Elsevier

Research Policy

Volume 51, Issue 9, November 2022, 104138
Research Policy

Does light touch cluster policy work? Evaluating the tech city programme

https://doi.org/10.1016/j.respol.2020.104138Get rights and content

Highlights

  • Evaluates the causal impact of a flagship UK technology cluster programme, using synthetic controls and placebo tests.

  • Uses rich workplace-level microdata from the Business Structure Database, alongside a range of other administrative sources.

  • The policy grew and densified the cluster, but has much weaker, partial effects on tech firm performance.

  • Most policy ‘effects’ began before rollout, raising questions about the programme's added value.

  • More careful policy design could make future interventions more effective.

Abstract

Cluster policies are popular with policymakers, but we know surprisingly little about their effectiveness. This paper evaluates the causal impact of a flagship UK technology cluster programme that uses ‘light touch’, market-orientated interventions. I build a simple framework and identify effects using synthetic controls plus placebo tests to handle programme endogeneity concerns. I implement this design on rich workplace-level microdata from the Business Structure Database, alongside a range of other administrative sources. I explore mechanisms through further tests for timing, cross-space variation, scaling and churn. The policy grew and densified the cluster, but has much weaker, partial effects on tech firm performance. I also find most policy ‘effects’ began before rollout, raising questions about the programme's added value. More careful policy design could make future interventions more effective.

Introduction

Clusters have been a well-known feature of urban economies since Marshall first identified them in 1918. A vast literature explores their determinants and characteristics (Duranton and Kerr, 2015). Cluster policy is more controversial: it is popular with policymakers but disliked by many academics (Tödtling and Trippl, 2005). Clusters – industrial districts of co-located, interacting firms – typically have market and co-ordination failures. In theory, public policy can improve cluster-level outcomes, outcomes for cluster participants or both. But clustering results from many firm and worker decisions; so market/co-ordination failures are complex; and this complexity may lead to policy failure (Duranton, 2011, Martin and Sunley 2003). The scale of these challenges is an empirical question. However, the literature evaluating cluster policies is small, and the set of robustly designed evaluations smaller still (see reviews by Duranton (2011), Urraya and Ramlogan (2013) and Chatterji et al (2014)).

Broadly speaking, we can distinguish three families of cluster policies. The first involves top-down, formal partnerships backed by grants or subsidies, usually generated through competitive calls for entry. Examples include French Local Productive Systems (Duranton et al, 2010, Martin et al, 2011) and Pôles de Compétitivité (Fontagné et al, 2013, Ben Abdesslem and Chiappini, 2019, Lucena-Piquero and Vicente, 2019); Japan's METI programmes (Nishimura and Okamuro, 2011); Innovation Network Denmark (DAMVAD, 2011) and German schemes such as BioRegio and BioProfile (Engel et al, 2013, Graf and Broekel, 2020) or Bavaria's High-Tech Offensive (Falck et al, 2010). These evaluations generally find positive impacts, although effect sizes are often modest. A second set of interventions centres on physical redevelopment, typically an ex-industrial neighbourhood in a city. For example, the 22@ cluster in Barcelona involved re-zoning the Poblenou area, extensive landscaping and construction, and incentives for new ‘knowledge-based’ firms (Viladecans-Marsal and Arauzo-Carod, 2012). The US ‘Innovation Districts’ movement advocates similar approaches (Katz and Wagner, 2014). These interventions are less well-studied, although the 22@ evaluation above found small shifts in industry composition . A third group involves light-touch, market-orientated interventions, mainly targeted at existing clusters. Building on work by Porter (1996; 2000) these programmes emphasise business support, improving firm-firm linkages, expanding market access and tackling other market/co-ordination failures. Examples include the Regional Innovation Cluster programme in the US (Yu and Jackson, 2011), and City Growth Strategies in the UK (McDonald et al., 2007). These tools are now often combined with place-branding interventions (Lundequist and Power, 2002; Markusen and Gadwa, 2010). These programmes are most commonly evaluated by participant surveys, with few or no quantitative impact evaluations to date (Urraya and Ramlogan, 2013).1

This paper develops a rigorous impact evaluation of a recent light-touch cluster policy. I study the UK flagship Tech City programme that launched in London in late 2010,2 and aimed to grow the cluster of technology companies (c. 2,800 firms in 2010) centred on Shoreditch and Old St roundabout. The cluster had been growing for years without direct policy input (Foord, 2013). It came to prominence in 2008 with a wave of media attention about 'Silicon Roundabout' (Butcher, 2013; Foord, 2013; Nathan et al., 2019). The Tech City initiative aimed to ‘accelerate’ the cluster by ‘going with the grain’ of the existing ecosystem (Cameron, 2010), combining business support, tax breaks, place-branding and network-building elements. While proponents argue the policy has been very successful (Mayor of London, 2014; Osborne and Schmidt, 2012), these claims have never been robustly tested.

Clusters involve both positive and negative feedback loops (Nathan and Overman, 2013). As they get larger and denser, agglomeration economies get stronger. However, such growth also raises crowding, and competition for market share. I argue that the Tech City policy could plausibly contribute to all three channels. As the cluster was growing pre-policy, and has continued to grow since, I need to identify any additional policy effects relative to the counterfactual of continued ‘organic’ development. To do this I apply theoretical frameworks developed by Arzaghi and Henderson (2008), Duranton (2011) and Kerr and Kominers (2015). I first explore economic changes in the area between 1997 and 2017. Next, I use difference-in-differences and synthetic controls on rich microdata to identify overall policy effects on cluster size, density and local tech firm performance. I explore mechanisms with four further pieces of evidence. I run placebo-in-time tests to identify effect timing; use treatment intensity analysis to explore within-cluster shifts; test for changes in high-growth tech firm activity; and run a before-and-after analysis of tech firm entry/exit patterns, for UK and foreign-owned firms.

I find that policy led to a larger and denser cluster, with an influx of hardware and software (‘digital tech’) companies, changing the composition of industry space. However, effects on firm performance are much less stable, with ‘digital content’ firms (such as those in media, marketing and webs services) gaining revenue/worker in some specifications, but no clear effect for digital tech businesses. I also find increased churn and spatial disruption. Finally, many policy ‘effects’ began when the cluster first came to media attention, rather than when the policy launched. Year-on-year outcome changes are often weaker in the latter period, and I find a negative policy effect for one group of firms. Consistent with theory, this suggests that the programme weakened the net benefits of cluster location (Duranton, 2011). Given the intention to ‘accelerate’ the cluster, the policy has had – at best – mixed impacts on the area and on the firms in it.

It is critical to develop a stronger evidence base for local economic development policies, including cluster programmes. This is the first quantitative impact evaluation of a light touch cluster policy that I am aware of.3 It complements other recent studies exploring clusters and local/regional economic performance, such as Delgado et al (2014) and Iammarino and McCann (2006). More broadly, the paper adds to the sparse cluster policy evaluation literature, and to a larger, related literature on economic area-based initiatives.4 The paper also adds to a small set of studies on London's post-industrial economic evolutions (see inter alia Hall (2000), Hamnett and Whitelegg (2007), Hutton (2008), Pratt (2009), and Harris (2012)).

Section snippets

Background

The Tech City area is located in a set of ex-industrial East London neighbourhoods between Islington, Tower Hamlets, Hackney and the City of London. It shares many characteristics of inner urban creative/technology districts such as Silicon Alley (New York) and SoMa (San Francisco) including a tight cluster shape, use of ex-industrial buildings, abundant social amenities and a gritty physical appearance (Zukin, 1995; Indergaard, 2004; Hutton, 2008). Cluster protagonists make extensive use of

Descriptive analysis

Since the introduction of the Tech City programme, the cluster has got both bigger and more expensive. There is tech firm growth in all parts of the zone (Fig. 2, top panel). At the same time, rents have risen relative to comparable submarkets (bottom panel). There is also extensive anecdotal evidence of displacement of smaller firms.14

Analytical framework

A cluster is a dynamic industrial production district. Four core dynamics condition its growth and change. As a cluster grows and gets denser, knowledge spillovers and other agglomeration economies raise participants’ productivity. But the costs of cluster location also rise with participant numbers, crowding some firms out of the district (Duranton 2011). A larger cluster also increases market competition between participants (Combes et al 2012). Competing land uses further influence costs,

Identification

Evaluating cluster programmes is inherently difficult because of the lack of clear comparators (Duranton, 2011). Cluster evaluations typically compare changes in the treated area/firms to some set of control areas/firms. Difference in differences gives a consistent average treatment effect on the treated (ATT), conditional on observables and on parallel pre-trends in treated and control groups. Causal inference requires that LSOA-specific time-varying unobservable characteristics affecting

Overall policy effects

Estimates of overall policy impact are given below. The next section explores what is driving these.

Fig. 4 shows results for cluster size. The left column shows changes in log digital tech and digital content firm counts in Tech City versus synthetic Tech City. The right column shows effect sizes for Tech City and 213 placebo units. Effect sizes are weighted by pre-treatment RMSPEs, so these graphs show relative effect size controlling for fit.

I find clear policy effects on digitech firm

Explaining policy effects

The analysis throws up three main findings. First, compared to a no-policy counterfactual, the policy raised firm and job counts, especially for digital tech activity, with weaker and less stable effects on digital content. Second, cluster density has increased, particularly job density. Third, this bigger, denser cluster does not always increase tech firm performance, with very unstable results.

Using Section 4’s framework, I explore the mechanisms behind these results. First, given the

Conclusions

This paper evaluates the causal impact of a flagship UK technology cluster programme, which uses ‘light-touch’, market-enabling interventions to ‘accelerate’ the cluster and firms within it. It is the first impact evaluation of this strand of cluster policy that I am aware of. I use rich microdata in a synthetic control setting to estimate overall impacts relative to continued ‘organic’ growth, and the mechanisms behind these.

I find that the policy substantively increased cluster size and

CRediT authorship contribution statement

Max Nathan: Conceptualization, Methodology, Software, Formal analysis, Validation, Visualization, Writing - original draft, Writing - review & editing, Project administration, Funding acquisition.

Declaration of Competing Interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Acknowledgements

Thanks to Maryann Feldman and three anonymous referees for invaluable comments. Thanks also to Simon Collinson, Anne Green, Neil Lee, Henry Overman, Maria Sanchez-Vidal, Rosa Sanchis-Guarner and Emmanouil Tranos for input, Francesca Arduini for code, Cushman & Wakefield for rents data, and to Martin Dittus, Kat Hanna, Sandra Jones, Natalie Kane, Andy Pratt, Matt Spendlove, Emma Vandore, Georgina Voss, Jess Tyrell, Rob Whitehead and Jonty Wareing for chat. Seminar participants at Birmingham,

References (78)

  • F Tödtling et al.

    One size fits all?: towards a differentiated regional innovation policy approach

    Res. Policy

    (2005)
  • A Abadie et al.

    Synthetic control methods for comparative case studies: estimating the effect of california's tobacco control program

    J. Am. Statist. Assoc.

    (2010)
  • A Abadie et al.

    Comparative politics and the synthetic control method

    Am. J. Political Sci.

    (2015)
  • P Aghion et al.

    The effects of entry on incumbent innovation and productivity

    Rev. Econ. Stat.

    (2009)
  • S. Appold

    Location patterns of US industrial research: mimetic isomorphism and the emergence of geographic charisma

    Reg. Stud.

    (2005)
  • WB. Arthur

    Competing technologies, increasing returns, and lock-in by historical events

    Econ. J.

    (1989)
  • M Arzaghi et al.

    Networking off Madison Avenue

    Rev. Econ. Stud.

    (2008)
  • S Athey et al.

    The state of applied econometrics: causality and policy evaluation

    J. Econ. Perspect.

    (2017)
  • S Becker et al.

    The impact of public employment: evidence from Bonn

    IZA Discussion Paper 11255

    (2018)
  • A Ben Abdesslem et al.

    Cluster policy and firm performance: a case study of the French optic/photonic industry

    Reg. Stud.

    (2019)
  • E Brynjolfsson et al.

    The Second Machine Age: Work, Progress, and Prosperity in a Time of Brilliant Technologies

    (2014)
  • M. Butcher

    Goodbye Mr Silva — a brief history of East London's ‘Tech City’

    TechCrunch

    (2013)
  • D. Cameron

    East end tech city speech

    4 November

    (2010)
  • E Cavallo et al.

    Catastrophic natural disasters and economic growth

    Rev. Econ. Stat.

    (2013)
  • A Chatterji et al.

    Clusters of Entrepreneurship and Innovation

    Innov. Policy Economy

    (2014)
  • P-P Combes et al.

    The productivity advantages of large cities: distinguishing agglomeration from firm selection

    Econometrica

    (2012)
  • E Currid et al.

    The geography of buzz: art, culture and the social milieu in Los Angeles and New York

    J. Econ. Geogr.

    (2010)
  • The impacts of cluster policy in Denmark - An impact study on behaviour and economical effects of Innovation Network Denmark

    (2011)
  • G. Duranton

    California dreamin’: The feeble case for cluster policies

    Rev. Econ. Anal.

    (2011)
  • G Duranton et al.

    The Economics of Clusters: Lessons from the French Experience

    (2010)
  • G Duranton et al.

    Nursery cities: urban diversity, process innovation and the life cycle of products

    Am. Econ. Rev.

    (2001)
  • G Duranton et al.

    The logic of agglomeration

    NBER Working Paper 21452

    (2015)
  • E Einio et al.

    The effects of spatially targeted enterprise initiatives: evidence from UK LEGI

    SERC Discussion Paper

    (2013)
  • D Engel et al.

    Does cluster policy Trigger R&D activity? evidence from german biotech contests

    Eur. Plann. Stud.

    (2013)
  • G. Faggio

    Relocation of public sector workers: the local labour market impact of the Lyons Review

    SERC Discusssion Paper 155

    (2015)
  • B Ferman et al.

    Cherry-Picking with Synthetic Controls

    (2018)
  • R. Florida

    Bohemia and economic geography

    J. Econ. Geogr.

    (2002)
  • L Fontagné et al.

    Cluster policies and firm selection: evidence from France

    J. Reg. Sci.

    (2013)
  • S Galiani et al.

    The synth_runner Package: Utilities to Automate Synthetic Control Estimation Using synth

    (2016)
  • Cited by (4)

    • Uncommon methods and metrics for local entrepreneurial ecosystems

      2022, Research Policy
      Citation Excerpt :

      The ways that both public policy and the business strategies of large firms influence local outcomes is a prime area for future research. Using microdata and synthetic control methods, Nathan (2022) estimates the mechanisms and overall impacts of government policies on organic growth. Nathan examines the effectiveness of a cluster policy – a flagship UK technology cluster program that uses light touch and market-oriented interventions.

    View full text