be supplemented with survey data, we also describe weaknesses in existing surveys. Finally, since a significant part of the data gap for measuring both levels and growth of economic activity reflects a lack of data sharing and data access, we also discuss these issues.

For much of its history, business data coverage in the United States was by and large limited to the agriculture and manufacturing sectors, reflecting where much of the economic activity in the country was located. Some of the earliest efforts by private-sector firms to count and characterize firms included those by R.G. Dun and Company, which produced credit reports on firms in the 1840s, with information on owners, ownership structure, firm performance, and financial transactions. From 1842 to 1890, the company amassed records on some 1.2 million firms (Madison, 1975). Dun & Bradstreet data extend back sporadically to the 1870s for some sectors of the economy.

On the government side, various censuses of manufacturing were conducted, some at the state level, prior to 1900. The 1810 Census of the Population was the first to include questions—albeit a highly limited set—on manufacturing. The Office of Business Economics, the precursor of BEA, began producing data on the number of firms by major industry in 1929 and continued to do so through 1963, at which time the program was discontinued due at least in part to concerns about data quality. By the mid-1950s, the quinquennial economic census was developed to the point that the Census Bureau’s Enterprise Statistics Program was able to produce statistics on firm size for selected industries. However, even with additional input from the bureau’s County Business Patterns—begun in 1946 but not available annually until 1964—data production was sporadic and incomplete, typically released with four- or five-year lags (Armington, 2004, p. 2).

It was not until the development of comprehensive business lists that the statistical agencies began generating high-quality descriptive data on something approaching the universe of U.S. employer businesses. Historically, most data collection by federal statistical agencies has been oriented toward a cross-sectional view of business activity. While progress has been made more recently at the statistical agencies to develop and improve statistics on business dynamics, there are still important topics that cannot be adequately explored due to data deficiencies.²

Figure 4-1 summarizes the distribution of business data coverage, mapping data sources in terms of the business and labor populations they cover