Library · Causal inference & statistics
What Fixed Effects Panel Methods got wrong, from 61 theses
Summary written by a model from the records below: The records document methodological challenges and diagnostic failures encountered when applying fixed effects panel estimators across diverse empirical settings. Researchers frequently report severe biases arising from staggered adoption rollouts, incidental parameter inconsistencies in non-linear models, dynamic endogeneity, and the loss of variation when absorbing entity-level characteristics. These records come from graduate theses, 2021 to 2026, each labeled with its degree where the repository states it, and each links to its thesis. They were extracted by language models reading the full text, so treat each as a lead to read, not a verdict.
Plus 5 problems the authors left open, listed at the end.
Failure modes, by theses reporting them
Select a mode to jump to its records.
Two-way fixed effects estimators introduce negative weighting bias under staggered adoption and treatment heterogeneity
Standard two-way fixed effects regressions generate biased estimates and sign reversals when treatment timing is staggered across cohorts with heterogeneous effects. Comparing newly treated units to already treated units creates forbidden comparisons and dynamic weighting distortions that force researchers to adopt stacked or modern robust estimators.
Considered and rejected
Considered and rejected: Rejected standard two-way fixed effects (TWFE) and event study (TWFE-DiD) estimators due to attenuation bias towards zero from staggered rollouts comparing newly treated to already treated units
Essays on income inequality in the United States · UT Austin · PhD
Considered and rejected
Considered and rejected: Rejected using unrefined or standard two-way fixed effects estimators for main panel analyses due to bias introduced by repeated and staggered treatments.
Essays on the Politics of International Law · Harvard · PhD
Tried and failed
two-way fixed effects difference-in-differences estimation applied to staggered policy adoption evaluation. Reason: pervasive negative weights on early-treated cohorts and failure of treatment homogeneity assumptions
Three Essays on the Economics of Agricultural Production Behavior, Renewable Natural Resources, and Welfare Dynamics · Cornell · PhD
Tried and failed
two-way fixed effects with staggered rollouts applied to policy impact estimation across panel data. Reason: negative weighting issues inherent to staggered treatment timing yielded biased, counter-intuitive sign estimates
ESSAYS ON ENVIRONMENTAL POLICIES IN THE TRANSPORTATION SECTOR · Cornell · PhD
Lost to a baseline
Two-Way Fixed Effects (TWFE) estimator overestimated treatment effects compared to Callaway & Sant'Anna in the presence of dynamic treatment effects and staggered rollout.
Essays on education and environmental economics · UT Austin · PhD
Considered and rejected
Considered and rejected: Rejected standard two-way fixed effects (TWFE) difference-in-differences due to treatment effect heterogeneity across adoption cohorts and negative weighting issues.
Essays on applied labour economics: personality traits, artificial intelligence and formalisation policies · Nottingham · PhD
Considered and rejected
Considered and rejected: Rejected standard two-way fixed-effects (TWFE) difference-in-differences models due to bias caused by dynamic/staggered treatment timing and negative weights comparing early-treated to late-treated units.
Three Essays On Economics of Marriage Law · Texas Tech · PhD
Considered and rejected
Considered and rejected: Rejected standard pooled Two-Way Fixed Effects (TWFE) due to bad comparisons and negative weights in staggered treatment settings.
Essays on the Impact and Effectiveness of Share Repurchase Regulations · Harvard · PhD
Considered and rejected
Considered and rejected: Rejected standard two-way fixed effects (TWFE) as primary causal estimator due to bias from negative weighting and dynamic treatment effect heterogeneity under staggered adoption.
Creator platform design · Boston University · PhD
Considered and rejected
Considered and rejected: Rejected Two-Way Fixed Effects (TWFE) and dynamic TWFE models due to negative weighting biases and unreliability when comparing treated to already-treated units under staggered policy timing.
The Implications of Public Policies on Health Economics · University of Nevada, Reno · PhD
Considered and rejected
Considered and rejected: Rejected using standard two-way fixed effects estimators for staggered policy timing due to known weighted average biases and negative weighting issues
Three Essays on Responsiveness to Higher Education Related Policies Addressing Inequality · Harvard · PhD
Considered and rejected
Considered and rejected: Rejected standard two-way fixed effects (TWFE) with staggered adoption in favor of stacked-by-event DiD due to negative weighting and dynamic treatment heterogeneity bias.
Three essays on applied microeconomics · Boston University · PhD
Considered and rejected
Considered and rejected: Standard Two-Way Fixed Effects (TWFE) event-study regression rejected due to dynamic treatment rollout, negative weighting, and cross-lag contamination issues under heterogeneous effects
THREE ESSAYS IN LABOR ECONOMICS · Cornell · PhD
Considered and rejected
Considered and rejected: Rejected relying primarily on two-way fixed effects (TWFE) / party and year fixed effects in main manifesto models due to interpretability issues and bias under treatment effect heterogeneity.
Dovish Reputation Theory: When Fighting To Demonstrate Resolve Backfires · Penn · PhD
Considered and rejected
Considered and rejected: Rejected relying solely on two-way fixed-effects OLS for staggered policy rollouts due to vulnerability to treatment effect heterogeneity and sign-reversals.
ESSAYS ON THE HUMAN CAPITAL AND OCCUPATIONAL CHOICES OF ADOLESCENTS · Cornell · PhD
Considered and rejected
Considered and rejected: Standard two-way fixed effects (TWFE) event study regressions for 3G rollout (rejected due to negative weighting and cohort treatment effect heterogeneity).
Essays in Development Economics · MIT · PhD
Considered and rejected
Considered and rejected: Rejected conventional two-way fixed effects (TWFE) for staggered solar policy adoption due to forbidden comparisons and negative weighting under heterogeneous effects.
Institutions and Operational Frictions in the Energy Transition · Georgia Tech · PhD
Considered and rejected
Considered and rejected: Rejected standard two-way fixed effects (TWFE) / Mean Treatment Effect DiD regressions as primary estimators due to severe bias under heterogeneous treatment effects and staggered timing.
Beyond Trade Losses: Sanctions and the Resilience of Productive Economies · Texas Tech · PhD
Considered and rejected
Considered and rejected: Rejected standard Two-Way Fixed Effects event study because leads/lags mechanically eliminate never-treated units and fail under treatment effect heterogeneity.
Technology, Violent Conflict, and the Determinants of Migration · Georgia Tech · PhD
Considered and rejected
Considered and rejected: Rejected standard Two-Way Fixed Effects (TWFE) with staggered treatment due to time-varying treatment effects leading to biased comparisons with already-treated units, adopting stacked difference-in-differences instead
Essays on infrastructure and urban development in developing countries · Boston University · PhD
Considered and rejected
Considered and rejected: Standard two-way fixed effects (TWFE) model as primary causal estimator, rejected because it forces a common treatment effect across provinces and time when treatment effects vary
The causal effect of carbon pricing on industrial energy consumption in Canada · University of Manitoba · Master's
Considered and rejected
Considered and rejected: Rejected standard two-way fixed-effects staggered DID due to bias from heterogeneous treatment effects across units/time and insufficient state-level sample sizes for date-specific cohorts.
Evaluation and implementation science methods for behavioral health interventions · Johns Hopkins · PhD
Considered and rejected
Considered and rejected: Rejected relying solely on standard Two-Way Fixed Effects (TWFE) without verifying robustness across modern heterogeneous treatment effect DiD estimators due to forbidden comparison bias.
ESSAYS IN PLATFORM GOVERNANCE · Penn · PhD
Tried and failed
two-way fixed effects regression with baseline controls applied to dynamic treatment effect estimation. Outcome: no signal. Reason: failed to capture dynamic treatment heterogeneity across units and time periods
REGIONAL ECONOMIC INEQUALITY OF CHINA: THE TREND AND INTERPRETATIONS FROM THE INSTITUTIONAL PERSPECTIVE · Cornell · PhD
Specification tests reject random effects models when unobserved unit heterogeneity correlates with explanatory variables
Panel analyses frequently evaluate whether entity-level unobserved characteristics violate orthogonality assumptions using Hausman specification diagnostics. When correlation between unobserved effects and covariates is confirmed, researchers reject random effects in favor of fixed effects to prevent omitted variable bias, though fixed effects are occasionally rejected when orthogonality holds to preserve degrees of freedom.
Tried and failed
random effects panel regression applied to executive compensation and institutional assets. Reason: fails to control for unobserved entity-level characteristics correlated with both outcome and explanatory variables
Essays in Applied Microeconomics · Cornell · PhD
Tried and failed
fixed effects panel regression applied to spatial energy burden variation. Reason: Hausman test indicated inability to accurately capture cross-sectional and temporal variations compared to random effects
New Approaches for Analyzing Distributional & Environmental Impacts of Energy & Climate Policies with Focus on Energy Affordability; Using Artificial Intelligence & Causal Inferenc · Georgia Tech · PhD
Considered and rejected
Considered and rejected: Fixed effects estimation rejected in favour of random effects based on Hausman test diagnostics.
The determinants of intrafamily ideological differentiation: Western European social democracy between 1990 and 2019 · LUISS · PhD
Considered and rejected
Considered and rejected: Random-effects model rejected in favor of fixed-effects panel regression based on the Hausman test.
Effect of income diversification on Canadian credit union performance. · Saskatchewan · Master's
Considered and rejected
Considered and rejected: Fixed effects model rejected in favor of random effects model based on the Hausman test for panel analysis.
SOCIAL DETERMINANTS OF HEALTH AND CHRONIC OBSTRUCTIVE PULMONARY DISEASE READMISSIONS: SYSTEM CHARACTERISTICS, COMMUNITY FACTORS, AND PRIMARY CARE. · Johns Hopkins · Doctorate
Considered and rejected
Considered and rejected: Rejected Random-Effects (RE) models in favor of Fixed-Effects (FE) estimators due to Hausman test results indicating unobserved firm-level time-invariant heterogeneity.
An Investigation into the Constructive Role of Task-based Conflict within the Agency Theory · Tübingen · PhD
Considered and rejected
Considered and rejected: Rejected random effects models across all performance and fee specifications based on significant Hausman test results favoring fixed effects.
The Impact of Audit Committees on Audit Quality and Firm Performance - The Case of Jordanian Listed Companies · De Montfort University · PhD
Considered and rejected
Considered and rejected: Rejected random-effects panel regression due to Hausman test confirming inconsistency versus fixed-effects estimator
Fragile Statehood and Military Aid (In)Effectiveness: An Assessment of United States Experiences in Lake Chad Basin · Carleton University · PhD
Considered and rejected
Considered and rejected: Random effects model rejected in favor of fixed effects regression for calculating asset-based income based on Hausman test results.
Dynamics in poverty and livelihood choices in rural Southeast Asia · Leibniz Universität Hannover · PhD
Considered and rejected
Considered and rejected: Rejected random effects in favor of fixed effects panel specification based on Hausman test results.
Counterterrorist Financing and the Evolution and Adaptation of Terrorist Tactics · Carleton University · PhD
Non-linear fixed effects specifications suffer from incidental parameter bias in short panel settings
Estimating fixed effects within non-linear models such as logit, probit, and stochastic frontier specifications produces inconsistent estimates and non-convergence in short panels. Authors consistently reject these non-linear estimators in favor of linear probability models or standard least squares to avoid incidental parameter problems.
Considered and rejected
Considered and rejected: Rejected using Logit/Probit models with fixed effects due to potential inconsistency from incidental parameter problems, selecting OLS with clustered robust standard errors.
Experimental Approaches to Strategy and Innovation · Harvard · Doctorate
Considered and rejected
Considered and rejected: Rejected estimating a fixed-effects probit model with an ordered dependent variable using panel data due to large and persistent bias of the fixed-effects estimator.
Three essays on post-retirement labor · Texas Tech · PhD
Considered and rejected
Considered and rejected: Rejected using logistic regression with fixed effects for binary panel outcomes due to incidental parameter bias, selecting OLS linear probability models instead.
Considered and rejected
Considered and rejected: Rejected logistic regression in favor of linear probability regression due to biased estimation properties in the presence of fixed effects
Incorporating Equity into Healthcare Decision Making Around New Technologies · University of Washington · PhD
Considered and rejected
Considered and rejected: Rejected non-linear fixed effects models (e.g., fixed-effects Logit/Probit) for primary specifications due to the incidental parameter problem, preferring linear probability models and OLS.
ESSAYS ON THE HUMAN CAPITAL AND OCCUPATIONAL CHOICES OF ADOLESCENTS · Cornell · PhD
Considered and rejected
Considered and rejected: Rejected standard logistic regression for the primary state-level policy evaluation in favor of linear probability fixed-effects models, because logistic estimators underestimate standard errors under high-dimensional fixed effects.
Examining the state policies and external organizational ties that affect women’s access to and engagement in substance use disorder treatment services · Boston University · PhD
Tried and failed
fixed effects logit regression applied to short panel discrete choice data. Outcome: no signal. Reason: Incidental parameters bias and severe overfitting in panels with short time dimensions.
Essays In Online Platform Operations · Penn · PhD
Considered and rejected
Considered and rejected: Rejected standard panel fixed effects and maximum likelihood logit with fixed effects on the full dataset due to computational intractability and incidental parameters bias from high freelancer growth relative to panel length
Essays In Online Platform Operations · Penn · PhD
Considered and rejected
Considered and rejected: Rejected True Fixed Effects Stochastic Frontier model due to severe incidental parameter bias when panel length is short (T < 10) relative to cross-sectional units.
Essays on vulnerability, finance and livelihoods of rural households in Southeast Asia · Leibniz Universität Hannover · PhD
Considered and rejected
Considered and rejected: Decided against standard fixed effects SFA due to incidental parameter bias contaminating variance parameters in short panels.
Eficiencia de las sucursales bancarias: el caso de un banco comercial español · University of Las Palmas de Gran Canaria · PhD
Considered and rejected
Considered and rejected: Rejected Greene (2005) True Fixed Effects (TFE) stochastic frontier model due to excessive parameters causing ML non-convergence on unbalanced panel data and incidental parameter bias.
Performance and risk in the Chinese regional ranking sector: a meta-frontier approach · Nottingham · PhD
Tried and failed
estimating high-dimensional fixed effects in nonlinear GMM applied to random-coefficient discrete choice demand estimation. Outcome: unstable. Reason: severe curse of dimensionality and inaccurate parameter estimates from excessive parameters
ESSAYS ON INDUSTRIAL ORGANIZATION AND FIRM DYNAMICS · Cornell · PhD
High-dimensional fixed effects absorb key variation in slowly changing and time-invariant variables
Adding granular unit, individual, or spatial fixed effects eliminates within-group variation needed to estimate the impacts of persistent characteristics. This lack of variation leads to statistically insignificant coefficients, loss of test power, and inability to evaluate non-time-varying regressors.
Tried and failed
unit fixed effects in panel regression applied to labor turnover and mobility models. Outcome: no signal. Reason: Adding individual fixed effects absorbed the variation, making the estimated effect statistically indistinguishable from zero
The Effects Of Racial Segregation And Isolation On Teacher Labor Market Outcomes: Evidence From Pennsylvania · Penn · PhD
Tried and failed
fixed effect and first-difference panel regression applied to slowly changing or time-invariant panel variables. Outcome: no signal. Reason: lack of within-panel variation led to statistically insignificant or wrongly signed coefficient estimates
Human capital, decision-making and performance · Georgia Tech · PhD
Tried and failed
instrumental variable regression with fine spatial fixed-effects applied to estimating historical property rights effects. Outcome: no signal. Reason: historical instrument lacked sufficient spatial variation to survive fine geographic fixed effects
The Economics of Deforestation in the Amazon · Harvard · PhD
Considered and rejected
Considered and rejected: Rejected using year fixed effects interacted with subdivision-specific trends because recreational utility varied at the zip code level, leaving insufficient variation across repeat sales
Three essays on the economics of water pollution control · UT Austin · PhD
Considered and rejected
Considered and rejected: Rejected fixed effects probit models because they omit or fail to estimate effects of non-time-varying or slowly varying characteristics (e.g., age, income, wealth) unless categorized.
Three essays on aging, risk tolerance, financial literacy, and financial satisfaction · Texas Tech · PhD
Static fixed effects models fail to account for dynamic endogeneity and generate Nickell bias
Static panel regressions and naive differencing break down when applied to dynamic processes characterized by inertia and lagged dependent variables. These specifications induce severe downward Nickell bias, transformation bias, and weak instrument problems in short time horizons.
Tried and failed
Static panel OLS and fixed effects estimation applied to translog cost function estimation. Outcome: worse than baseline. Reason: omits dynamic unobserved productivity shocks and endogeneity, causing severe downward bias in returns to scale
Understanding the costs of urban transportation using causal inference methods · Imperial · PhD
Considered and rejected
Considered and rejected: Rejected relying on first-difference panel regressions because misspecification of lead/lag structures severely biased and flipped coefficient signs; selected levels fixed-effects regression instead.
The Impact of Roads on Agglomeration · University of Washington · PhD
Tried and failed
first-differencing to remove fixed effects applied to dynamic panel data estimation. Reason: introduces transformation bias from omitted dynamics unless the autoregressive parameter is exactly one
High-Dimensional Statistics for Causal Inference and Panel Data · MIT · PhD
Considered and rejected
Considered and rejected: Rejected OLS, Fixed Effects, and Difference-GMM estimators due to severe weak instrument bias and failure to address dynamic panel endogeneity in persistent growth time series.
Foreign Investment and Institutional Effectiveness in Resource-Dependent Economies · University at Buffalo · PhD
Tried and failed
static panel regression omitting lagged outcomes applied to panel data with dynamic effects. Reason: omitting past outcomes introduces severe dynamic bias, even under randomized treatment assignment
High-Dimensional Statistics for Causal Inference and Panel Data · MIT · PhD
Omitting fixed effects produces severe omitted variable bias in pooled cross-sectional regressions
Classical pooled regressions fail because unobserved individual, city, firm, or spatial heterogeneity leaks into the error term. Omitting fixed effects leads to severely biased elasticity estimates, incorrect parameter signs, and distorted policy evaluation coefficients.
Tried and failed
omitting industry-time fixed effects in panel regression applied to firm-level labor supply elasticity estimation. Reason: omitted variable bias from industry-wide demand shocks biased elasticity estimates downward
Essays in Labor and Finance · MIT · PhD
Considered and rejected
Considered and rejected: Rejected classical pooled OLS regression in favor of Fixed Effects / Random Effects panel regression to prevent spatial heterogeneity leakage into the error term.
Integrated approaches for understanding the operating environments of water and wastewater infrastructure systems in times of population dynamics · UT Austin · PhD
Considered and rejected
Considered and rejected: Rejected pooled OLS for Cobb-Douglas production function estimation due to omitted variable bias (F-test for individual fixed effects F=167.1, p < 0.001).
STRATEGIES OF TECHNOLOGY ADOPTION FOR MANAGING IT SKILLED LABOR · Penn · PhD
Considered and rejected
Considered and rejected: Rejected pooled OLS regression because omitting city fixed effects produced incorrect positive signs on dependency ratio due to heterogeneity
THE INFLUENCE OF POPULATION AGEING ON HOUSE PRICES IN CHINA · Dalhousie · Master's
Considered and rejected
Considered and rejected: Rejected using conventional cross-sectional demographic spatial regression models due to spatial granularity tradeoffs, privacy limitations, lagged census data, and unobserved spatial fixed effects.
Essays on Risk Avoidance Behaviors · University of Washington · PhD
Differential pre-trends violate parallel trends assumptions in event study fixed effects designs
Two-way fixed effects difference-in-differences designs fail when pre-treatment lead coefficients show statistically significant differences across groups. These violations demonstrate confounding trends such as regional catch-up growth and cause models to fail placebo falsification tests.
Tried and failed
difference-in-differences with two-way fixed effects applied to evaluating policy effects on crime rates. Reason: violation of parallel trends assumption demonstrated by failed pre-trend placebo tests
Three essays on economics of crime and immigration · Texas Tech · PhD
Tried and failed
two-way fixed effects event study design applied to estimating policy effects on charter enrollment. Reason: statistically significant pre-treatment leads violated parallel trends assumption
The Effects of Centralized School Assignment on Public School Enrollment · MIT · Master's
Tried and failed
two-stage least squares with two-way fixed effects applied to panel data with differential pre-trends. Reason: unobserved regional catch-up growth trends confounded the instrumental variable estimates
Essays on the Effects of Immigration on Labor Markets · MIT · PhD
Left open by the authors
Problems the authors named and did not get to.
Left open
Disaggregate regional fixed effects into smaller subregions in OLS panel regressions modeling farm legal business organization shares using USDA Census data. Blocker: None
Factors Impacting Legal Business Organization · Texas Tech · Master's
Left open
Estimate patent citation network regressions with lagged dependent variables using autoregression estimators instead of fixed effects estimators. Blocker: None
Mean field games: theory, approximations, and applications · Imperial · PhD
Left open
Analyze 2SLS identification and estimation rates under increasing-domain asymptotics with community fixed effects eliminated via differencing. Blocker: None
Weak Identification and Network Measurement Error in Peer Effects Estimation · MIT · PhD
Left open
Estimate whether public and private welfare expansions act as complements or substitutes under heightened electoral competition using fixed effects panel regressions. Blocker: None
When Voice Leads to Exit: Democracy, Development, and Private Provision · MIT · PhD
Left open
Extend the directed network jackknife bias correction method to multi-period panel networks with time fixed effects and dynamic splitting. Blocker: None
Essays in Econometrics · MIT · PhD
Who has worked on this
Authors of the public theses cited on this page, most-cited first.
- Sylvia Klosin · MIT · PhD · 2025
High-Dimensional Statistics for Causal Inference and Panel Data - Christa Deneault · Cornell · PhD · 2023
ESSAYS ON THE HUMAN CAPITAL AND OCCUPATIONAL CHOICES OF ADOLESCENTS - Jiding Zhang · Penn · PhD · 2022
Essays In Online Platform Operations - Victoria Colette Aglar · Georgia Tech · PhD · 2026
Institutions and Operational Frictions in the Energy Transition - Matias Alejandro Golman · Nottingham · PhD · 2026
Essays on applied labour economics: personality traits, artificial intelligence and formalisation policies - Rimvydas Ragauskas · Texas Tech · PhD · 2026
Beyond Trade Losses: Sanctions and the Resilience of Productive Economies - Amirali Kamalirad · University of Manitoba · Master's · 2026
The causal effect of carbon pricing on industrial energy consumption in Canada - Shun Yiu · Penn · PhD · 2026
ESSAYS IN PLATFORM GOVERNANCE - Rachyl Kitten · Texas Tech · Master's · 2026
Factors Impacting Legal Business Organization - William Wei Wang · MIT · PhD · 2025
Weak Identification and Network Measurement Error in Peer Effects Estimation - Elliot Christian Tobin · Harvard · PhD · 2025
Essays on the Impact and Effectiveness of Share Repurchase Regulations - Ahmet Gulek · MIT · PhD · 2025
Essays on the Effects of Immigration on Labor Markets
Checking a claim in this area?
We can run the same search on any method or claim. If nothing turns up, we will say so, and that proves nothing on its own.