When they come into existence, for example, startup firms create, on average, three millions jobs per year. Many of these jobs are lost and new ones are created, but employment at the moment of startup is not reached again in subsequent years.
They are looking at the ever-fascinating data set on business dynamics. The whole paper is interesting, although the language and methods may seem unfamiliar to many economists.
The future of empirical macro is in the new labor dynamics (JOLTS) and business dynamics data sets. I predict that if the CBO, Mark Zandi, and other traditional macro modelers continue to ignore these data sets, they will be stuck in a time warp, like somebody who doesn't know how to use email or a cell phone.
I also think that these data sets will change macroeconomic theory. I think they will bring more people around to thinking of economic activity as patterns of sustainable specialization and trade, rather than as aggregates of spending.