This paper explores the relationship between long memory and regime switching in econometric models. The authors argue that these two phenomena are closely related and can be easily confused, especially when only a small amount of regime switching occurs. They examine this relationship in several models, including a simple mixture model, Engle and Lee's stochastic permanent break model, and Hamilton's Markov switching model. The authors show analytically that stochastic regime switching can be mistaken for long memory, even asymptotically, if only a "small" amount of regime switching occurs. A Monte Carlo analysis supports the relevance of the theory and provides additional insights. The paper begins by discussing the theoretical and empirical literatures on long memory and regime switching, which have evolved largely independently. The authors then consider alternative definitions of long memory and the mechanisms for generating long memory. They review the mechanisms for generating long memory that have been stressed in the literature, which are very different from those that they develop and therefore provide interesting contrast. The authors then consider several simple models of structural change, or more precisely, stochastic regime switching, and show when and why they produce realizations that appear to have long memory. They present a Monte Carlo analysis that verifies the predictions of the theory and produces additional insights. They conclude that long memory and structural change are easily confused, and that this confusion is not merely a theoretical curiosity, but rather a distinct possibility in routine empirical economic and financial applications. The paper also discusses related work and provides additional perspective on their results.This paper explores the relationship between long memory and regime switching in econometric models. The authors argue that these two phenomena are closely related and can be easily confused, especially when only a small amount of regime switching occurs. They examine this relationship in several models, including a simple mixture model, Engle and Lee's stochastic permanent break model, and Hamilton's Markov switching model. The authors show analytically that stochastic regime switching can be mistaken for long memory, even asymptotically, if only a "small" amount of regime switching occurs. A Monte Carlo analysis supports the relevance of the theory and provides additional insights. The paper begins by discussing the theoretical and empirical literatures on long memory and regime switching, which have evolved largely independently. The authors then consider alternative definitions of long memory and the mechanisms for generating long memory. They review the mechanisms for generating long memory that have been stressed in the literature, which are very different from those that they develop and therefore provide interesting contrast. The authors then consider several simple models of structural change, or more precisely, stochastic regime switching, and show when and why they produce realizations that appear to have long memory. They present a Monte Carlo analysis that verifies the predictions of the theory and produces additional insights. They conclude that long memory and structural change are easily confused, and that this confusion is not merely a theoretical curiosity, but rather a distinct possibility in routine empirical economic and financial applications. The paper also discusses related work and provides additional perspective on their results.