Editor's note: This is the fifth column in a series that challenges misconceptions about entrepreneurship. Myth: When measuring share of small business employment, the terms "small firm" and "small establishment" mean the same thing. Reality: To determine the share of employment small businesses account for in the U.S., I've always used data from the Office of Advocacy of the Small Business Administration which provides data produced by the U.S. Census on small business employment. The Census data on the SBA Web site shows that, in 2006 (the latest year available), 50.2% of U.S. employment lies in businesses with fewer than 500 employees. But recently I began looking at data from payroll provider Automatic Data Processing (ADP), which uses payroll data to track U.S. employment. ADP's data shows that the share of U.S. employment in businesses with less than 500 employees is more than 30 percentage points higher. In 2006, the ADP data showed that 82.9% of U.S. employment was in businesses with less than 500 employees. Huh? A 32.7% gap in the share of employment in businesses with fewer than 500 employees is much too large to be the result of just some slight difference in measurement. So something else must be going on. What's Behind the NumbersTo figure out what could explain the differences, I took a look at what the two sources are measuring. Both are comparing employment in businesses of less than 500 employees to the overall number of people employed and both exclude employment on farms. So it's not the size of the businesses or the exclusion of agriculture that's the cause of the difference. It's also not their labor force figures. The two sources' estimates of the number of people employed aren't that far off each other—119,917,000 for the SBA in 2006 and 113,475,000 for ADP in 2006. Even if we assume that every employee counted by the SBA and missed by ADP was employed in a large business, ADP's estimates would show that small businesses accounted for 77.2% of U.S. employment, whereas the Census/SBA estimates would show that small businesses only accounted for 50.2% of U.S. employment. The use of payroll data is also not the explanation. Census measures employment on the basis of payroll tax records, using employer identification numbers to identify businesses. ADP gets its estimates from "aggregated and anonymous payroll data that represents approximately 400,000 of ADP's 500,000 U.S. business clients" which are then extrapolated to the overall population. So both sources are using payroll data to measure employment. Different definitions of "business"The difference, it turns out, lies in how the two groups define "businesses." ADP is measuring establishments, whereas Census is measuring firms. And small establishments, it turns out, are very different than small firms. According to the U.S. Census, "an establishment is a single physical location where business transactions take place and for which payroll and employment records are kept." Firms are "groups of one or more establishments under common ownership or control." Many more Americans work in establishments with fewer than 500 employees than in firms with fewer than 500 employees because a lot of establishments are part of large firms. For example, the Gap (GPS) outlet at your local mall is considered an establishment if it has fewer than 500 employees, but it is considered part of a firm if it has more than 500 employees. So if we measure employment at establishments with fewer than 500 employees, everyone working at the Gap outlets in different malls around the country would be considered employed in small establishments. However, if we measure employment at firms with fewer than 500 employees, everyone working at those different Gap outlets (as long as they weren't franchised) would be considered employed at large firms. The difference between the establishment and firm data is corroborated by looking at County Business Patterns, a Census Bureau effort to measure establishments. County Business Patterns shows small establishment employment numbers very similar to those shown by ADP. In 2007, the Census data shows 79.7% of employment was in establishments with fewer than 500 employees, while ADP shows that 83.1% of employment was in "businesses" with fewer than 500 employees. Skewed Employment StatisticsPutting the data together, it is clear that a lot of people work in small establishments that are part of large firms. Unfortunately, most people looking at the different sources of data don't know this, especially since the different sources are saying that they are measuring "businesses." And, unless people look at both sources at the same time, they probably don't even know they are different. But the difference does make it difficult to discuss a basic fact about small business—the share of employment it accounts for. There are probably a number of reasons why measuring small establishments is problematic when seeking to understand small businesses. But here's just one: suppose policymakers want to put in place policies to increase small business employment if small business isn't accounting for an increasing share of total employment. If policymakers look at establishment data, they will get the wrong answer about what to do. The share of employment in small firms has been constant in recent years, while the share of employment in small establishments has been rising. So looking at establishment numbers would give policymakers the impression that small business is accounting for an increasing share of employment. A fact that turns out to be untrue.