Local randomization is a technique used in statistical analysis and experimental design where random assignment is applied only to a subset of units or over a small range of the running variable. It allows researchers to approximate a randomized control trial within a specific subset of a population, making it useful in settings where full randomization is infeasible or unethical.