by Erik A Knoder
Oregon Employment Department
Our standard analysis of employment by firm size shows that although there are a large number of small firms, most employment is in large firms. But, as with much in statistics, the result of any analysis depends on the assumptions made at the beginning. Changing the definition of firm-size classes can show that most employment is in smaller firms. Which is correct? Do small or large businesses provide most jobs?
The Oregon Employment Department traditionally assigns firms to one of eight size classes based on their employment: 1-4, 5-9, 10-19, 20-49, 50-99, 100-249, 250-500, and 500+ employees. These size classes are the same as the ones used by the federal government. It is clear that these size classes are unequal in size range. The first class has a range of four, the second has a range of five, the third class has a range of 10, and so on. The final class ranges from 500 to more than 10,000 employees. Given this, it is hardly surprising that the larger class sizes often contain more employment. The larger the size range, the more firms will be included within that range. Graph 1 shows the distribution of Oregon’s employment using these traditional size classes.
The data used in Graph 1 show that small firms with either one to four employees or five to nine employees supply the fewest jobs of any size class. Big businesses, with more than 500 employees each, supply the most jobs of any size class. The general picture suggested by this analysis is that although there are many small businesses, they don’t contribute as much to Oregon’s total employment as do the smaller number of large businesses.
This conclusion is true, but only if we accept these size classes as correct. If we decided that firms with one to four employees and firms with five to nine employees should be in one size class then this new one-to-nine size class would immediately become the second-largest size class, with about 216,000 employees. This size class also would now be comparable in range to the traditional second size class (10-19 employees). This traditional second size class contains only 138,674 employees.
We could continue this process of making all the size classes the same size and count the employment in them. For the sake of simplicity, classes with a range of 10 employees will be chosen for this exercise, starting with sizes 1-10, 11-20, 21-30, and so on. Class 1-10 will have 234,487 employees; class 11-20 will have 130,394 employees; and class 21-30 will have 81,493 employees. The number of employees in each of these size classes falls fairly steadily as the size class increases. For example, the size class 501-510 has only 3,526 employees. This is because there were only a handful of businesses in this size range.
This distribution of employment by equal size class tends to present a quite different picture than Graph 1 with its unequal size classes. Both distributions are correct, they simply slice the data in different ways. In Graph 2 it appears that small businesses account for the lion’s share of Oregon’s employment and larger businesses contribute much less. The alert reader will note however that the right-hand side of the distributions seems to be missing, and it is. Confidentiality restrictions prevent the publication of data for Graph 2 for size classes with more than about 600 employees.
The exception to the rule of fewer employees in the larger size classes occurs when the size class is in the thousands, which is not shown in Graph 2. In these cases a single business is so large that its employment is substantial even if it is the only business in its size class. In fact, Oregon has some businesses with more than 10,000 employees, so it is true that large businesses contribute substantially to Oregon’s employment.
One disadvantage to using small size classes, such as in the example above with a range of 10 employees, is that confidentiality requirements prohibit the publication of employment if there are fewer than three businesses in a size class. This is a common occurrence for a range of 10 employees when firms have more than 600 employees. Another problem is that selecting the range for size classes is arbitrary, how should we decide which size businesses should be lumped together?
One way of dealing with these problems is to use a size class of one, allowing each size of firm to be its own class. In this case all firms with one employee make up the first size class, firms with two employees make up the second size class, and so on. As firms get larger some size classes will not exist. For example, there were no firms with exactly 1,415 employees in April, 2012 so that size class would be empty. In fact, although some Oregon firms have more than 10,000 employees, there were only about 650 different sizes of firms.
Graph 3 shows how Oregon’s employment is distributed among all sizes of businesses, with each size of business being its own size class. First, note that the axes of the graph have no scales. This is to protect the confidentiality of firms in the size classes with fewer than three firms. Only the shape of distribution can be shown. Second, the distribution of employment is U shaped; Oregon’s employment is concentrated into small firms and large firms.
The left side of Graph 3 is similar to the left side of Graph 2. It shows that much of Oregon’s employment is in relatively small firms. In fact, half of Oregon’s employment was in firms with 94 employees or fewer and half in larger firms.
The right side of Graph 3 is what could not be shown in Graph 2 – the distribution of employment in large firms. As noted above, Oregon has a small number of large firms that contribute significantly to its total employment. Their employment, however, is still less than the employment provided by their counterparts at the other end of the scale.
There is no inherently correct way to express whether most of Oregon’s employment is provided by small, medium or large businesses. It depends almost entirely what is meant by small, medium and large. For some purposes the traditional division into eight unequal size groups might be the best way of analyzing employment. For other uses different sized groups might provide better information. Analysts at the Oregon Employment Department can provide this analysis on a custom basis as long as confidentiality requirements are met.