This thesis investigates the theoretical definition, econometric estimation, and empirical realities of labor-market power. Its main chapter focuses on US manufacturing, finding that manufacturing workers received $1.10 cents of each marginal unit of production in 1973 but only 48 cents in 2014. I find that labor-market power is strongly correlated with adoption of ICT and automation technology. My estimates imply that labor-market power is significantly more important than markup power in US manufacturing. In the second chapter, I show that when current production-based "ratio estimators" of market power return values different from 1, they imply either model misspecification or input frictions (such as labor market power or labor adjustment costs). I argue that the latter interpretation is more plausible in practice, and that researchers and policymakers should largely reinterpret existing work using these estimators as measuring input frictions. Finally, in the third chapter I propose an estimation method that produces unbiased and consistent estimates of labor-market and markup power by flexibly modeling markups as a specified function of observables and fixed effects. My modified two-step estimator is simple in concept and implementation, requiring less onerous assumptions than popular proxy production estimators. In sum, this thesis argues that labor-market power is an important feature of the US labor market, and provides theoretical support and estimation techniques to study it further.