Our new project will explore the economic effects of diverse teams and workplaces — and the wider role of urban diversity — on entrepreneurship, innovation and productivity in the UK.
I’ll be working with brilliant colleagues Jon Reades at CASA, Tom Kemeny (QMUL), Ceren Ozgen (Birmingham), Anna Rosso (Insubria) and Anna Valero (LSE), and helped out by a suitably high-powered project advisory group. Plus two new team members: more on that below.
The full project outline is below, but here’s a summary of what we will do:
- We’ll use a mix of web data from the Diffbot knowledge graph, UK and EU administrative microdata and a range of machine learning tools, to build rich new worker-firm panels — something that’s extremely challenging to do at scale in the UK.
- We’ll then use this platform to test causal effects of diversity on various measures of economic performance at both firm and urban level.
- Critically, we should be able to explore multiple dimensions of diversity — gender, ethnicity, nationality and country of birth — something that’s also proven hard to do in the past.
- As well as a number of academic papers and short-form outputs, we’ll also be making a version of the platform available post-project as a public resource for other researchers through the UK Data Service.
- The grant starts in July 2022 and runs for three years.
This project’s aim is to explore the economic effects of diverse teams and workplaces — and the wider role of urban diversity — specifically, on entrepreneurship and firm-level innovation and productivity in the UK. These are important issues that are under-explored, especially in the UK, largely because of data challenges. And exploring these issues in the way we set out will make a valuable contribution to the huge, ongoing public debates on equalities, diversity and inclusion — both in the UK and across the world.
Our project will combine administrative microdata, novel online data sources and frontier methods in econometrics and data science. Specifically, we will match a range of individual-level data from the Diffbot knowledge graph to companies, then to administrative firm-level data (the Business Structure Database, plus patents and other information). Working in secure settings, we will use name analysis tools to probabilistically identify gender and ethnicity, and would also gather information on nationality and country of birth. We will focus the resulting panels on sectors where we’re confident our worker data has good coverage — likely to be strategically important industries like tech, finance and business services — and run our data through multiple quality checks. We will use various tools to get closer to causality, including instrumental variable strategies and using policy ‘shocks’ such as a) Brexit and subsequent policy events, and b) recent UK gender pay gap legislation. We will also deploy a robust set of technical safeguards to ensure individuals’ privacy, publishing only non-disclosive results.
The project will develop new knowledge in an important but under-researched set of topics. In the process it would also build a unique data platform that other researchers could use in the future. We will work together with leading industry, policy and civil society stakeholders with expertise on relevant concepts, data/methods and policy agendas. These enable the project to directly contribute to economic policymaking on productivity and its drivers, including the UK’s emerging levelling-up agenda, while also informing business decision-making and speaking to important and ongoing wider public conversations.
The project will generate a series of linked outputs:
- Three research papers, covering links between gender and ethnic diversity (and their intersections) and firm-level productivity, innovation and entrepreneurship. These would be published as working papers on high-profile platforms, then submitted to peer-reviewed journals.
- Additional non-technical, short-form content for each paper — blogposts / essays / features on well-read platforms and outlets, policy briefings, media comment; inviting our network / community to co-author or directly contribute whenever possible;
- The underlying data platform, which (subject to permissions) we will make available to other researchers as a safeguarded data asset;
- The wider network / community of researchers and practitioners we will build through the co-production process.