Fixed Effects

Individual staffers at international organizations matter for policy output: mission chiefs at the IMF bring heterogeneous ideological biases to their jobs and those biases shape the policies the organization imposes.

This paper develops a dynamic discrete choice framework for multiple attitudes to identify which attitudes are most influential within an attitude system and to distinguish sources of stability and instability. The model accommodates ordinal attitudinal measures, time-varying gaps between interviews, and a multiple-factor structure to capture time-invariant individual traits. The authors apply the approach to five core political attitudes and to issue attitudes about government spending and immigration, preserving qualifications about the roles of permanent heterogeneity and transitory shocks.

This article asks how two types of large-scale conflict—major interstate wars and major civil wars—relate to children’s health, measured by infant mortality rates. Using country-year analysis across 1950–2007, the author examines both immediate and cumulative associations between time spent at war and infant mortality, and emphasizes what this evidence says about violations of children’s right to health without making causal claims beyond the study’s design.

Pepinsky, Goodman, and Ziller (2024, American Political Science Review, PGZ) reevaluate a recent study on the long-term consequences of concentration camps in Germany and conclude that controlling for contemporary (post-treatment) state heterogeneity yields unbiased estimates of camps' effects on present-day outgroup intolerance.

This study asks whether high school civic education influences voter turnout in adulthood by merging detailed academic transcript information on social studies and civics coursework with a large-scale, longitudinal survey of a nationally representative cohort of adolescents. The authors present a sequence of regression models: initial specifications show apparent associations between civics coursework and later turnout, but the associations shrink after adding individual and family controls and disappear in family fixed-effects models that compare siblings. The analysis concludes that any causal effect of high school civic instruction on adult voting is at best very small, and the study finds no evidence that courses emphasizing service learning, civic skills, or political issues increase turnout.

Multilevel models (MLMs) with random effects and fixed effects (FE) models with cluster-robust standard errors are two common choices for grouped data. This paper clarifies how the approaches work, documents widespread practical problems, and shows how simple adjustments reconcile the two methods.

Ethnic divisions in rapidly urbanizing Kenya have long been debated. "First-generation" modernization theories expected urban migration to reduce ethnic identification and improve intergroup trust, while "second-generation" perspectives predicted the opposite.

When estimating grouped data with a binary dependent variable and fixed effects, linear (OLS) and logit specifications can produce different results. The apparent reason is that logit excludes groups where the outcome is all zeros or all ones, while the linear specification retains all groups.

Conditioning on observed covariates is a standard strategy for reducing confounding in non-experimental causal inference, but some covariates can increase rather than decrease bias. This article develops a clearer way to think about when and why that happens by decomposing omitted-variable bias into three distinct parts and studying the mechanisms that produce bias increases when conditioning.