The Dot-Com Boom and Bust in the Context of Regional and Sectoral Changes

Abstract

This paper examines the effects of the dot-com boom and bust on firm survival, migration and growth in other sectors. Dot-com expansion revived a slowing central city economy but also raised issues regarding displacement of manufacturing, distribution, social services and other non-profits from San Francisco's “South of Market”. This study uses the National Establishment Time Series (NETS) database to examine how the growth of dot-com businesses affected San Francisco and surrounding counties. We find that start-ups were central to San Francisco's job growth. Relocations also were a significant aspect of job change. Dot-com growth was positively associated with propensity for pre-boom information and professional services establishments and for non-high-tech manufacturing or distribution establishments to move, while arts, social services and non-profits remained in the city.

Keywords:

• Dot-com boom
• multimedia
• new economy
• relocation
• regional change
• agglomeration
• San Francisco Bay Area

Notes

¹ US Bureau of Economic Analysis (BEA), Regional Economic Information System data.

² An establishment's first year is the year in which it first appears in the Dun and Bradstreet database. For all establishments starting business after 1989, this is the establishment's first year in business. First year is listed as 1989 for all establishments in business prior to 1990.

³ Firm detail from OneSource, a global business browser at http://globalbb.onesource.com/homepage.aspx

⁴ This can give several kinds of errors. New franchises may be “stand-alone” establishments although under license to the parent corporation. Two firms that merge may get a new Duns number. For example, in the year that Chevron and Texaco merged, the new headquarters appears as a new establishment with new Duns number. Attempts to identify these situations by creating a table counting the number of establishments per headquarters per year did not succeed in filtering out some of these errors. In the Chevron case, this new headquarters identification code was not quickly transferred to other Chevron establishments. While NETS reports the number of related firms and “kids” for every establishment, this is based on the establishment's full history in the database, rather than a particular year. To eliminate all establishments with one or more relatives, then, would be to eliminate the most successful of new start-ups that grew quickly and expanded. A different issue is that some branch start-ups are new innovative activities. It is not unusual for a tech company to begin turning a new innovation into a commodity under the umbrella of the old organization, spinning it off to a new organization when it passes early stages of development.

⁵ As pointed out by Neumark et al., (2006), this is not the only possible definition of move. We use alternative forms of analysis based on employment growth to examine whether growth shifted outside of San Francisco.

⁶ Three types of errors are possible with this definition. Dun and Bradstreet does not update every establishment every year. In the most recent year, establishments that have already “died” may remain in the database. The reverse can also happen, an establishment missed in one year may disappear from the database, to be resurrected in a year or two. An establishment could also disappear if it moves out of the country. We compared the 2007 and 2005 databases for the first two error types and found less than a 1 per cent discrepancy. Analyses run on the two different databases gave virtually identical results.

⁷ The models use rents relative to the USA. We also ran versions with rents relative to California and to Alameda County rents, with similar results.