One of the first questions when doing a national business survey is the question of what businesses to use as the population for the survey. This is one which many organizations, like ourselves, struggle with in an ongoing basis. While we are a private organization, and as such, could not hope of getting the "gold standard" of frames, a government business register, even with the United States government and within the European Union directorates, few agencies are able to access Census, Bureau of Labor Statistics, or Internal Revenue Service pr similar business lists for the purposes of carrying out surveys. As such, federal agencies and private organizations like Kauffman are forced to go down a different route to obtain business lists which can, if not carefully considered, impact the overall quality of the research. But beyond this, an additional cost is born by taxpayers, businesses, and others more generally, because when using a non-governmental register of businesses, it becomes much more complicated to match survey responses to other administrative data. With this, we end up with a lot of independent surveys, few of which can actually be matched together, causing the surveys to be longer than really necessary and knowledge gained about different issues to be stymied. There are a host of privacy issues here which I am not going to address, but anyone seriously concerned with this topic should look at that topic separately.
Private organizations have popped up to provide these lists, some that are specific to different industries, such as Corptech
, which claims to cover high-tech businesses, while others like Dunn and Bradstreet
report to provide full industry coverage on the national level with international coverages that are very country-specific. And even companies like Dunn and Bradstreet allow others to repackage and sell their data, such as what Don Walls does with the NETS database
But how do these different lists compare to federal government lists? Private companies provide the data in a much quicker fashion and I would say the general consensus is that this benefit is offset by more messiness in the data. While some messiness is involved in any data set, as researchers are increasingly using micro data sets and not just aggregated tabulations, non-random messiness can become a problem. Many of these companies didn't start collecting this data for research purposes but more for marketing or credit checks. As the Internet has blown up, and companies have come to realize the value in some of this data, more products have appeared over time. Some work was done in the 1990s by entrepreneurship scholars to test the coverage of the private sources, but the reality is that not much has been done, to my knowledge, in at least ten years looking at a systematic comparison of advantages and disadvantages of some of these business registers. From conversations with others about the NETS database, as an example, we know that the last ten years have brought about a lot of changes in the population of businesses these private companies are able to find and include in their registers. This can create new opportunities but also real challenges for researchers looking to use the longitudinal component of these data.
I am aware of at least two projects that are underway in this arena that might be able to provide insights into some of the different private business registers (as well as governmental business registers). The first, and it appears furthest along of the projects, is comparing Census data to InfoUSA. I will be attending the 2009 Joint Statistical Meetings (JSM) in Washington, DC, in August and hope to attend the session at which this comparison will be presented
. The second project which I have heard alluded to several times is that Census is matching the California file of businesses from NETS to its business register to study similarities and differences. Both of these efforts should be informative to researchers and I will try to follow-up with additional details of the outcomes of this research in future posts. At the JSM, other register issues will be discussed including some international examples such as that of Finland
. Nordic countries, by most accounts, have the most robust registers so that might be an interesting section. If other projects are underway that are comparing private and governmental registers, I would appreciate an email