microeconometrics

Microeconometrics is a branch of econometrics that analyzes micro-level data—such as information on individuals, households, or firms—to understand economic behavior and decision making. It emphasizes causal inference and the treatment of heterogeneity across units and contexts, with attention to issues of endogeneity, selection, and measurement error. The discipline develops and applies models grounded in microeconomic theory to estimate relationships at the level of decision making rather than aggregate outcomes.

Common methods include panel data models (fixed effects, random effects) to exploit within-unit variation; instrumental variables and two-stage least squares to address endogeneity; and limited dependent variable models (logit, probit, tobit) for discrete or censored outcomes. Count data models (Poisson, negative binomial), duration (survival) models, and discrete choice models are used for outcomes such as labor supply, consumer demand, or procurement choices. Nonlinear, sample-selection, and duration models (e.g., Heckman selection model) address selection bias; quantile regression captures effects across the outcome distribution. More recently, causal inference techniques such as differences-in-differences, regression discontinuity, and natural experiments are employed to identify effects when randomized experiments are not feasible.

Data sources include cross-sectional surveys, panel surveys, administrative records, and experimental or quasi-experimental data. Estimation methods include maximum likelihood, generalized method of moments, and Bayesian approaches. Researchers assess identification, robustness, and external validity, and often combine structural models with reduced-form specifications to test economic theory. Applications span labor economics (wage equations, employment effects), consumer demand and price effects, education and health, development, and firm productivity.

Microeconometrics intersects with but remains distinct from macroeconometrics by prioritizing micro-level variation and causal mechanisms. It has a rich methodological literature and widely used software in R, Stata, and Python.