This thesis studies explanations for the existence of currency anomalies and time-varying parameter effects of exchange-rate predictability within the context of currency markets. The thesis comprises three essays. The first essay examines the relation between currency anomalies and academic research. Using real-time data, currency anomalies are profitable during in-sample and out-of-sample periods both before and after transaction costs, but trading profits decrease substantially after the publication of academic research. The decline is greater for anomalies with larger in-sample period profits and lower arbitrage costs, and signal performance decays quickly. This finding is consistent with the idea that academic research draws trading attention to currency anomalies. The second essay relates currency anomalies and foreign exchange analysts, where mispricing is systematically found related to mistakes and changes in analysts’ currency forecasts. In particular, analysts expect anomaly payoffs that are too low or even negative compared to actual anomaly profits. While analysts’ mistakes decrease after anomaly publication and analysts update their forecasts to incorporate lagged anomaly information, trading profits from mispricing are more than three times those using analysts’ forecasts. These results are consistent with a behavioral explanation for currency anomalies. The third essay constructs bilateral measures of cyclical external imbalances to predict exchange-rate returns in a framework that allows for the parameters of the forecasting regression to vary over time. A strategy using bilateral measures of cyclical external imbalances exhibits high economic value relative to the random walk benchmark in short horizons of one quarter ahead. Predictive regressions employing constant parameter models are found to be inferior to time varying coefficient parameter models, suggesting a dynamic relationship between bilateral cyclical imbalances and exchange-rate returns that vary over time.