Should I stay or should I...? UK creative clusters and the EU referendum
In the Referendum of the 23rd of June, 52% of voters opted to leave the EU. How did people in creative clusters vote?
Answering this question is important for two reasons.
First, and according to many accounts, the referendum result reflected a ‘split’ between those parts of the UK population who feel the benefits of a globalised, knowledge intensive economy, and a larger group of people who don’t see the situation that way (see for example, these articles in the Centre for Cities LSE Brexit blogs, and this blog by Stian Westlake here at Nesta).
This split between voters may be influenced by the industrial landscape around them - when they look at their city or town, can they see sectors able to generate good jobs or not? If the creative industries can make local communities feel more economically empowered and optimistic, then they can play an important part in government attempts to rebalance the UK's economy, addressing the needs and aspirations of those parts of the country that feel left behind.
Second, the voting decisions of people working in the creative industries represent, at least in part, a ‘revealed preference’ between different policy regimes: one based on free trade in Europe and open access to talent, and a less clear set of alternatives selected by most people who voted in the EU Referendum. By learning about how the creative industries voted, we are learning about their hopes and concerns about Brexit. The government should take this information into account when it manages its negotiations with the EU, given the increasing importance of the creative industries for the UK economy.
What's the starting hypothesis?
My hypothesis is that areas with a stronger creative industries presence voted Remain, and that this link will hold even after we control for other local factors, such as average levels of employment, education or ethnic diversity in the population.
This prior sounds reasonable: Polls of the sector by the Creative Industries Federation suggest strong support for Remain among respondents. The creative industries have grown fast in recent years, and they have a strong export orientation. As Stian Westlake points out in this blog, workers in innovative, intangible-intensive sectors such as the creative industries tend to have personality traits such as openness to new experiences that, according to opinion surveys, correlate with a propensity to vote remain. A Nesta paper by Bakhshi, Frey and Osborne shows that creative industries workers in the UK are less at risk of having their jobs automated, a technology trend that has worsened the economic prospects of workers in many other industries.
Findings: There is a positive link between creative industries clustering and a location's propensity to vote Remain
The chart below plots “creative industries clustering” (the relative importance of creative industries businesses in the local economy compared to the UK average) in the horizontal axis, and an area’s “Remain Propensity” (total of population voting remain over total voting leave) in the vertical axis.
A score above 1 in creative clustering indicates that an area is relatively specialised in the creative industries. A score above 1 in Remain propensity indicates that more people in the area voted to Remain in the EU than to leave.
The chart shows a positive link between creative industries clustering and a location’s propensity to vote remain: 70% of those areas that had a creative clustering score above 1 voted to Remain, while 70% of those areas that had a creative clustering score below 1 voted to Leave. We also see a clearly higher propensity to vote Remain in Scotland (blue circles) than in England (orange circles), but even in Scotland, those locations with a stronger creative industries presence (Edinburgh and Glasgow) leaned more strongly towards Remain.
Potentially confounding factors: Creative clustering is stronger in highly educated, wealthier and more diverse areas
The relationship between creative clustering and voting for Remain could be driven by other ‘confounding’ factors - say, if the creative industries tend to locate in areas with higher levels of education, higher salaries or more ethnic diversity, which as we know from other research tended to vote Remain.
The ‘heatmap’ below shows the correlation between Remain voting, creative clustering and some of those factors. Warmer (orange-red) colours represent positive correlations (i.e. the areas that score positively in one variable tend to score positively in the other), while cooler (blue) colours represent negative ones (areas tend to score positively in one variable tend to score negatively in the other).
The chart confirms some things we already know: locations where a high proportion of the population has at least a university degree or equivalent qualification strongly tended to vote remain (correlation coefficient r=0.64), while the opposite is true for areas where many people don’t have any qualifications (r=-0.17). Areas with higher employment rates and annual gross income also tended to vote remain (r=0.17 and r=0.34). The correlation between creative business clustering and propensity to vote remain is positive but not as high (r=0.13). The correlation between propensity to vote remain and the share of the population between 16 and 64 which is non-white according to the 2011 census doesn’t appear to be significant (r=-0.01).
Also, in line with what I said above, creative clustering correlates positively with average education, employment rates, annual gross pay and share of non-white population between 16-64 years. As expected, creative industries tend to cluster in highly educated, well-paying, more ethnically diverse areas.
Multivariate analysis: There is a significant link between creative clustering and Remain share of the vote even after we control for other local factors
Does all the above mean that the link between creative clustering and an area’s propensity to vote remain is completely explained by other local factors? To check this, I have estimated an Ordinary Least Square (OLS) model of the link between creative business clustering and propensity to vote Remain while controlling for those factors.
The findings support the idea of a positive link between creative clustering and share of remain in the Referendum vote even after we take other important factors into account: on average, and all else equal, being in the top 13% of creative clustering (one standard deviation above the mean) meant a boost in the share of remain in the Referendum vote of almost 3% - not an insignificant amount in a vote this close (see summary of regression outputs below; a more detailed regression table is included in the online notebook).
The analysis in this blog suggest that there is a significant, substantial link between creative industries presence in a location, and its share of Remain votes in the EU referendum. The relationship holds after controlling for other potentially confounding factors. This finding fits with the idea that those places in the UK that have a stronger stake on an open, creative and innovative economy preferred to stay inside the EU, while those locations that felt less economically - and creatively - enfranchised opted out.
This raises important policy issues:
- How can we encourage the development of creative clusters across the UK, as a way of making local economies more resilient, heal the economic rifts that exist between different parts of the country, and encourage a more optimistic outlook about an economy where creativity and innovation are becoming more important every day?
- How can the UK minimise the negative impact of exiting the EU on sectors with high growth potential like the creative industries, which thrive in open markets and unfettered access to creative talent?
Our Geography of Creativity report sets out some actions - and releases new data - to support this agenda:
- Accelerate the development of creative clusters outside of London and the South East
- Continue efforts to share the benefits of London’s status as a global creative industries hub across the UK
- Local Enterprise Partnerships and universities should consider what more they can do to address the strengths and weaknesses within their particular area, such as an over reliance on large firms or growing links between graduate talent pools and creative clusters
- Networks of UK creative industries should strive to maintain their global reach
The recently created Creative Industries Council working group looking the impact of Brexit should play a central role in this process.
Analytically, there are many other interesting questions that we could pursue. In particular, it would be interesting to determine the relative importance of different mechanisms explaining the link between creative industries clustering and a propensity to vote Remain. To what extent was this driven by the voting decisions of individual creative industries workers, by voters in the wider social and economic networks of the creative industries, or more diffusely, by local atmospheres of tolerance that, according to the literature, are in part generated by the creative industries themselves?
We will look into some of these issues in follow-up research, and keep you posted about what we find.
This blog received helpful comments from George Windsor, Antonio Lima, James Gardiner and Hasan Bakhshi.
Annex 1: Data sources and definitions
The measure of creative clustering used in this blog is based on official definitions of the creative industries, and official business registry data for 2014. There is more information in our Geography of Creativity report, which you can download here. The data is available here.
The dependent variable (referendum outcomes) and control variables (employment and education data from the Annual Population Survey, income data from the Annual Survey of Hours and Earnings, and ethnic diversity data from the 2011 Census) were respectively extracted from the Electoral Commission website, and from the ONS Nomis website. In several instances (Electoral Commission, Census and ASHE data), this data wasn’t available at the Travel To Work Area (TTWA) level where we had creative clustering data. In those cases, we extracted data at the local authority district level and allocated it to TTWAs (or calculated weighted medians) using a lookup based on the distribution of Local Authority postcodes over TTWAs. All the code and data used in the analysis is available in an online otebook here.
All the analysis was performed using Python. For more information about specific packages, check the online notebook.
Annex 2: Regression summary
 There is more information about the data, and a link to an online notebook describing the analysis, in the data annex at the bottom.
 The size of the circles measures the levels of creative industries employment in the area (logged, to prevent London, which employs around 40% of the UK’s creative workforce from completely dominating the chart.)
 When estimating econometrics models with spatial data there is the risk that some observations may influence each other in a way that violates standard assumptions and biases the results. In our case, this would be happening if one area’s propensity to vote remain, or to have strong creative clusters influenced what happens in its neighbours. I have run some diagnostics suggesting that such ‘spatial autocorrelation’ is indeed present in the data. I have estimated a model that controls for these ‘spatial dependencies’, without big changes in the findings - see the online notebook for a summary). In the notebook, I also report the findings after ‘winsorising’ all variables, that is, after replacing potential outlier values above the top and bottom 1% with the values of the top and bottom 1%. The results stay very similar.