Spinoffs and clustering

• DOI: http://doi.org/10.1111/1756-2171.12130
• Date: 01 May 2016
• Author: Russell Golman,Steven Klepper
• Published date: 01 May 2016

RAND Journal of Economics

Vol.47, No. 2, Summer 2016

pp. 341–365

Spinoffs and clustering

Russell Golman∗

and

Steven Klepper∗∗

Geographic clustering of innovativeindustries is associated with the entry and success of spinoff

firms. We develop a model to explain the multiple empirical patterns regarding cluster growth

and spinoff formation and performance, without relying on agglomeration externalities. Clus-

tering naturally follows from spinoffs locating near their parents. In our model, firms grow and

spinoffs form through the discoveryof new submarkets based on innovation. Rapid and successful

innovation creates more opportunities for spinoff entry and drives a region’s growth. Our model

provides baseline estimates of levels of agglomeration that can be attributed to this process of

innovation and spinoff formation.

1. Introduction

Geographic clustering of people and organizations is a fact of modern economic life.

At the aggregate level, around half the world’s population is located in cities. At the industry

level, Ellison and Glaeser (1997) and Duranton and Overman (2005) show that in the modal

manufacturing industry in the United States and United Kingdom, respectively, plants are more

clustered geographically than would be expected if they located randomly. These simple facts

have widely been interpreted to reflect some sort of advantage of clustering. Wages and prices

are higher in cities and in industry clusters such as Silicon Valley (Rosenthal and Strange, 2004;

Puga, 2010). Consequently, businesses in clusters must enjoy some kind of advantages to be

competitive.

These advantages appear to extend well beyond the natural advantages some regions have

for certain types of industries, such as the weather favoring the location of the movie industry

in Hollywood (Ellison and Glaeser, 1999). New economic geography models (Fujita, Krugman,

and Venables, 1999) feature the role that costs of transporting goods play in inducing people and

businesses to cluster in cities. Models of industry clustering commonly feature ideas proposed

∗Carnegie Mellon University; rgolman@andrew.cmu.edu.

∗∗Steven Klepper passed away in 2013. He is dearly missed.

Wethank Serguey Braguinsky, Peter Thompson, Erik Stam, and two anonymous referees for helpful comments. We also

thank Aaron Bramson, the Complexity Research Corporation, and Steven Wangfor valuable research assistance. Klepper

gratefully acknowledges support from the National Science Foundation, Science of Science and InnovationPolicy, grant

no. SES0965451.

C2016, The RAND Corporation. 341

342 / THE RAND JOURNAL OF ECONOMICS

by Marshall (1890) about how clustering gives rise to agglomeration economies benefiting all

firms located in clusters. These economies are related to the pooling of labor, the colocation of

suppliers and producers, and localized spillovers of technological knowledge(Jaffe, Trajtenberg,

and Henderson, 1993; Duranton and Puga, 2004). Although it is difficult to test the mechanisms

underlying clustering, numerous studies find evidence consistent with the advantages of clusters,

such as firms in clusters performing better, entry being concentrated in clusters, and firms in

different industries clustering to a greater degree the more they share similar types of labor,

inputs, and knowledge (Audretsch and Feldman, 2004; Rosenthal and Strange, 2004; Ellison,

Glaeser, and Kerr, 2010; Greenstone, Hornbeck, and Moretti, 2010; Puga, 2010).

In recent years new evidence has emerged about how industry clusters evolve that calls out

for explanation. Studies of the origins of firms in the automobile, tire, semiconductor, disk drive,

and biotherapeutics industries, all of which were innovative in their time and evolved to be highly

clustered, reveala similar pattern. The regions where these industries ultimately clustered initially

had one or at most a few related early successful firms. What distinguished them from other

regions that also had successful early producers was that subsequentlythey g rewthrough entry of

new firms that were mainly spinoffs descended from their earlysuccessful producers. (We define

spinoffs as entrants founded byindividuals who previously workedfor incumbent fir ms in the same

industry. We refer to the firm where the primary founder last worked as the spinoff’s “parent.”) In

a number of instances, the early successful firms receded, but the region nonetheless prospered,

propelled forward by successive generations of spinoffs. Geographic modellers are increasingly

recognizing the importance of entrepreneurship for clustering (Glaeser, Rosenthal, and Strange,

2010), but mainstream theories of clustering generally abstract from such forces and thus cannot

readily address the accumulating evidence connecting clustering and spinoffs. Many studies of

spinoffs in the last 10 years havefocused on innovative manufacturing industries, where clustering

is also prominent (Feldman, 1994; Audretsch and Feldman, 2004), and the growing body of

empirical evidence regarding spinoff formation and performance is now attracting increased

attention in its own right (see Klepper, 2009b).

The main purpose of this article is to organize the empirical evidence on spinoffs and

clustering across industries and to develop a model to explain these stylized facts. Why do firms

agglomerate, especially in innovative industries? Why is so much of the entry driving the growth

of agglomerative clusters coming in the form of spinoffs? Also, why are these spinoffs typically

more successful, often becoming the industry leaders? We construct a simple theory of spinoffs

that is related to firms growing over time through the discovery of new submarkets based on

innovation. Spinoffs are assumed to locate close to their parents, as has been commonlyfound, in

accordance with more general findings that new firms of all kinds tend to locate close to where

their founders have previouslyworked and resided (Figueiredo, Guimaraes, and Woodward, 2002;

Romanelli and Feldman, 2006; Stam, 2007). We show that spinoffs (locating near their parents)

naturally generate clustering.

Our model explains why the growth of an industry in a particularly concentrated region

typically is marked by spinoffs, even if they only capture profits that would have gone to their

parents. A simple insight powers the model. Innovation begets more innovation in a positive

feedback cycle (Arthur, 1990; Danneels, 2002), creating more new profit opportunities for more

successful firms. When this dynamic gets going, rapid innovation opens the door for spinoffs to

enter, driving the entire region’s growth. Innovative capabilities are not directly observable, but

when we see clusters of spinoffs, we get a strong signal about their innovative potential.

We posit that a firm’s innovations build on the expertise it already has. Evolving inno-

vative capabilities reflect organizational learning (Mitchell, 2000). Our model of new business

opportunities growing out of existing ones is similar in spirit to recombinant growth (Weitzman,

1998), endogenous technological change (Klette and Kortum, 2004), combinatorial technological

evolution (Arthur, 2009), and creative development (Feinstein, 2015). As in Klette and Kortum

(2004), in particular, business units within a firm beget new business units. The new feature

here is that newly created business units may be spun off into independent firms. As innovation

C

The RAND Corporation 2016.