publications

2026

1. JEEA

   Exclusive Secrets

   with Giacomo Calzolari and Alireza Naghavi

   Journal of the European Economic Association (forthcoming), 2026

   When trade-secret protection is weak, manufacturers keep suppliers exclusive to limit who else they can sell to in order to protect sensitive know-how. Evidence from global automotive-parts contracts, where this link is strongest for moderately specialised components.

   Abs DOI SSRN CEPR Blog Code Slides

   We investigate the effects of trade secret (TS) protection on supplier exclusivity in vertical industrial relations. We show that manufacturers limit their suppliers’ buyer networks under weak TS protection to safeguard sensitive information while encouraging supplier investment. Using worldwide automotive parts contracts data, we find that weak TS protection constrains suppliers to serve only a few Original Equipment Manufacturers (OEMs), sometimes leading to exclusivity, especially for parts with intermediate relationship-specificity. This association strengthens with greater OEM bargaining power, aligning with theoretical insights. We validate these findings by examining supplier attributes, including size, cartel involvement, and ownership structure.

2. PNAS

   AI-Assisted Teams Outperform AI-Led Teams but Not Human-Only Teams in Assessing Research Reproducibility in Quantitative Social Science

   with Abel Brodeur, David Valenta, and 260 others

   Proceedings of the National Academy of Sciences, 2026

   A 288-researcher experiment testing whether AI can verify published research. AI assistants matched human teams only when humans stayed in the loop. Left to work alone, they reproduced just 37% of findings. Contributing author.

   Abs DOI Preprint

   Large Language Models (LLMs) such as ChatGPT are transforming how scientists conduct and validate research, offering promise as tools to improve scientific reproducibility. However, computational reproducibility and error detection remain expensive and labor-intensive. We experimentally test how collaboration between researchers and LLM assistants influences the reproduction of quantitative social science findings across different levels of AI autonomy. We randomly assigned 288 researchers to 103 teams working under three conditions: human-only, AI-assisted (using ChatGPT as a collaborative tool), or AI-led (ChatGPT operating with minimal human oversight). Teams reproduced published results from leading social science journals, detected coding errors, and proposed robustness checks. Human-only and AI-assisted teams achieved comparable reproduction rates (94% vs. 91%) and performed similarly on most outcomes, except human-only teams identified significantly more major coding errors. Both substantially outperformed AI-led teams, which achieved only a 37% reproduction rate, detected fewer errors across all categories, proposed weaker robustness checks, and required more time. This autonomous approach, however, likely represents only a lower bound of AI capabilities. Despite rapid model advances, expert human judgment currently remains indispensable for reliable empirical verification. While AI assistance did not degrade most outcomes, it provided no measurable advantages and was associated with reduced detection of major errors. However, the 37% autonomous reproduction rate indicates that AI could provide value in settings where scale or cost constraints preclude human review of papers, even though general-purpose LLMs offer no immediate advantages for human-supervised verification.

3. RP

   Do Firms Report More University Knowledge Sourcing When Informed About Engagement Channels? Evidence from a Field Experiment

   with Ammon Salter

   Research Policy, 2026

   Simply reminding firms of the many ways they can work with universities makes them 34% more likely to report universities as a source of innovation, with implications for how innovation surveys are designed.

   Abs DOI PDF Supp Code

   Does providing information about a wide range of university engagement channels influence the likelihood that companies report that they source knowledge from universities in their innovative efforts? To address this question, we embedded a field experiment in a multi-wave survey on university engagement that randomly exposed firms to information about a variety of engagement channels before asking about knowledge sources for innovation. We find that exposure to these channels significantly increases the likelihood of firms reporting universities as a knowledge source. The magnitude of this effect is substantial, with the propensity to report sourcing knowledge from universities increasing by 34% in our preferred estimate. This increase is concentrated among firms rating universities as of ‘low’ or ‘medium’ importance and among those firms with some university engagement. We discuss the implications for understanding university-industry engagement and the design of innovation surveys.

4. EI

   Mitigating Technology Gaps’ Contribution to International Income Inequality

   with Franco Mariuzzo and Junjun Zhang

   Economic Inquiry, 2026

   Revisits and extends a 2023 study on how gaps in countries’ capacity to innovate fuel global income inequality. The link holds up, but how strong it looks depends on which countries are compared and how innovation is measured.

   Abs DOI Code

   This replication study revisits a 2023 analysis by Sampson on the impact of international technology gaps on income inequality, trade, and wages. We extend the original framework by using broader patent metrics from PATSTAT, expanding the country sample and time frame, and testing alternative calibrations. Our results confirm that cross-country differences in innovation efficiency contribute significantly to wage and income disparities. However, the strength of this relationship varies with country composition and the choice of innovation measure. Broader patent indicators yield weaker inequality effects, highlighting the importance of measurement choices in assessing technology’s role in global income distribution.

2025

1. WP

   The Private Costs of Legal Uncertainty: Evidence from the Unified Patent Court

   2025

   Working paper

   Estimates the economic cost of legal uncertainty using the introduction of the EU’s Unified Patent Court. Rather than accept an untested jurisdiction, many patentees chose the established national route, forgoing savings of roughly EUR 100 million, or about EUR 10,000 per patent.

   Abs Poster Slides

   "Legal certainty" signifies the certainty of economic agents about the meaning and enforceability of legal rules and the procedures for resolving legal disputes. It has been considered a cornerstone of market economies for centuries, yet it remains challenging to quantify. This paper leverages the introduction of the Unitary Patent (UP) system in the European Union as an exogenous shock that exposed patentees to uncertainty about the new Unified Patent Court (UPC). While the reform aims to reduce the cost of maintaining patents across multiple European countries, the established system of national validations (EP) remains available. Patentees can thus "pay" to avoid UPC jurisdiction by opting for national validations. We calculate that patentees who retained the national route despite possible savings forfeited approximately EUR 100 million, while those who accepted simultaneous UPC jurisdiction left EUR 65 million in savings "on the table". On a per-patent basis, this equates to around EUR 10,000 per grant and EUR 25,000 per patentee. For patentees with a long history of maintaining European patents, I estimate that the average patentee requires fee savings of EUR 4,000-6,000 per patent to be indifferent between the UP and EP routes. I disaggregate part of these costs of the UPC route into expectations about litigation costs, litigation outcomes, and the importance of patent licensing, and find considerable heterogeneity in the remaining baseline cost between patentees from different UP member countries.

2. WP

   Competition and the Strategic Disclosure of Innovation: Theory and Evidence from Patent Applications

   with Bernhard Ganglmair

   2025

   CRC TR 224 Discussion Paper No. 664

   A firm with a pending patent application must decide whether to disclose it early. Doing so reveals information that shapes how rivals respond, both in how aggressively they compete in the product market and in whether they scale back or step up their own R&D. Whether it pays off depends on the competitive setting, with evidence from patent disclosures in company press releases.

   Abs Preprint Slides

   We develop a duopoly signaling model in which an innovative leader strategically announces a pending patent application to influence the follower’s behavior, both in the product market (in the short run) and in R&D (with effects in the long run). Our model captures different competitive structures by representing market competition in reduced form, thereby encompassing competition in both strategic substitutes and strategic complements with varying intensities. Extending the conventional wisdom of the optimality of disclosure under strategic substitutes, our model predicts some disclosure for low competition intensities under strategic complements when the follower’s R&D project is technologically less similar to the leader’s, and no disclosure even under strategic substitutes when competitors’ R&D exhibits high similarities. We provide empirical support for the model’s core predictions by identifying patent disclosures in press releases and using a technique from the corporate finance literature to measure the nature of market competition.

3. JEL

   The Sources of Researcher Variation in Economics

   with Nick Huntington-Klein, Claus C. Pörtner, Ian McCarthy, and 146 others

   Journal of Economic Literature (revise and resubmit), 2025

   146 research teams independently answer the same economics question with the same data. Their headline estimates vary surprisingly little, but data cleaning, more than modelling, explains the differences that remain. Contributing author.

   Abs SSRN NBER

   We have 146 research teams of economists independently use the same data source to answer the same question about the causal effect of a policy. Each team performs the task three times, first with free choice of how to answer the question, second with all researchers required to use a shared research design, and third with pre-cleaned data and a shared research design. We find considerable variation across researchers in reported sample sizes, sample definitions, and modeling choices, although estimated effects show muted variation. Levels of researcher variation were most heavily influenced by data preparation and cleaning, with much smaller roles for researcher background, modeling and estimation decisions, and exposure to peer review. This implies that data preparation and cleaning should receive considerably more attention in researcher training and the evaluation of research.

2022

1. WP

   The Impact of COVID-19 on Industry Interactions with Universities: The Case of the UK

   with Alan Hughes, Michael Kitson, and Ammon Salter

   2022

   Working paper (under revision)

   COVID-19 disrupted the ties between firms and universities unevenly. Companies with the broadest university networks were hit hardest, while those with prior collaboration experience and larger patent portfolios proved more resilient.

   Abs

   This paper explores the impact of the COVID-19 pandemic on the interactions between firms and universities. Although it might be expected that restrictions during COVID-19 made interactions between business firms and universities more challenging, there has been little evidence on this relationship. Drawing on a large-scale survey carried out in the UK during the pandemic which focused on pre-pandemic business interactions with universities matched to secondary data, we explore what factors shape whether firms’ interactions with universities were negatively affected by the crisis. We find that firms with broad interactions and with different university partners were more negatively impacted, whereas those firms with pre-existing university collaboration experience and a larger stock of patents were less negatively affected. We explore the implications of these findings for policies to support business-university interactions in the post-pandemic era.

2026

1. Chapter

   Capabilities That Attract Frictions: Young Firms and Startup Policy in Central Asia

   In National Innovation Systems in Central Asia, 2026

   Book chapter (under review)

   Draws on firm surveys from four Central Asian economies to show that young firms are held back less by red tape than by weak capabilities, and that building those capabilities can paradoxically attract more red tape, except where liberalisation (as in Uzbekistan) has loosened the link.

   Abs

   Central Asian governments have introduced a range of startup policies, but their impact on young firms remains unclear. This chapter uses World Bank Enterprise Survey microdata from Kazakhstan, the Kyrgyz Republic, Tajikistan, and Uzbekistan (2008–2024) to assess young firms’ capability development and the evolution of entry frictions. Findings show that Central Asian firms face friction levels similar to peers in other post-Soviet and post-socialist economies but have fewer capabilities, indicating that capability constraints are the main barrier. Firm-level regressions reveal that capable young firms tend to operate in high-friction environments, consistent with institutional-voids theory. However, this effect disappears when frictions are measured at the industry level, which instead suggests an attract-frictions mechanism: activities such as certification, exporting, and credit applications increase firms’ exposure to inspectors and rent-seeking, thereby raising friction levels. This link between capabilities and frictions discourages investment in productive capabilities, especially among young firms. Uzbekistan represents an exception; its post-2016 liberalisation appears to have weakened this link, enabling broader capability accumulation and yielding the most favourable young-firm outcomes in the region. These results suggest that capability-support interventions will be more effective if paired with measures that reduce regulatory exposure for firms investing in capabilities.

2024

1. I4R

   A Comment on Sampson (2023)

   with Farasat Bokhari, Franco Mariuzzo, and Junjun Zhang

   I4R Discussion Paper Series, 2024

   Reproduces Sampson (2023) and stress-tests it. The headline claim that technology gaps drive inequality survives, but softens once the choice of developing countries and innovation measure is examined closely.

   Abs HTML

   In their paper, Sampson (2023) introduces a theoretical framework and conducts empirical testing to elucidate the impact of gaps in countries’ innovative efficiencies on income, wages, and trade dynamics. We successfully replicate the paper’s findings by running the provided codes, and confirm the absence of any coding errors in the process. We also provide an extensive battery of robustness checks, which confirms the resilience of their results. We then scrutinize two key aspects of their study: the choice of developing countries and the innovation measure employed. The outcomes of this refined analysis partly temper the original paper’s message of technology gaps driving inequality, underscoring the need for additional research in this domain.

2022

1. Report

   The Changing State of Business-University Interactions in the UK

   with Alan Hughes, Michael Kitson, Ammon Salter, and Robert Hughes

   2022

   A national survey of 3,823 UK companies mapping how, and how much, they exchange knowledge with universities, with comparisons stretching back to 2005.

   Abs PDF Code Poster Website

   This report presents the results of an online survey of UK companies which took place in 2020-2021. The survey was designed to set out and assess the extent and nature of company knowledge exchange interactions with the university sector. It covers the three-year period ending in March 2020 prior to the Covid-19 pandemic. The survey also includes questions relating to the subsequent impact of the pandemic on knowledge exchange patterns. The survey covers 3823 companies in all sectors, regions and countries of the UK and the full range of employment sizes from micro-firms employing less than 10 people to the largest public listed corporations. We use a matched sample of respondents to this survey and to a previous survey to examine changes over the period 2005-2021.