We find that analyst forecasts are more optimistic and have larger errors near holidays, but more pessimistic and have smaller errors when there is a disaster with significant fatalities. These results are neither explained by sentiment associated with contemporaneous economic conditions, nor by under-reaction or over-reaction to more bad news released on days immediately before weekends or holidays. Overall, our results are consistent with the notion that when analysts are in a positive (negative) mood, they generally make more positively (negatively) biased and less (more) accurate forecasts.