This paper quantitatively assesses whether firm-specific shocks can drive the U.S. business cycle. Firm-specific shocks to the largest firms can directly contribute to aggregate fluctuations whenever the firm size distribution is fat-tailed giving rise to the granular hypothesis. I use a novel, comprehensive data set compiled from administrative sources that contains the universe of firms and trade transactions, and find that the granular hypothesis accounts at most for 16 percent of the variation in aggregate sales growth. This is about half of that found by previous studies that imposed Gibrat’s law where all firms are equally volatile regardless of their size. Using the full distribution of growth rates among U.S. firms, I find robust evidence of a negative relationship between firm-level volatility and size, i.e. the size-variance relationship. The largest firms (whose shocks drive granularity) are the least volatile under the size-variance relationship, thus their influence on aggregates is mitigated. I show that by taking this relationship into account the effect of firm-specific shocks on observed macroeconomic volatility is substantially reduced. I then investigate several plausible mechanisms that could explain the negative size-variance relationship. After empirically ruling out some of them, I suggest a “market power” channel in which large firms face smaller price elasticities and therefore respond less to a given sized productivity shock than small firms do. I provide direct evidence for this mechanism by estimating demand elasticities among U.S. manufactures. Lastly, I construct an analytically tractable framework that is consistent with several empirical regularities related to firm size.
|US Census Bureau Center for Economic Studies Paper