Headline results

One answer per paper: the headline result of each of the 54 papers below, with the caveat, the evidence and the citation that belong to it. Where the paper reports a publication-bias-corrected number, that is the one shown.

What correction and best practice did to each number One dot per meta-analysis, placed by how far its corrected or best-practice estimate sits from the average estimate that literature reported. 21 of the 25 moved toward zero and 4 away from it; the median revision is -51%. -100% -50% +0% +50% +100% +150% no change median -51% intertemporal substitution. Reported estimates average 0.5; corrected or best practice 0.0145. -97%. The site's headline, 0.3-0.4, is the corrected value for a narrower subsample: micro estimates for asset holders. What is plotted is Table 1's corrected mean for the WHOLE literature, which is the quantity this figure measures and which the paper reports as essentially zero. The two ends also treat the tails differently: the mean of 0.5 excludes estimates above 10 in absolute value, while the regression keeps all 2,735 because precision weighting already gives the extremes little weight. daylight saving time. Reported estimates average -0.34; corrected or best practice -0.01. -97%. risk aversion. Reported estimates average 7.5; corrected or best practice 1. -87%. The paper gives economics and finance separately. This is economics -- 590 estimates from 58 studies, the value its abstract leads with -- against the paper's own uncorrected mean for the same subsample. Finance moves further: 1-3 corrected against 45 uncorrected. tuition and enrollment. Reported estimates average -0.19; corrected or best practice -0.037. -81%. The paper's headline answer is verbal - the elasticity is 'close to zero' - so the number here is its best-practice estimate from Table 6 rather than a figure it puts in the abstract. The Table 6 figure is a synthetic study: publication characteristics are set at their sample maxima, the standard error and OLS at their minima, and the endogeneity control at its maximum. excess sensitivity. Reported estimates average 0.37; corrected or best practice 0.11. -70%. social cost of carbon. Reported estimates average 411; corrected or best practice 134. -67%. 134 is the LARGEST of the paper's corrected means, and 411 the uncorrected mean of the same estimates -- the pair the paper itself puts in one sentence. Taking the largest corrected value understates the correction, which is the conservative direction. capital-labor substitution. Reported estimates average 0.9; corrected or best practice 0.3. -67%. size premium. Reported estimates average 5.08; corrected or best practice 1.72. -66%. gasoline income elasticity. Reported estimates average 0.28; corrected or best practice 0.1. -64%. The paper gives two horizons. This is the short run, the one its abstract leads with, and the only one whose corrected value and comparator come from the same whole sample. discount rates. Reported estimates average 0.8; corrected or best practice 0.33. -59%. gender diversity ESG. Reported estimates average 0.28; corrected or best practice 0.125. -55%. plotted at 0.125, the midpoint of the 0.08 to 0.17 the paper reports for the whole sample; both ends are its own figures. The 0.12 the page quotes is the best-practice estimate for most of the world, and the paper reports 0.276 for the Middle East and about zero for Southeast Asia, so it does not cover the same firms as an average over every estimate. gasoline demand. Reported estimates average -0.691; corrected or best practice -0.31. -55%. The paper gives two horizons. This is the long run, the one its abstract names first, against the paper's own simple mean for the same horizon; the short run moves the same way. climate sensitivity. Reported estimates average 3.27; corrected or best practice 1.6. -51%. The site's headline for this paper is a range, 1.4-2.3 C, which is the span across specifications; the figure uses the paper's own stated best estimate of 1.6 C. Frisch elasticity. Reported estimates average 0.5; corrected or best practice 0.25. -50%. shareholder activism. Reported estimates average 1.49; corrected or best practice 0.75. -50%. The paper gives a corrected range, 0% to 1.5%, and names no central value, so the midpoint is plotted. The ten estimators behind that range average 0.60%, so the midpoint understates the correction rather than overstating it. border effect. Reported estimates average 3.03; corrected or best practice 1.76. -42%. water income elasticity. Reported estimates average 0.256; corrected or best practice 0.15. -42%. The paper says the elasticity is 'approximately 0.15 or less', an upper bound rather than a point, so 0.15 understates the correction. electricity demand. Reported estimates average -0.267; corrected or best practice -0.16. -40%. The paper gives two horizons. This is the short run, the value its abstract leads with, against the paper's own raw mean for the same row; the long run moves the same way. beauty premium. Reported estimates average 5; corrected or best practice 3. -40%. both ends are the paper's own approximate figures, and 3% is the conservative end: it says other techniques cut the premium further. The 1.1% the page quotes is a different quantity -- it controls for cognitive ability and reweights occupations towards US employment shares, and the paper says its subgroups do not average to the overall mean. hedge fund alpha. Reported estimates average 0.36; corrected or best practice 0.335. -7%. The paper's headline is a 30-40 basis point range; plotted here is the mean of its own twelve corrected estimates, 0.335%, which it states in Table 7, against its own unconditional sample mean of 0.36%. COVID learning loss. Reported estimates average -0.127; corrected or best practice -0.12. -5%. finance and growth. Reported estimates average 0.15; corrected or best practice 0.199. +33%. The paper's headline is verbal — an 'authentic positive link' — so the number here is the FAT-PET effect from its Table 2 rather than a figure it puts in the abstract. native-immigrant substitution. Reported estimates average 13; corrected or best practice 22. +69%. skill substitution. Reported estimates average 1.8; corrected or best practice 4. +122%. Armington elasticity. Reported estimates average 1.56; corrected or best practice 3.8. +144%. correction shrinks the effect correction enlarges it
What correction and best practice did to the number. One dot per meta-analysis, placed by how far its corrected or best-practice estimate sits from the mean the literature reported. Red: smaller in absolute magnitude. Green: larger. 21 of the 25 moved toward zero and 4 away from it. The median revision is -51%. The 14 drawn as rings are approximate pairs. Vertical position carries no meaning. The dots are stacked only to keep them apart.
How this figure is built, and which papers it leaves out

The index is (|corrected| − |mean|) / |mean|, the same relative revision as Table 3 of Gechert et al. (2025), which applies it to 24 literatures mostly by other researchers.

The comparator is the paper’s own uncorrected mean wherever it states one, which 23 of the 25 do. Otherwise it is the average of that literature’s estimates in the data on this site, winsorised at 1%. The paper’s own number is preferred because the harmonised table keeps only estimates that report a usable standard error, so a mean computed from it can rest on a subset of what the paper analysed.

If you distrust means, there is a second reading. 2 of the 25 comparators are computed here instead of quoted from a paper, and those are winsorised means. Use the median of those literatures’ estimates instead and the overall median revision becomes -51%, with the number moving upward unchanged at 4. The swap is confined to those rows on purpose: where a paper states its own mean, its corrected value need not be in the same units as the estimate column at all, so a median taken from that column would not be a comparator. The check covers only the estimates this site holds, under the same standard-error selection.

The 14 rings are pairs that are approximate in a stated way: the central value of a range, or an upper bound (7), the horizon the paper leads with (3), a number read from a results table where the headline is verbal (3), the subsample the paper leads with (1). Hover a ring and it says which.

25 of the 54 papers qualify. Of the other 29, 8 have no single literature effect to correct: methods papers, an experiment, and the review that supplies the companion figure. The rest answer in words rather than a number, or measure the corrected effect as a different quantity from the one their estimate-level data holds, or give a headline that is not a correction at all, or sit against a comparator so near zero that the ratio is noise; one was dropped because two defensible comparators disagreed about which way it moved. The rule takes no account of which way a paper moved, and every one of those 29 is written down with its reason, individually, in correction_ratios.json.

What best practice means here. Several of these papers do not simply average their estimates. They evaluate their model at the literature’s preferred values, for recent data, more observations, and better-cited outlets, and report that as the figure a careful study would have produced. It is the construction these authors use, it is what Gechert et al.’s corrected column is built from, and in each case it is the paper’s own headline conclusion. The rows concerned name it in the spec.

Most of these papers correct with estimators that shrink toward zero. What the figure adds is the size of each move.

Figures last assembled 2026-08-11. Each is the paper's own headline result. The paper's own page carries the rest.

Read the caveat before you reuse a number. Definitions, samples and units differ across literatures, and a corrected mean is still an estimate. The paper's own page carries the full abstract, and the data and code where they exist. Read it there when the number matters. The same table is available as a spreadsheet: estimates.csv, which also carries the sentence in the paper each figure came from.

Macroeconomics

How much do house prices fall after a monetary policy rate hike?

Effect of monetary policy on house prices: corrected for publication bias and misspecification, a 1.2% fall in house prices per 1-percentage-point policy rate rise, peaking after two years (correcting for publication bias alone gives 0.23%, but conditioning on best practice offsets it), based on 1,555 estimates from 237 impulse responses in 37 studies (Ehrenbergerova et al. 2023, IMF Economic Review).

Cite as Dominika Ehrenbergerova, Josef Bajzik, Tomas Havranek (2023), "When Does Monetary Policy Sway House Prices? A Meta-Analysis." IMF Economic Review 71, 538-573.

Paper page · PDF · 10.1057/s41308-022-00185-5

How do structural reforms affect growth in transition countries?

Effect of structural reforms on economic growth: reforms in transition countries cost growth in the short run but raise it strongly in the long run, and that holds after correcting for publication bias, based on 60 studies with 537 estimates (Babecky and Havranek 2014, Economics of Transition).

Cite as Jan Babecky & Tomas Havranek (2014), "Structural Reforms and Growth in Transition: A Meta-Analysis." Economics of Transition 22(1), pp. 13-42.

Paper page · PDF · 10.1111/ecot.12029

What is the elasticity of intertemporal substitution in consumption?

Elasticity of intertemporal substitution in consumption: corrected for publication bias, 0.3-0.4 (micro estimates for asset holders; macro mean zero), based on 2,735 estimates from 169 published studies (Havranek 2015, Journal of the European Economic Association).

Cite as Tomas Havranek (2015), "Measuring Intertemporal Substitution: The Importance of Method Choices and Selective Reporting." Journal of the European Economic Association 13(6), 1180-1204.

Paper page · PDF · 10.1111/jeea.12133

What is the elasticity of substitution between capital and labor?

Elasticity of substitution between capital and labor: the meta-analytic estimate is 0.3 (best-practice specification; the mean reported estimate is 0.9), based on 3,186 estimates from 121 studies (Gechert et al. 2022, Review of Economic Dynamics).

Cite as Sebastian Gechert, Tomas Havranek, Zuzana Irsova, and Dominika Kolcunova (2022), "Measuring Capital-Labor Substitution: The Importance of Method Choices and Publication Bias." Review of Economic Dynamics 45, 55-82.

Paper page · PDF · 10.1016/j.red.2021.05.003

How large is the excess sensitivity of consumption to anticipated income?

Excess sensitivity of consumption to anticipated income changes: the estimate corrected for aggregation and publication bias is 0.11, against a mean reported estimate of 0.37, based on about 3,000 estimates from 144 studies (Havranek and Sokolova 2020, Review of Economic Dynamics).

Cite as Tomas Havranek & Anna Sokolova (2020), "Do Consumers Really Follow a Rule of Thumb? Three Thousand Estimates from 144 Studies Say `Probably Not'." Review of Economic Dynamics 35(1), 97-122.

Paper page · PDF · 10.1016/j.red.2019.05.004

What is the optimal long-run inflation rate?

Optimal long-run inflation rate: the meta-analytic estimate is about 0.6% per year (calibration-dominated model literature), based on 777 estimates from 116 primary studies (Opatrny et al. 2026).

Cite as Matej Opatrny, Martin Opatrny, Tomas Havranek, Zuzana Irsova, Mojmir Hampl (2026), "Optimal Inflation Rate: A Meta-Analysis." Charles University, Prague. Available at meta-analysis.cz/inflation.

Paper page · PDF

Do prices rise after a monetary policy tightening?

The price puzzle (prices rising after monetary tightening): it disappears once publication and misspecification biases are corrected, and prices fall instead, bottoming out 0.33% below baseline six months after a 1-percentage-point rate rise, based on about 1,000 estimates from 70 studies (Rusnak et al. 2013, Journal of Money, Credit and Banking).

Cite as Marek Rusnak, Tomas Havranek, & Roman Horvath (2013), "How to Solve the Price Puzzle? A Meta-Analysis." Journal of Money, Credit and Banking 45(1), 37-70.

Paper page · PDF · 10.1111/j.1538-4616.2012.00561.x

How long does monetary policy take to affect prices?

Transmission lag of monetary policy to prices: the meta-analytic estimate is 29 months on average, but 25-50 months in developed economies against 10-20 in post-transition ones; the factor explaining that gap is financial development, with greater development meaning slower transmission (prices bottom out 0.9% below baseline after a 1-percentage-point hike), based on 67 published studies (Havranek and Rusnak 2013, International Journal of Central Banking).

Cite as Tomas Havranek & Marek Rusnak (2013), "Transmission Lags of Monetary Policy: A Meta-Analysis." International Journal of Central Banking 9(4), pp. 39-76.

Paper page · PDF

Why does intertemporal substitution differ across countries?

Why intertemporal substitution varies across countries: the meta-analytic finding is income and asset market participation are the most effective factors in explaining the cross-country heterogeneity, based on 2,735 estimates from 169 published studies covering 104 countries (Havranek et al. 2015, Journal of International Economics).

Cite as Tomas Havranek, Roman Horvath, Zuzana Irsova, and Marek Rusnak (2015), "Cross-Country Heterogeneity in Intertemporal Substitution." Journal of International Economics 96(1), 100-118.

Paper page · PDF · 10.1016/j.jinteco.2015.01.012

International economics

What is the Armington elasticity between domestic and foreign goods?

Armington elasticity (substitution between domestic and foreign goods): the meta-analytic estimate is 2.5-5.1, median 3.8, based on 3,524 estimates (Bajzik et al. 2020, Journal of International Economics).

Cite as Josef Bajzik, Tomas Havranek, Zuzana Irsova, and Jiri Schwarz (2020), "Estimating the Armington Elasticity: The Importance of Study Design and Publication Bias." Journal of International Economics 127, 103383.

Paper page · PDF · 10.1016/j.jinteco.2020.103383

How much do national borders reduce international trade?

Border effect on international trade: the meta-analytic finding is one-third reduction in trade, based on 1,271 estimates from 61 studies (Havranek and Irsova 2017, IMF Economic Review).

Cite as Tomas Havranek and Zuzana Irsova (2017), "Do Borders Really Slash Trade? A Meta-Analysis." IMF Economic Review 65(2), 365-396.

Paper page · PDF · 10.1057/s41308-016-0001-5

What determines the size of horizontal FDI spillovers?

Determinants of horizontal FDI spillovers: the main result is that spillovers are zero on average but depend systematically on the technology gap and ownership structure; encouraging joint ventures with investors holding a smaller technology edge raises the average spillover by about 0.3, based on 1,205 estimates screened against 43 candidate determinants (Irsova and Havranek 2013, World Development).

Cite as Zuzana Irsova & Tomas Havranek (2013), "Determinants of Horizontal Spillovers from FDI: Evidence from a Large Meta-Analysis." World Development 42, 1-15.

Paper page · PDF · 10.1016/j.worlddev.2012.07.001

Does adopting the euro increase trade between member countries?

Effect of the euro on trade: no detectable effect once publication bias is corrected, while other currency unions raise trade by more than 60% (Havranek 2010, Review of World Economics).

Cite as Tomas Havranek (2010), "Rose Effect and the Euro: Is the Magic Gone?" Review of World Economics, 146(2), pp. 241-261.

Paper page · PDF · 10.1007/s10290-010-0050-1

How much do FDI spillovers boost Czech firm productivity?

FDI spillovers to domestic firm productivity (Czech Republic): the meta-analytic estimate is +11% productivity per 10-percentage-point rise in foreign presence (as of 2018; 19% for joint ventures), based on 332 estimates (Hampl et al. 2020, Applied Economics).

Cite as Mojmir Hampl, Tomas Havranek, and Zuzana Irsova (2020), "Foreign Capital and Domestic Productivity in the Czech Republic: A Meta-Regression Analysis." Applied Economics 52(18), 1949-1958.

Paper page · PDF · 10.1080/00036846.2020.1726864

Is the forward premium puzzle real in currency markets?

Forward premium puzzle (slope from regressions of spot on forward rates): corrected for publication bias, 0.23-0.45 for developed and 0.95-1.16 for emerging currencies (unbiasedness hypothesis predicts slope of 1), based on 3,643 estimates from 91 studies (Zigraiova et al. 2021, European Economic Review).

Cite as Diana Zigraiova, Tomas Havranek, Zuzana Irsova, and Jiri Novak (2021), "How Puzzling Is the Forward Premium Puzzle? A Meta-Analysis." European Economic Review 134, 103714.

Paper page · PDF · 10.1016/j.euroecorev.2021.103714

Do FDI spillovers benefit domestic suppliers and buyers?

Vertical FDI spillovers to suppliers and buyers: the estimate corrected for publication and misspecification bias is about 9% higher productivity for suppliers per 10-percentage-point rise in foreign presence, small for buyers and none within the sector, based on 3,626 estimates (Havranek and Irsova 2011, Journal of International Economics).

Cite as Tomas Havranek & Zuzana Irsova (2011), "Estimating Vertical Spillovers from FDI: Why Results Vary and What the True Effect Is." Journal of International Economics, 85(2), pp. 234-244.

Paper page · PDF · 10.1016/j.jinteco.2011.07.004

Financial economics

Does bank competition threaten financial stability?

Effect of bank competition on financial stability: corrected for publication bias, about 0, based on 598 estimates from 31 studies (Zigraiova and Havranek 2016, Journal of Economic Surveys).

Cite as Diana Zigraiova and Tomas Havranek (2016), "Bank Competition and Financial Stability: Much Ado About Nothing?" Journal of Economic Surveys 30(5), 944-981.

Paper page · PDF · 10.1111/joes.12131

Does board gender diversity improve ESG performance?

Effect of board gender diversity on ESG performance: the meta-analytic estimate is about 0.12 ESG points per 1-pp diversity increase (higher in the Middle East, about zero in Southeast Asia), based on 533 estimates from 106 studies (Hozova et al. 2026).

Cite as Karolina Hozova, Tomas Havranek, Zuzana Irsova (2026), "Do Female Directors Raise ESG Ratings? A Meta-Analysis." Charles University, Prague. Available at meta-analysis.cz/esg.

Paper page · PDF

Does financial development boost economic growth?

Effect of financial development on economic growth: the meta-analytic estimate is positive but widely varying; studies ignoring endogeneity overstate it, and the effect weakens after the 1980s and in poorer countries, based on 1,334 estimates from 67 studies (Valickova et al. 2015, Journal of Economic Surveys).

Cite as Petra Valickova, Tomas Havranek, and Roman Horvath (2015), "Financial Development and Economic Growth: A Meta-Analysis." Journal of Economic Surveys 29(3), 506-526.

Paper page · PDF · 10.1111/joes.12068

Do remittances boost economic growth?

Effect of remittances on economic growth: the meta-analytic finding is positive but economically small (growth-enhancing in Asia but not Africa), based on 538 estimates from 95 studies (Cazachevici et al. 2020, World Development).

Cite as Alina Cazachevici, Tomas Havranek, and Roman Horvath (2020), "Remittances and Economic Growth: A Meta-Analysis." World Development 134, 105021.

Paper page · PDF · 10.1016/j.worlddev.2020.105021

How much alpha do hedge funds generate?

Hedge fund alpha (monthly risk-adjusted return): corrected for publication bias, 30-40 basis points per month (predominantly net of fees: 984 of the 1,019 estimates are net; the companion study puts the best-practice figure near zero once recent data and study design are imposed), based on 1,019 alpha estimates from 74 studies (Yang et al. 2024, Journal of Economic Surveys).

Cite as Fan Yang, Tomas Havranek, Zuzana Irsova, and Jiri Novak (2024), "Is research on hedge fund performance published selectively? A quantitative survey." Journal of Economic Surveys 38(4): 1085-1131.

Paper page · PDF · 10.1111/joes.12574

How large is the size premium in stock returns?

Size premium in stock returns: corrected for publication bias, 1.72% per year (smallest vs. largest capitalization quintile), based on 1,746 estimates from 102 published studies (Astakhov et al. 2019, Journal of Economic Surveys).

Cite as Anton Astakhov, Tomas Havranek, and Jiri Novak (2019), "Firm Size and Stock Returns: A Quantitative Survey." Journal of Economic Surveys 33(5), 1463-1492.

Paper page · PDF · 10.1111/joes.12335

How much do stock prices move after shareholder activism?

Stock-price reaction to shareholder activism: corrected for publication bias, 0% to 1.5%, based on 1,973 estimates from 67 studies (Bajzik et al. 2025, Corporate Governance: An International Review).

Cite as Josef Bajzik, Tomas Havranek, Zuzana Irsova, Jiri Novak (2025), "Does Shareholder Activism Create Value? A Meta-Analysis." Corporate Governance: An International Review 33(5): 1039-1061.

Paper page · PDF · 10.1111/corg.12637

Does stronger rule of law relate to a larger size premium?

Trust, rule of law, and the size premium in stock returns: the meta-analytic finding is stronger rule of law, larger size premium (association in reported estimates, not causal), based on 1,613 estimates from 105 studies and 31 countries (Schwarz et al. 2026).

Cite as Jiri Schwarz, Tomas Havranek, Zuzana Irsova, Jiri Novak (2026), "Trust, Rule of Law, and the Size Premium: Evidence from a Meta-Analysis." Anglo-American University and Charles University, Prague. Available at meta-analysis.cz/trust.

Paper page · PDF

What explains the variation in hedge fund alpha?

Variation in hedge fund alphas: the main result is that 9 of 34 study characteristics consistently explain it, above all whether alpha is measured gross or net of fees, a gap of 0.439 percentage points a month; expected alphas under best practice are close to zero for all common strategies, based on 1,019 estimates from 74 studies (Yang et al. 2026).

Cite as Fan Yang, Tomas Havranek, Zuzana Irsova, and Jiri Novak (2026), "What Matters in Explaining the Variation in Hedge Fund Performance?" Charles University, Prague. Available at meta-analysis.cz/alphas.

Paper page · PDF

Labor and education economics

Does smaller class size improve student achievement?

Effect of class size on student achievement: the meta-analytic estimate is about 0 (except in the STAR experiment and classes under 15 students), based on 2,819 estimates from 66 studies (Opatrny et al. 2025, Journal of Labor Economics).

Cite as Matej Opatrny, Tomas Havranek, Zuzana Irsova, Milan Scasny (2025), "Publication Bias and Model Uncertainty in Measuring the Effect of Class Size on Achievement." Journal of Labor Economics.

Paper page · PDF · 10.1086/737989

Does working while in school hurt academic outcomes?

Effect of working while in school on educational outcomes: the meta-analytic estimate is about 0, slightly negative (more substantive for dropout decisions), based on 861 estimates from 69 studies (Kroupova et al. 2024, Economics of Education Review).

Cite as Katerina Kroupova, Tomas Havranek, Zuzana Irsova (2024), "Student Employment and Education: A Meta-Analysis." Economics of Education Review 100, 102539.

Paper page · PDF · 10.1016/j.econedurev.2024.102539

How substitutable are native and immigrant workers?

Elasticity of substitution between native and immigrant workers: corrected for publication bias, about 17 (baseline regional specification; simple bias correction implies 22), based on 1,091 estimates from 41 studies (Kantova et al. 2026).

Cite as Klara Kantova, Tomas Havranek, Zuzana Irsova, Jiri Schwarz (2026), "The Elasticity of Substitution between Native and Immigrant Labor: A Meta-Analysis." Charles University, Prague. Available at meta-analysis.cz/migrant.

Paper page · PDF

How substitutable are skilled and unskilled workers?

Elasticity of substitution between skilled and unskilled labor: the meta-analytic estimate is 4, with a lower bound of 2 and smaller values in developing countries, based on 682 estimates from 77 studies (Havranek et al. 2024, Review of Economics and Statistics).

Cite as Tomas Havranek, Zuzana Irsova, Lubica Laslopova, and Olesia Zeynalova (2024), "Publication and Attenuation Biases in Measuring Skill Substitution." Review of Economics and Statistics 106 (5), 1187-1200.

Paper page · PDF · 10.1162/rest_a_01227

What is the Frisch elasticity of labor supply?

Frisch elasticity of labor supply (intertemporal substitution in hours worked): the meta-analytic estimate is about 0.25 for total hours (0.2 on the intensive margin, near zero on the extensive), based on over 700 estimates for each margin (Elminejad et al. 2023, Review of Economic Dynamics).

Cite as Ali Elminejad, Tomas Havranek, Roman Horvath, and Zuzana Irsova (2023), "Intertemporal Substitution in Labor Supply: A Meta-Analysis." Review of Economic Dynamics 51, 1095-1113.

Paper page · PDF · 10.1016/j.red.2023.10.001

How much learning did students lose from COVID-19 school closures?

Learning loss from COVID-19 school closures: the meta-analytic estimate is about -0.12 SD, roughly 30% of a school year (Luskova et al. 2026).

Cite as Martina Luskova, Nino Buliskeria, Ali Elminejad, Tomas Havranek, Zuzana Irsova, Stepan Jurajda, Marek Kapicka (2026), “Publication Bias and P-Hacking in the Effect of COVID-19 on Learning.” Charles University, Prague. Available at meta-analysis.cz/learning.

Paper page · PDF

Does raising tuition reduce university enrollment?

Tuition-enrollment elasticity (effect of tuition on university enrollment): corrected for publication bias, about 0, based on 443 estimates from 43 studies (Havranek et al. 2018, Oxford Bulletin of Economics and Statistics).

Cite as Tomas Havranek, Zuzana Irsova, and Olesia Zeynalova (2018): "Tuition Fees and University Enrolment: A Meta‐Regression Analysis". Oxford Bulletin of Economics and Statistics 80(6): 1145-1184.

Paper page · PDF · 10.1111/obes.12240

Micro and experimental economics

How large is the beauty premium in earnings?

Beauty premium in earnings: the meta-analytic estimate is 1.1% (95% CrI -0.8% to 3.0%; controlling for cognitive ability; large only for sex workers), based on 1,159 estimates in 67 studies (Irsova et al. 2025).

Cite as Zuzana Irsova, Tomas Havranek, Kseniya Bortnikova, František Bartoš (2025), "Meta-Analysis of Field Studies on Beauty and Professional Success." Charles University, Prague. Available at meta-analysis.cz/beauty.

Paper page · PDF

What is the coefficient of relative risk aversion?

Coefficient of relative risk aversion: corrected for publication bias, 1 in economics, 2-7 in finance (from consumption Euler equation studies), based on 1,021 estimates from 92 studies (Elminejad et al. 2025, Journal of Economic Surveys).

Cite as Ali Elminejad, Tomas Havranek, and Zuzana Irsova (2025), "Relative Risk Aversion: A Meta-Analysis." Journal of Economic Surveys, 39(5): 2315-2333.

Paper page · PDF · 10.1111/joes.12689

Do financial incentives improve performance?

Effect of financial incentives on performance: corrected for publication bias, about 0 (field contexts; lab and loss framing modestly positive), based on 2,193 estimates from 88 economics experiments (Cala et al. 2026, Journal of Political Economy Microeconomics).

Cite as Petr Cala, Tomas Havranek, Zuzana Irsova, Martina Luskova, Jindrich Matousek, and Jiri Novak (2026), "Financial Incentives and Performance: A Meta-Analysis of Experiments in Economics." Journal of Political Economy Microeconomics. https://doi.org/10.1086/743543

Paper page · PDF · 10.1086/743543

Does physical exercise improve cognition?

Effect of physical exercise on cognition: corrected for publication bias, SMD 0.227 for general cognition, 0.027 for memory and 0.012 for executive function (wide prediction intervals spanning negative and positive effects), based on 2,239 effect-size estimates from 215 meta-analyses (Bartoš et al. 2025).

Cite as František Bartoš, Martina Lušková, Kseniya Bortnikova, Karolína Hozová, Klara Kantova, Zuzana Irsova, Tomas Havranek (2025), "Effect of Exercise on Cognition, Memory, and Executive Function: A Study-Level Meta-Meta-Analysis Across Populations and Exercise Categories," available at osf.io/preprints/psyarxiv/qr8e2_v1.

Paper page · 10.31234/osf.io/qr8e2_v1

How strong is habit formation in consumption?

Habit formation in consumption: the meta-analytic estimate is 0.4 (macro 0.6, micro 0.1; mean reported estimate, with no correction for publication bias), based on 597 estimates from 81 published studies (Havranek et al. 2017, European Economic Review).

Cite as Tomas Havranek, Marek Rusnak, and Anna Sokolova (2017), "Habit Formation in Consumption: A Meta-Analysis." European Economic Review 95, 142-167.

Paper page · PDF · 10.1016/j.euroecorev.2017.03.009

How impatient are people in experiments?

Individual discount rate (impatience measured in experiments): corrected for publication bias, 33% per year, based on 927 estimates from 56 studies (Matousek et al. 2022, Experimental Economics).

Cite as Jindrich Matousek, Tomas Havranek, and Zuzana Irsova (2022), "Individual Discount Rates: A Meta-Analysis of Experimental Evidence." Experimental Economics 25(1), 318-358.

Paper page · PDF · 10.1007/s10683-021-09716-9

Energy and environmental economics

What does the published literature imply about climate sensitivity?

Climate sensitivity (warming from a doubling of CO2): corrected for publication bias, 1.4-2.3 degrees Celsius (all 48 published estimates from 16 studies, literature search closed March 2014; the IPCC AR6 Working Group I report, published six years later in 2021, gives a best estimate near 3 C), based on 48 estimates from 16 studies (Reckova and Irsova 2015, Energy and Environment).

Cite as Dominika Reckova and Zuzana Irsova (2015), "Publication Bias in Measuring Anthropogenic Climate Change." Energy and Environment 26 (5), 853-862.

Paper page · PDF · 10.1260/0958-305x.26.5.853

Does daylight saving time reduce electricity use?

Electricity savings from daylight saving time: the best-practice estimate is essentially zero, 0.01% savings, against a 0.34% simple average of reported estimates (savings during the days DST applies; the paper finds no publication bias in this literature, so this is a methodological correction), based on 162 estimates from 44 studies (Havranek et al. 2018, Energy Journal).

Cite as Tomas Havranek, Dominik Herman, and Zuzana Irsova (2018), "Does Daylight Saving Save Electricity? A Meta-Analysis." Energy Journal 39(2), 35-61.

Paper page · PDF · 10.5547/01956574.39.2.thav

How much does gasoline demand rise with income?

Income elasticity of gasoline demand: corrected for publication bias, 0.1 short run, 0.23 long run (Havranek and Kokes 2015, Energy Economics).

Cite as Tomas Havranek and Ondrej Kokes (2015), "Income Elasticity of Gasoline Demand: A Meta-Analysis." Energy Economics 47, pp. 77-86.

Paper page · PDF · 10.1016/j.eneco.2014.11.004

How much does water demand rise with income?

Income elasticity of water demand: corrected for publication bias, about 0.15 or less, based on 307 estimates (Havranek et al. 2018, Land Economics).

Cite as Tomas Havranek, Zuzana Irsova, and Tomas Vlach (2018), "Measuring the Income Elasticity of Water Demand: The Importance of Publication and Endogeneity Biases." Land Economics 94(2): 259-283.

Paper page · PDF · 10.3368/le.94.2.259

Does having abundant natural resources hurt long-run economic growth?

Natural resource curse (effect of resource richness on long-run growth): the meta-analytic finding is weak support for a curse, based on 605 estimates from 43 studies (Havranek et al. 2016, World Development).

Cite as Tomas Havranek, Roman Horvath, and Ayaz Zeynalov (2016), "Natural Resources and Economic Growth: A Meta-Analysis." World Development 88, 134-151.

Paper page · PDF · 10.1016/j.worlddev.2016.07.016

How much does electricity demand fall when prices rise?

Price elasticity of electricity demand: corrected for publication bias, -0.16 short run, -0.38 long run (best-identified short-run estimates statistically indistinguishable from zero), based on 4,720 estimates from 462 studies (Kudela et al. 2026).

Cite as Peter Kudela, Tomas Havranek, Zuzana Irsova, Anna Kudelova, Vojtech Sikl (2026), “Electricity demand has not become more price-responsive despite ninety years of technological change.” Charles University, Prague. Available at meta-analysis.cz/electricity.

Paper page · PDF

How much does gasoline demand fall when prices rise?

Price elasticity of gasoline demand: corrected for publication bias, -0.09 short run, -0.31 long run (Havranek et al. 2012, Energy Economics).

Cite as Tomas Havranek, Zuzana Irsova, and Karel Janda (2012), "Demand for Gasoline is More Price-Inelastic than Commonly Thought," Energy Economics 34(1), pp. 201-207.

Paper page · PDF · 10.1016/j.eneco.2011.09.003

What is the social cost of carbon?

Social cost of carbon: corrected for publication bias, 0-134 USD per metric ton of carbon (2010 prices, emission year 2015; per ton of carbon, not CO2), based on 809 estimates from 101 studies (Havranek et al. 2015, Energy Economics).

Cite as Tomas Havranek, Zuzana Irsova, Karel Janda, and David Zilberman (2015), "Selective Reporting and the Social Cost of Carbon." Energy Economics 51, 394-406.

Paper page · PDF · 10.1016/j.eneco.2015.08.009

Meta-research methods

Are meta-analyses of correlations biased?

Bias in meta-analyses of correlations: the methodological finding is that conventional fixed- and random-effects estimators are biased; the UWLS+3 and Hunter-Schmidt estimators correct this (Stanley et al. 2025, Research Synthesis Methods).

Cite as T. D. Stanley, Hristos Doucouliagos, and Tomas Havranek (2025): "Reducing the biases of the conventional meta-analysis of correlations." Research Synthesis Methods 16(1): 42-59.

Paper page · PDF · 10.1017/rsm.2024.5

Are meta-analyses of partial correlations biased?

Bias in meta-analyses of partial correlations: the methodological finding is that all such meta-analyses are biased; the UWLS+3 weighted average corrects them (Stanley et al. 2024, Research Synthesis Methods).

Cite as T. D. Stanley, Hristos Doucouliagos, and Tomas Havranek (2024): "Meta-analyses of partial correlations are biased: Detection and solutions." Research Synthesis Methods 15: 313-325.

Paper page · PDF · 10.1002/jrsm.1704

How far do corrected meta-analytic effects move from conventional wisdom?

How far corrected effects move from conventional wisdom: the methodological finding is that corrected effects are typically 45-60% smaller, much closer to zero, and sometimes switch sign, based on a systematic review of influential economics meta-analyses against their fields' conventional wisdom (Gechert et al. 2025, Journal of Economic Surveys).

Cite as Sebastian Gechert, Bianka Mey, Matej Opatrny, Tomas Havranek, T. D. Stanley, Pedro R. D. Bom, Hristos Doucouliagos, Philipp Heimberger, Zuzana Irsova, Heiko J. Rachinger (2025), “Conventional wisdom, meta-analysis, and research revision in economics.” Journal of Economic Surveys 39, 980-999. 10.1111/joes.12630

Paper page · PDF · 10.1111/joes.12630

How should researchers conduct a modern meta-analysis?

How to run a modern meta-analysis: a concise, nontechnical, step-by-step guide for practitioners, treating publication bias, p-hacking, and heterogeneity as problems every meta-analyst must confront (Irsova et al. 2024, Journal of Economic Surveys).

Cite as Zuzana Irsova, Hristos Doucouliagos, Tomas Havranek, and T. D. Stanley (2024): "Meta-Analysis of Social Science Research: A Practitioner’s Guide." Journal of Economic Surveys 38(5): 1547-1566.

Paper page · PDF · 10.1111/joes.12595

How sensitive are meta-analysis results to outlier handling?

Sensitivity of meta-analysis results to outlier handling: the methodological finding is median shift at most 0.047 (Cohen's d); significance flips in 11.5% of cases (across four pre-registered outlier treatments and two estimators), based on 358 behavioral science meta-analyses (Havranek et al. 2026).

Cite as Tomas Havranek, Zuzana Irsova, Martina Luskova, T. D. Stanley (2026), "Do decisions about outliers and influential effects matter? Evidence from 358 behavioral science meta-analyses." Charles University, Prague. Available at meta-analysis.cz/outliers.

Paper page · PDF

Can p-hacked precision distort meta-analysis weighting?

Spurious precision in meta-analysis (the MAIVE estimator): the methodological finding is estimated precision can be p-hacked, which can undermine inverse-variance weighting; MAIVE instruments variance with sample size (Irsova et al. 2025, Nature Communications).

Cite as Zuzana Irsova, Pedro Bom, Tomas Havranek, and Heiko Rachinger (2025): "Spurious Precision in Meta-Analysis of Observational Research." Nature Communications 16, 8454. doi.org/10.1038/s41467-025-63261-0

Paper page · PDF · 10.1038/s41467-025-63261-0

Does multi-agent AI debate improve feedback on research papers?

Whether multi-agent AI debate improves feedback on research papers: probably not, at least for meta-analyses in economics; authors of 44 meta-analyses preferred a single frontier-model pass to two multi-agent debate tools, based on a pre-registered, identity-masked experiment (Havranek and Irsova 2026).

Cite as Tomas Havranek, Zuzana Irsova (2026), "Does Multi-Agent Debate Improve AI Feedback on Research Papers?" arXiv:2607.14713. Available at meta-analysis.cz/debate.

Paper page · PDF · https://arxiv.org/abs/2607.14713

Does the choice of meta-analysis method change the answer for partial correlations?

Whether the estimator choice matters for partial correlations in practice: the methodological finding is that small-sample biases are negligible and Fisher's z helps with neither them nor publication selection bias; UWLS and Hunter-Schmidt give smaller, arguably less biased means than random effects, based on 172 economics meta-analyses with 63,730 partial correlations from 6,191 papers (Stanley et al. 2026).

Cite as T. D. Stanley, Petr Cala, Hristos Doucouliagos, Zuzana Irsova, and Tomas Havranek (2026), "Do methods matter in the meta-analysis of partial correlation coefficients?" Charles University, Prague. Available at meta-analysis.cz/pcc_survey.

Paper page · PDF