The Bug in SBA's Calculations:

To get a better fix on why the regression fallacy, sorting firms by size class in a dynamic situation, produces an unacceptable result, let's look at a hypothetical case.

Assume a three-year employment history of four firms, as shown below:

Year Firm 1 Firm 2 Firm 3 Firm 4

Y1 450 550 200 800

Y2 550 450 800 200

Y3 450 550 200 800

Using the Small Business Administration's definition of small and large firms as those employing fewer than 500 workers, and those employing 500 or more workers, respectively, we have two large and two small firms in both the first and third years--in other words, the structure of our little economy has remained unaltered. Moreover, each firm has the same number of workers at the end of the period as it had at the beginning.

Now let's look at how the SBA's Office of Advocacy would score the game. When Firm 1 increases its employment from 450 to 550, small business is credited with creating 100 jobs. When that same firm decreases its employment from 550 to 450, large business is charged with destroying 100 jobs. Adding up the gains and losses for all of the four firms, SBA would erroneously credit small firms each year with generating 700 jobs, and large firms with destroying 700 jobs.

Obviously, this 700 figure measures not job generation, but job volatility.