Affiliations: [a] Graduate School of Management, Rutgers University,
180 University Avenue, Newark, NJ 07102, USA | [b] U.S. Census Bureau, Room 2232, FOB 4, Washington, DC
20233, USA
Note: [1] The opinions expressed are those of the authors and do not represent the
views of Rutgers University or the Bureau of the Census. The authors would
like to thank Robert McGuckin, Tim Dunne, Michael Ikeda, Sang Nguyen, and
Mark Doms for their helpful comments on an earlier draft. As usual,
responsibility for remaining errors rests solely with the authors.
Abstract: Understanding the structural relationship of production technologies across
manufacturing industries is vital for analyzing dynamic economic activity
because firms and establishments often change industries in response to economic
conditions. Most researchers use the Standard Industrial Classification (SIC)
system, with its hierarchical 2, 3 and 4-digits, to identify these changes and
build the underlying relationships across industries. However, as discussed in
Andrews and Abbott (1988), the SIC is replete with problems, including the lack
of an single, overall guiding principle for the classification. This paper expands on our exploratory work using clustering techniques in Abbott
and Andrews (1990). It builds on the idea of using continuous measures of
technological distance between industries and establishments based on their
production technologies. As illustration, we develop distance measures between
existing 4-digit industries and show what an “optimal” hierarchical structure
might look like using these methodologies. While this approach yields results which are similar to the SIC, there are
important differences between the two classifications in terms of the industrial
categories which emerge and the amount of information lost in the process of
aggregation. Thus, we feel justified in concluding that a lot can be learned
about the dynamic interactions between firms and establishments by looking at
industry and establishment classification in a less rigid fashion.
