How this work has been used

The papers collected here have been cited more than 10,000 times according to Google Scholar. They and the data behind them have been used in academic research, in policy analysis at national and international institutions, in central bank publications, and in teaching. Selected examples follow, with links to the documents.

Independent use of the data

David Slichter and Nhan Tran use eleven datasets from this site in a study of whether prestigious journals publish more accurate estimates. Those eleven supply 502 of the 871 primary studies in their 24-literature sample. They find almost no relationship between journal rank and accuracy: the chance that an estimate from a top journal is closer to the truth than one from a mid-ranked journal is about 51%. The paper is accepted at the Journal of Political Economy Microeconomics; the current version is IZA Discussion Paper 17960.

Dimitrije Ruzic reanalyses the capital-labor substitution dataset at the estimate level, 3,186 estimates from 121 studies, using its coding of how each estimate was identified. Estimates from the first-order condition for labor average 0.85, those from the condition for capital 0.45. He argues the gap is real, because external inputs substitute more readily for labor than for capital, and confirms it in the US integrated production accounts for 1963 to 2016. Where the meta-analysis treats omission of the capital condition as a source of upward bias, he argues the two conditions identify different responses once external inputs are allowed (Review of Economics and Statistics, 2026).

Samangi Bandaranayake, Kuntal Das, and Robert Reed reproduce the bank competition meta-analysis from its data and code, recode every study from scratch, extend it to 35 studies it did not include, and re-run it under a different estimator. All four confirm the main finding, that competition has an economically negligible effect on financial stability. The paper’s other findings are, in their words, “less supportive”: meta-regression is sensitive to how the data are coded and to the choice of estimator (Journal of Economic Surveys, 2020).

Christopher Carroll, Edmund Crawley, Jiri Slacalek, Kiichi Tokuoka, and Matthew White open Sticky Expectations and Consumption Dynamics (American Economic Journal: Macroeconomics, 2020) with the habit formation meta-analysis’s Figure 2, reproduced with attribution, and their replication archive ships its data file unchanged. The figure shows the gap the meta-analysis found, macro studies near 0.6 and micro studies near zero, which their conclusion calls “arguably the most important puzzle in the microfoundations of aggregate consumption dynamics”.

Witold Wiecek includes the exercise and cognition dataset, 2,239 estimates from 215 meta-analyses, in BEAR (Benchmarks of Empirical Accuracy in Research), an open-source database that brings more than twenty metascience datasets from medicine, psychology, economics, and other fields into one common format. BEAR grew out of the data assembled for A statistical case for qualified scientific optimism by Erik van Zwet, Andrew Gelman, and Wiecek; that paper analyses sixteen of the datasets.

The student employment dataset ships with RoBMA, the Bayesian model-averaging package for meta-analysis on CRAN, as one of its worked examples: 861 estimates from 69 studies, with meta-analysis.cz given as the source.

Academic use

The papers are cited for different reasons: a benchmark value, a bias-corrected estimate, evidence on why estimates differ from study to study, or a method for detecting publication bias. Thomas Sargent and John Stachurski cite the cross-country meta-analysis of intertemporal substitution in Dynamic Programming (Cambridge University Press) for “0.5 as a plausible average value for international studies, with rich countries tending slightly higher”. Simon Johnson, Lukasz Rachel, and Catherine Wolfram calibrate to the same literature in A Theory of Price Caps on Nonrenewable Resources, American Economic Review 116(7): 2711-2753, calling it “the mean value in the influential meta-study of Havranek, Horvath, Irsova, and Rusnak (2015)”. Both take the average of published estimates. Others need the average corrected for selective reporting. Michael Best, James Cloyne, Ethan Ilzetzki, and Henrik Kleven, estimating the same elasticity from mortgage notches in the Review of Economic Studies, take it from the companion paper, which “conducts a meta analysis of the existing literature and finds estimates centered around 0.3-0.4, after controlling for publication bias”.

Tomohiro Hirano and Joseph Stiglitz cite that paper, the companion paper, and the capital-labor substitution meta-analysis in Overlapping generations models, multiplicity of steady states and momentary equilibria, and economic fluctuations, Oxford Review of Economic Policy. Michael Kremer, Gautam Rao, and Frank Schilbach cite the publication-bias paper in Behavioral development economics, Handbook of Behavioral Economics. David Card, Jochen Kluve, and Andrea Weber cite it in their own meta-analysis of active labor market programs in the Journal of the European Economic Association, as a contrast to the funnel asymmetry they do not find; Card’s Berkeley graduate course reproduced its funnel plot in its 2016 lecture on intertemporal labor supply, Lecture 5 of Economics 250A. Five of these authors are Nobel laureates: Sargent, Johnson, Stiglitz, Kremer, and Card.

Papers on what artificial intelligence will do to growth need a value for the elasticity of substitution between capital and labor, because values above one would let capital replace labor without limit. The capital-labor meta-analysis puts it well below one. Charles Jones and Christopher Tonetti take it as their upper bound in Past Automation and Future A.I.: “the meta analysis of Gechert et al. (2022) finds a mean estimate of 0.3 … We therefore choose σ = 0.2, smaller than all these estimates.” Philip Trammell and Anton Korinek open Economic Growth under Transformative AI with the same restriction: “today, labor and capital are at least weakly gross complements: their elasticity of substitution is not greater than one.” Tomohiro Hirano and Alexis Akira Toda cite it for the same reason in Bubble Necessity Theorem, Journal of Political Economy 133(1).

Eric Leeper, Nora Traum, and Todd Walker check their own estimates against the habit-formation meta-analysis in Clearing Up the Fiscal Multiplier Morass, also in the American Economic Review: their habit parameters are high, but “within the 90 percent bands for external habits that Havranek, Rusnak, and Sokolova’s (2017) meta study reports”. Diego Comin, Danial Lashkari, and Marti Mestieri calibrate the elasticity of intertemporal substitution to 0.5 in Econometrica, citing the publication-bias paper. Alonso Alfaro-Urena, Isabela Manelici, and Jose Vasquez cite the vertical spillovers meta-analysis in The Effects of Joining Multinational Supply Chains, Quarterly Journal of Economics, on why estimated spillovers vary so widely.

The work is cited outside economics as well. Science cites the gasoline price elasticity in Tracking the global footprint of fisheries, as one of the comparisons for the fuel price elasticity it estimates for the world fishing fleet. PNAS cites the social cost of carbon meta-analysis in a synthesis of what that cost is, the publication-bias paper in Declining CO2 price paths, and the borders meta-analysis in a study of how legible international borders are. Other papers here are cited in Nature Climate Change, Nature Energy, Nature Communications, Nature Water, Nature Human Behaviour, and Nature Reviews Psychology. In Nature Ecology & Evolution, a paper on exaggeration bias and selective reporting in ecology cites the resource curse meta-analysis as a precedent for the diagnosis. In Nature Water, an account of why the cost of drought falls unevenly across households takes the income elasticity of residential water demand from the water demand meta-analysis: “water use increases roughly 1% for every 10% increase in household income”.

In policy institutions

SAGE, the computable general equilibrium model the US Environmental Protection Agency uses to analyse regulation, calibrates its intertemporal-substitution parameter from the cross-country meta-analysis. The elasticity in that model is 1/η, so the documentation takes the meta-analytic mean for the United States and inverts it: “in a recent review of over 1,400 estimates of the elasticity of intertemporal substitution for the United States, Havranek et al. (2015) find a mean value of 0.6. Based on this evidence, we set the value of η to 1.66” (SAGE Model Documentation 2.1.1, 2024).

The Congressional Research Service uses the corrected figure rather than the average. Reviewing whether tax policy can raise saving, it reports the meta-analytic mean of 0.6 for the United States, then: “Havranek, a coauthor of this meta-analysis, subsequently published the basic (worldwide) results after correcting for estimated publication bias. The correction indicates that the elasticity for macro aggregate studies is zero (as Hall originally found). In the basic case (without selecting across studies for other characteristics), the elasticity for micro studies (which were about a quarter of the studies) was 0.2. This study suggests that an elasticity of zero to 0.2 might be in order” (Can Tax Policy Increase Saving?, 2024). Its review of the models used for dynamic scoring does the same for the Frisch elasticity of labor supply, reporting the uncorrected 0.5 and then the zero that survives correction for publication bias and identification (Dynamic Scoring for Tax Legislation, 2025).

The Congressional Research Service has cited the daylight saving meta-analysis in the report it prepares for the US Congress in every version since 2018, most recently the July 2026 update. The European Parliamentary Research Service used the same paper in the ex-post impact assessment of the EU summer-time directive written for the European Parliament’s Legal Affairs Committee, calling it “a seemingly methodologically sound literature review of the available research”. The House of Commons Library cites the same paper first in the energy section of the briefing it prepares for the UK Parliament: “In 2018, a large meta-analysis (a study combining the results of previous studies) found that daylight saving was associated with an average 0.34% reduction in electricity use, and that energy savings were larger in countries further from the equator. Excluding lower-quality studies from the analysis reduced energy savings to near zero” (British Summer Time, 2026). Spain took the same result to the Council. Asking it in October 2025 to resume negotiations on abolishing seasonal clock changes, the note Spain put to the Council of the European Union’s Energy Council says: “a 2018 meta-analysis of 44 studies confirms that the overall energy impact of DST is effectively zero” (Council document 14195/25).

The flagship reports use the results too. The IMF’s World Economic Outlook of April 2023 uses the meta-analysis of transmission lags in its box on the speed of monetary transmission, and as the source of one of its figures: “a meta-analysis of 67 published studies covering 30 different economies (Havranek and Rusnak 2013) finds that the effect of a tightening on prices takes an average of about three years to reach its trough”. The Bank of England cites the same paper in its Monetary Policy Report of May 2024. The IMF’s Fiscal Monitor of April 2016 cites the horizontal spillovers meta-analysis on where productivity spillovers from foreign investment actually appear, and the World Bank’s Making Global Value Chains Work for Development returns to the vertical spillovers meta-analysis in three separate chapters. The European Commission’s review of structural reforms in Spain summarises the borders meta-analysis in detail, down to the 33% mean reduction in international trade it reports. The BIS Annual Economic Report 2024, on central bank financial results, cites Hampl and Havranek’s review of central bank financial strength and inflation for finding “no systematic evidence that central bank financial strength affects inflation outcomes”.

Each document below was published by the institution named and cites a paper from this site.

Some institutional reviews follow the methods rather than the results. An IMF meta-analysis of macroprudential policy across more than 6,000 estimates cites four of these papers and says its “paper selection approach is similar to Havranek and Sokolova (2020) and Havranek and Irsova (2011)” (IMF Working Paper 20/67). A World Bank policy research working paper on agglomeration economies runs its publication-bias tests “following Havranek, 2015” and averages models the way the habit-formation meta-analysis does (Policy Research Working Paper 9730). A European Commission Joint Research Centre report on how farm subsidies capitalise into farmland prices runs its own meta-analysis, of 841 estimates from 26 studies, under the reporting standards written here. Its method section opens: “In line with Havránek et al. (2020), we systematically review the empirical literature of the past three decades on the capitalisation of agricultural subsidies into land prices” (JRC126423). The OECD’s 2025 review of what climate policies do to emissions lists the gasoline meta-analysis among the syntheses of how fuel use responds to fuel prices (IFCMA Papers), and OECD authors cite both spillovers meta-analyses in the International Productivity Monitor.

Central banks

John Williams, then President of the Federal Reserve Bank of San Francisco, cited the meta-analysis of transmission lags in nine speeches in 2015 and 2016, always on the same point: “Milton Friedman famously taught us that monetary policy has long and variable lags. Research shows it takes at least a year or two for it to have its full effect” (Dancing Days Are Here Again, 2015, is one of them). Mary Daly, his successor, has cited it in three more, most recently at the European Banking Congress in 2023.

Each document below was published by the bank named.

From 2015 to 2019 Havranek was full-time Advisor to the Board of the Czech National Bank, and the opinions he wrote for the Board argue from meta-analytic averages. The opinion on Situation Report 6/2017 makes the case for a steeper path of rate increases partly on the ground that “the average of published estimates of transmission in the Czech Republic suggests that a change in rates has its greatest effect on prices after 15 months”, citing the meta-analysis of transmission lags. The Bank has since released those opinions, and they are here in English translation beside the Czech originals it published: 2/2016, 6/2016, 6/2017, and 8/2018.

Teaching and software

Havranek taught a course on research synthesis at the University of Canterbury as an Erskine Fellow in 2025, taught a two-day course for policy economists at the Swedish Agency for Growth Policy Analysis in Stockholm in August 2025 (slides, code and data), and gave a three-hour course on meta-analysis at the University of Osaka in 2026. The written form of the same material is the practitioner’s guide and the MAER-Net reporting guidelines. MAIVE, the estimator that reduces the bias arising when reported precision is spurious, is on CRAN and runs in a browser at EasyMeta.org.

Tomas Havranek and Zuzana Irsova maintain the site at the Institute of Economic Studies, Charles University, Prague. Corrections are welcome: tomas.havranek@fsv.cuni.cz and zuzana.irsova@fsv.cuni.cz.