This paper studies the interplay between the sectoral allocation of credit and longrun economic development. We document new Financial Kuznets Facts: as economies grow, (i) the share of manufacturing credit relative to value added falls, (ii) the share of real estate credit rises, and (iii) the reliance on and price of real estate collateral increase. A two-sector structural change model with collateral constraints explains these patterns through an economic channel (a rise in manufacturing productivity) and a financial channel (a relaxation of real estate financing constraints). We provide several pieces of reduced-form evidence that the financial channel asymmetrically increases real estate credit over development. Further, new data reveal that the liberalization of directed credit policies is associated with a reallocation of credit from manufacturing to real estate, as such policies historically prioritized manufacturing. Finally, we document that manufacturing credit predicts higher long-run growth, while real estate credit predicts lower growth, suggesting that the allocation of credit may matter for understanding long-run growth.
@misc{dai_muller_verner_2025,
author = {Dai, Paul and M{\"u}ller, Karsten and Verner, Emil},
title = {Credit Allocation, Collateral, and Economic Development},
year = {2025},
howpublished = {\url{https://ssrn.com/abstract=5079393}}
}
Using a new dataset on sectoral credit exposures in 115 economies from 1940 to 2014, we provide evidence that corporate debt plays a key role in explaining macroeconomic boom-bust cycles, financial crises, and sluggish recoveries. We find that: (i) corporate debt accounts for two-thirds of aggregate credit expansion and three-quarters of nonperforming loans during downturns; (ii) expansions in corporate debt predict crises, conditional on expansions in household credit; (iii) firm credit growth backed by real estate collateral and cash flows is linked to future crises; (iv) dispersion in firm credit growth predicts crises; and (v) financial crises following booms in corporate debt are associated with slower recoveries.
@techreport{NBERw32225,
title = {Corporate Debt, Boom-Bust Cycles, and Financial Crises},
author = {Ivashina, Victoria and Kalemli-\"Ozcan, \c{S}ebnem and Laeven, Luc and M\"uller, Karsten},
institution = {National Bureau of Economic Research},
type = {Working Paper},
series = {Working Paper Series},
number = {32225},
year = {2024},
month = mar,
doi = {10.3386/w32225},
url = {http://www.nber.org/papers/w32225}
}
Bank runs are a central concern for financial stability, yet systematic empirical evidence remains scarce. We construct a novel historical dataset of bank runs, covering 184 countries since 1800 by combining narrative evidence from 503 sources with statistical indicators of aggregate deposit contractions. We find that: (i) the unconditional likelihood of a bank run is 1.9%; (ii) systemic runs—those accompanied by aggregate deposit outflows—are associated with output losses of 9% over five years, more than after non-systemic runs or deposit contractions alone; (iii) these losses persist even when banks are well capitalized and there is no evidence of fundamental triggers, banking crises, or widespread bank failures; (iv) central banks and deposit insurance are linked to a lower probability of runs becoming systemic, while liability guarantees coincide with smaller output losses.
@TechReport{JKMS2024,
author = {Jamilov, Rustam and K\"onig, Tobias and M\"uller, Karsten and Saidi, Farzad},
title = {{Two Centuries of Systemic Bank Runs}},
year = {2024},
month = aug,
institution = {C.E.P.R. Discussion Papers},
type = {CEPR Discussion Papers},
number = {19382},
url = {https://ideas.repec.org/p/cpr/ceprdp/19382.html}
}
Using a newly-constructed panel dataset of U.S. states from 1863 to 2022 that combines bank balance sheets, real economic activity, and a systematic survey of all major chronologies of U.S. financial crises, we document the following facts: (i) financial crises are followed by a 6% decline in state-level output, (ii) output losses vary substantially across states, (iii) the severity of output losses is predictable with local contractions in deposits or wholesale liabilities, and with the incidence of bank failures, (iv) a composite measure of local financial distress, combining narrative evidence with statistical indicators, predicts state-level output losses of 3%, and (v) the share of states experiencing local financial distress predicts national output beyond a binary crisis indicator. These findings suggest that studies of systemic crises may underestimate the frequency and costs of financial distress.
@article{crises2025,
author = {Hoon, Joseph and Liu, Chang and Payne, Jonathan and M\"{u}ller, Karsten and Zheng, Zhongxi},
title = {The Costs of Financial Crises in the United States},
journal = {Working Paper},
year = {2025}
}
We construct a novel dataset of 60 macroeconomic time series at the U.S. state level, spanning from the 1863 to the present, based on digitizing and harmonizing 113 historical sources. Equipped with these data, we estimate an annual index of state-level economic activity over nearly 160 years. This index aligns closely with official indicators such as state GDP and unemployment when available. Using this measure of economic activity, we uncover several new facts about state-level business cycles: (1) there is substantial heterogeneity across states in both cyclical dynamics and their underlying drivers; (2) business cycles have become more synchronized since World War II; and (3) downturns have become shorter and recoveries quicker over time.
@article{datapaper2025,
author = {Hoon, Joseph and Liu, Chang and M\"{u}ller, Karsten and Zheng, Zhongxi},
title = {{U.S. State-Level Business Cycles Since the Civil War}},
journal = {Working Paper},
year = {2025}
}
The Global Macro Database is an open-source, continuously updated dataset of macroeconomic statistics that unifies and extends existing resources. By harmonizing and integrating data from 32 major contemporary sources—including the IMF, World Bank, and OECD—with historical records from 78 additional datasets, we construct comprehensive annual time series for 46 variables across 243 countries. This database covers global macroeconomic trends from the origins of modern data collection to projected estimates for 2030. Using this extensive database, we study the long-run output losses of financial crises and global temperature shocks, two applications in which historical time series are a crucial input. Our findings show that financial crises are associated with statistically detectable contractions in real GDP for five decades into the future, which are considerably larger than previously estimated. Temperature shocks also predict real GDP contractions up to 30 years ahead, especially in emerging economies.
@techreport{NBERw33714,
title = {The Global Macro Database: A New International Macroeconomic Dataset},
author = {M\"uller, Karsten and Xu, Chenzi and Lehbib, Mohamed and Chen, Ziliang},
institution = {National Bureau of Economic Research},
type = {Working Paper},
series = {Working Paper Series},
number = {33714},
year = {2025},
month = apr,
doi = {10.3386/w33714},
url = {http://www.nber.org/papers/w33714}
}
We study the role of Export Credit Agencies—the predominant tool of modern industrial policy—on exports and firm investment by using the effective shutdown of the Export-Import Bank of the United States (EXIM) from 2015–2019 as a natural experiment. We document sizable real effects of the shutdown: a $1 reduction in EXIM trade financing reduces exports by approximately $4.50. EXIM-dependent firms experience a contraction in total revenues, investment, and employment. EXIM's shutdown has the largest effects for exporters facing financing frictions and selling to markets with high contractual frictions, indicating a plausible underprovision of trade financing by private financial institutions. Consistent with these findings, we find that the shutdown increased the misallocation of capital because it particularly affected firms with a higher ex-ante marginal revenue product of capital.
@techreport{NBERw32019,
title = {EXIM's Exit: The Real Effects of Trade Financing by Export Credit Agencies},
author = {Kabir, Poorya and Matray, Adrien and M\"uller, Karsten and Xu, Chenzi},
institution = {National Bureau of Economic Research},
type = {Working Paper},
number = {32019},
year = {2024},
month = jan,
doi = {10.3386/w32019},
url = {http://www.nber.org/papers/w32019}
}
Using plausibly exogenous variation in regional Twitter adoption in the United States, we show that a 10% increase in social media usage causes a 2.5% rise in stock ownership. Consistent with lowering the costs of acquiring information, Twitter has larger effects in counties with low pre-existing stock market knowledge, improves knowledge about asset returns, and leads to a decline in the number of financial advisors. Social media also boosts interest in volatile "meme stocks" favored by retail investors. Our findings highlight the unique influence of social media on household portfolio decisions, distinct from other modern information technologies.
@TechReport{repec:cpr:ceprdp:18445,
type = {CEPR Discussion Papers},
institution = {C.E.P.R. Discussion Papers},
author = {M\"uller, Karsten and Pan, Yuanyuan and Schwarz, Carlo},
title = {Social Media and Stock Market Participation},
year = {2023},
month = sep,
number = {18445},
doi = {None}
}
Social media platforms are often credited with empowering grassroots movements in the pursuit of political freedoms. In this paper, we show how social media can also be exploited by political elites to undermine democratic institutions, using the January 6th, 2021 Capitol insurrection as a case study. We present three main findings. First, by exploiting plausibly exogenous variation in Twitter usage, we document that social media exposure predicts participation in the Capitol attack, donations for anti-democratic causes, beliefs in election fraud, and support for the January 6th rioters. Second, Donald Trump's tweets questioning the election's integrity were followed by spikes in "Stop the Steal" activity on Twitter and pro-Trump donations originating from high Twitter usage counties. Third, the insurrection and Trump's account deletion were followed by a decrease in the public expression of toxic political and "Stop the Steal" messaging by pro-Trump users on Twitter, but had little effect on privately held beliefs about the election outcome and pro-Trump donations.
@techreport{MullerSchwarz2026_SocialMediaVsDemocracy,
author = {M{\"u}ller, Karsten and Schwarz, Carlo and Shen, Zekai},
title = {{Social Media vs. Democracy}: Evidence from the January 6th Insurrection},
year = {2026},
month = feb,
doi = {10.2139/ssrn.4296306},
url = {https://ssrn.com/abstract=4296306}
}
Social media companies are under scrutiny for the prevalence of hateful content on their platforms, but there is scarce empirical evidence of the consequences of regulating such content. We study this question with a particular focus on anti-refugee hate crime in the context of the "Network Enforcement Act" (NetzDG) in Germany, which mandates major social media companies to remove hateful posts within 24 hours. Using a difference-in-differences strategy, we find that the law was associated with a 4% reduction in the toxicity of refugee-related tweets by far-right social media users. Further, we show that the NetzDG reduced anti-refugee hate crimes in municipalities with more far-right social media users.
@article{JimenezMuellerSchwarz2023,
title = {{The Effect of Content Moderation on Online and Offline Hate: Evidence from Germany's NetzDG}},
author = {Jim\'enez Dur\'an, Rafael and M\"uller, Karsten and Schwarz, Carlo},
year = {2022},
url = {https://ssrn.com/abstract=4230296}
}
An extensive literature studies interactions of stock market anomalies using double-sorted portfolios. But given hundreds of known candidate anomalies, examining selected interactions is subject to a data mining critique. In this paper, we conduct a comprehensive analysis of all possible double-sorted portfolios constructed from 102 underlying anomalies. We find hundreds of statistically significant anomaly interactions, even after accounting for multiple hypothesis testing. An out-of-sample trading strategy that invests in the top backward-looking double-sort strategy generates equal-weighted (value-weighted) monthly average returns of 4% (2.7%) at an annualized Sharpe ratio of 2 (1.38), on par with state-of-the-art anomaly-based machine learning strategies.
@article{MuellerSchmickler2021,
title = {Interacting Anomalies},
author = {M\"uller, Karsten and Schmickler, Simon},
journal = {Review of Asset Pricing Studies},
volume = {15},
number = {2},
pages = {162--216},
year = {2025}
}
We study the relationship between credit expansions, macroeconomic fluctuations, and financial crises using a novel database on the sectoral distribution of private credit for 117 countries since 1940. We document that, during credit booms, credit flows disproportionately to the non-tradable sector. Credit expansions to the non-tradable sector, in turn, systematically predict subsequent growth slowdowns and financial crises. In contrast, credit expansions to the tradable sector are associated with sustained output and productivity growth without a higher risk of a financial crisis.
@article{MuellerVerner2024,
author = {Karsten M\"uller and Emil Verner},
title = {{Credit Allocation and Macroeconomic Fluctuations}},
journal = {The Review of Economic Studies},
year = {2024},
volume = {91},
number = {6},
pages = {3645--3676}
}
We study how social media affects election outcomes. We exploit variation in the number of Twitter users across U.S. counties based on early adoption among participants of the 2007 South by Southwest (SXSW) festival—a key event in Twitter's rise to popularity. We show that this variation, which remains predictive of Twitter use a decade later, is unrelated to electoral outcomes before the platform's mass adoption. Our results suggest that exposure to Twitter lowered the Republican vote share in the 2016 presidential election but had limited effects on turnout and vote shares in House and Senate races as well as previous presidential elections.
@article{FujiwaraMuellerSchwarz2024,
author = {Thomas Fujiwara and Karsten M\"uller and Carlo Schwarz},
title = {{The Effect of Social Media on Elections: Evidence from The United States}},
journal = {Journal of the European Economic Association},
year = {2024},
volume = {22},
number = {3},
pages = {1495--1539}
}
This paper examines the relationship between entertainment television and political outcomes, focusing on Donald Trump's reality TV show "The Apprentice" and its impact on his electoral success.
@article{MuellerSchwarz2024Apprentice,
author = {M\"uller, Karsten and Schwarz, Carlo},
title = {From Apprentice to President? Entertainment TV and US Elections},
journal = {The Leadership Quarterly},
volume = {35},
number = {3},
pages = {101758},
year = {2024}
}
Do politics matter for macroprudential policies? I show that changes in macroprudential regulation exhibit a predictable electoral cycle in the run-up to 221 elections across 58 countries from 2000 through 2014. Policies restricting mortgages and consumer credit are systematically looser before elections, particularly during economic expansions. Consistent with theories of opportunistic political cycles, this pattern is stronger when election outcomes are uncertain, regulators are closely tied to politicians, and institutions are poor.
@article{MuellerElectoralCycles,
author = {M\"uller, Karsten},
title = {{Electoral Cycles in Macroprudential Regulation}},
journal = {American Economic Journal: Economic Policy},
year = {2023},
volume = {15},
number = {4},
pages = {295--322}
}
We study whether social media can amplify anti-minority sentiments with a focus on Donald Trump's political rise. Using an instrumental variable strategy based on Twitter's early adopters at the South by Southwest festival in 2007, we find that higher Twitter use in a county is associated with a sizeable increase in anti-Muslim hate crimes after the 2016 presidential primaries. Trump's tweets about Muslims predict increases in xenophobic tweets by his followers, cable news mentions of Muslims, and hate crimes on the following days.
@article{Mueller2019Trump,
title = {From Hashtag to Hate Crime: Twitter and Anti-Minority Sentiment},
author = {M\"uller, Karsten and Schwarz, Carlo},
journal = {American Economic Journal: Applied Economics},
year = {2023},
volume = {15},
number = {3},
pages = {270--312}
}
This paper estimates the effect of bankruptcy court caseload on access to credit by exploiting firms' plausibly exogenous exposure to the largest recorded drop in court backlog in the United States following the 2005 consumer bankruptcy reform. I show that a drop in court congestion reduces the time firms spend in bankruptcy and increases recovery values, which is priced into credit spreads and loan maturities. Consistent with a shock to credit supply, less congested courts increase firm leverage but leave default risk unchanged.
@article{MULLER2021,
title = {{Busy Bankruptcy Courts and the Cost of Credit}},
journal = {Journal of Financial Economics},
year = {2022},
volume = {143},
number = {2},
pages = {824--845},
author = {Karsten M\"uller},
doi = {https://doi.org/10.1016/j.jfineco.2021.08.010}
}
This paper investigates the link between social media and hate crime. We show that antirefugee sentiment on Facebook predicts crimes against refugees in otherwise similar municipalities with higher social media usage. To establish causality, we exploit exogenous variation in the timing of major Facebook and internet outages. Consistent with a role for "echo chambers," we find that right-wing social media posts contain narrower and more loaded content than news reports. Our results suggest that social media can act as a propagation mechanism for violent crimes by enabling the spread of extreme viewpoints.
@article{MuellerSchwarzJEEA2020,
author = {M\"uller, Karsten and Schwarz, Carlo},
title = {{Fanning the Flames of Hate: Social Media and Hate Crime}},
journal = {Journal of the European Economic Association},
volume = {19},
number = {4},
pages = {2131--2167},
year = {2021},
doi = {10.1093/jeea/jvaa045}
}
How do changes in banking regulation affect the syndicated loan market? Because branch networks and loan syndication both enable banks to diversify geographical credit risk, we investigate the staggered implementation of the Riegle–Neal Interstate Branching and Banking Efficiency Act of 1994. Exploiting that the act only changed the legal framework for out-of-state commercial banks, we find that branching deregulation decreased syndicated loan issuance but spurred bilateral lending to corporations.
@article{keil_mueller_2020,
title = {Bank Branching Deregulation and the Syndicated Loan Market},
volume = {55},
number = {4},
journal = {Journal of Financial and Quantitative Analysis},
publisher = {Cambridge University Press},
author = {Keil, Jan and M\"uller, Karsten},
year = {2020},
pages = {1269--1303}
}
The County Business Pattern ("CBP") files contain employment and establishment counts for detailed industry codes covering all counties in the United States. The contribution of this project is to digitize, clean, and prepare the CBP files from 1946–1974. We also apply the methods developed in Eckert, Fort, Schott, and Yang (2020a) to impute missing employment observations in the raw data. We provide three digital data products for public use.
@techreport{NBERw30578,
title = {The Early County Business Patterns Files: 1946-1974},
author = {Eckert, Fabian and Lam, Ka-leung and Mian, Atif R and M\"uller, Karsten and Schwalb, Rafael and Sufi, Amir},
institution = {National Bureau of Economic Research},
type = {Working Paper},
number = {30578},
year = {2022},
month = oct,
doi = {10.3386/w30578}
}
Do financial stability risks originate in the growth of lending to particular sectors? Should financial regulators target individual sectors, such as mortgages, and should such tools vary over time? In this paper, I discuss what recent research can tell us about these questions and discuss some new evidence based on a dataset on sectoral credit for 120 countries from 1940 to 2014.
@article{Mueller2019Alloc,
title = {{Does the Sectoral Allocation of Credit Matter for Financial Stability Risks?}},
author = {Karsten M\"uller},
year = {2022}
}