Do Cities Substitute for Internal Firm Resources? A Study of Advanced Internet Technology Adoption Chris Forman Avi Goldfarb Shane Greenstein.

Do Cities Substitute for Internal Firm Resources? A Study of Advanced Internet Technology Adoption

Chris Forman

Avi Goldfarb

Shane Greenstein

Location, Firm Resources, and Internet Investment Location

Are the costs of Internet investment lower in large cities? Firm Resources

How do resources available through internal channels shape adoption of frontier IT?

Substitution between cities and internal firm resources How do internal and external channels substitute for one

another in reducing the costs of adopting frontier IT?

Broad Overview

Examine how internal and external channels shape the adoption of new IT Examine how external channels shape Internet

adoption Examine how local establishment capabilities and

mobile firm capabilities shape adoption Measure the extent of substitution between

External and internal channels Local and mobile firm channels

Why care?

Contributes to understanding the factors that nurture technological development and their mobility

Further builds on our research program to alter the conversation about the Digital Divide How does the affect of location on Internet use

vary across economic units?

Bottom line

External resources found in cities play a significant role in decreasing the adaptation costs of complex Internet technology

Internal channels also play a role This is true for both establishment capabilities and mobile

firm capabilities Internal and external channels substitute for one

another in lowering the costs of adopting frontier Internet

Moreover, establishment capabilities and firm capabilities are substitutes

Implications

Large companies in major cities are the biggest adopters of all. Single establishment companies in isolated locations do worst of all. If these types of investments matter for competitive survival, then

the single establishments in isolated locations are at a disadvantage

Establishments from major companies in isolated locations do better than those not from major companies, but not as well as single-establishment companies located in cities A major company can be in an isolated location and draw on

resources from its other locations, but that will mostly, not fully, makes up for poor locations

Large single establishment companies in isolated locations do better than small single-establishment companies in large locations Organizational capabilities are only partially mobile

Local Supply Conditions for IT Projects

IT Projects atEstablishment 2

Other projects atEstablishment 1

IT Project at Establishment 1

Physical infrastructureLabor poolingConsultant availabilityKnowledge Spillovers

Shared human capitalShared physical capitalEconomies of Scale and ScopeLearning economiesKnowledge Spillovers

Shared human capitalShared physical capitalEconomies of Scale and ScopeLearning economiesKnowledge Spillovers

How Internal and External Channels Influence Diffusion

Location and Internet Adoption We examine adoption of Internet technologies that

involve within-firm communication and where benefits of adopting do not depend on location We label these technologies within-establishment Internet

(WEI) As a result, we expect urban leadership to hold:

adoption costs decrease as population size and density increase Our results are robust to other measures of Internet

investment

Hypotheses: Internal Capabilities and Adoption H1A: Firms with greater organizational

capabilities will be more likely to adopt Internet technology at any of their establishments

H2A: Establishments with greater establishment capabilities will be more likely to adopt Internet technology

Hypotheses: Interactions

H1B: The sensitivity of Internet adoption to increases in location size will be declining in the internal organizational capabilities found in other establishments within the same firm.

H2B: The sensitivity of Internet adoption to increases in location size will be declining in the internal establishment capabilities.

H3: Establishment Capability and Organizational Capability are substitutes

Understanding the hypotheses

Ado

ptio

n R

ate

Organizational Capability Establishment Capability

Establishment in urban area

Establishment in rural area

Data

Survey of 86,879 business establishments in the US with 100+ employees in 3Q-4Q,2000 Only private, non-farm Approx half of all such establishments in US Two-thirds of US labor force work in such

establishments Detailed information about IT capabilities.

Data Endogenous Variables

Within-Establishment Internet Involves Internet protocols in the input and output of data to

and from business applications software. E.g. ERP, CRM. Robustness:

CEI adoption, Internet language use, PC server adoption

Exogenous Variables Capabilities

# Programmers Factor analysis of programmers, employees, software

City dummy (MSA population >500,000) Controls for industry (NAICS dummies), multi-

establishment firm, establishment employment

Econometric Methodology

Probit model of Internet adoption At the level of the establishment Interpret as “Net benefit” to an establishment of

Internet adoption Weight by 1999 County Business Patterns

Generally already representative of industry/location of US business establishments, so weighting not too important to the results.

Econometric Assumptions Location is predetermined (exogenous to

technology adoption) Decisions are made at the establishment

level. Capabilities are exogenous to the adoption

decision. Robustness checks examine these

assumptions

Main Results

Present results using number of programmers, results using composite measure of capabilities are qualitatively similar

Direct Effects: all results are statistically and economically significant (average adoption rate 11.9%) Establishments located in cities are 1.3% more likely to

adopt (urban leadership) One SD increase in log of establishment programmers

increases likelihood of adoption by 3.6% One SD increase in log of organization programmers

increases likelihood of adoption by 0.45%

Main Results

Interaction Effects: All results are statistically and economically significant One SD increase in log of establishment programmers decreases the likelihood of adoption by 1.1% for establishments located in cities

One SD increase in log of organization programmers decreases the likelihood of adoption by 0.9% for establishments located in cities

While establishment programmers have a much stronger direct effect on adoption, substitution between cities and internal capabilities is similar at establishment and organization level

Table 2: Main ResultsDirect Effect Only Direct Effect and

Interaction Effect

CapabilityDefined byProgrammers

Capability Defined byFactors

Capability Defined by Programmers

Capability Defined by Factors

OC 0.0021** 0.0062 ** 0.0054 ** 0.0118 **

EC 0.0361** 0.0303 ** 0.0457 ** 0.0550 **

OC * City -0.0038 ** -0.0069 *

EC* City -0.0108 ** -0.0266 **

City 0.0129 ** 0.0180 ** 0.0256 ** 0.0188 **

Multiestablishment Firm Dummy

0.0154 ** 0.0217 ** 0.0149 ** 0.0213 **

Log(Establishment Employment)

0.0313 ** 0.0309 **

Observations 86871 86871 86871 86871

LL -24550.40 -25914.41 -24528.56 -25861.03

Table 3: Predictions Using Organizational Capabilities

Low Organization Capability

Medium Organization Capability

High Organization Capability

Low density location

13.44% 15.00% 17.23%

High density location

18.44% 18.97% 19.69%

Table 3: Predictions Using Establishment Capabilities

Low Establishment Capability

Medium Establishment Capability

High Establish Capability

Low density location

13.44% 17.55% 27.87%

High density location

18.44% 22.17% 30.76%

Substitution between establishment and organization capabilities There is significant substitution between

establishment and organization capabilities No matter how we measure them, the

interaction of establishment and organization capabilities is negative and significant

A one SD increase in the log of organization programmers will decrease the marginal effect of establishment capabilities by 0.42%

Table 5: Are Establishment Capabilities and Organizational Capabilities Substitutes?

Capability Defined by Programmers

Capability Defined by Factors

OC 0.0069** 0.0089** 0.0111** 0.0144**

EC 0.0446** 0.0527** 0.0335** 0.0566**

OC*EC -0.0033** -0.0036** -0.0066** -0.0084**

OC*City -0.0026* -0.0042

EC*City -0.0098** -0.0251**

OC*EC*City 0.0005 0.0023

City 0.0174** 0.0184** 0.0091** 0.0198**

Multiestablishment Firm Dummy 0.0197** 0.0194** 0.0096** 0.0095**

Log(Establishment Employment) 0.0351** 0.0347**

Observations 86871 86871 86872 86872

LL -25823.42 -25778.86 -25199.06 -25185.64

Robustness

Results are not isolated in any particular industry

Results are robust to Instruments for EC, OC, EC*City, and OC*City Subset of establishments that explicitly claim in

the survey that they conduct the adoption decision Different city definitions Different adoption measures Controls for competition

Conclusions

External resources found in cities decrease the costs of technology adoption

Internal resources decrease the costs of technology adoption. IT capabilities are mobile within firms.

These resources substitute for each other Therefore, complementary resources found in cities

will be most important for single-establishment firms that do not have such internal resources Implication is that geographic digital divide in business

Internet use, where it exists, will be most apparent in small firms