This paper examines the impact of sentiment and frequency of news articles on monthly and 12-month stock returns in 51 countries. I show that the results of Calomiris and Mamaysky (2019), who found significant predictive value of news variables, hold when corrected for severe overlapping data issues present in their research. However, large variation in predictive value is revealed by country-level Ordinary Least Squares (OLS), Least Angle Regressions (LARS), and Weighted Least Squares (WLS) procedures. Furthermore, I find that United States’ data is on average better at predicting a country’s stock returns than data of the country itself