# meta-analysis.cz

> Methods, data, code, and papers for meta-analysis, by Tomas Havranek and Zuzana Irsova of Charles University, Prague, and their co-authors. The applied corpus is mainly economics and the social sciences; the methods and tools, among them MAIVE, are for use in any field. Each paper page links the full-text PDF and, for most papers, the dataset and estimation code.

> Maintained by Tomas Havranek (in Czech Tomáš Havránek) and Zuzana Irsova (Zuzana Iršová, earlier work indexed as Zuzana Havránková) at the Institute of Economic Studies, Charles University, Prague. Both are Professors of Economics and Associate Editors responsible for meta-analysis at the Journal of Economic Surveys. Havranek sat on the European Research Council's Advanced Grant panel for economics and finance in 2020, the first Czech economist to do so. Irsova is the first author of the paper that introduced MAIVE (Nature Communications, 2025) and was elected to Academia Europaea in 2023. Irsova's website and social profiles: https://www.irsova.com/ ; LinkedIn https://www.linkedin.com/in/zuzanairsova/ ; Bluesky https://bsky.app/profile/irsova.bsky.social ; X https://x.com/zuzanairsova . Who they are, with awards and rankings: https://meta-analysis.cz/about/

> **Licence: everything on this site is CC BY 4.0** (https://creativecommons.org/licenses/by/4.0/) — the papers, the datasets, their CSV and Parquet conversions, the pooled table, the codebooks, and this documentation. You may use, adapt, and redistribute any of it, including commercially and including as training data for machine-learning models. The only condition is attribution: cite the source paper for a dataset, and the collection as DOI 10.5281/zenodo.21773678.

## Papers

- [Does Shareholder Activism Create Value? A Meta-Analysis](https://meta-analysis.cz/activism/): Meta-analysis of 1,973 estimates from 67 studies finds shareholder activism raises stock prices by 0% to 1.5% after correcting for publication bias.
- [What Matters in Explaining the Variation in Hedge Fund Performance? A Meta-Analysis](https://meta-analysis.cz/alphas/): Meta-analysis of 1,019 estimates from 74 studies identifies which factors explain the variation in hedge fund performance; net performance is near zero. Published as "What Matters in Explaining the Variation in Hedge Fund Performance?".
- [A Meta-Analysis of the Trade Cost Elasticity](https://meta-analysis.cz/armington/): Meta-analysis of 3,524 estimates finds the Armington trade elasticity is 2.5-5.1 after correcting for publication bias. Published as "Estimating the Armington Elasticity: The Importance of Study Design and Publication Bias".
- [Beauty and Professional Success: A Meta-Analysis](https://meta-analysis.cz/beauty/): A meta-analysis of 1,159 estimates: the beauty premium in earnings and productivity is small once publication bias and cognitive ability are addressed. Published as "Meta-Analysis of Field Studies on Beauty and Professional Success".
- [A Meta-Analysis of Horizontal Spillovers from FDI](https://meta-analysis.cz/bma/): Meta-analysis of 1205 estimates finds horizontal FDI spillovers average zero, shaped by technology gap and ownership structure. Published as "Determinants of Horizontal Spillovers from FDI: Evidence from a Large Meta-Analysis".
- [Do Borders Really Slash Trade? A Meta-Analysis](https://meta-analysis.cz/border/): Meta-analysis of 1,271 estimates from 61 studies finds methodological choices shrink the true border effect on trade.
- [Central Bank Equity as an Instrument of Monetary Policy](https://meta-analysis.cz/cbequity/): Central bank equity as an unconventional policy tool, with a meta-analysis of 176 estimates finding that central bank financial strength has not historically correlated with inflation.
- [Class Size and Student Achievement: A Meta-Analysis](https://meta-analysis.cz/class/): A meta-analysis of 2,819 estimates from 66 studies finds little publication bias and a negligible class size effect on student achievement. Published as "Publication Bias and Model Uncertainty in Measuring the Effect of Class Size on Achievement".
- [Publication Bias in Measuring Anthropogenic Climate Change](https://meta-analysis.cz/climate/): Meta-analysis of 48 estimates finds publication bias inflates climate sensitivity; corrected range is 1.4-2.3°C.
- [Bank Competition and Financial Stability: Much Ado About Nothing?](https://meta-analysis.cz/competition/): A meta-analysis of 598 estimates from 31 studies finding little relationship between bank competition and financial stability.
- [The sources of contagion risk in a banking sector with foreign ownership](https://meta-analysis.cz/contagion/): Extreme-value analysis of the Czech, Polish, Slovak, and Turkish banking sectors from 1994 to 2013 finds contagion risk from a foreign parent bank to its subsidiary is substantially smaller than between two local banks.
- [Conventional wisdom, meta-analysis, and research revision in economics](https://meta-analysis.cz/conventional_wisdom/): A systematic review of influential economics meta-analyses against the conventional wisdom they tested: after correcting for observable biases, effects are typically much closer to zero and sometimes switch sign, with a relative reduction of 45-60%.
- [Reducing the biases of the conventional meta-analysis of correlations](https://meta-analysis.cz/correlations/): Simulations show UWLS+3 and PET-PEESE reduce small-sample and publication-selection bias in meta-analyses of correlations.
- [Should Inflation Measures Used by Central Banks Incorporate House Prices? The Czech National Bank's Approach](https://meta-analysis.cz/cpi_housing/): The Czech National Bank's supplementary inflation index, the CPIH, gives prices of existing dwellings a 14% weight; compared with the CPI it would have called for looser policy in 2009, 2010 and 2012 and somewhat tighter policy in 2015 and 2016.
- [Banking, debt, and currency crises: Early warning indicators for developed countries](https://meta-analysis.cz/crisis_ews/): A new quarterly dataset of banking, debt, and currency crises in 40 developed countries from 1970 to 2010: banking crises usually come first, cost the most, and follow booms in credit and activity. A later and substantially different version was published in the Journal of Financial Stability in 2014 (doi.org/10.1016/j.jfs.2014.07.001).
- [Banking, debt, and currency crises in developed countries: Stylized facts and early warning indicators](https://meta-analysis.cz/crisis_jfs/): A quarterly dataset of banking, debt, and currency crises in 40 developed countries from 1970 to 2010: banking crises usually come first and cost the most, and booms in credit and activity precede them.
- [Does Multi-Agent Debate Improve AI Feedback on Research Papers?](https://meta-analysis.cz/debate/): Preregistered evaluation of whether multi-agent AI debate improves the quality of AI-generated feedback on research papers.
- [Individual Discount Rates: A Meta-Analysis of Experimental Evidence](https://meta-analysis.cz/discrate/): Meta-analysis of experimental discount-rate estimates finds publication bias and a corrected mean annual rate of 0.33.
- [Does Daylight Saving Save Electricity? A Meta-Analysis](https://meta-analysis.cz/dst/): A meta-analysis of 162 estimates from 44 studies finds daylight saving time saves essentially no electricity: 0.01% after correction, against a 0.34% simple average of reported estimates.
- [Does daylight saving time save electricity? Evidence from Slovakia](https://meta-analysis.cz/dst_slovakia/): Hourly Slovak data for 2010-2017 put the electricity savings from daylight saving time at 1% of annual consumption, or below 0.5% when extrapolated from a meta-analysis of other countries.
- [Publication Bias in Measuring the Impact of Tuition on Enrollment](https://meta-analysis.cz/education/): Meta-analysis of 443 estimates finds the tuition-enrollment elasticity is near zero once publication bias is corrected. Published as "Tuition Fees and University Enrolment: A Meta‐Regression Analysis".
- [Publication Bias in Measuring Intertemporal Substitution](https://meta-analysis.cz/eis/): Meta-analysis of 2,735 EIS estimates finds strong selective reporting inflates the elasticity of intertemporal substitution. Published as "Measuring Intertemporal Substitution: The Importance of Method Choices and Selective Reporting".
- [Estimating the Effective Lower Bound on the Czech National Bank's Policy Rate](https://meta-analysis.cz/elb/): The effective lower bound on the Czech National Bank's policy rate is estimated at slightly below -1%, in the interval (-2.0%, -0.4%), and negative rates alone would not have matched the exchange rate commitment.
- [Price Elasticity of Electricity Demand: A Meta-Analysis](https://meta-analysis.cz/electricity/): Meta-analysis of the price elasticity of electricity demand: no increase in price responsiveness despite ninety years of technological change. Published as "Electricity demand has not become more price-responsive despite ninety years of technological change".
- [Do Female Directors Raise ESG Ratings? A Meta-Analysis](https://meta-analysis.cz/esg/): Meta-analysis of 533 estimates from 106 studies on whether female board directors raise corporate ESG ratings.
- [A Meta-Analysis of the Trade Effect of the Euro](https://meta-analysis.cz/euro/): Meta-analysis finds the euro's trade-promoting effect is insignificant once publication bias is corrected for. Published as "Rose Effect and the Euro: Is the Magic Gone?".
- [A Meta-Analysis of the Excess Sensitivity of Consumption to Anticipated Income Changes](https://meta-analysis.cz/excess_sensitivity/): Meta-analysis of 144 studies finds little excess sensitivity of consumption once micro data, publication bias, and liquidity constraints are addressed. Published as "Do Consumers Really Follow a Rule of Thumb? Three Thousand Estimates from 144 Studies Say ‘Probably Not’".
- [Effect of Exercise on Cognition, Memory, and Executive Function: A Study-Level Meta-Meta-Analysis Across Populations and Exercise Categories](https://meta-analysis.cz/exercise/): Meta-meta-analysis of 1,933 estimates from 201 meta-analyses finds publication-bias-adjusted exercise effects on general cognition and executive function smaller than previously reported.
- [Foreign Capital and Domestic Productivity in the Czech Republic: A Meta-Regression Analysis](https://meta-analysis.cz/fdi/): Meta-analysis of 332 estimates finds FDI spillovers raise Czech domestic firm productivity, larger for joint ventures.
- [Financial Development and Economic Growth: A Meta-Analysis](https://meta-analysis.cz/finance_growth/): A meta-analysis of 1334 estimates from 67 studies finds financial development has a positive, heterogeneous effect on economic growth.
- [How Puzzling Is the Forward Premium Puzzle? A Meta-Analysis](https://meta-analysis.cz/forward/): Meta-analysis of 3,643 estimates from 91 studies finds forward rate unbiasedness holds better than commonly thought once biases are corrected.
- [Publication and Identification Biases in Measuring the Intertemporal Substitution of Labor Supply](https://meta-analysis.cz/frisch/): Meta-analysis of over 700 estimates finds publication bias inflates the Frisch labor supply elasticity, with a corrected total hours elasticity near 0.25. Published as "Intertemporal Substitution in Labor Supply: A Meta-Analysis".
- [A Meta-Analysis of the Income Elasticity of Gasoline Demand](https://meta-analysis.cz/gasoline/): Meta-regression corrects for publication bias, finding gasoline demand income elasticity of 0.1 short run and 0.23 long run. Published as "Income Elasticity of Gasoline Demand: A Meta-Analysis".
- [A Meta-Analysis of the Price Elasticity of Gasoline Demand](https://meta-analysis.cz/gasoline_price/): Meta-analysis shows published gasoline demand elasticities are exaggerated twofold due to publication selection bias. Published as "Demand for Gasoline is More Price-Inelastic than Commonly Thought".
- [Meta-Analysis of Social Science Research: A Practitioner’s Guide](https://meta-analysis.cz/guidelines/): Step-by-step guidelines for modern meta-analysis in social sciences, covering publication bias, p-hacking, heterogeneity, and AI-assisted workflows. Published as "Meta-analysis of social science research: A practitioner’s guide".
- [Habit Formation in Consumption: A Meta-Analysis](https://meta-analysis.cz/habits/): Meta-analysis of 597 estimates finds habit formation averages 0.4, with macro data yielding larger estimates than micro data.
- [Publication Bias in Hedge Fund Performance: A Meta-Analysis](https://meta-analysis.cz/hedge/): Quantitative survey of 1,019 hedge fund alpha estimates from 74 studies finds little evidence of publication selection bias. Published as "Is research on hedge fund performance published selectively? A quantitative survey".
- [When Does Monetary Policy Sway House Prices? A Meta-Analysis](https://meta-analysis.cz/house_prices/): A meta-analysis of 1,555 estimates from 37 studies across 45 countries finds publication bias exaggerates the effect of interest rates on house prices.
- [Financial Incentives and Performance: A Meta-Analysis of Experiments in Economics](https://meta-analysis.cz/incentives/): Meta-analysis of 2,193 estimates finds financial incentives have near-zero corrected effect on performance in most field settings.
- [Optimal Inflation Rate: A Meta-Analysis](https://meta-analysis.cz/inflation/): Meta-analysis of 777 estimates from 116 studies finds an optimal long-run inflation rate near 0.6% per year, below common two-percent central-bank targets.
- [Transmission Lags of Monetary Policy: A Meta-Analysis](https://meta-analysis.cz/lags/): A meta-analysis of 67 studies finds monetary policy transmission lags average 29 months, longer in developed than transition economies.
- [Leading indicators of crisis incidence: Evidence from developed countries](https://meta-analysis.cz/leading_indicators/): Bayesian model averaging over crises in EU and OECD countries from 1970 to 2010 identifies domestic housing prices, share prices, and credit growth as the leading indicators of what a crisis costs.
- [Publication Bias and P-Hacking in the Effect of COVID-19 on Learning: A Meta-Analysis](https://meta-analysis.cz/learning/): Meta-analysis of COVID-19 learning loss: bias-corrected RoBMA and MAIVE estimates converge on about -0.12 SD, roughly 30% of a school year. Published as "Publication Bias and P-Hacking in the Effect of COVID-19 on Learning".
- [Meta-Analysis Instrumental Variable Estimator](https://meta-analysis.cz/maive/): MAIVE, an instrumental-variable estimator that corrects meta-analysis for spurious precision when standard errors are estimated and can be p-hacked. Published as "Spurious precision in meta-analysis of observational research".
- [The Elasticity of Substitution between Native and Immigrant Labor: A Meta-Analysis](https://meta-analysis.cz/migrant/): Meta-analysis of 1,091 estimates finds native-immigrant labor substitution elasticity of about 17 under best practice, against an uncorrected mean of 13.
- [Do decisions about outliers and influential effects matter? Evidence from 358 behavioral science meta-analyses](https://meta-analysis.cz/outliers/): Pre-registered analysis of 358 behavioral science meta-analyses: alternative outlier and influence handling treatments barely move the mean but reverse statistical significance in 11.5% of cases.
- [Bank Efficiency and Interest Rate Pass-Through: Evidence from Czech Loan Products](https://meta-analysis.cz/passthrough/): Czech loan and deposit products show a pass-through from market rates that was close to complete before the financial crisis and much weaker after it; banks paying high deposit rates charge high loan markups, and cost-efficient banks delay their responses, smoothing rates for clients.
- [Meta-analyses of partial correlations are biased: Detection and solutions](https://meta-analysis.cz/pcc/): Shows all partial-correlation meta-analyses are biased and proposes UWLS+3 to correct small-sample bias.
- [Do methods matter in the meta-analysis of partial correlation coefficients?](https://meta-analysis.cz/pcc_survey/): A survey of 172 economics meta-analyses covering 63,730 partial correlations from 6,191 papers: small-sample biases are negligible in practice, Fisher's z reduces neither small-sample nor publication selection bias, and UWLS and Hunter-Schmidt give smaller mean estimates than random effects.
- [The Power of Science: Statistical Power in Published Research Across Five Disciplines](https://meta-analysis.cz/power/): About 748,000 estimates from meta-analyses in five disciplines show published research underpowered for small effects everywhere; for medium effects, median power passes 80 percent in psychology, economics and political science but not in environmental science or medicine.
- [How to Solve the Price Puzzle? A Meta-Analysis](https://meta-analysis.cz/price_puzzle/): Meta-analysis of 1,000 VAR estimates shows the price puzzle stems from model misspecification, not a true price effect.
- [Structural Reforms and Economic Growth: A Meta-Analysis](https://meta-analysis.cz/reforms/): Meta-analysis of 60 studies finds structural reforms in transition countries carry short-run costs but strong long-run growth benefits. Published as "Structural Reforms and Growth in Transition: A Meta-Analysis".
- [Remittances and Economic Growth: A Meta-Analysis](https://meta-analysis.cz/remittances/): Meta-analysis of 538 estimates finds remittances have a small positive, publication-bias-corrected effect on growth, stronger in Asia than Africa.
- [Reproducibility and Robustness of Economics and Political Science Research](https://meta-analysis.cz/reproducibility/): A 347-author reproduction of 110 articles in twelve leading economics and political science journals: more than 85% of claims were computationally reproducible, 72% of significant estimates stayed significant under re-analysis, and the median reproduced effect was 99% of the published one.
- [Natural Resources and Economic Growth: A Meta-Analysis](https://meta-analysis.cz/resource_curse/): Meta-analysis of 605 estimates finds weak support for the resource curse once publication bias and methods are addressed.
- [Relative Risk Aversion: A Meta-Analysis](https://meta-analysis.cz/risk/): A meta-analysis of 1,021 estimates from 92 studies finds publication-bias-corrected relative risk aversion near 1 in economics and 2-7 in finance.
- [Selective Reporting and the Social Cost of Carbon](https://meta-analysis.cz/scc/): Meta-analysis of 809 SCC estimates from 101 studies finds selective reporting creates an upward bias, especially in peer-reviewed journals.
- [Measuring Capital-Labor Substitution: The Importance of Method Choices and Publication Bias](https://meta-analysis.cz/sigma/): Meta-analysis of 3,186 estimates from 121 studies finds the capital-labor elasticity of substitution is 0.3, far below the literature's average of 0.9.
- [A Meta-Analysis of the Relation Between Firm Size and Stock Returns](https://meta-analysis.cz/size/): First meta-analysis of the size premium finds strong publication bias and a corrected effect of 1.72% in annual returns. Published as "Firm Size and Stock Returns: A Quantitative Survey".
- [Publication and Attenuation Biases in Measuring Skill Substitution](https://meta-analysis.cz/skill/): Meta-analysis finds publication bias dominates in skill substitution estimates, implying a mean elasticity of 4.
- [Adjusting for Publication Bias Reveals Evidence Against Social Comparison As a Behaviour Change Technique Across the Behavioural Sciences](https://meta-analysis.cz/social_comparison/): Re-analysis of a meta-analysis of 79 randomized controlled trials: once publication bias is adjusted for, the effect of social comparison as a behaviour change technique disappears and the data show moderate evidence against it.
- [A Meta-Analysis of Vertical Spillovers from FDI](https://meta-analysis.cz/spillovers/): A meta-analysis of 3,626 estimates finds vertical FDI spillovers to suppliers are economically significant but spillovers to buyers small. Published as "Estimating Vertical Spillovers from FDI: Why Results Vary and What the True Effect Is".
- [Survey Article: Publication Bias in the Literature on Foreign Direct Investment Spillovers](https://meta-analysis.cz/spillovers_bias/): A quantitative survey of 3,626 estimates from 57 studies of FDI spillovers: publication selection appears only in peer-reviewed journals, only among the estimates authors call most important, and weakens over time.
- [Student Employment and Education: A Meta-Analysis](https://meta-analysis.cz/students/): Meta-analysis of 861 estimates finds student employment has a negative but economically small effect on educational outcomes.
- [Why Does Intertemporal Substitution Vary Across Countries? A Meta-Analysis](https://meta-analysis.cz/substitution/): Meta-analysis of 2,735 estimates across 104 countries links intertemporal substitution to income and stock market participation. Published as "Cross-Country Heterogeneity in Intertemporal Substitution".
- [Dynamic Elasticities of Tax Revenue: Evidence from the Czech Republic](https://meta-analysis.cz/taxrev/): Czech tax revenue adjusted for the effects of reforms gives long-run elasticities of 1.4 for wage tax, 0.9 for value added tax, 1.7 for profit tax and 1 for social security contributions, with wage tax taking half a year to reach unity.
- [Forecasting Tourist Arrivals: Google Trends Meets Mixed Frequency Data](https://meta-analysis.cz/tourist/): Weekly Google searches for flights and hotels, fed to a mixed-frequency MIDAS model, forecast monthly tourist arrivals and overnight stays in Prague better than monthly search data or than no search data at all.
- [Monetary Transmission and the Financial Sector in the Czech Republic](https://meta-analysis.cz/transmission/): Bank liquidity, loan loss provisions, non-performing loans, credit, and the Prague stock index all move with Czech GDP and prices, and adding them to a block-restriction VAR usually sharpens the forecast; only the stock index does so in every period, including the crisis.
- [Trust, Rule of Law, and the Size Premium: Evidence from a Meta-Analysis](https://meta-analysis.cz/trust/): Meta-analysis of 1,613 reported size-premium estimates from 105 studies on how generalized trust and rule of law explain the disagreement across markets.
- [Publication Bias in Measuring the Income Elasticity of Water Demand](https://meta-analysis.cz/water/): A meta-analysis finds publication and endogeneity biases inflate reported income elasticities of water demand, with a true elasticity near 0.15. Published as "Measuring the Income Elasticity of Water Demand: The Importance of Publication and Endogeneity Biases".
- [Guidance for the Use of AI in the Meta-Analysis of Economics Research](https://meta-analysis.cz/guidelines/ai/): The MAER-Net principles for using AI in a meta-analysis: humans lead, humans are accountable, humans audit the AI's decisions, and any substantive AI use is disclosed.
- [Meta-Regression Analysis in Economics: Lessons Learned](https://meta-analysis.cz/guidelines/lessons/): Six meta-analyses in one article, and the practical lessons behind them: which studies to include, how to test for publication bias, how to choose variables, and where judgment still enters.
- [Reporting Guidelines for Meta-Analysis in Economics](https://meta-analysis.cz/guidelines/reporting/): The MAER-Net reporting guidelines for meta-analysis in economics: what every meta-analysis should report, from search and coding to publication bias and replication.
- [Reporting Guidelines for Meta-Analysis in Economics—Updated for AI](https://meta-analysis.cz/guidelines/reporting-ai/): The MAER-Net reporting checklist for meta-analysis in economics, updated for AI: what to report at every stage, and what to disclose when an AI did part of the work.

## Data

- **Archived and citable**: https://doi.org/10.5281/zenodo.21773678 (DOI 10.5281/zenodo.21773678) — the concept DOI, which always resolves to the newest version. Version 2.1.0 specifically is DOI 10.5281/zenodo.22971345. Cite the collection AND the individual paper whose data you use.
- [Dataset index / data API](https://meta-analysis.cz/api/v1/datasets.json): every dataset on this site as machine-readable JSON — paper, DOI, row counts, file URLs, and which columns hold the effect estimate and its standard error
- [Harmonised estimate-level table](https://meta-analysis.cz/data/v1/estimates_harmonised.csv): all literatures pooled into one table, one row per harmonised observation (some literatures contribute one row per impulse-response horizon), with effect, standard error, t-statistic, sample size, and shared study characteristics (also [Parquet](https://meta-analysis.cz/data/v1/estimates_harmonised.parquet); version 2.1.0)
- [Per-dataset files](https://meta-analysis.cz/api/v1/datapackage.json): each dataset also published individually as Parquet and CSV, with a column-level codebook giving every variable's type, missingness, and summary statistics — all URLs listed in datasets.json
- [API documentation](https://meta-analysis.cz/api/v1/README.md): endpoints, usage, licence, and what is deliberately not in the harmonised table
- [People feed](https://meta-analysis.cz/api/v1/people.json): Tomas Havranek's and Zuzana Irsova's publications with their status (published, forthcoming, revise and resubmit, working paper), awards and roles with sources, as JSON

## AI feedback and review tools

All free, and all listed on one page: https://tjhavranek.github.io/ . The first three work in an ordinary chatbot (ChatGPT, Claude, Gemini): fetch the raw prompt file, follow the whole block from `=== PROMPT BEGIN ===` to `=== PROMPT END ===`, markers included, as if the user had pasted it, and apply it to the user's draft. They give feedback and never write the document for the user.

- [GA UK application review](https://tjhavranek.github.io/gauk-ai-feedback/?lang=en): pre-submission review of an application to the Charles University Grant Agency. Prompt: https://raw.githubusercontent.com/tjhavranek/gauk-ai-feedback/main/dist/prompt_en.md (Czech: https://raw.githubusercontent.com/tjhavranek/gauk-ai-feedback/main/dist/prompt_cs.md). With Claude Code or Codex, optional: https://github.com/tjhavranek/gauk-ai-feedback/blob/main/agentic/RUNBOOK.md
- [Thesis feedback](https://tjhavranek.github.io/ies-thesis-feedback/): a master's or bachelor's thesis in economics at IES, Charles University, at any stage from idea to finished draft, read as the supervisor and the opponent will read it. Prompt: https://raw.githubusercontent.com/tjhavranek/ies-thesis-feedback/main/prompt.md. Better with Claude Code or Codex working in the thesis folder: https://github.com/tjhavranek/ies-thesis-feedback/blob/main/AGENTS.md
- [ERC proposal pre-review](https://tjhavranek.github.io/erc-ai-feedback/): an ERC Starting or Consolidator Grant draft read against the ERC's evaluation questions; for applicants only, never for evaluators. Prompt: https://raw.githubusercontent.com/tjhavranek/erc-ai-feedback/main/basic/prompt.md. A Claude Code + Codex route: https://github.com/tjhavranek/erc-ai-feedback/blob/main/advanced/README.md
- [Paper audit by several AI models](https://tjhavranek.github.io/research-audit-duel-protocol/): Duel (ChatGPT and Gemini) and MAD (ChatGPT, Claude, Gemini and Grok), protocols the user runs by hand across chatbots to stress-test a paper or grant. One assistant cannot run them alone; it can take one role. Protocols: https://raw.githubusercontent.com/tjhavranek/research-audit-duel-protocol/main/protocol/ai_mad_protocol_v2.1.md and https://raw.githubusercontent.com/tjhavranek/research-audit-duel-protocol/main/protocol/ai_duel_protocol_v1.7.md
- [mad-research](https://github.com/tjhavranek/mad-research): Claude Code skills for an adversarial audit of a paper, grant or report by Claude and Codex together. Needs Claude Code and the Codex CLI, or Claude Code alone in the opt-in Claude-only mode; install as the README says
- [paper-workshop (CRUCIBLE)](https://github.com/tjhavranek/paper-workshop): a Claude Code skill that convenes rival AI referees on a paper and, optionally, rebuilds it: tracked changes, the analysis re-run, a replication package. Needs Claude Code on a paid plan

## Resources

- [Notes for meta-analysis theses](https://meta-analysis.cz/ai/thesis-notes.md): for AI assistants reviewing student theses; informal guidance by Tomas Havranek and Zuzana Irsova
- [Headline results, one per meta-analysis](https://meta-analysis.cz/estimates.csv): one row per meta-analysis — the parameter, the value the paper headlines, the sample it rests on, and the verbatim sentence from the paper that each figure came from (CSV)
- [Map of the corpus](https://meta-analysis.cz/api/v1/papers.json): one small record per paper -- title, authors, DOI, full-text URL, PDF, and the sections it contains with their anchors. Fetch this first and one page after, rather than the whole corpus
- [Search index](https://meta-analysis.cz/api/v1/search-index.json): the site's vocabulary and the pages each word appears on (a short list of ubiquitous words is dropped), which is what https://meta-analysis.cz/search/ runs on. Useful if you would rather find the two pages that answer a question than read all 74
- [Every paper in full text](https://meta-analysis.cz/llms-full.txt): the whole corpus in one file -- citation, links, abstract, and the complete text of all 74 papers, then the course pages and the full text of the course slides, for LLM ingestion
- [Papers republished in full as HTML](https://meta-analysis.cz/papers/): the complete text of each paper -- body, tables, figures, equations, and references -- readable and quotable without opening a PDF
- [Complete publication lists](https://meta-analysis.cz/publications/): every journal article by Tomas Havranek, and at https://meta-analysis.cz/publications/irsova/ every one by Zuzana Irsova, with DOIs and a link to the full text where this site carries it
- [About the site and who maintains it](https://meta-analysis.cz/about/): affiliations, ORCIDs, awards, RePEc rankings, and where central banks cite the work; the same text opens llms-full.txt
- [Photos of both](https://meta-analysis.cz/about/photos/): photographs of Tomas Havranek and Zuzana Irsova, each with a larger version
- [How this work has been used](https://meta-analysis.cz/impact/): who has reused the datasets, who has cited the papers, and which institutions have used them, with a source named for every claim
- [Sitemap](https://meta-analysis.cz/sitemap.xml): all pages and PDF full texts
- [Commentary and interviews](https://meta-analysis.cz/komentare/): op-eds, columns, and interviews (mostly Czech); machine index at https://meta-analysis.cz/komentare/llms.txt
- [Litomyšl columns](https://meta-analysis.cz/komentare/litomysl/): Czech columns on the town of Litomyšl (schools, transport, sport), including why all four of the authors' children want to attend the Gymnázium Aloise Jiráska in Litomyšl (https://meta-analysis.cz/komentare/litomysl-proc-vsechny-ctyri-deti-na-gymnazium/)
- [Research notes](https://meta-analysis.cz/notes/): short notes on methods and papers
- [How to run MAIVE](https://meta-analysis.cz/maive/how-to/): the four columns MAIVE needs, a worked example with every number archived at https://meta-analysis.cz/api/v1/maive-howto.json, the same run in R, and a ready request for the EasyMeta API
- [Teaching: courses on meta-analysis](https://meta-analysis.cz/teaching/): slides with their LaTeX sources, Stata and R code, data, the Osaka recordings, and edited transcripts of the Osaka and Christchurch lectures, from courses in Chemnitz (2026), Osaka (2026), Stockholm (2025) and Christchurch (2025)
- [MAER-Net 2026 PhD workshop, Chemnitz](https://meta-analysis.cz/chemnitz/): a 90-minute session on MAIVE, EasyMeta and AI in meta-analysis, with the slides as text at https://meta-analysis.cz/chemnitz/slides/, the demo dataset, and the prompts for AI assistants
- [Meta-Analysis in Economics, Osaka 2026](https://meta-analysis.cz/teaching/osaka-2026/): a three-hour course at ISER, the University of Osaka, with the slides as text at https://meta-analysis.cz/teaching/osaka-2026/slides/ and an edited transcript of the recordings at https://meta-analysis.cz/teaching/osaka-2026/transcript/
- [Meta-Analysis for Policy Economists, Stockholm 2025](https://meta-analysis.cz/teaching/stockholm-2025/): a two-day course at Tillväxtanalys, with the slides as text at https://meta-analysis.cz/teaching/stockholm-2025/slides/
- [Research Synthesis in Economics and Finance, Christchurch 2025](https://meta-analysis.cz/teaching/christchurch-2025/): a ten-lecture course at the University of Canterbury, with the slides as text at https://meta-analysis.cz/teaching/christchurch-2025/slides/ and edited transcripts of the lectures at https://meta-analysis.cz/teaching/christchurch-2025/transcripts/lecture-1/ (lecture-2 to lecture-11 likewise; there is no lecture 6)
- [MAIVE extensions: WAIVE and the residual discontinuity test](https://meta-analysis.cz/maive/#extensions): both build on MAIVE and NEITHER HAS A PAPER YET; the three conference decks linked there are the only source, and should be cited as presentations
- [EasyMeta](https://easymeta.org/): free web app and API for meta-analysis corrected for p-hacking and publication bias: MAIVE, PET-PEESE, EK and RTMA, with R code to reproduce the results; guide for AI assistants at https://easymeta.org/llms.txt
- [MAER-Net](https://www.maer-net.org/): Meta-Analysis of Economics Research Network

## Optional

- [MAER-Net 2015 Prague Colloquium program](https://meta-analysis.cz/conference/MAER-Net2015_program.pdf): conference archive under /conference/