For data geeks only: sometimes data can mislead.

The takeaway: Employment by industry statistics may misrepresent employment depending on the data set in use. Based on County Business Pattern data, one posits that the Census Bureau 1) places R&D life science workers in the biotech and all-purpose R&D industries and 2) places life science workers at company headquarters in the management of companies industry. The Quarterly Census of Employment and Wages (QCEW) data from the Bureau of Labor Statistics (BLS), on the other hand, appears to place them all in manufacturing. As a result, the QCEW data understates R&D workers and, by extension, the R&D activity in the state.

An interesting outcome of the federal government's fractured statistical system became evident conducting this analysis. There are several statistical agencies dispersed through the federal departments--for example, the Census Bureau and the Bureau of Economic Analysis are in the Department of Commerce whereas the BLS resides within the Department of Labor. Census, by law, cannot share much of its data with other agencies.

It turns out that Census and the BLS have different establishment lists. An establishment is a location, an address, for a business (or nonprofit or government office). Data for employment and wages, for example, are collected at the establishment level and aggregated by industry, county, state, ownership type, etc. In the main, BLS developed their list based on state workforce agencies' unemployment insurance filings (and not all establishments have to file for unemployment insurance). The Census list is initially derived from the Internal Revenue Service as a source of business information. Census also uses administrative records from the Social Security Administration, in addition to information from the BLS and their own annual Company Organization Survey.

Still awake? There are differences in the collection, scope, data definitions and reference periods for the above sources. One particularly interesting difference is that the definition of an active establishment varies depending on the data set. In the aggregate, the number of BLS and Census establishments differed by about 8...