10.2139/ssrn.2832991

Abstract

We introduce a granular-origins approach to macro-nowcasting using accounting data and deliver three main messages. First, we show that the largest listed firms in each sector of the U.S. economy effectively capture most of the variation in the earnings innovations of all listed firms. Indeed, we find evidence of diminishing marginal benefits from adding more firms in the aggregation process beyond the twenty-five largest firms in each sector. Second, we construct a real-time index of earnings innovations based on the twenty-five largest firms in each sector and find that our index is incrementally relevant for nowcasting current-quarter GDP growth. Third, we find that even though professional nowcasters fully incorporate the nowcasting content of salient variables known to anticipate contemporaneous economic activity, they do not fully incorporate information embedded in our index of earnings innovations. As a result, their nowcast errors of GDP growth are predictable based only on our index. In out-of-sample tests, we document substantial improvements in terms of mean-squared prediction error using our index of earnings innovations, especially when macro uncertainty is high and professional nowcasters disagree the most. Overall, our paper provides evidence that accounting data embeds information that is incrementally relevant for macro-nowcasting and that professional nowcasters can gain timely insights about the current state of the U.S. economy. Our granular-origins approach shows that such insights can be gained in a cost-effective way by focusing on the subset of a few large firms across sectors of the economy.