, policies you can’t vote for - before choosing the party to give your vote to. That can be a real challenge. 

What we’re asking now, is can we distil our preferences down to just one party? To what extent can a single political party ever represent our views? Given that our current voting system means we can choose only one party, if most of us don’t agree with a single party across all issues, does that mean there will always be an aspect of mis-representation when we vote?  

To understand this, we looked at the data from our 2019 election survey. 

Analysis of the 2019 survey data

As you would expect, security and data protection are fundamental to the work we do. Not only has access to the data discussed in this article been tightly controlled, but the raw survey data itself has been structured in a manner that ensures respondent anonymity is preserved.

How many issues do we compare policies for?

Vote for Policies received responses from a little over 1 million people in the 2019 survey, with close to 90% of those living in England, 6% in Scotland and 4% in Wales. All but a handful of constituencies were represented, though unsurprisingly responses tended to be clustered around the major towns and cities.

The Vote for Policies survey (see Figure 1) asks respondents to select 1-15 of the policy issues that matter to them, such as Education, Housing and so on. For each issue selected a series of “policy sets” is presented, one per party. The respondent chooses from these sets ‘blind’ and, once all selections are made, the parties behind those policy choices are revealed.

Figure 1: Our election survey helps users compare policies across different issues, without knowing which party they belong to.

As the chart in Figure 2 shows, respondents tended towards selecting 5 or 6 issues. The 15-issue group is an obvious outlier. It’s driven, we think, by the design of the survey, whereby the user interface allows respondents to ‘select all’ with a single click.

Number of respondents by number of issues selected. Based on 1,079,640 respondents
Figure 2: Number of respondents by number of issues selected. Based on 1,079,640 respondents.

At the other extreme we have an apparent group of single issue voters.

We’ll return to these ‘outliers’ shortly. But before we focus on the numbers, let’s review the common issue and party/policy selections.

Which policy issues dominate?

Whilst the highest ranking issue selected for the 1-issue respondents is Brexit, in all other groups - that’s 2 issues up to 14 - ‘Health / NHS’ is the top ranking issue selected (in the case of 15 issues, all issues rank equally of course). Brexit drops out of the top 3 altogether once we get to respondents selecting 6 or more issues.

Top ranked issues by number of issues selected
Figure 3: Reading the top row: of those who selected just 1 issue in the survey, 26% selected Brexit as that issue.

So the issues that matter seem broadly consistent across these cohorts: Health, Brexit, Education and Jobs. What about the parties behind the policies that resonated with these respondents? Is there a link between the number of issues you select and the parties whose policies you are likely to identify with?

The numbers below suggest not. Although at first glance support appears to flip from Conservatives to Labour as the issue count rises, the actual percentage of respondents within each cohort is quite similar - certainly for the top 2 parties in each case. So the margins are fine.

Top ranked parties by number of issues selected
Figure 4: Reading the bottom row: of those who selected 15 issues in the survey, 86% selected (once or more) policies of the Green Party.

How many parties do we choose? 

Let’s now return to the concept of ‘mis-representation’. A general election prompts us to choose one party’s set of policies over all others. Does this reflect a natural tendency to identify with a single party’s manifesto, or does it obscure common ground across party lines?

To answer this we sought to measure the average number of parties respondents chose. But ‘average’ is a broad concept, and the distribution of our survey data demonstrates this. In our 15 cohorts, as mentioned, we have two extremes which appear to be genuine outliers.

Number of respondents by number of issues selected
Figure 5: Number of respondents by number of issues selected. Based on 1,079,640 respondents.

Removing the outliers

We think those who chose 15 did so because of survey design. We think it’s reasonable to conclude that the survey itself is what led to so many respondents selecting that option; we think a small minority of actual voting decisions are made based on such a wide set of policy issues.

At the other end, the issue of the day - Brexit - dominated the single-issue respondents, forming 26% of the policy choices for that group. The next highest Brexit ranking was 16% for the 2-issue respondents (compare this to 30% on ‘Health / NHS’ for the same cohort). So a particularly hot topic at the time of the survey has, we think, inflated the size of the 1-issue group.

So in the rest of this post, we’re setting these two extremes aside (179,069 15-issue respondents and 41,907 1-issue respondents). When we do so the distribution is far more symmetrical, albeit with some skew.

Number of respondents by number of issues selected (outliers removed)
Figure 6: Number of respondents by number of issues selected. Based on 858,664 respondents (1-issue and 15-issue respondents removed).

So with outliers removed, we can say this population of 858,664 respondents selected 6 issues on average (median 6, mode 6, mean 6.8). The 6 most frequently selected policy issues overall were ‘Health / NHS’, ‘Education’, ‘Environment’, ‘Jobs / Work’, ‘Brexit’ and ‘Crime’ (note this happens to be true with or without the outliers included).

Now let’s consider how these respondents’ issue selections mapped to parties. We’re not concerned with the parties themselves - just the extent to which a respondent’s preferences map naturally to a single party.

Figure 7 summarises this. Each column groups the respondents by the number of issues they selected (2 to 14). The colours indicate the number of parties those selections map to (1-6). The height of each coloured bar represents the percentage of respondents in that group. So, for example, one can see that of those respondents selecting just two issues, around a quarter selected the policies of one party. The remaining three quarters chose two parties.

Percentage of respondents whose policy selections map to 1, 2, 3, 4, 5 or 6 parties
Figure 7: Percentage of respondents whose policy selections map to 1, 2, 3, 4, 5 or 6 parties. Based on 858,664 respondents (1-issue and 15-issue respondents removed).

Moving from left to right, ‘adding’ another issue leads to a proportion of respondents ‘adding’ another party (though that proportion quickly drops off with each additional issue until of course we hit the 6-party limit).

So it seems fair to conclude that when we make party choices on policy grounds alone, one party is rarely representative of our preferences, and indeed 3 or 4 (yellow and green above) is the norm. Here’s the same chart with absolute numbers rather than percentages.

Number of respondents whose policy selections map to 1, 2, 3, 4, 5 or 6 parties
Figure 8: Number of respondents whose policy selections map to 1, 2, 3, 4, 5 or 6 parties. Based on 858,664 respondents (1-issue and 15-issue respondents removed).

Comparing averages

We can try to summarise this distribution as an average number of parties selected. The mode and median, again, are the same: 3 parties. But this is very marginal - 35% selected 3 parties, while 33% selected 4. And 47% of respondents selected policies from 4 or more parties. The mean on the other hand (3.42), while a ‘theoretical’ value, takes this spread into account.

So, with all the assumptions and caveats already mentioned thus far, we can summarise that the ‘average’ respondent bases their political decisions on around 6 issues, and seems best represented when selecting policies from approximately three and a half separate political parties. 

Contrast this with the 5.4% percent of all respondents who selected policies only from one party. And after removing the outliers (1-issue and 15-issue respondents) the percentage goes down even further - to just 1.9%.

What the data tells us

Like all data, it’s only as good as the analysis applied to it. We have to take into consideration that our survey data is not based on a nationally representative sample, and also that averages are, by their very nature a simplification of group behaviour. But there is still a lot to gain from a dataset of this size. For Vote for Policies, our interest is in what this means specifically for delivering our mission of 90% turnouts by 2030, with every citizen able to make an informed choice.

For governments to be effective, they must be representative. That means more of us voting so that our politicians engage with the full, diverse range of voices and views across the country. If our voting system allows us only one preferred party (as our ‘First Past The Post’ system does), for most of us that means there is an inherent compromise in the choice we make.

At Vote for Policies, we want to give people the information and motivation to vote, and keep voting because they see value in doing so. If we are asking people to engage with a democratic process which is unlikely to be truly representative then we need to recognise this too. It will never live up to the sense of fairness that we expect from our institutions, and it will erode our sense of efficacy - the feeling that our actions can have an impact and that institutions respond to our needs. 

The more we experience a system that doesn’t live up to its values, the less we’ll expect. We’ll become less inclined to engage in shaping and upholding our democracy. 

Whatever voting system we use, one of its strengths must be in accommodating how we align with multiple parties across different issues. Without this, we are accepting a system of mis-representation. This is why we need to understand the role of the voting system in our efforts to increase participation. 

We don’t have research that tells us how high a priority the voting system is for different voter groups - especially those less likely to vote. But we know enough to say we need to bring it into the discussion, and make sure we give everyone the right information to make up their own minds. Certainly when you compare different voting systems, it’s clear there are options that provide a less polarised representation of votes. This has to be a consideration when trying to increase public motivation and willingness to vote. 

By helping more people to engage in our democracy, we can help ensure that it becomes fairer, and that it lives up to our values for generations to come.

We all want to see a fair system that does the right thing by all members of our society. By helping more people to engage in our democracy - with free access to the information we need - we can help ensure that it becomes fairer, and that it lives up to our values for generations to come.


For more information on different voting systems, visit Make Votes Matter and the Electoral Reform Society.