Tomas Havranek

Tomas Havranek

Professor of Economics
Charles University, Prague

I am an economist who uses meta-analysis to find out how research conclusions change once we correct for selective reporting and biases in study design.

How our meta-analyses revised the published evidence

How our meta-analyses revised the published evidence 44 meta-analyses. Revision in absolute magnitude: 100 × (|corrected| / |reported| − 1). 31 effects shrank toward zero, 11 grew away from zero, and two are plotted as unchanged. Median −41%. Three off-scale points: price puzzle +394%, structural reforms +631%, Czech FDI productivity +826%. FDI spillover determinants and habit formation are at zero by the source's rule. Each point links to its paper. Horizontal position encodes revision; vertical stacking has no meaning. Snapshot: 25 September 2026. shrank grew unchanged no change median −41% −100% −50% 0% +50% +100% +150% Change from the average reported estimate (%) intertemporal substitution −97% beauty premium −40% finance and growth +33% skill substitution +122% intertemporal substitution: reported 0.5, corrected 0.0145 (−97%). Whole-literature corrected mean; the 0.3–0.4 on its paper page is for asset holders. daylight saving time: reported −0.34, corrected −0.01 (−97%) risk aversion: reported 7.5, corrected 1 (−87%) Approximate comparison; see the linked paper for details. tuition and enrollment: reported −0.19, corrected −0.037 (−81%) Approximate comparison; see the linked paper for details. euro and trade: reported 0.038, corrected 0.00899 (−76%) Approximate comparison; see the linked paper for details. excess sensitivity: reported 0.37, corrected 0.11 (−70%) social cost of carbon: reported 411, corrected 134 (−67%) Approximate comparison; see the linked paper for details. capital-labor substitution: reported 0.9, corrected 0.3 (−67%) resource curse: reported −0.078, corrected 0.026 (−67%). The correction reversed the sign. size premium: reported 5.08, corrected 1.72 (−66%) gasoline income elasticity: reported 0.28, corrected 0.1 (−64%) Approximate comparison; see the linked paper for details. remittances and growth: reported 0.103, corrected 0.039 (−62%) Approximate comparison; see the linked paper for details. correlation meta-bias: reported 0.128, corrected 0.05 (−61%) Approximate comparison; see the linked paper for details. discount rates: reported 0.8, corrected 0.33 (−59%) gender diversity ESG: reported 0.28, corrected 0.125 (−55%) Approximate comparison; see the linked paper for details. gasoline demand: reported −0.691, corrected −0.31 (−55%) Approximate comparison; see the linked paper for details. exercise and cognition: reported 0.451, corrected 0.208 (−54%) Approximate comparison; see the linked paper for details. climate sensitivity: reported 3.27, corrected 1.6 (−51%) Approximate comparison; see the linked paper for details. Frisch elasticity: reported 0.5, corrected 0.25 (−50%) shareholder activism: reported 1.49, corrected 0.75 (−50%) Approximate comparison; see the linked paper for details. border effect: reported 3.03, corrected 1.76 (−42%) water income elasticity: reported 0.256, corrected 0.15 (−42%) Approximate comparison; see the linked paper for details. electricity demand: reported −0.267, corrected −0.16 (−40%) Approximate comparison; see the linked paper for details. beauty premium: reported 5, corrected 3 (−40%). Conservative end of the paper's range; the 1.1% on its paper page is a different quantity. financial incentives: reported 0.051, corrected 0.031 (−39%) Approximate comparison; see the linked paper for details. optimal inflation: reported 0.74, corrected 0.61 (−18%) Approximate comparison; see the linked paper for details. class size: reported −0.25, corrected −0.21 (−16%) Approximate comparison; see the linked paper for details. policy transmission lag: reported 33.5, corrected 29.2 (−13%) hedge fund alpha: reported 0.36, corrected 0.335 (−7%) Approximate comparison; see the linked paper for details. COVID learning loss: reported −0.127, corrected −0.12 (−5%) house prices: reported −1.2, corrected −1.154 (−4%) Approximate comparison; see the linked paper for details. FDI spillover determinants: reported −0.002, corrected 0.021 (0%). No publication bias worth correcting; plotted at zero. habit formation: reported 0.4, corrected 0.4 (0%). No publication bias worth correcting; plotted at zero. forward premium puzzle: reported −0.602, corrected 0.611 (+1%). The correction reversed the sign. vertical FDI spillovers: reported 0.88, corrected 0.94 (+7%) Approximate comparison; see the linked paper for details. finance and growth: reported 0.15, corrected 0.199 (+33%) Approximate comparison; see the linked paper for details. native-immigrant substitution: reported 13, corrected 22 (+69%) bank competition: reported −0.012, corrected 0.022 (+83%). The correction reversed the sign. skill substitution: reported 1.8, corrected 4 (+122%) working while studying: reported −0.017, corrected −0.038 (+124%) Approximate comparison; see the linked paper for details. Armington elasticity: reported 1.56, corrected 3.8 (+144%) price puzzle: reported −0.067, corrected −0.331 (+394%). Horizon-dependent comparison. +394% structural reforms: reported −0.052, corrected −0.38 (+631%). Small comparator base. +631% Czech FDI productivity: reported −0.119, corrected 1.1 (+826%). Partly a projection to 2018. The correction reversed the sign. +826% How our meta-analyses revised the published evidence 44 meta-analyses. Revision in absolute magnitude: 100 × (|corrected| / |reported| − 1). 31 effects shrank toward zero, 11 grew away from zero, and two are plotted as unchanged. Median −41%. Three off-scale points: price puzzle +394%, structural reforms +631%, Czech FDI productivity +826%. FDI spillover determinants and habit formation are at zero by the source's rule. Each point links to its paper. Revision increases from top to bottom; horizontal stacking has no meaning. Snapshot: 25 September 2026. shrank grew unchanged Change from the average reported estimate (%) no change median −41% −100% −50% 0% +50% +100% +150% intertemporal substitution −97% beauty premium −40% finance and growth +33% skill substitution +122% intertemporal substitution: reported 0.5, corrected 0.0145 (−97%). Whole-literature corrected mean; the 0.3–0.4 on its paper page is for asset holders. daylight saving time: reported −0.34, corrected −0.01 (−97%) risk aversion: reported 7.5, corrected 1 (−87%) Approximate comparison; see the linked paper for details. tuition and enrollment: reported −0.19, corrected −0.037 (−81%) Approximate comparison; see the linked paper for details. euro and trade: reported 0.038, corrected 0.00899 (−76%) Approximate comparison; see the linked paper for details. excess sensitivity: reported 0.37, corrected 0.11 (−70%) social cost of carbon: reported 411, corrected 134 (−67%) Approximate comparison; see the linked paper for details. capital-labor substitution: reported 0.9, corrected 0.3 (−67%) resource curse: reported −0.078, corrected 0.026 (−67%). The correction reversed the sign. size premium: reported 5.08, corrected 1.72 (−66%) gasoline income elasticity: reported 0.28, corrected 0.1 (−64%) Approximate comparison; see the linked paper for details. remittances and growth: reported 0.103, corrected 0.039 (−62%) Approximate comparison; see the linked paper for details. correlation meta-bias: reported 0.128, corrected 0.05 (−61%) Approximate comparison; see the linked paper for details. discount rates: reported 0.8, corrected 0.33 (−59%) gender diversity ESG: reported 0.28, corrected 0.125 (−55%) Approximate comparison; see the linked paper for details. gasoline demand: reported −0.691, corrected −0.31 (−55%) Approximate comparison; see the linked paper for details. exercise and cognition: reported 0.451, corrected 0.208 (−54%) Approximate comparison; see the linked paper for details. climate sensitivity: reported 3.27, corrected 1.6 (−51%) Approximate comparison; see the linked paper for details. Frisch elasticity: reported 0.5, corrected 0.25 (−50%) shareholder activism: reported 1.49, corrected 0.75 (−50%) Approximate comparison; see the linked paper for details. border effect: reported 3.03, corrected 1.76 (−42%) water income elasticity: reported 0.256, corrected 0.15 (−42%) Approximate comparison; see the linked paper for details. electricity demand: reported −0.267, corrected −0.16 (−40%) Approximate comparison; see the linked paper for details. beauty premium: reported 5, corrected 3 (−40%). Conservative end of the paper's range; the 1.1% on its paper page is a different quantity. financial incentives: reported 0.051, corrected 0.031 (−39%) Approximate comparison; see the linked paper for details. optimal inflation: reported 0.74, corrected 0.61 (−18%) Approximate comparison; see the linked paper for details. class size: reported −0.25, corrected −0.21 (−16%) Approximate comparison; see the linked paper for details. policy transmission lag: reported 33.5, corrected 29.2 (−13%) hedge fund alpha: reported 0.36, corrected 0.335 (−7%) Approximate comparison; see the linked paper for details. COVID learning loss: reported −0.127, corrected −0.12 (−5%) house prices: reported −1.2, corrected −1.154 (−4%) Approximate comparison; see the linked paper for details. FDI spillover determinants: reported −0.002, corrected 0.021 (0%). No publication bias worth correcting; plotted at zero. habit formation: reported 0.4, corrected 0.4 (0%). No publication bias worth correcting; plotted at zero. forward premium puzzle: reported −0.602, corrected 0.611 (+1%). The correction reversed the sign. vertical FDI spillovers: reported 0.88, corrected 0.94 (+7%) Approximate comparison; see the linked paper for details. finance and growth: reported 0.15, corrected 0.199 (+33%) Approximate comparison; see the linked paper for details. native-immigrant substitution: reported 13, corrected 22 (+69%) bank competition: reported −0.012, corrected 0.022 (+83%). The correction reversed the sign. skill substitution: reported 1.8, corrected 4 (+122%) working while studying: reported −0.017, corrected −0.038 (+124%) Approximate comparison; see the linked paper for details. Armington elasticity: reported 1.56, corrected 3.8 (+144%) price puzzle: reported −0.067, corrected −0.331 (+394%). Horizon-dependent comparison. +394% structural reforms: reported −0.052, corrected −0.38 (+631%). Small comparator base. +631% Czech FDI productivity: reported −0.119, corrected 1.1 (+826%). Partly a projection to 2018. The correction reversed the sign. +826%
Each dot is one meta-analysis. Most effects shrink (median −41%); some grow. All results and caveats: meta-analysis.cz/results.

Selected research

More research

If you only have time for two of my papers, try the ones on spurious precision and class size.

  1. Financial Incentives and Performance: A Meta-Analysis of Experiments in Economics2026 · with Petr Cala, Zuzana Irsova, Martina Luskova, Jindrich Matousek, and Jiri NovakJournal of Political Economy Microeconomics, forthcoming.
  2. Publication Bias and Model Uncertainty in Measuring the Effect of Class Size on Achievement2026 · with Matej Opatrny, Zuzana Irsova, and Milan ScasnyJournal of Labor Economics, forthcoming.
  3. Reproducibility and Robustness of Economics and Political Science Research2026 · with Abel Brodeur and others (one of 347 authors)Nature 652, 151–156.
  4. Spurious Precision in Meta-Analysis of Observational Research2025 · with Zuzana Irsova, Pedro Bom, and Heiko RachingerNature Communications 16, 8454.
  5. Publication and Attenuation Biases in Measuring Skill Substitution2024 · with Zuzana Irsova, Lubica Laslopova, and Olesia ZeynalovaReview of Economics and Statistics 106(5), 1187–1200.
  6. Measuring Intertemporal Substitution: The Importance of Method Choices and Selective Reporting2015Journal of the European Economic Association 13(6), 1180–1204.

Papers, data and code are at meta-analysis.cz.
Our meta-analysis app is at easymeta.org.

About

I am Professor of Economics at the Institute of Economic Studies, Charles University, and an Associate Editor for meta-analysis at the Journal of Economic Surveys. I also teach at the Faculty of International Relations, Prague University of Economics and Business, and am a Research Affiliate of CEPR and of METRICS, the Meta-Research Innovation Center at Stanford.