This technical working paper by James H. Stock and Mark W. Watson examines the prevalence of structural instability in macroeconomic time series relations and evaluates the effectiveness of various adaptive forecasting techniques in handling such instability. The study uses a sample of 76 U.S. monthly postwar macroeconomic time series, comprising 5700 bivariate forecasting relations. Formal tests for instability, including Nyblom's test, CUSUM tests, and break point tests, are applied to assess parameter stability. The results indicate widespread instability in both univariate and bivariate autoregressive models. However, adaptive forecasting models, particularly time-varying parameter (TVP) models, show limited success in exploiting this instability to improve forecasts over fixed-parameter or recursive autoregressive models. The paper concludes that while adaptive models occasionally perform well, they generally fail to outperform simpler models in terms of out-of-sample forecast accuracy, suggesting that the class of models considered may not be effective in modeling and exploiting structural instability in macroeconomic applications.This technical working paper by James H. Stock and Mark W. Watson examines the prevalence of structural instability in macroeconomic time series relations and evaluates the effectiveness of various adaptive forecasting techniques in handling such instability. The study uses a sample of 76 U.S. monthly postwar macroeconomic time series, comprising 5700 bivariate forecasting relations. Formal tests for instability, including Nyblom's test, CUSUM tests, and break point tests, are applied to assess parameter stability. The results indicate widespread instability in both univariate and bivariate autoregressive models. However, adaptive forecasting models, particularly time-varying parameter (TVP) models, show limited success in exploiting this instability to improve forecasts over fixed-parameter or recursive autoregressive models. The paper concludes that while adaptive models occasionally perform well, they generally fail to outperform simpler models in terms of out-of-sample forecast accuracy, suggesting that the class of models considered may not be effective in modeling and exploiting structural instability in macroeconomic applications.