Researchers and practitioners employ a variety of time-series processes to forecast betas, using either short-memory models or implicitly imposing infinite memory. We find that both approaches are inadequate: beta factors show consistent long-memory properties. For the vast majority of stocks, we reject both the short-memory and difference-stationary (random walk) alternatives. A pure long- memory model reliably provides superior beta forecasts compared to all alternatives. Finally, we document the relation of firm characteristics with the forecast error differentials that result from inadequately imposing short-memory or random walk instead of long-memory processes.