Last weekend in The New York Times, business columnist Floyd Norris wrote that new data from the Bureau of Labor Statistics "challenge" the notion that small companies create more jobs over time than large companies. Indeed, according to these numbers, from 1990 to 2011, employment in big firms (defined in the dataset as those with more than 500 employees) increased by 29 percent, while it rose by only 10.5 percent in firms with 49 or fewer employees.
The debate over how job creation breaks down by firm size is one with which the Kauffman Foundation is familiar and we have put out our own papers on this topic as well as worked with those who are knee-deep in the data. It shouldn't be a surprise to anyone that looking at job creation solely through the lens of firm size gives an incomplete picture. Elsewhere, the important distinction between firm age and firm size has been adequately underscored, so we won't dwell on it here. Here are just a few points about the Norris column that reveal some of the nuance under the numbers, while confirming that aggregate data are true--mostly.
First, as Norris points out, lumping all job creation together over a long period of time conceals moves of companies between size categories, whether from growing firms or seasonal variation. Second, these job creation numbers obviously include acquisitions, which will usually be reflected only in additions to large company employment numbers and subtraction from small company employment. Not always, of course, but usually. Third, we have known for a while, and discuss this is a forthcoming paper, that a larger and larger share of employment has been moving to large companies over time. Today, the share of total employment in the largest companies, those with over 10,000 employees, is the largest it has ever been in the United States.
Fourth, aggregate numbers mask all the variation beneath the surface, which is really where the real action of reallocation and productivity and economic growth takes place. Other recent work, for example, demonstrates that job churn--moving from one job to another--represents two-thirds of hiring in any given year. According to the Business Dynamics Statistics series from the Census Bureau, the "excess reallocation rate" has averaged around 30 percent since 1977--this is the gross job creation rate plus the gross job destruction rate, less the net job creation rate, capturing the persistently high level of churn in the U.S. economy. I don't know the breakdown of job-to-job flows in terms of firm size, but my guess would be that it is biased, if slightly, toward big firms. In any case, simply counting job growth by firm employment size over a twenty-year period misses what is really important about job creation and economic activity.
Finally, as with net job creation by firm age, part of the size calculation is a function of the math. There is a definitional limit on the share of job growth that can be accounted for by small and medium-sized companies. They can only account for job growth up to 49 and 499 employees, respectively. In contrast, there is an unlimited size level for job growth in big companies. (Theoretically, at least: if we use actually firm size numbers we might say that job growth in big companies can occur within a band of 500 employees and 1 million employees, roughly the size of Walmart.) Job growth can occur, along the vector of size at least, in two ways: more employees at existing firms, and more firms of certain sizes. Within the first, there is a statistical limit by which small companies can comprise relative job growth.
So, as with any statistic, the data Norris cites are true, but for certain reasons that, once understood, reveal a more nuanced--and far more interesting--picture of economic change.