Europe’s Innovation Gap and the Cost of Failure
An Exchange on Dismissal Costs, Firm Size and the Evidence for Labour-Market Reform. By Oliver Coste, Pietro Galeone, and Giorgio Presidente
In 2025, Oliver Coste, tech entrepreneur and president of FESDI, and Yann Coatanlem, economist and entrepreneur, argued that employment protection makes failure too costly for disruptive innovation.
Their IEP Bocconi and HEC Paris paper, Cost of Failure, Disruptive Innovation and Targeted Flexicurity, drew on about 250 restructuring plans to link high employee-related costs to weak European investment in tech and biotech.
These findings prompted further research at IEP Bocconi, given the lack of reliable cross-country data on restructuring costs.
The resulting September 2026 working paper, Restructuring Costs and Innovation, draws on a survey of 200 firms across four countries.
Pietro Galeone, head of the Labor and Social Policy Unit at IEP Bocconi, Giorgio Presidente, head of its Enterprises, Technology and Growth Unit, and Luca Di Casola, an economics MRes/PhD candidate at Queen Mary University of London, designed the survey to measure the magnitude and composition of restructuring costs and collect direct evidence from firms on the obstacles to innovation.
They find lower dismissal costs than earlier estimates. Firms’ responses indicate that restructuring costs, while an important constraint particularly for high-tech firms, are generally not perceived as the main barrier to innovation in the sample. Firms under stricter dismissal rules show no greater tendency to rank them as such.
Coste has written a short note in response to the new working paper. We publish it below, together with the authors’ reply.
by Oliver Coste
The IEP Bocconi pilot survey Restructuring Costs and Innovation (working paper, September 2026; commentary, 13 October 2026) concludes that dismissal costs are “an order of magnitude” lower than previously estimated, and that restructuring costs are not the main barrier to innovation. Neither conclusion follows from the data it presents, for three reasons.
1. The sample is not representative of the firms that do tech R&D. Tech R&D is primarily the work of large groups. In the United States, the 268 large tech companies of the EU Industrial R&D Scoreboard account for close to 90% of tech R&D and employ 22,000 people on average1.
The key disruptive innovations of the last twenty years (the smartphone, the cloud, AI GPUs, autonomous vehicles) came from large groups.
The survey’s respondents are mostly small firms: half have fewer than 250 employees, and 92.5% fewer than 5,000. Investigating innovation in tech with this sample is like studying innovation in the car industry by interviewing dealerships and EV start-ups, and leaving out car manufacturers. It is not a relevant sample for assessing restructuring costs and their impact on innovation.
2. The gap with earlier estimates has another explanation. The paper suggests that the estimates of Coatanlem & Coste (2025) were inflated by non-employee costs (asset write-downs, facility closures) bundled into the restructuring charges disclosed in annual reports. They were not.
The 2025 study explicitly reported employee-related costs only (section 2.1), and discarded every case where that share could not be isolated. The gap between the two sets of figures is the logical consequence of the different sizes of the companies in the two samples, not of what was counted2.
3. The question is put to firms that have already adapted. Firms operating in Germany and Italy are, by construction, those whose business model survives strict employment protection, even in tech. Asking them whether that protection is their main obstacle is like asking oil companies in the Gulf whether the lack of rain hurts their business. Their answer will be no, and it will be accurate for them. The relevant question is a different one, put to a relevant sample: how do large tech companies decide where to make their major investments, country by country?
The pilot is welcome as a first attempt at firm-level measurement of restructuring costs, and the authors are right that country-level indices are a crude proxy. But its methodology reflects a widespread European assumption: that tech is made of start-ups.
That was the world of the 1980s. In 2026, tech is made of large groups. They thrive outside Europe and are badly missed within it, for reasons that need to be rigorously investigated, with the firms concerned in the sample.
FESDI would welcome contributing to the design of that larger survey.
1 EU Industrial R&D Scoreboard 2025, USA, ICT producers + ICT Services: 268 companies, total of 421 bn euros R&D investment in 2024, employee count: median at 4,100, average at 22,000. US Startups R&D investment, estimated at one third of annual fundraising, source Dealroom 2023 - 2025.
2 Employment protection laws always provide flexibility for small companies and start-ups: light procedures for individual dismissals, and collective-dismissal procedures that apply only above headcount and dismissal thresholds. In addition, small firms benefit from a general tolerance of interim, contractor and self-employment arrangements that is not extended to large firms.
This is enough to explain the gap between the restructuring costs found in the pilot’s sample (9.6 months of salary in Italy, mandatory plus voluntary) and those of large groups (48 months in Italy).
by Pietro Galeone and Giorgio Presidente
We thank Oliver Coste for engaging with our paper.
We agree that restructuring costs are a barrier to innovation in Europe. The key question is how much of these costs can be attributed to dismissal regulation. On this point, the evidence remains insufficient to support the policy conclusions drawn by Coatanlem and Coste (2025).
Below, we provide a point-by-point response to his note.
- The sample is not representative of the firms that do tech R&D
Correct, the sample was designed to measure restructuring costs across different types of firms, not among top R&D spenders. Our sample deliberately covers firms of different sizes and across sectors.
By contrast, the sample used by Coatanlem and Coste (2025)—hereafter CC—disproportionately represents banks, insurers, telecommunications companies, and unionized former public utilities. These firms typically employ workers with long tenure and are themselves unlikely to undertake risky innovation projects given the sector they are in. Moreover, given their political salience, they are also more likely to offer generous pension and early-retirement packages, which can substantially inflate voluntary dismissal costs – which CC does not distinguish from mandatory dismissal costs.
More broadly, the note (and CC) assumes that breakthrough innovation must come from large firms. But that is a hypothesis to be tested, not a sampling rule.
Although our paper is not about the issue of who innovates more, we observe that there are many examples of small firms disrupting their own industry, including in Europe.
BioNTech and Mistral are cases in point. Less anecdotally, a long-standing literature examines the relative contributions of young (which happen to be small) firms and large incumbents to innovation (e.g. Acs and Audretsch, 1988; Akcigit and Kerr, 2018).
Large incumbents account for a substantial share of aggregate R&D expenditure and patenting, but this mostly reflects their greater size.
The relevant distinction is between the volume of innovative activity and its intensity (e.g. in relation to sales), productivity, and impact.
Young firms (which happen to be small) may contribute disproportionately to radical innovations relative to their economic weight. OECD microdata show that, among firms performing R&D, small firms—particularly young ones—invest a larger share of sales in research than established firms (Appelt et al., 2022).
Evidence from US firms also indicates that innovation productivity, measured by patent citations per dollar of R&D, declines with firm size and age (Plehn-Dujowich, 2013).
More recent evidence based on linked US inventor and firm data suggests that, although large incumbents generate more patents, their inventions tend to be less impactful: patents produced by inventors at established large firms receive approximately 35% fewer citations than those produced at young firms (Goldschlag, 2026).
Taken together, these findings suggest that young firms might play a disproportionate role in generating breakthrough innovation, even though large incumbents dominate aggregate R&D expenditure and patenting. So, it is not clear at all that one should discard a priori small firms from the sample.
- The gap with earlier estimates has another explanation
Unlike stated in the note, we do not claim that the estimates in CC include non-employee costs. Our paper explicitly states that these costs are excluded.
Our point concerns our own data: survey-based estimates expressed in monthly-salary equivalents fall within the range reported by CC only when we focus on the largest restructuring plans and include all cost categories.
The note correctly points out that EPL exempts some small firms and start-ups. However, the employment thresholds at which dismissal regulations become binding are well below 250 employees. Half of the firms in our sample exceed that threshold. Firm size alone therefore cannot explain the gap between CC’s estimates and our survey-based measures.
Sample composition, however, can (see previous point).
- The question is put to firms that have already adapted
Survivor bias is the note’s best criticism. The pattern in our data — restructuring costs cited as a barrier more by US than EU firms, and more by small than large firms — is what selection predicts.1
On that margin the cost of EPL is risk not taken i.e. projects that firms choose not to pursue. Such forgone opportunities cannot be observed in filings-based estimates.
Surveys can ask firms about projects they decided against, but establishing whether EPL caused those decisions remains difficult. Both CC’s estimates and ours capture restructuring costs among firms that chose to incur them, rather than the projects that never materialized.
Yet on the margin we do observe, the data neither single out dismissal regulation nor support the magnitude of CC’s estimates — on which the urgency of their proposed EPL reforms largely rests.
Among high-tech firms — those for which adaptation is costliest — the share ranking restructuring costs as their main barrier differs little by regime (figure below).
Moreover, if dismissal rules were the main reason restructuring costs more in Italy than in the United States, the gap should be widest in mandatory costs and close to nil in non-employee costs — write-offs, machinery, building works — which EPL does not regulate.
Instead, the Italy/US ratio is 2.9 for mandatory, 2.8 for voluntary and 3.4 for non-employee costs. Whatever makes restructuring expensive in Italy scales every component alike; a cost working mainly through dismissal law would not.
This reinforces the main message of our paper: restructuring costs are a real barrier to innovation in Europe, particularly for high-tech firms, but they can have many drivers.
The existing evidence does not establish that dismissal regulation is the main one.
Better measurement of the different components of restructuring costs and their underlying determinants is therefore needed before drawing policy conclusions.
References
Acs, Z. J., and Audretsch, D. B. (1988). Innovation in Large and Small Firms: An Empirical Analysis. American Economic Review, 78(4), 678–690.
Akcigit, U., and Kerr, W. R. (2018). Growth through Heterogeneous Innovations. Journal of Political Economy, 126(4), 1374–1443.
Appelt, S., Bajgar, M., Criscuolo, C., and Galindo-Rueda, F. (2022). Micro-data Based Insights on Trends in Business R&D Performance and Funding: Findings from the OECD microBeRD+ Project. OECD Science, Technology and Industry Working Papers, No. 2022/04. OECD Publishing, Paris.
Bartelsman, E. J., Gautier, P. A., and De Wind, J. (2016). Employment Protection, Technology Choice, and Worker Allocation. International Economic Review, 57(3), 787–825.
Coatanlem, Y., and Coste, O. (2025). Cost of Failure, Disruptive Innovation and Targeted Flexicurity: More Evidence Supporting Targeted Reforms. IEP@BU Working Paper Series, Bocconi University.
Goldschlag, N. (2026). Older and Large Firms Produce More Patents but Startups’ Inventions Are More Impactful. America Counts, U.S. Census Bureau, January 22.
Klenow, P. J., and Li, H. (2021). Innovative Growth Accounting. NBER Macroeconomics Annual, 35, 245–295.
Kortum, S., and Lerner, J. (2000). Assessing the Contribution of Venture Capital to Innovation. RAND Journal of Economics, 31(4), 674–692.
Plehn-Dujowich, J. (2013). Product Innovations by Young and Small Firms. U.S. Small Business Administration, Office of Advocacy, Research Report No. 408.
Saint-Paul, G. (2002). Employment Protection, International Specialization, and Innovation. European Economic Review, 46(2), 375–395.
Samaniego, R. M. (2006). Employment Protection and High-Tech Aversion. Review of Economic Dynamics, 9(2), 224–241.
IEP Bocconi does not express opinions of its own. The opinions expressed in this publication are those of the authors. Any errors or omissions are the responsibility of the authors.
