ArticlePDF Available

TRANSFERRING KNOWLEDGE BY TRANSFERRING INDIVIDUALS: INNOVATIVE TECHNOLOGY USAGE AND ORGANIZATIONAL PERFORMANCE IN MULTI-UNIT FIRMS

Authors:

Abstract

Transferring individuals who possess relevant knowledge from one organizational unit to another – a form of resource redeployment – may help to overcome impediments to knowledge transfer. Despite the promise of this mechanism, which often occurs through intra-firm geographic mobility, relatively little research has examined how the knowledge and expertise of individuals interacts with the organizational resources of the units to which individuals move. This study examines whether intra-firm geographic mobility improves organizational performance by providing a conduit for the transfer of knowledge, while accounting for the interaction between individual knowledge and factors at the organization-unit level of analysis. We analyze the performance effects of the transfer of engineers who have expertise in innovative process technologies. The results from a large multinational company show that the innovative process technology-related expertise of an individual engineer who moves to a new organizational unit is positively associated with the performance of that unit, suggesting that intra-firm geographic mobility improves organizational performance by providing a conduit for the transfer of knowledge. The results also show that the technology-related knowledge of engineers is a substitute for organization-level factors when a unit uses only technologies with which it is already familiar, whereas the technology-related knowledge of engineers is a complement to organization-level factors when units introduce new technologies. Thus, individuals who bring novel expertise to their organizational units through intra-firm mobility may be important vehicles for organizational learning and building new competences, helping to diffuse best practices.
TRANSFERRING KNOWLEDGE BY TRANSFERRING INDIVIDUALS:
INNOVATIVE TECHNOLOGY USAGE AND ORGANIZATIONAL
PERFORMANCE IN MULTI-UNIT FIRMS
Christian Stadler
1
, Constance E. Helfat
2
, Gianmario Verona
3
ACCEPTED FOR PUBLICATION AT ORGANIZATION SCIENCE
Acknowledgements: This paper benefited from the comments of seminar participants at Harvard
University, LUISS Business School, the University of North Carolina, Chapel Hill, and the University of
Texas at Austin, and from a presentation at the Academy of Management annual meeting. The paper also
benefited from the expert help of Paul Wolfson with Stata coding.
Keywords: knowledge transfer, resource redeployment, strategic human capital, microfoundations,
organizational learning, process technology, employee mobility, multinational firms
1
Warwick Business School
2
Tuck School of Business at Dartmouth
3
Bocconi University
TRANSFERRING KNOWLEDGE BY TRANSFERRING INDIVIDUALS:
INNOVATIVE TECHNOLOGY USAGE AND ORGANIZATIONAL
PERFORMANCE IN MULTI-UNIT FIRMS
ABSTRACT
Transferring individuals who possess relevant knowledge from one organizational unit to another a form
of resource redeployment may help to overcome impediments to knowledge transfer. Despite the promise
of this mechanism, which often occurs through intra-firm geographic mobility, relatively little research has
examined how the knowledge and expertise of individuals interacts with the organizational resources of the
units to which individuals move. This study examines whether intra-firm geographic mobility improves
organizational performance by providing a conduit for the transfer of knowledge, while accounting for the
interaction between individual knowledge and factors at the organization-unit level of analysis. We analyze
the performance effects of the transfer of engineers who have expertise in innovative process technologies.
The results from a large multinational company show that the innovative process technology-related
expertise of an individual engineer who moves to a new organizational unit is positively associated with
the performance of that unit, suggesting that intra-firm geographic mobility improves organizational
performance by providing a conduit for the transfer of knowledge. The results also show that the
technology-related knowledge of engineers is a substitute for organization-level factors when a unit uses
only technologies with which it is already familiar, whereas the technology-related knowledge of engineers
is a complement to organization-level factors when units introduce new technologies. Thus, individuals
who bring novel expertise to their organizational units through intra-firm mobility may be important
vehicles for organizational learning and building new competences, helping to diffuse best practices.
Keywords: knowledge transfer, resource redeployment, strategic human capital, microfoundations,
organizational learning, process technology, employee mobility, multinational firms
Knowledge has the beneficial property that it can be reused in different businesses and locations
within the firm (Argote, 2013) in order to generate competitive advantage (Kogut and Zander, 1992; Grant,
1996). Reusing knowledge in this way entails the transfer of knowledge from one part of the organization
to another, which in turn promotes organizational learning (Argyris and Schon, 1978; Argote and Ingram,
2000). However, firms face the difficulty that knowledge does not always transfer easily. Research has
documented that the transfer of knowledge about process technologies across geographically-dispersed
organizational units is time consuming and costly with varying effectiveness (Argote, Beckman, and Epple,
1990; Galbraith, 1990; Mansfield et al., 1983; Teece, 1977). Research has also shown that firms may face
difficulties in transferring best practices between organizational units, including units in different locations
(Szulanski, 1996, 2003). Impediments to knowledge transfer pose a challenge even for multinational firms
(Gupta and Govindarajan, 2000; Jensen and Szulanski, 2004; Zander and Kogut, 1995), whose primary
advantage comes from the internal transfer of intangible assets especially knowledge across borders
(Hymer, 1960; Kogut and Zander, 1993).
To address this puzzle, in this paper we focus our attention on knowledge related to human capital
(Becker, 1964) and specifically on the transfer of individuals within an organization. Transferring
individuals who possess relevant knowledge from one organizational unit to another a form of resource
redeployment (Helfat and Eisenhardt, 2004; Sakhartov and Folta, 2014) may help to overcome
impediments to knowledge transfer (Argote and Ingram, 2000; Hocking, Brown, and Harzing, 2004; Kogut
and Zander, 1993). Despite the potential benefits of this mechanism, which often occurs through intra-firm
geographic mobility (Choudhury, 2019), a limited amount of research has examined the effect on
organizational performance.
In assessing this effect, though, it is important to consider the interaction between the knowledge
of individuals and organization-level factors within the units to which individuals move. In fact, research
on strategic human capital has argued that there are likely to be interaction effects between factors at the
individual and organizational levels of analysis (Nyberg et al., 2014). Therefore, the total effect on
organizational performance of individuals and organization-level factors depends not only on their
2
2
independent effects, but also on whether the two are complements or substitutes (Crocker and Eckhardt,
2014).
Our study investigates whether intra-firm geographic mobility improves organizational
performance by providing a conduit for the transfer of knowledge, and in particular whether the knowledge
of individuals who move between units and factors at the organization-unit level of analysis are
complements or substitutes. More specifically, we analyze the performance effects of the within-firm
transfer of individuals who have expertise in innovative process technologies. Although process
innovations that lower costs or raise the quality of output are important in many sectors of the economy
(Tidd and Bessant, 2018), difficulties in implementing new process technologies often impede their
effectiveness and raise costs in the short-term (Argote, 2013). Because transferring individuals who have
prior experience using innovative process technologies also transfers their expertise with respect to these
technologies, these individuals may help organizational units overcome difficulties encountered when
implementing the technologies (Galbraith, 1990). For this reason, process technology usage provides an
especially relevant context for understanding the role of individuals in intra-firm knowledge transfer.
Given the relatively small amount of prior research on the effect on organizational performance of
intra-firm knowledge transfer through the transfer of individuals, rather than propose formal hypotheses,
we ask research questions. To establish a baseline for the analysis, we first ask whether the transfer of
individuals across organizational units improves performance by providing a conduit for the transfer of
knowledge about innovative process technologies. Then we ask whether the innovative process technology
knowledge of individuals who move between units is a complement or substitute for organization-unit level
factors in the effect on the performance of organizational units.
We take these questions one step further because research on strategic human capital has suggested
that whether individual human capital and organization-level factors are substitutes or complements may
depend on contextual factors (Crocker and Eckhardt, 2014). In the context of process technologies, it is
well established that performance is worse when organizations first introduce new technologies, due to
difficulties in applying unfamiliar techniques (Argote, 2013). We therefore examine whether the answers
3
3
to the two questions posed above differ when units introduce new innovative process technologies versus
when units use only innovative technologies that they have used in the past.
Our study analyzes unique proprietary data from a major multinational oil company that is well-
known as a leader in the internal development and application of innovative upstream oil technologies. The
frequent transfer of engineers between organizational units allows us to separate the effect on organizational
performance of individual expertise from factors at the organization unit-level.
The results show that the innovative process technology-related expertise of an individual engineer
who moves to a new organizational unit is positively associated with the performance of that unit,
suggesting that intra-firm geographic mobility improves organizational performance by providing a conduit
for the transfer of knowledge. However, the results also show that there is a substitution effect between the
technology-related knowledge of engineers and organization-level factors when the unit uses only
technologies with which it is already familiar, suggesting that some of the knowledge that new employees
bring about these technologies is redundant. Conversely, the technology-related knowledge of engineers is
a complement to organization-level factors when units introduce new technologies. Thus, individuals who
bring novel expertise to their organizational units through intra-firm mobility may be important vehicles
for organizational learning and building new competences, helping to diffuse best practices. These findings
are consistent with a prior study showing that intra-firm geographic mobility leads to greater patenting
within R&D units and may aid in knowledge transfer (Choudhury, 2016). The results in this study are also
consistent with prior work suggesting that inter-firm mobility helps to diffuse knowledge (for a review, see
Mawdsley and Somaya, 2016).
Beyond the literatures on knowledge transfer and employee mobility, our findings have
implications for the literature on microfoundations, which has argued that individuals (or other factors) at
a lower level of analysis within the organization may moderate the effects of factors at a higher level of
analysis (Felin, Foss, Heimeriks, and Madsen, 2012). The evidence in this study helps to provide an
improved understanding of the conditions under which such moderation is positive versus negative. This
evidence is also relevant for the literature on strategic human capital that is concerned with interactions
4
4
between individuals and organization-level factors (Nyberg et al., 2014). Our findings further inform
research on knowledge transfer in multinational firms (Gupta and Govindarjan, 2000; Hocking et al., 2004;
Kogut and Zander, 1993) by providing evidence that the extent to which firms benefit from intra-firm
mobility as a conduit for knowledge transfer depends on whether or not the expertise of individuals
complements factors at the organization-unit level. Finally, this study contributes to research on resource
redeployment (Folta, Helfat, and Karim, 2016) by investigating the performance effects of redeploying
knowledge within the firm through the redeployment of human capital.
Background and Research Questions
Beginning with Wright’s (1936) study of aircraft manufacturing, numerous studies of the
introduction of new process technologies have established that costs per unit of output decline with
cumulative output, often referred to as the experience curve (Argote, 2013). Research has shown that the
decline in costs arises in part from organizational learning after new technologies are introduced,
including through improvements in the configuration of the production process (Levitt, List, and
Syverson, 2013) and on-the-job learning by workers (Thompson, 2001) (see Thompson (2012) for a
comprehensive review). Firms may also seek to lower the initial high costs to organizational units of
using new process technologies by transferring knowledge from other organizational units that have
experience using the technologies (Galbraith, 1990).
When firms seek to transfer knowledge about process technologies, they may encounter
difficulties that arise from the characteristics of the knowledge itself. In particular, causal ambiguity and
lack of a proven track record for the knowledge in question may impede its transfer (Szulanski, 1996).
The use of process technology tends to entail tacit knowledge (Galbraith, 1990), leading to causal
ambiguity about how to use it, which in turn makes the knowledge more difficult to transfer (Szulanski,
1996). In addition, recipient units may resist the transfer of knowledge that lacks a proven track record
5
5
(Szulanski, 1996). This is likely to impede the transfer of knowledge about innovative process
technologies.
Intra-firm mobility may help to overcome these barriers because the movement of experts to new
organizational units and new locations also moves their knowledge (Argote and Ingram, 2000; Hocking et
al., 2004; Choudhury, 2019). Individuals who have previously used innovative process technologies bring
tacit knowledge about their application, which can help to reduce causal ambiguity. In addition, intra-firm
mobility may improve knowledge transfer because recipients are less likely to resist an expert source
(Szulanski, 1996). Experts who move across units may also help to counteract the lack of absorptive
capacity on the part of recipients (Szulanski, 1996) by providing advice on the use of technological
resources (Choudhury, 2019). Given the potential benefits of knowledge transfer through individuals,
organizations may select individuals for their expertise when moving them to another unit. Thus, the
performance effects of knowledge transfer through intra-firm mobility may come from both “selection”
(of individuals who have transferable knowledge) and “treatment” (transfer of knowledge).
A limited amount of prior research has provided evidence that individuals with specialized
expertise can help to transfer knowledge about technology use within an organization. In a study of stent
utilization in hospitals, Greenwood, Agarwal, Agarwal, and Gopal (2019) found that physicians with
specialized expertise the skills and knowledge that individuals accumulate through learning in a
particular domain (Simon, 1991) adopted new practices more quickly and helped to diffuse these
practices within the organization. In addition, Galbraith (1990) found that plants that implemented a new
manufacturing technology had faster recovery of lost productivity when an engineering team from within
the company that was familiar with the technology was transferred to the plants on at least a temporary
basis.
In light of the foregoing research, to establish a baseline for our analysis, we ask:
RQ 1: Does the technology-related knowledge of individuals who move to another organizational
unit have a positive effect on organizational performance, over and above organization-level
factors?
6
6
To answer this question, we focus on individuals with expertise in innovative process technologies gained
through prior experience, as part of their human capital (Becker, 1964).
Beyond the independent effect on organizational performance of the mobility of individuals with
specialized expertise, the full effect depends on whether the expertise of individuals and organization-
level factors are complements or substitutes (Crocker and Eckhardt, 2014; Nyberg et al., 2014). Individual
expertise and organization-level factors are complements when the marginal effect on organizational
performance of the combined use of individual expertise and organizational resources, including
aggregate knowledge and skills at the organization-unit level, is greater than the sum of their separate
effects on performance (Crocker and Eckhardt, 2014; Rothaermel and Hess, 2007). This may occur when
the workflow in a unit depends on the knowledge of particular individuals (Crocker and Eckhardt, 2014).
The complementarity of individual expertise and organizational resources, however, is not a foregone
conclusion. If the knowledge held by an individual and an organizational unit are equifinal, the two may
produce similar outcomes (Hess and Rothaermel, 2011; Rothaermel and Hess, 2007). If using one of them
also marginally decreases the benefit for organizational performance of using the other, then the two are
substitutes; this may occur when they are used simultaneously for the same purpose (Hess and
Rothaermel, 2011).
To assess the interactive effect of individuals on organizational performance, we ask:
RQ2: Is any effect on organizational performance of the technology-related knowledge of
individuals who move to another unit a complement or substitute for factors at the organizational
level?
If they are complements, this magnifies any positive independent effect on organizational performance of
moving individuals who have technology-related knowledge. Conversely, if they are substitutes, this
detracts from any positive independent effect of moving individuals who have technology-related
knowledge.
There is limited evidence regarding interaction effects on organizational performance of the
knowledge of individuals and organization-level factors in the context of new technology. Based on a
7
7
measure of human capital aggregated to the firm level, research has found that the total amount of higher
skilled cognitive labor within the firm and the use of information technology at the firm level are
complements in their impact on firm productivity (e.g., Bresnahan, Brynjolfsson, and Hitt, 2002). At the
individual level of analysis, Choudhury, Starr, and Agarwal (2020) also found in an experimental setting
that the use of machine learning technology and the domain-specific expertise of individuals are
complements. In contrast, Rothaermel and Hess (2007) found a negative interaction effect on firm
innovative output in the global pharmaceutical industry of the number of star scientists (a proxy for
expertise) and firm R&D expenditures (a proxy for R&D capability). These findings indicate that
sometimes individuals’ knowledge of technology and organization-level factors are complements and other
times they are substitutes, but we lack a systematic understanding of the reasons for this difference.
We investigate a contingency that may shed light on the conditions under which individual
expertise and organization-level factors are complements or substitutes: whether the knowledge of mobile
individuals relates to technologies that are familiar or new to organizational units. The literature discussed
above regarding knowledge transfer concerns knowledge and technologies that are new to organizational
units. When a unit introduces new technologies, the knowledge that individuals bring about these
technologies may not only contribute directly to organizational performance, but also the unit’s workflow
may depend more heavily on the new knowledge that resides in individuals. As a result, individual
knowledge and factors at the organization-unit level may be complements. When units do not introduce
new technologies and rely only on technologies with which they are already familiar, the knowledge that
individuals bring with them may still contribute to organizational performance. However, in this situation
some of the knowledge that individuals bring may be redundant, potentially resulting in substitution
between individual knowledge and factors at the organization-unit level. Our setting and data allow us to
investigate this contingency, by asking the following question:
RQ3: If the knowledge of mobile individuals relates to technologies that are familiar rather than
new to organizational units, does this affect the complementarity versus substitution of this
knowledge with organization-level factors?
8
8
Empirical Setting
Description of OILCOMP
We explore whether the transfer of technology-related knowledge through the transfer of
individuals who have expertise in innovative process technologies contributes to organizational
performance in a context of substantial economic importance: the well drilling operations of one of the
largest oil companies in the world. OILCOMP conducts business in all segments of the oil and gas industry.
The company is a leader within the industry in the internal development and application of innovative
process technologies. The firm is also highly decentralized, and is organized into units according to country
of operation. These units enjoy a high degree of independence. OILCOMP became highly international
many years ago at a time when poor communication technology made it difficult to direct operations in
disparate countries from headquarters. Hence, the company developed a structure of country subsidiaries
that adjust to local conditions.
The independence of the country subsidiaries the organizational units in our analysis applies to
the use of innovative technology. The R&D unit at headquarters develops new technologies and can offer
support for innovative technology use by disseminating information and providing advice on
implementation to the country units, but headquarters cannot force the units to use innovative technologies.
In addition, the country units have their own priorities, and may lead their own technology initiatives, which
can overlap with similar initiatives elsewhere in the company. The firm also has a system that transfers
engineers frequently between countries, which helps to transfer knowledge across organizational units.
Empirically, the movement of engineers between units allows us to separate the expertise of individual
engineers, gained through the use of innovative technologies in other units, from factors connected to the
use of innovative technologies at the organizational level of analysis.
Well Drilling and Innovative Technology
9
9
Oil companies first conduct exploration activities to try to discover new sources of oil, the
companies then conduct development activities to prepare fields for production, and finally the companies
start production (Stadler, 2011; Stadler, Helfat, Verona, 2013). In the first phase, companies drill
exploration wells based on seismological studies that provide a preliminary understanding of the local
subsurface geology. If the companies discover oil and gas, they undertake further analysis to decide how
and where to drill development wells. Well drilling is a highly complex process that consists of distinct
activities. For example, during drilling, the wellbore pressure must be monitored, the equipment must be
lifted, the borehole must be cased and cemented, and a drill-bit must be deployed. When oil companies
intend to drill, they must decide which technology to use for each activity in the drilling process, such as
which drill bit and casing to use and how to monitor pressure during drilling operations.
For each activity in the drilling process, firms can choose either an established or an innovative
technology. Each innovative technology applies to only one activity in the drilling process and each
innovative technology covers a different activity in the process, so in theory a firm could use all of the
innovative technologies when drilling a well. None of the innovative technologies is superior to the others
because each technology applies to a different activity in the drilling process, and no activity dominates the
others in importance. In addition, each activity is potentially a bottleneck in the drilling process because
failure to complete all of the activities will make it impossible to finish drilling the well. Moreover, due to
the complexity of geologic formations, firms cannot always tell which technologies are likely to work best
until they start drilling. Thus, no specific technology is more important than another, and it is the number
of innovative technologies that are used which affects drilling performance.
In general, firms use an innovative technology for more difficult drilling conditions when they
expect that an established technology will not perform as well. The use of an innovative technology reduces
the amount of time that it takes to drill a well, so firms do not need to pay for personnel, equipment, and
materials for as long a time period, which reduces costs. However, an innovative technology is also more
expensive to deploy than an established technology due to higher material and equipment costs. This means
that in deciding whether to utilize an innovative technology, firms must balance cost savings from a shorter
10
10
drilling time against the higher cost of the technology. The decision is not straightforward due to the
difficulty of ascertaining which technologies will work best. In addition, firms that are unfamiliar with an
innovative technology may be uncertain about how well it will perform. Firms therefore may hesitate to
use an innovative technology if they are not sure whether improved performance will offset the additional
expense. Moreover, if firms have never used a particular innovative technology before, they may have to
bear additional costs such as the costs of initial missteps in applying the technology, as suggested by the
experience curve literature on the introduction of new process technologies discussed earlier.
If a firm decides to use an innovative technology, it is generally because on balance the firm expects
an innovative technology to reduce well drilling costs. As noted above, firms can use more than one
innovative technology per well and all technologies are equally important. It follows that the greater is the
number of innovative technologies used to drill a well, the greater is the effect on well drilling performance.
Innovative Technologies in OILCOMP
To better understand the role of individuals and organizational units in technology usage in
OILCOMP, we interviewed 22 engineers involved in the adoption of three types of innovative technologies
[Swellables, Managed Pressure Drilling (MPD), and Soft Torque Rotary System (STRS)] as well as other
innovative technologies. Despite the potential cost advantages of using innovative technologies, not all
units are equally familiar with these technologies. In addition, our interviews indicated that country units
are often skeptical that more costly innovative technologies will improve performance, and managers and
well engineers may be reluctant to introduce an additional element of risk into operations already
characterized by uncertainty. For example, one senior well engineer commented that: “…our operating
arms…don’t…even want new technology unless it’s absolutely proven beneficial.”
Our conversations also indicated that country units have distinct, locally-composed strategies with
respect to technology use, and there is heterogeneity in how welcoming the units are to new technology.
Path dependency, the track record of old technology, and unit size appear to be at play here. For example,
the first operational applications of STRS took place in a unit that had been a pioneer in the original
(predecessor) Soft Torque technology and strongly believed in its potential. Units were also more open to
11
11
trying a new technology when more established technology simply did not deliver. For example, a deep-
water project drilled nine unsuccessful sidetracks (secondary wellbores away from the original wellbore)
using conventional approaches, before it was prepared to try MPD, a more innovative technology that
turned out to be suitable. In addition, our informants mentioned that units which drill a large number of
wells have a greater ability to experiment with new technologies in several wells without harming the
overall performance of the unit.
Within OILCOMP, the decision of which technologies to use is reached by consensus within an
organizational unit, involving the set of engineers working in an oil field where drilling takes place together
with the business manager responsible for oil field operations and the head of technology for the unit. The
set of engineers is also responsible for implementing the deployment of technologies in the oil field in
consultation with the business manager and the head of technology for the unit. Thus, technology usage
within a unit is an organization unit-level phenomenon.
4
The expertise of an individual engineer can also have an additional effect on well drilling
performance. Our informants indicated that an individual engineer’s expertise may be especially important
during the deployment of an innovative technology after it has been chosen for use by the unit.
Implementation of innovative technologies in well drilling depends substantially on tacit knowledge, and
the expertise of individual engineers who have had prior experience with the technology can be helpful.
One example involves Swellables, a technology where plastic components expand when they are in contact
with water, shutting out sections of the well where water rather than oil enters the wellbore. In this instance,
a well that was drilled “barefoot” (without a casing or liner) provided ideal conditions to try using the
technology. The prior experience of an individual engineer turned out to be invaluable in deploying the
technology effectively, and the well became highly productive at a lower cost than would have been
possible using established technology.
Internal Mobility of Engineers
4
Individual engineers move in and out of organizational units frequently and operational needs differ over time depending on the
number and types of wells to be drilled, so units do not have dedicated teams.
12
12
OILCOMP has a sophisticated human resource management system, and runs an internal labor
market for engineers: country units post job openings and engineers seeking to transfer between country
units apply for jobs. Engineers are expected to move frequently between units in order to advance in their
careers. Where engineers move and when depends on a variety of factors. National regulations affect which
units are seeking to hire engineers, as some governments impose quotas on how many foreign specialists
the units can employ. The attractiveness of a location in terms of quality of life also matters to engineers,
and OILCOMP takes this into consideration by offering higher pay in unattractive locations. Engineers may
also consider the chances of future promotion in their job choices, as some locations often larger country
units - are of strategic importance to the company and therefore receive more attention from top
management. Finally, personal fit and qualifications for a particular job matter. This might include expertise
on innovative technologies, although in our interviews this was not mentioned as a factor. Nevertheless, in
the empirical analysis, we account for the possibility that the assignment of engineers to organizational
units based on innovative technology experience could affect well drilling performance.
Applying the Research Questions in this Study to OILCOMP
The data from OILCOMP make it possible to investigate the extent to which the expertise of
individual engineers with respect to innovative technologies contributes to organizational performance,
over and above the impact of organization-level factors connected with innovative technology usage. We
further investigate whether there is an interactive effect between the technology-related knowledge of
individuals and organizational usage of innovative technology, and if so, whether the two are complements
or substitutes. Given OILCOMP’s decentralized structure, organizational performance and organization-
level factors are measured at the level of the organizational unit. To assess the impact on organizational
performance, we use a standard measure of performance in the industry, well drilling costs per meter.
As noted earlier, firms in the upstream oil industry use innovative technologies in an effort to lower
costs, and no one innovative technology is more important than another. Instead, it is the number of
innovative technologies used by an organizational unit that affects how much costs are reduced (all else
equal, such as the number of wells drilled). The technology-related expertise of an individual engineer may
13
13
have an additional effect on the well drilling performance of an organizational unit. Substantial research
has documented that greater prior experience of individuals in a particular domain results in greater
individual expertise (see Greenwood et al., 2019). Therefore, we measure the expertise of individual
engineers with respect to innovative technology usage as their prior experience using innovative
technologies. Given that units cannot tell ahead of time which technologies will prove most useful and
given that no technologies are more important than others, an engineer has greater potential to improve the
drilling capabilities of the unit and thereby decrease costs when he or she has expertise in a larger
number of innovative technologies. To assess whether individual-level expertise and organization-level
factors are complements or substitutes (or neither) with respect to their effect on organizational
performance, we interact individual-level innovative technology expertise with organization-unit-level
innovative technology usage.
As discussed earlier, performance is generally worse when organizations first introduce new
process technologies, due to difficulties in applying unfamiliar techniques (Argote, 2013). Thus, it is not
surprising that units in OILCOMP are often reluctant to try new technologies due to uncertainty about how
well the technologies will work. The decentralized structure of OILCOMP exacerbates this uncertainty,
because knowledge does not transfer easily between units. Therefore, when units try new technologies, they
may incur costs due to missteps in applying the technologies. This in turn may affect the impact on
performance of innovative technology use by a unit. If the costs associated with new technology
introduction are high enough, well drilling costs could even increase rather than decrease. In addition,
research shows that new technologies may require individuals to perform different tasks when
implementing the technology (Choudhury, Starr, and Agarwal, 2020), which suggests that the expertise of
individual engineers during the deployment of new technologies may affect performance. We therefore
investigate whether the effects on performance of the number of innovative technologies used by a unit, the
innovative technology-related expertise of an individual engineer, and the interaction between them differs
when units introduce innovative technologies that they have not used previously versus using innovative
technologies that the units have used in the past.
14
14
In bringing our data to bear on these questions, we control for a large number of additional factors
that are likely to affect well drilling costs per meter at the level of both the organizational unit and the
individual engineer. These factors include other types of drilling-related expertise of individual engineers,
the overall ability of each engineer, the technology-related expertise of other engineers in each
organizational unit, characteristics of each unit that might affect well drilling performance, and unit fixed
effects (to control for more versus less able units overall), which otherwise might provide alternative
explanations of our findings. We do not control for engineer fixed effects. The transfer of knowledge across
units through the movement of engineers does not depend on or require changes in the expertise of
individual engineers over time. In addition, controlling for engineer fixed effects would make it impossible
to ascertain the effect of the time-invariant portion of an engineer’s expertise on the performance of the unit
to which the engineer has moved, which would be problematic given that engineers’ expertise changes
slowly over time.
As noted earlier, the effects of moving individuals with specialized expertise may reflect both
selection (the hiring of individuals with innovative technology expertise) and treatment (the application of
individuals’ expertise within their units). In our main analysis, we investigate the treatment effect insofar
as possible within the limits our data using instrumental variables. Then, in a robustness analysis, we
investigate whether accounting for selection effects using a Heckman type of analysis that allows for
endogenous regressors with fixed effects alters the results of our main analysis. We next explain our data,
variables, and empirical methods.
Data, Variables, and Methods
Data
Our analysis draws on three different types of data to create a nine-year longitudinal dataset
covering the years 2004 to 2012. The first set of data on OILCOMP’S well drilling and innovative
technology usage covers all wells drilled (7135 wells) between 2000 and 2012. None of the innovative
15
15
technologies, which were relatively new to the industry, was used by OILCOMP prior to 2000. We merged
these data with a second dataset that contains information on all previous job assignments for every engineer
employed by OILCOMP in 2015. For each job assignment of an engineer, the data include the location
(country unit), the start and end date of the assignment, the individual’s engineering specialization, age,
nationality, level of seniority, the date when the individual first joined the company, as well as the
company’s assessment of the individual’s potential for future promotions. The data cover 33,642
assignments, with the earliest one starting in 1979. Finally, we obtained data on oil prices from the U.S.
Energy Information Agency (EIA).
We applied several criteria in merging these data to create our dataset. First, we included only job
assignments from the 22 country units where drilling operations take place. Second, we included only job
assignments during the time period for which we also have well drilling and technology usage data. Third,
because OILCOMP’s data on engineers are organized by job assignment, we organized the merged dataset
according to the job assignment of an individual engineer. We matched each job assignment with data on
well drilling and innovative technology usage in an engineer’s organizational unit during the time period
of an assignment, as well as oil price data.
5
Fourth, to measure an engineer’s prior experience with
innovative technologies and well drilling conditions, we used data from the engineer’s previous job
assignments within the company. Prior to an engineer’s first (earliest) job assignment in the merged panel,
we have no information about his or her previous exposure to innovative technologies. As a result, it is not
possible to measure an engineer’s prior experience with innovative technologies for the engineer’s first job
assignment in the dataset. Therefore, we dropped observations for the first (earliest) job assignment for
each engineer, although data from the first job assignment are used to construct engineer-specific
explanatory and control variables for subsequent job assignments.
5
As noted above, OILCOMP’s engineer-level data are structured according to individual assignments. To capture the prior
exposure of individual engineers to well drilling environments and innovative technologies, we included wells that were in progress
in the country of assignment during the period of each individual assignment. This included wells for which drilling started prior
to the engineer’s arrival but finished either during the assignment or extended beyond it, as well as wells for which drilling started
during the engineer’s assignment.
16
16
Applying these criteria reduced our final sample to 3,208 observations, covering the years 2004 to
2012. Each observation is a job assignment of a single engineer in a single organizational (country) unit.
The sample includes 633 individuals who on average have 5 assignments, with a standard deviation of 3.8,
and a minimum of 1 and a maximum 20 assignments. Forty-five percent of the engineers have a managerial
role, largely as supervisors of well drilling projects (30 percent of all engineers), but 15 percent of the
engineers are in a more senior role. The mean length of an assignment is 198 days.
Despite the richness of our dataset, we note three limitations. First, our dataset includes only
engineers working for OILCOMP in 2015. This raises the question of whether there are systematic
differences between engineers who were hired more recently and those who had worked for the company
throughout the sample period. An examination of the data shows that the mean values for the number of
days spent on an assignment, tenure in OILCOMP, seniority level, and the company’s assessment of an
engineer’s future potential are similar in the first and last year of our merged dataset. This suggests that
there are not systematic differences over time in the sample of engineers on these dimensions.
Secondly, the first year in the panel dataset is 2004 and data on individual engineers’ technology
usage in prior assignments begins in the year 2000. Because job assignments in later years allow for more
experience accumulation subsequent to the year 2000, we control for the last year of each assignment and
for the number of days of each assignment, which in combination control for the first year of each
assignment. In addition, an engineer is likely to have been exposed to a larger number of innovative
technologies the longer that he or she has worked for the company, all else equal. For this reason, we control
for the total length of time that an engineer had worked for OILCOMP prior to each assignment. Engineers
that had previously worked for other companies may also have been exposed to innovative technologies.
Therefore, we control for whether an engineer had worked outside of OILCOMP during his or her career.
Finally, the structure of the well, technology, and individual career data reflects technology data on
the country unit level. In OILCOMP’s decentralized system, country units have responsibility for
operational decisions, including which technologies to deploy. Although units make drilling decisions for
17
17
each oil field separately, unfortunately data are not available that would make it possible to conduct
sensitivity analyses at the project (oil field) or well level.
Variables
We define our variables at the job assignment level that is, during the time period of an assignment
of an engineer to an organizational unit because this is how the OILCOMP personnel data are structured.
Therefore, we measure the performance of an organizational unit during each job assignment.
6
Other
variables are also measured at the job assignment level.
Dependent Variable
We use the standard measure of well drilling performance in the industry, drilling cost per meter.
Because this measure is highly skewed, we log transformed it: Log ( 𝑇𝑜𝑡𝑎𝑙 𝑐𝑜𝑠𝑡𝑠 𝑜𝑓 𝑎𝑙𝑙 𝑤𝑒𝑙𝑙𝑠 𝑑𝑟𝑖𝑙𝑙𝑒𝑑
𝑇𝑜𝑡𝑎𝑙 𝑛𝑢𝑚𝑏𝑒𝑟 𝑜𝑓 𝑚𝑒𝑡𝑒𝑟𝑠 𝑑𝑟𝑖𝑙𝑙𝑒𝑑 ). The
total costs of all wells drilled in a unit during an assignment includes all costs of operations.
7
The total
number of meters drilled is the combined depth in meters of all wells drilled during an assignment.
Independent Variables
Unit innovative technology use: The categorization of innovative technologies comes from
OILCOMP, which uses a common classification in the industry developed by Rushmore Reviews, a large
provider of data services to the upstream oil industry. There are 11 innovative technologies: low rheology
mud, pressure managed drilling (MPD), onshore operation centers for control of drilling processes,
slimhole, intelligent drill pipe, expandable casing, bi-centred bit, lightweight cement, dual derrick, pressure
monitoring while drilling, and other. OILCOMP did not use intelligent drill pipe technology in the drilling
operations in our sample. Given that we are interested in how many different innovative technologies a unit
uses (controlling for the number of wells drilled), we count the number of innovative technologies used at
6
Although we lack data at the oil field or well level, an engineer may work on more than one well or oil field in a unit during an
assignment. Nevertheless, measuring performance at the unit level rather than the oil field or well level makes it more difficult to
pick up an effect on performance of the expertise of an individual engineer.
7
These include the cost of general overhead, base operations, staff overhead, incentive payments, well logging, transportation,
materials, materials supply, marine vessel support, marine supply base, port facility, warehousing, well design and programming,
site survey, well preparation and reinstatement, oil rig mobilization and demobilization, well completion, and well test operations.
18
18
least once by the unit during an engineer’s job assignment.
8
The maximum number of technologies used by
a unit at least once during a job assignment was seven.
Engineer innovative technology experience: This variable is a count of the number of different
innovative technologies used at least once during any of an engineer’s prior assignments. The variable is
calculated similarly to the unit innovative technology use variable except that the number of technologies
is based on the past assignments of an individual engineer. Although individual engineers may not have
worked on every well drilled in their units during prior assignments, our interviews indicated that engineers
exchange information frequently within the units and engineers are exposed to the different well drilling
technologies used in their units.
Control Variables
Co-worker high level of innovative technology experience: In order to tease out the impact of
individual experience with innovative technologies, we need to control for the prior experience of co-
workers within the unit. We first calculated the total number of innovative technologies that each co-worker
(i.e., another engineer working in the same unit as the focal engineer during the time of the focal engineer’s
assignment) had experience with in prior assignments, and summed these for all co-workers. Because this
measure is highly correlated with engineer innovative technology experience, we created a binary (0,1)
variable indicating whether the amount of co-worker experience was in the top quartile. Hence, this variable
captures whether co-workers had substantial prior experience with innovative technology.
9
We also control for several engineer-specific characteristics that could affect well drilling
performance, including measures of experience other than experience with innovative technologies.
10
Experience in headquarters (HQ) location is a binary variable indicating whether an engineer had prior
8
As explained earlier, it is the number of innovative technologies that are used that determines well drilling outcomes. We do not
use an alternate variable that counts the total number of times that innovative technologies were used because this does not measure
the number of technologies that were used. In addition, the total number of innovative technologies that were used is highly
correlated with the size of each unit, which introduces the potential for multicollinearity to affect the coefficient estimates.
9
Using the total number of innovative technologies that co-workers had experience with does not change the results for the
explanatory variables and their interaction, but this co-worker experience variable has an implausible negative relationship with
unit performance. This is likely due to multicollinearity, which can cause the signs of coefficient estimates to switch.
10
The control variables do not include education as is common in the human capital literature, because virtually all of the engineers
in our sample have the same type of engineering degree and there is little variance in the level of education.
19
19
experience in a headquarters location. Because OILCOMP’s main technology development center as well
as several initiatives that support technology are located at its headquarters, we would expect that prior
headquarters experience would increase the chances that an engineer had been exposed to innovative
technologies. At the same time, our interviews also suggested that units might resist headquarters-driven
technology initiatives. Country units might therefore receive engineers with headquarters experience with
some reluctance, making them less effective.
OILCOMP specifies which skills (unconnected to innovative technologies) are needed for a
particular job. For drilling operations, there are four relevant job skill specifications: well engineering,
completion and integration, other engineering, and commercial.
11
For each job assignment, we created
binary variables for each of the last three to indicate whether the skill was a requirement for the assignment,
and therefore whether the engineer would have possessed the skill at the start of the assignment. The job
skill specification of well engineering is the omitted dummy variable.
We control for two types of engineer experience with different geological and geophysical
conditions. Offshore and onshore experience is a binary variable that indicates whether an engineer had
prior experience with both offshore and onshore operations versus only one type. Experience in different
reserve types is a count variable that captures how many different types of reserves an engineer has
previously worked with. OILCOMP identifies five different reserve types, namely, conventional, shale,
high pressure, high temperature, and salt formations.
The amount of time that an individual has spent working as an engineer can contribute to the
engineer’s general ability. We therefore control for the firm-specific experience of each engineer, using the
11
Other engineering includes Civil & Structural Engineering, Electrical Engineering, Engineering Management, Engineering
Materials and Corrosion, Engineering Mechanical Static, Engineering Mining, Engineering Offshore Structures, Engineering
Pipelines, Engineering Production Automation, Control & Operations, Quality Engineering, Subsea, Maintenance Downstream,
Maintenance Upstream, Operational HSSE, Petroleum Engineering, Petrophysics, Reservoir Engineering, Process Engineering
Exploration & Production, Process Engineering Oil & Gas Production Systems, Process Engineering Production Control &
Optimization, Process Safety, Production & Manufacturing Management, Production & Manufacturing Operations, Production &
Manufacturing Product Quality Control, Production Operations Upstream, Production Technology & Production Chemistry,
Project Engineering, Project Services. Commercial includes Asset Management, Business Administration and Support, Business
Information and Data, Commercial Business to Business (B2B), Commercial Upstream, Contracting and Procurement, Corporate
Affairs, Finance Controlling and Accounting, Finance Internal Audit, Gas Commercialisation, General Management Country,
General Management Global, General Management Zonal, Real Estate, Business Analysis, Management & PE, Economics &
Scheduling.
20
20
variable total time in firm. This variable is a count of the number of days since an engineer first joined
OILCOMP, as of the start of each job assignment. Because this variable is skewed, we log transformed it.
More experience in OILCOMP might enable engineers to make more appropriate decisions with regard to
drilling operations. As noted earlier, this variable further controls for experience with innovative
technologies within OILCOMP prior to the start of an assignment. It is also possible that longer tenure in
the firm might lead to inertia and a reluctance to change course. Longer tenured engineers may also be paid
more, raising costs. In addition, we constructed a binary variable, prior experience outside of firm, that
captures whether an engineer had prior experience outside of OILCOMP. This variable controls for the
possibility that an engineer might have had more experience than the total time in firm variable captures.
To further control for the general ability of the engineer, we use two variables. OILCOMP’s HR
system identifies the current level of seniority (i.e., rank) within the company for each engineer and the
potential level of seniority that the engineer is forecast to eventually attain. OILCOMP typically identifies
the seniority potential of each employee soon after the person begins work at the company, which is an
assessment of the ability of the employee, and revises it over time as the company gains more information
about the employee. To some extent, the seniority potential of an engineer creates a self-fulfilling prophecy,
because more attractive jobs may be easier to get for those viewed as having higher potential. This variable
has a value of 1 for individuals with the potential to reach seniority level 1, a value of 2 for individuals with
the potential to reach seniority levels 2 or 3, a value of 3 for individuals with the potential to reach seniority
levels 4 and 5, and a value of 4 for individuals who might reach seniority levels 6, 7, or 8. Seniority level
indicates the current level of seniority (rank), the value of which varies from 1 to 14, with 1 being the most
senior and 14 the least senior. Drilling operations do not involve top executives and the most senior
engineers in our sample are classified as 4. Only 1.34% of the observations in our sample fall into this
category.
We control for assignment-specific factors at the level of the individual as well. The length of an
assignment might affect an engineer’s impact on performance, so we control for the number of days on
assignment. A longer assignment permits an engineer to develop a more intimate knowledge of a unit and
21
21
the local geology. At the same time, we know from our interviews that OILCOMP sees the transfer of
engineers as a substantial part of their development. It is possible that less successful engineers find it harder
to get new job offers and therefore stay in each assignment longer.
In addition to control variables specific to individual engineers and their co-workers, we control
for characteristics of the unit that might affect well drilling performance. First, we include controls for the
scale of operations. The number of wells drilled captures the number of opportunities that a unit has to
experiment with new technologies. If a unit drills many wells and an innovative technology is used in one
of them but it does not perform as expected, this has less impact on the unit’s overall performance than in
a smaller unit. In addition, a larger number of wells may have the advantage of economies of scale. Because
the number of wells does not fully capture the scale of operations, due to differences in costs of individual
wells, we also control for total well cost, the total amount spent on drilling in a unit during an engineer’s
assignment. We log transformed both variables as they are highly skewed.
We also control for whether the unit used a top international drilling contractor. Most firms use
drilling contractors to supplement their operating capabilities. For example, in OILCOMP, it may not be
cost effective for smaller units to maintain their rigs in-house. However, at OILCOMP drilling contractors
are not involved in the choice of which technologies to use. The largest and best-known contractors may
be more skilled at conducting drilling operations, potentially reducing well drilling costs, but their services
may also be more expensive because these contractors are in a stronger bargaining position due to their size
and expertise. We first created a binary variable for each well that indicated whether one of the world’s 10
largest drilling contractors was used. Then we added up the binary variables for all wells drilled during an
assignment, creating a count variable.
Exploration well is a count of the number of exploration wells drilled during an assignment.
Exploration wells, which are drilled in search of new oil, are generally more challenging and costly to drill
because the firm cannot rely on prior knowledge of the geology. At the same time, firms might plan more
carefully because they are aware of the challenges. Similarly, offshore rig is a count of the number of wells
22
22
drilled offshore. These wells are typically more difficult and costly to drill than onshore wells but again,
firms might plan more carefully when drilling these types of wells.
Oil firms measure the number of total non-productive days when they halt drilling operations for
an oil field, due to difficulties encountered during drilling or for economic reasons unrelated to drilling
performance. For example, it may turn out that a different oil field holds more immediate promise in terms
of reserves or a firm may decide to wait for a more favorable oil price before it continues to drill. A larger
number of non-productive days (across all oil fields in a unit) affects both the number of meters drilled and
total well drilling costs, because no drilling costs are incurred and no meters are drilled during a halt to
drilling for an oil field.
Finally, we include unit-level (country) fixed effects to account for time-invariant factors specific
to each unit that might affect drilling costs but are not accounted for by the other control variables, such as
the time-invariant overall drilling capability of the unit. We also include oil price in the first year of each
assignment, measured as the year-end U.S. refiner acquisition cost of crude oil.
12
When the price of crude
oil is high, firms will make more resources available for drilling. We expect the impact on drilling efficiency
to be negative. Because higher oil prices make it possible to make profits while drilling more costly wells,
firms are likely to drill more wells in more difficult operational environments. We further included a dummy
variable for the last year of an assignment in order to control for temporal effects other than oil prices,
including the effect mentioned earlier that assignments in later years would allow more time for engineers
to accumulate experience with innovative technologies.
Methods
In our base regressions, we use OLS estimation with unit (country) fixed effects. We first estimate
the model without the interaction between unit innovative technology use and engineer innovative
technology experience, and then we include the interaction term. In addition, we estimate two stage least
squares (2SLS) instrumental variables regressions with unit fixed effects to account for potential
12
Refiner acquisition cost reflects how much a firm can earn from selling a barrel of crude oil to refiners in the U.S., and is a
composite price for imported and domestic oil.
23
23
endogeneity associated with the explanatory variables. Although we control for many unit-level and
engineer-level factors, there may still be omitted variables that are correlated with our variables of interest.
In addition, there may be reverse causality from well drilling performance to innovative technology use,
particularly if high costs lead a unit to use a larger number of innovative technologies in an effort to reduce
costs. Thus, both of the explanatory variables could be subject to endogeneity, as might the interaction of
these variables. Later in the analysis we also investigate the possibility of sample selection bias if an
organizational unit takes innovative technology into account when hiring engineers.
In the instrumental variables model without the interaction term, we use four instruments for the
potentially endogenous variables of unit innovative technology use and engineer innovative technology
experience. In the instrumental variables model with the interaction term, we include an additional
instrument for the interaction term. These instrumental variables are likely to be correlated with the
variables that measure a unit’s innovative technology use or the innovative technology expertise of an
individual engineer, but are unlikely to have a direct effect on well drilling performance during the time
period when drilling occurs, as explained below. As reported later, the Hansen J test of overidentifying
restrictions provides support for this logic.
The first instrument is the number of times that a unit used innovative technologies in the year
immediately prior to the start of a job assignment in the unit. The frequency of innovative technology use
in the year prior to the start of an assignment should not directly affect well drilling costs during an
assignment because different wells are usually involved. How often a unit uses innovative technologies,
and the number of innovative technologies that a unit uses, change substantially from one year to the next
depending on the wells being drilled at the time.
13
However, the frequency of innovative technology usage
in the prior year may be correlated with the current number of innovative technologies used by each unit if
units that used innovative technologies more frequently in the recent past are prone to use a larger number
of innovative technologies in the near future.
13
Because this instrument reflects technology use in only one year, and technology use varies from year to year within a unit, the
instrument does not capture the accumulated experience of a unit using innovative technologies.
24
24
The second instrument is a binary variable for whether an engineer had experience drilling complex
wells in prior assignments. Complex wells are those for which drilling is not strictly vertical. Different
complex wells use different combinations of technologies due to differences in the unique geological
formations in which the wells are drilled. The uniqueness of geological formations involved in complex
wells means that experience with complex wells does not transfer well between assignments, so an
engineer’s prior experience drilling complex wells is unlikely to directly affect current well drilling costs.
However, innovative technologies are more likely to be used in drilling complex wells. Therefore, prior
experience with complex wells might be correlated with an engineer’s prior experience with innovative
technologies, and might affect costs indirectly through the engineer’s innovative technology expertise.
A third instrument captures the extent to which an engineer held managerial positions prior to the
current assignment. It is a count of the total number of the engineer’s prior assignments that had a
managerial job title above the level of supervisor, which is a low-level managerial position. Greater prior
managerial responsibility might have provided greater exposure to innovative technologies, and therefore
may be correlated with an individual engineer’s innovative technology experience and expertise. Prior
managerial responsibility might also affect whether an individual currently holds a managerial position, but
we control for a current managerial position through the seniority level variable. Beyond these factors, it is
unlikely that prior managerial experience would directly affect current well drilling costs independent of
other types of well drilling experience that are already included in the model.
14
The fourth instrument is a count of the number of times that an engineer had previously used the
innovative technology with which he or she had the most prior experience, a measure of the engineer’s
innovative technology specialization. This is likely to be correlated with the number of innovative
technologies in which an engineer has expertise, since both variables will tend to increase as an engineer’s
prior well drilling experience increases. However, because no one technology is more important than
another and because the technology in which an engineer has the most expertise is no more likely to be
14
Although it might seem that prior managerial experience could sensitize an engineer to the need to control costs, this is
unlikely because most of the managerial positions in our data do not involve direct budgetary responsibility.
25
25
used than any other technology, the depth of an engineer’s prior experience with a specific technology
would not have a systematic effect on current well drilling costs. Finally, in the model that includes the
interaction term between the two explanatory variables, we include a fifth instrument consisting of the
interaction between prior unit innovative technology usage and engineer managerial experience. The
models are estimated using xtreg (OLS fixed effects) and xtivreg2 (2SLS fixed effects) in Stata 16.
Results
Table 1 reports descriptive statistics, and Table 2 reports correlation coefficients for the full sample.
It is notable that units used relatively few innovative technologies per assignment; the average value is 1.50,
with a minimum of zero and a maximum of seven. The mean value of engineer innovative technology
experience is considerably higher, with a mean value of 3.52, a minimum of zero and a maximum of 10.
Table 3 reports the results for the OLS and instrumental variables models. In Table 3, models 1 and 2 report
results for the OLS regressions estimated with robust standard errors clustered at the unit level.
15
Models 3
and 4 in Table 3 report results for the 2SLS instrumental variables regressions with robust standard errors.
A Hausman type test confirms that the three variables are endogenous.
16
In what follows, we focus on the 2SLS instrumental variables results, because the OLS results are
subject to endogeneity. We also concentrate on the results that include the interaction term in model 4,
given our primary interest in understanding whether individual and organization level effects are
complements or substitutes. The estimated coefficient on the interaction term between the two explanatory
variables is statistically significant, indicating that the level of one variable moderates the strength of the
relationship of the other variable with respect to performance. In model 4, a test for under-identification
rejects under-identification (Kleibergen-Papp rk LM statistic value of 49.667, p-value=0.000). A test for
weak identification using the Cragg-Donald Wald F statistic and the Stock-Yogo critical values rejects the
15
The table does not report a constant term for the instrumental variable models 3 and 4 because the xtivreg2 routine does not
estimate a constant term for models with fixed effects (see https://www.stata.com/statalist/archive/2012-07/msg00078.html).
16
It is not possible to estimate cluster robust standard errors using 2SLS due to the structure of the data, because some units have
a small number of observations. The Hausman type test was performed using the endog option in the xtivreg2 routine in Stata.
26
26
hypothesis of weak identification based on 5% maximal relative bias at the 5% level of significance (the
threshold for no more than 5% maximal bias is 9.53, which is exceeded by the value of 10.32 for the Cragg-
Donald Wald F statistic). A test for over-identification fails to reject the null hypothesis that the over-
identifying restrictions are valid (Hansen J statistic value of 1.55, p=0.461).
17
Thus, the test fails to reject
the hypothesis that the instruments are independent of the errors in the second-stage performance
regression.
Several of the control variables in model 4 are statistically significant. The estimated coefficients
for the variables for the scale of well drilling the total number of wells drilled and total costs of well
drilling have large magnitudes in addition to high statistical significance. Drilling a larger number of wells
is associated with lower costs per well, suggestive of economies of scale. In addition, higher total drilling
costs are associated with higher costs per well. Other statistically significant drilling-related variables
include the use of a top drilling contractor, which is associated with higher costs, consistent with the ability
of these contractors to command higher prices for their services. Additional statistically significant drilling-
related variables include the number of wells drilled offshore, which is positively associated with costs per
meter, and the number of non-productive days, which is negatively associated with costs, as expected. Some
control variables at the level of the individual engineer are also significant. Prior experience with different
types of reserves is associated with lower costs per meter, as is co-worker innovative technology experience.
Seniority level (which is reverse coded such that more junior engineers have a higher value of the variable)
is also associated with lower costs. It is possible that all else equal more junior engineers have a larger
impact on drilling costs because they are more directly involved in drilling operations and therefore make
important implementation decisions. Having both onshore and offshore experience as well as length of time
employed by the company are associated with higher costs per meter. The estimated coefficient for onshore
and offshore experience may reflect that individuals with broader experience are paid more and are
17
Model 3 passes the underidentification test (Kleibergen-Papp rk LM statistic value of 173.436, p=0.000), the weak identification
test (based on Cragg-Donald Wald statistic of 51.492, which exceeds the Stock-Yogo critical value of 11.04 for 5% maximal bias
and p<0.05), and the overidentification test (Hansen J statistic value of 3.276, p=0.1943).
27
27
therefore more costly. The positive coefficient estimate for length of time employed by the company may
reflect less up-to-date skills or higher pay for longer-tenured employees. In addition, experience in the
headquarters location and the number of days on an assignment are associated with higher costs.
For the explanatory variables, both innovative technology use by the unit and innovative
technology experience of an individual engineer are statistically significant, and are associated with lower
costs per meter before accounting for the interaction effect. The estimated coefficient on unit innovative
technology use indicates that the use of one additional innovative technology is associated with a 15.4%
reduction in costs per meter (exp(0.1435)=1.154). At an average total drilling cost of approximately 290
million dollars for a unit during an assignment, a 15.4% reduction in costs amounts to an average of 44.7
million dollars. The estimated coefficient on engineer innovative technology experience indicates that an
engineer’s experience with one additional innovative technology is associated with a 4.2% decrease in costs
per meter (exp(0.0413)=1.042), or $12.2 million dollars on average per engineer during an assignment.
Given that an engineer has an average of five assignments in our sample, the total costs savings per engineer
for additional innovative technology experience are substantial.
The interaction between these two variables is also statistically significant, but the coefficient is
positive, indicating that the interaction is negatively associated with performance (costs per meter are
higher). Figure 1 displays the estimated interaction effect graphically. The results indicate that individual
engineer innovative technology expertise and unit innovative technology use are substitutes: a higher level
of individual engineer experience with innovative technology reduces the strength of the relationship
between unit innovative technology use and lower costs, and vice versa.
Robustness Analysis for the Full Sample
The analysis thus far provides evidence that the innovative technology expertise of an individual
engineer has an effect on well drilling costs over and above factors at the organizational level of analysis.
However, an organizational unit often has more than one engineer working on a well and the impact of an
individual engineer may vary depending on the number of engineers who are working on a well. The
number of engineers per well tends to vary depending on drilling conditions, and there are generally fewer
28
28
engineers per well for wells that are easier to drill. Thus, having fewer engineers per well does not mean
that an engineer would necessarily have a larger effect on well drilling costs. However, to assess whether
the results thus far hold up when there are fewer engineers per well, we conducted a separate analysis of
assignments for which the average number of engineers per well during an assignment was below the mean
of 7.3. Table 3 reports the results of this subsample analysis in models 5 (OLS) and 6 (2SLS), which include
the interaction term.
18
The results are consistent with those in the full sample.
The results in the full sample could also be subject to sample selection bias if organizational units
take innovative technology into account when hiring engineers. Although important considerations in hiring
an engineer are likely to include factors such as whether the engineer meets the job skill specification for
the position, expertise in innovative technologies could also affect the hiring decision. For example, because
a unit cannot predict prior to drilling which innovative technologies will be most useful, the unit may prefer
greater breadth of expertise among its engineers. Therefore, when making a hiring decision, a unit might
take into account the number of innovative technologies in which its engineers currently have expertise. A
unit that has expertise in fewer innovative technologies among its engineers might seek to hire engineers
who have greater breadth of experience.
We conduct a formal test for selection bias in the presence of endogenous regressors using the
procedure developed by Semykina and Wooldridge (2010) for panel data with fixed effects. As explained
in the online appendix, the analysis entails estimating a first stage probit selection model that controls for
fixed effects, and then estimating a second stage 2SLS model with fixed effects that tests for sample
selection bias. The dependent variable in the probit model is a binary variable indicating, for each engineer
employed by the company in each year, whether the engineer was hired into a unit during that year or stayed
in their current position. As an instrument in the probit model, we use the innovative technology expertise
of the engineers in a unit in the year prior to hiring, prior year engineer innovative technology experience
18
Model 6 passes the underidentification test (Kleibergen-Papp rk LM statistic value of 73.059, p=0.000), the weak identification
test (based on Cragg-Donald Wald statistic of 17.649, which exceeds the Stock-Yogo critical value of 9.53 for 5% maximal bias
and p<0.05), and the overidentification test (Hansen J statistic value of 2.58, p=0.2752).
29
29
per unit. The instrumental variable is constructed as the count of the number of different innovative
technologies used in the past by all engineers working in a unit in the prior year. The 2SLS coefficient
estimates for our explanatory variables reported in the online appendix are very similar to those reported
earlier, and there is no evidence of sample selection bias.
The robustness analyses support the original results that the interaction of individual expertise and
organization-level factors has a negative effect on organizational performance, and that the two are
substitutes. Next we examine whether the results depend on whether units introduce new technologies.
Comparison of Assignments With and Without the Introduction of New Technology
The organizational units in our sample differ in whether they use only innovative technologies that
they have used in the past and are therefore already familiar with, or whether they also use innovative
technologies with which they are unfamiliar. In 48% of the assignments, units introduced at least one new
innovative technology that they had not used before. When units introduce new technologies, organizational
units and individual engineers may have difficulty implementing innovative technologies with which they
are unfamiliar, and this may affect the main effects of these two factors as well as the interaction effect.
To investigate the impact of new technology introduction, we split the sample into two subsamples:
1) assignments during which the units adopted at least one new innovative technology, and 2) assignments
during which the units adopted no new innovative technologies. In our setting, when units introduced new
technologies, they tended to use more innovative technologies in total. The mean value of innovative
technology use is higher for assignments in the first sample (2.29 innovative technologies, of which 1.33
are newly introduced) than in the second sample (0.783 innovative technologies). Thus, much of the
difference in the average number of technologies that are used in each sample is accounted for by newly
introduced technologies. In contrast, average engineer innovative technology experience is similar between
the two subsamples, with a mean value of 3.55 for the first sample and 3.50 for the second sample. This
suggests that individual engineers on average are equally familiar with innovative technologies in the two
samples, and that whether units introduce new technologies is not correlated with engineers’ expertise.
30
30
Instead, as explained earlier, engineers’ innovative technology expertise may help units implement new
technologies.
Table 1 reports descriptive statistics for the two subsamples, and the online appendix reports
correlation coefficients for each subsample. Table 4 reports the OLS and instrumental variables results for
the two subsamples. For ease of presentation, we report results only for the models that include the
interaction term. In what follows, we concentrate on the instrumental variables results, which are reported
in model 8 (new technology introduction) and model 10 (no new technology introduction).
19
Figures 2 and
3 display the estimated interaction effects for the two models graphically.
For the subsample that had no new technology introductions (model 10), for the three variables of
interest, the signs of the coefficient estimates are the same as those for the full sample and are highly
statistically significant. The coefficient estimate on unit innovative technology use is approximately twice
that in the full sample, the coefficient estimate on engineer innovative technology experience is
approximately half that in the full sample, and the coefficient estimate on the interaction term is
approximately the same as in the full sample. The positive coefficient on the interaction effect suggests that
individual engineer technology-related experience and unit innovative technology use are substitutes.
In contrast, in model 8 for the subsample when units introduce new technologies, the signs on all
three coefficients are reversed, and the coefficient on engineer innovative technology experience is not
statistically significant at conventional levels. The estimated coefficients on unit technology use are
statistically significantly different between the two subsamples (p<0.00015), and the coefficients on the
interaction term differ as well (p<0.036), but the coefficients do not differ significantly for engineer
innovative technology experience (tests were performed using the metaparm routine in Stata).
19
In the new technology introduction subsample, one country unit has only one observation. Because the xtivreg2 routine does
not include a constant term, the routine dropped this observation from the analysis. In model 8, as an instrument we interact prior
unit technology use with prior innovative technology specialization per engineer rather than managerial experience, because the
latter interaction does not perform well as an instrument in the new technology introduction subsample. Both models 8 and 10
pass the underidentification test (Kleibergen-Papp rk LM statistic value of 20.041with p=0.0002 in model 8 and a value of 27.075
with p=0.0000 in model 10), the weak identification test (Cragg-Donald Wald statistic of 7.441 in model 8 and 6.870 in model
10, which exceed the Stock-Yogo critical value of 6.61 in models 8 and 10 for 10% maximal bias and p<0.05), and the
overidentification test (Hansen J statistic value of 1.322 with p=0.5164 in model 8 and a value of 1.033 with p=0.5967 in model
10).
31
31
In model 8, the positive and statistically significant coefficient on unit innovative technology use
indicates that the use of an additional innovative technology is associated with higher costs per meter. This
suggests that, consistent with prior literature, performance may suffer when units introduce new process
technologies due to difficulties in deploying unfamiliar techniques (Argote, 2013). The lack of statistical
significance on the coefficient for engineer innovative technology experience also suggests that engineers
who have had experience with a larger number of technologies, and are therefore more likely to have prior
experience with the newly introduced technologies, may have less difficulty. In addition, the negative
coefficient on the interaction term suggests that the interaction between individual engineer technology
experience and unit innovative technology use is positively related to performance (lower costs per meter),
and the two are complements. When units introduce new technologies, the greater the number of innovative
technologies with which an engineer has experience, then the greater is the likelihood that an engineer has
prior knowledge of the newly introduced technologies which could help the unit implement the
technologies more effectively. Notably, during assignments when units introduced new technologies,
individual engineers on average had prior experience with 0.93 of the newly introduced innovative
technologies used by their units, out of an average of 1.33 innovative technologies introduced per unit.
Discussion and Conclusion
Knowledge is a crucial source of competitive advantage and knowledge transfer is a fundamental
practice for achieving competitive advantage. Individual human capital represents a natural object of
transfer. In fact, prior research has argued that firms can overcome impediments to knowledge transfer
through the transfer of individuals across units, including through intra-firm geographic mobility. However,
a relatively small amount of empirical research has investigated the effects on performance, which depend
importantly on the interactive effects of the expertise of individuals and factors at the organization-unit
level. Given the relatively small amount of prior research on the effect on organizational performance of
intra-firm knowledge transfer through the transfer of individuals, rather than propose formal hypotheses, in
32
32
this study we asked three interrelated research questions.: (1) whether intra-firm transfer of knowledge
through the transfer of individuals with innovative technology-related expertise improves the performance
of the organizational units to which the individuals move, (2) whether the knowledge of these individuals
is a complement or substitute for organization-level factors in the effect on organizational performance, and
(3) whether the answers to these questions depend on the use of new versus familiar technologies.
Our study provides several findings of note. First, the transfer of knowledge through the transfer of
employees to geographically-dispersed organizational units reduces costs substantially, both as a
percentage of costs and in dollar value. Although the engineers in this study are skilled employees, they are
neither stars, nor at the top of the organization, nor especially powerful for other reasons. This finding offers
support for the argument in the microfoundations (Felin, Foss, and Ployhardt, 2015) and strategic human
capital (Nyberg et al., 2014) literatures that individuals have effects on organizational performance beyond
the effects of organizational capabilities. The results also show at a granular level that multinational firms
can benefit significantly from the transfer of knowledge through the internal transfer of employees.
In addition, our findings shed light on the mixed evidence on whether the expertise of individuals
is a complement or substitute for organization level factors by unpacking conditions under which each may
hold. Although it seems natural to assume that individual expertise complements organizational factors in
the effect on organizational performance, instead we find that these are substitutes in the full sample of job
assignments. Further investigation reveals that this result is driven by the subsample where organizational
units do not introduce new innovative technologies. When units rely entirely on innovative technologies
with which they are already familiar, the innovative technology expertise of engineers may be duplicative
to some extent, leading to a substitution effect. In this situation, the use of one input to production such
as familiar technology marginally decreases the benefit for organizational performance of another input
such as an engineers technology-related expertise (Hess and Rothaermel, 2011).
In contrast, the innovative technology expertise of individual engineers complements a unit’s use
of innovative technologies in the subsample where units introduce new technologies. This finding is
consistent with the suggestion in the strategic human capital literature that individual and organization-level
33
33
factors are complements when the workflow depends on the knowledge of individuals (Crocker and
Eckhardt, 2014), as may occur during well drilling when individual engineers have expertise in deploying
an innovative technology that the unit has not used in the past.
Prior research has proposed that whether individual human capital and organization-level factors
are substitutes or complements may depend on the pre-existing competences of the firm and its relationship
to the human capital of individuals (Rothaermel and Hess, 2007). Little subsequent research has
investigated this issue, and our findings suggest that individual and organization-level factors may have a
complementary effect on organizational performance when the organization seeks to acquire new
competences about which individuals possess relevant knowledge. Thus, individuals who bring novel
expertise to their units may be important vehicles for breaking organization-level routines, promoting
organizational learning, and building new competences. In addition, our study provides further evidence
that the movement of experts helps to diffuse knowledge within the organization (Greenwood et al., 2019).
These findings have implications for the microfoundations literature (Felin et al., 2015) and the strategic
human capital literature (Nyberg et al., 2014) by showing that individual expertise may be especially
important when organizational routines need to be changed or new competences built. These findings are
also relevant for multinational firms, which often transfer employees across geographic units, and for
multidivisional firms that redeploy resources such as personnel with specific technical expertise across
businesses (Levinthal and Wu, 2010).
Limitations and future research
Our study has several limitations. Most notably, the data come from a single large company in one
industry, and we do not know the extent to which the findings generalize. Studies of other firms in other
industries would be helpful. In addition, although the study benefited from unique data on the technology-
related experience of individual engineers, the data did not enable us to conduct an analysis of technology
use and performance at the project (well or oil field) level. As a consequence, we could not link individual
expertise to the performance of specific projects on which the individuals worked. Project level data would
be especially helpful in future research on intra-firm mobility, knowledge transfer, and resource
34
34
redeployment. In addition, the setting in our study did not lend itself to an analysis of teams (as opposed to
organizational units or single individuals), which would be helpful to investigate in future research.
Additional research would also be helpful to further disentangle the selection and treatment effects
of knowledge transfer through individuals, which is particularly important for the literature on intra-
organizational mobility as well as for the literature on resource redeployment. Although we used
instrumental variables to control for the endogeneity of individuals’ knowledge and expertise (the
treatment), the instruments may have limitations. Future research could look for exogenous shocks that
would help to further isolate the treatment effect. We also found that the selection effect of an organizational
unit hiring an individual based on the unit’s need for additional technology-related expertise did not affect
the estimates of the treatment effect. However, there are other selection effects that we were unable examine
due to data limitations, such as the selection of individuals into particular projects and the selection of
specific technologies for projects. Project level data would be helpful in order to gain a better understanding
of these types of selection issues.
Finally, our findings suggest the importance for managerial practice of taking the interaction
between individual and organization-level factors into account. Whether firms obtain the full benefit of the
expertise of individuals who are transferred across units depends on whether this expertise and
organization-level factors in the recipient unit are complements or substitutes with respect to organizational
performance. This may depend on contextual factors such as whether individuals bring knowledge about
novel techniques that can help units break with the past and engage in organizational learning.
REFERENCES
Argyris C, Schon D (1978) Organizational Learning A Theory of Action Perspective. Reading, MA Addison-Wesley.
Argote L (2013) Organizational Learning: Creating, Retaining and Transferring Knowledge, 2nd ed. (Springer, New
York).
Argote L, Beckman SL, Epple D (1990) The persistence and transfer of learning in industrial settings. Management
Science 36(2): 140-154.
Argote L, Ingram P (2000) Knowledge transfer: a basis for competitive advantage in firms. Organizational Behavior
and Human Decision Processes 82(1): 150-169.
Becker G (1964) Human Capital: A Theoretical and Empirical Analysis, with Special Reference to Education.
(Chicago, The University of Chicago Press).
Bresnahan TF, Brynjolfsson E, Hitt LM (2002) Information technology, workplace organization, and the demand for
skilled labor: Firm-level evidence. Quarterly Journal of Economics 117(1): 339-376.
35
35
Choudhury P (2016) Return migration and geography of innovation in MNEs: a natural experiment of knowledge
production by local workers reporting to return migrants. Journal of Economic Geography 16(3): 585-610.
Choudhury P (2019) Intra-firm geographic mobility: value creation mechanisms and future research directions.
Advances in Strategic Management, forthcoming.
Choudhury R, Starr E, Agarwal R (2020) Machine learning and human capital complementarities: experimental
evidence on bias mitigation. Strategic Management Journal 41(8): 1381-1411.
Crocker A, Eckhardt R (2014) A multilevel investigation of individual- and unit-level human capital
complementarities. J. Management 40(2): 509-530.
Felin T, Foss NJ, Heimeriks KH, Madsen TL (2012) Microfoundations of routines and capabilities: individuals,
processes, and structure. Journal of Management Studies 49(8): 1351-1374.
Felin T, Foss NJ, Ployhart RE (2015) The microfoundations movement in strategy and organization theory. Acad
Management Annals 9(1): 575-632.
Folta TB, Helfat CE, Karim S (2016) Examining resource redeployment in multi-business firms. In Folta TB, Helfat
CE, Karim S (eds.) Resource Redeployment and Corporate Strategy (Advances in Strategic Management)
35: 1-17.
Galbraith CS (1990) Transferring core manufacturing technologies in high-technology firms. California Management
Review 32(4): 56-70.
Grant RM (1996) Toward a knowledge-based theory of the firm. Strategic Management Journal 17(S2): 109-122.
Greenwood BN, Agarwal R, Agarwal R, Gopal A (2019) The role of individual and organizational expertise in the
adoption of new practices. Organization Science, 30(1): 191-213.
Gupta AK, Govindarajan V (2000) Knowledge flows within multinational corporations. Strategic Management
Journal 21(4): 473-496.
Helfat CE, Eisenhardt KM (2004) Inter-temporal economies of scope, organizational modularity, and the dynamics
of diversification. Strategic Management Journal 25(13): 1271-1232.
Hess AM, Rothaermel FT (2011) When are assets complementary? Star scientists, strategic alliances, and innovation
in the pharmaceutical industry. Strategic Management Journal 32: 895-909.
Hocking JB, Brown M, Harzing A-W (2004) A knowledge transfer perspective of strategic assignment purposes and
their path-dependent outcomes. International Journal of Human Resource Management 15(3): 565-586.
Hymer S (1960) The International Operations of National Firms: A Study of Direct Foreign Investment. (Doctoral
dissertation, Massachusetts Institute of Technology, Cambridge, MA).
Jensen RJ, Szulanski G (2004) Stickiness and the adaptation of organizational practices in cross-border knowledge
transfers. Journal of International Business Studies 35(6): 508-523.
Kogut B, Zander U (1992) Knowledge of the firm, combinative capabilities, and the replication of technology.
Organization Sci. 3(3): 383-397.
Kogut B, Zander U (1993) Knowledge of the firm and the evolutionary theory of the multinational corporation.
Journal of International Business Studies 24: 625-645.
Levinthal DA, Wu B (2010) Opportunity costs and non-scale free capabilities: profit maximization, corporate scope,
and profit margins. Strategic Management Journal 31: 780-201.
Levitt SD, List JA, Syverson C (2013) Toward an understanding of learning by doing: evidence from an automobile
assembly plant. Journal of Political Economy 121(4): 643-681.
Mansfield E, Romeo A, Schwartz M, Teece D, Wagner S, Brach P (1983) New findings in technology transfer,
productivity and economic policy. Research Management 26(2): 11-20.
Mawdsley JK, Somaya D (2016) Employee mobility and organizational outcomes: an integrative conceptual
framework and research agenda. J. Management 42(1): 85-113.
Mundlak Y (1978) On the pooling of time series and cross section data. Econometrica 46: 69-85.
Nyberg AJ, Moliterno TP, Hale D, Lepak D (2014) Resource-based perspectives on unit-level human capital: a review
and integration. J. Management 40: 316-346.
Rothaermel FT, Hess AM (2007) Building dynamic capabilities: innovation driven by individual-, firm-, and network-
level effects. Organization Sci. 18(6): 898-921.
Sakhartov AV, Folta TB (2014) Resource relatedness, redeployability, and firm value. Strategic Management Journal
35(12): 1781-1797.
Semykina A, Wooldridge JM (2010) Estimating panel data models in the presence of endogeneity and selection.
Journal of Econometrics 157: 375-380.
Simon, HA (1991) Bounded rationality and organizational learning. Organization Science 2(1): 125-134.
Stadler C (2011) Process innovation and integration in process-oriented settings: the case of the oil industry. J. Product
Innovation Management 28(s1): 44-62.
36
36
Stadler C, Helfat CE, Verona G (2013) The impact of dynamic capabilities on resource access and development.
Organization Sci. 24(6): 1782-1804.
Szulanski G (1996) Exploring internal stickiness: impediments to the transfer of best practice within the firm. Strategic
Management Journal 17 (Winter Special Issue): 27-43.
Szulanski G (2003) Sticky Knowledge: Barriers to Knowing in the Firm. (SAGE Publications, Thousand Oaks, CA).
Teece DJ (1977) Technology transfer by multinational firms: the resource cost of transferring technological know-
how. The Economic Journal 87: 242-261.
Thompson P (2001) How much did the liberty shipbuilders learn? New evidence for an old case study. Journal of
Political Economy 109(1): 103-137.
Thompson P (2012) The relationship between unit cost and cumulative quantity and the evidence of organizational
learning-by-doing. Journal of Economic Perspectives 26(3): 203-224.
Tidd J, Bessant JR (2018) Managing Innovation: Integrating Technological, Market, and Organizational Change, 6th
ed. (John Wiley & Sons, Hoboken, NJ).
Wooldridge JM (2002) Econometric Analysis of Cross Section and Panel Data. (The MIT Press, Cambridge, MA).
Wright TP (1936) Factors affecting the costs of airplanes. J. Aeronautical Sci. 3: 122-128.
Zander U, Kogut B (1995) Knowledge and the speed of the transfer and imitation of organizational capabilities: an
empirical test. Organization Sci. 6: 76-92.
Figure 1. Interaction Term Full Sample
Figure 2. Interaction Term Figure 3. Interaction Term
New Technology Introduction Subsample No New Technology Introduction Subsample
37
37
Table 1. Descriptive Statistics
Full Sample
New Tech Introduction Subsample
No New Tech Introduction Subsample
Mean
Std.Dev.
Min
Max
Mean
Std.Dev.
Min
Max
Mean
Std.Dev.
Min
Max
Total well cost per meter
9449.38
5833.66
661.95
42655.53
9342.36
3243.25
723.41
20111.63
9546.23
7437.04
661.95
42655.53
log(Total well cost per meter)
8.91
0.78
6.50
10.66
9.06
0.46
6.58
9.91
8.78
0.97
6.50
10.66
Experience in HQ
0.24
0.43
0
1
0.15
0.36
0
1
0.32
0.47
0
1
Exp. with different reserve types
4.10
1.19
0
5
4.28
0.93
0
5
3.94
1.37
0
5
Offshore and onshore experience
0.38
0.49
0
1
0.36
0.48
0
1
0.40
0.49
0
1
Days on assignment
186.26
178.40
1
1613
180.04
169.68
4
1613
191.88
185.82
1
1400
log(Days on assignment)
4.81
0.96
0.69
7.39
4.79
0.94
1.61
7.39
4.82
0.98
0.69
7.24
Seniority level
7.33
1.06
4
11
7.43
1.03
4
11
7.24
1.08
4
10
Seniority potential
2.85
0.75
1
4
2.87
0.75
1
4
2.84
0.74
1
4
Total time in firm
1600.01
1386.29
1
10637
1496.02
1324.89
1
8761
1694.12
1433.49
1
10637
log(Total time in firm)
6.93
1.14
0.69
9.27
6.90
1.05
0.69
9.08
6.96
1.23
0.69
9.27
Prior experience outside of firm
0.57
0.50
0
1
0.55
0.50
0
1
0.58
0.49
0
1
Size (# of wells drilled)
10.02
12.01
1
120
12.24
9.39
1
120
8.01
13.66
1
98
log(# of wells drilled)
2.04
0.82
0.69
4.80
2.39
0.64
0.69
4.80
1.72
0.83
0.69
4.60
Top international drilling contractor
1.95
2.98
0
29
2.17
2.68
0
16
1.75
3.22
0
29
Total well cost (million $)
289.54
252.53
5.74
1701.38
412.13
259.25
12.30
1701.38
178.59
186.46
5.74
1383.71
log(Total well cost)
5.21
1.08
1.91
7.44
5.75
0.87
2.59
7.44
4.72
1.01
1.91
7.23
Exploration well
0.96
1.71
0
18
1.31
1.74
0
11
0.65
1.61
0
18
Offshore rig
6.62
7.89
0
83
9.72
7.69
0
83
3.81
6.96
0
63
Total non-productive days
111.19
114.62
0.40
737.06
145.60
135.83
0.85
737.06
80.06
79.35
0.40
461.84
log(Total non-productive days)
4.27
1.01
0.34
6.60
4.61
0.87
0.62
6.60
3.95
1.03
0.34
6.14
Oil price in start year
81.33
18.52
24.93
106.65
75.80
15.53
33.31
106.65
86.34
19.56
24.93
106.65
Job spec: completion/integration
0.29
0.45
0
1
0.33
0.47
0
1
0.25
0.43
0
1
Job spec: commercial
0.02
0.13
0
1
0.02
0.13
0
1
0.02
0.12
0
1
Job spec: other engineering
0.04
0.20
0
1
0.04
0.19
0
1
0.04
0.21
0
1
Co-worker high level innovative
technology experience
0.26
0.44
0
1
0.24
0.43
0
1
0.27
0.44
0
1
Unit innovative technology use
1.50
1.42
0
7
2.29
1.46
1
7
0.78
0.91
0
3
Engineer innov technology exp.
3.52
2.06
0
10
3.55
1.53
0
10
3.50
2.45
0
10
38
38
Table 2. Correlation Coefficients Full Sample
* p<0.05 Variables in log form: days on assignment, total time in firm, size, total well cost, total non-productive days
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22
1 Total well cost per meter 1.00
2 Experience in HQ 0.07* 1.00
3 Exp with different reserve types 0.29* 0.18* 1.00
4 Off/onshore exp. -0.13* -0.08* 0.2* 1.00
5 Days on assignment -0.08* 0.08* -0.09* 0.01 1.00
6 Seniority level 0.11* -0.01 -0.17* -0.13* -0.06* 1.00
7 Seniority potential -0.01 0.03 -0.04* -0.13* 0.02 0.19* 1.00
8 Total time in firm -0.01 0.01 0.55* 0.23* 0.00 -0.38* -0.04* 1.00
9 Prior exp.outside firm -0.07* 0.04* 0.12* 0.04* 0.07* -0.32* 0.39* 0.14* 1.00
10 Size (# of wells drilled) -0.14* -0.33* 0.07* 0.00 0.34* -0.03 0.04* 0.04* 0.03 1.00
11 Top drilling contractor 0.11* 0.03 0.07* -0.09* 0.36* 0.00 0.02 -0.06* 0.06* 0.42* 1.00
12 Total well cost 0.55* -0.3* 0.26* -0.07* 0.21* 0.05* 0.01 -0.03 0.00 0.7* 0.41* 1.00
13 Exploration well -0.06* -0.13* 0.06* -0.17* 0.16* -0.04* 0.04* -0.09* 0.12* 0.51* 0.43* 0.41* 1.00
14 Offshore rig 0.21* -0.23* 0.08* -0.19* 0.27* 0.01 0.05* 0.01 0.01 0.7* 0.5* 0.68* 0.46* 1.00
15 Total non-productive days -0.08* -0.31* 0.01 0.27* 0.29* -0.03 -0.03 -0.07* 0.1* 0.59* 0.33* 0.55* 0.38* 0.31* 1.00
16 Oilprice in start year -0.04* 0.1* -0.15* 0.12* 0.1* 0.00 -0.01 0.03 -0.11* -0.24* -0.1* -0.26* -0.31* -0.16* -0.13* 1.00
17 Job spec.: comp./ integration 0.07* 0.09* 0.06* -0.06* 0.05* 0.13* 0.2* 0.07* 0.2* 0.01 0.00 0.03 -0.06* 0.03 -0.06* -0.02 1.00
18 Job spec.: commercial 0.01 -0.03 0.03 0.06* 0.03 -0.06* -0.07* 0.05* 0.02 0.01 0.01 0.01 -0.04* 0.01 0.02 -0.01 -0.08* 1.00
19 Job spec.: other engineering 0.05* -0.05* -0.07* -0.05* -0.04* 0.09* 0.05* -0.07* 0.00 0.01 0.04* 0.06* 0.09* 0.04* 0.02 -0.07* -0.13* -0.03 1.00
20 Co-worker high level innov tech exp 0.27* -0.32* 0.17* 0.02 -0.13* 0.03 -0.02 0.07* -0.07* 0.02 -0.05* 0.25* -0.06* 0.11* -0.12* 0.12* 0.02 0.01 0.01 1.00
21 Unit innov technology use 0.25* -0.17* 0.2* 0.06* 0.03 0.05* -0.02 -0.12* 0.08* 0.28* 0.19* 0.54* 0.42* 0.22* 0.44* -0.3* -0.01 -0.01 0.1* 0.25* 1.00
22 Engineer innov technology exp 0.26* 0.34* 0.66* 0.37* -0.07* -0.22* -0.04* 0.46* 0.14* -0.08* -0.04* 0.13* -0.14* -0.03 0.00 0.06* 0.08* 0.02 -0.05* 0.24* 0.2* 1.00
39
39
Table 3. Organizational Performance: Log(Total Well Cost per Meter)
M1: OLS
M2: OLS
M3: IV
M4: IV
M5: OLS
M6: IV
Experience in HQ
-0.0034
0.0129
0.0500*
0.0980***
-0.0144
0.0618*
(0.0292)
(0.0229)
(0.0223)
(0.0280)
(0.0168)
(0.0314)
Experience with different reserve types
-0.0241*
-0.0228*
-0.0153***
-0.0120**
-0.0140*
-0.0061
(0.0091)
(0.0084)
(0.0041)
(0.0043)
(0.0061)
(0.0043)
Offshore and onshore experience
0.0258*
0.0287*
0.0413***
0.0453***
0.0152
0.0348**
(0.0123)
(0.0135)
(0.0109)
(0.0117)
(0.0117)
(0.0116)
Days on assignmenta
0.0144
0.0150
0.0135***
0.0160***
0.0069
0.0058
(0.0125)
(0.0124)
(0.0039)
(0.0041)
(0.0190)
(0.0050)
Seniority level
-0.0017
-0.0029
-0.0040
-0.0080*
-0.0024
-0.0048
(0.0042)
(0.0039)
(0.0029)
(0.0031)
(0.0043)
(0.0027)
Seniority potential
-0.0028
-0.0025
-0.0024
-0.0011
-0.0007
-0.0004
(0.0036)
(0.0042)
(0.0038)
(0.0039)
(0.0034)
(0.0036)
Total time in firma
0.0188
0.0185*
0.0245***
0.0221***
0.0098
0.0148***
(0.0093)
(0.0087)
(0.0036)
(0.0039)
(0.0048)
(0.0033)
Prior experience outside of firm
0.0112
0.0083
0.0130*
0.0026
0.0030
0.0041
(0.0071)
(0.0066)
(0.0061)
(0.0068)
(0.0071)
(0.0063)
Size (# of wells drilled)a
-0.9492***
-0.9503***
-0.9442***
-0.9459***
-0.8623***
-0.8632***
(0.0816)
(0.0801)
(0.0142)
(0.0146)
(0.0697)
(0.0156)
Top international drilling contractor
0.0035
0.0043*
0.0027*
0.0050***
0.0032
0.0034*
(0.0024)
(0.0019)
(0.0014)
(0.0015)
(0.0027)
(0.0014)
Total well costa
0.8482***
0.8477***
0.8430***
0.8455***
0.7313***
0.7365***
(0.0966)
(0.0954)
(0.0112)
(0.0114)
(0.0821)
(0.0134)
Exploration well
-0.0053
-0.0010
-0.0091*
0.0085
-0.0015
-0.0011
(0.0109)
(0.0085)
(0.0041)
(0.0065)
(0.0081)
(0.0046)
Offshore rig
0.0054
0.0053
0.0055***
0.0051***
0.0033
0.0031***
(0.0045)
(0.0045)
(0.0010)
(0.0010)
(0.0032)
(0.0008)
Total non-productive daysa
-0.0602
-0.0595
-0.0582***
-0.0547***
0.0065
0.0084
(0.0513)
(0.0514)
(0.0079)
(0.0080)
(0.0306)
(0.0064)
Oil price in start year
-0.0005
-0.0005
-0.0003
-0.0004
-0.0003
-0.0002
(0.0007)
(0.0007)
(0.0002)
(0.0002)
(0.0010)
(0.0002)
Job specification: completion and
-0.0092*
-0.0084*
-0.0068
-0.0044
-0.0102
-0.0083
integration
(0.0039)
(0.0037)
(0.0056)
(0.0059)
(0.0065)
(0.0054)
Job specification: commercial
0.0103
0.0145
0.0094
0.0243
0.0054
0.0099
(0.0104)
(0.0101)
(0.0191)
(0.0206)
(0.0046)
(0.0143)
Job specification: other engineering
0.0086
0.0112
0.0077
0.0177
-0.0023
-0.0012
(0.0075)
(0.0075)
(0.0136)
(0.0139)
(0.0051)
(0.0115)
Co-worker high level of innov tech exper
-0.0474
-0.0538
-0.0404***
-0.0589***
0.0092
0.0036
(0.0521)
(0.0507)
(0.0097)
(0.0114)
(0.0311)
(0.0104)
Unit innovative technology use
-0.0373
-0.0656**
-0.0315*
-0.1435***
-0.0337⁺
-0.0603**
(0.0284)
(0.0221)
(0.0126)
(0.0364)
(0.0170)
(0.0230)
Engineer innov technology experience
0.0008
-0.0073*
-0.0161**
-0.0413***
-0.0017
-0.0269**
(0.0056)
(0.0034)
(0.0054)
(0.0093)
(0.0041)
(0.0096)
Unit technology x engineer technology
0.0066*
0.0228***
0.0051**
0.0116**
(0.0026)
(0.0067)
(0.0018)
(0.0045)
Constant
6.7210***
6.7492***
6.8862***
(0.1996)
(0.1956)
(0.1891)
Year Dummies
YES
YES
YES
YES
YES
YES
Unit (Country) Fixed Effect
YES
YES
YES
YES
YES
YES
Observations
3208
3208
3208
3208
2408
2408
R2
0.851
0.853
0.848
0.838
0.871
0.865
Adjusted R2
0.850
0.852
0.846
0.835
0.869
0.862
Standard errors in parentheses; ⁺ p<0.10, * p<0.05, ** p<0.01, *** p<0.001 a variable in log form
40
40
Table 4. New Technology and No New Technology Introduction Subsamples
New Tech Intro
No New Tech Intro
M7: OLS
M8: IV
M9: OLS
M10: IV
Experience in HQ
0.0158
-0.0124
-0.0122
0.0359
(0.0235)
(0.0347)
(0.0151)
(0.0264)
Experience with different reserve types
0.0000
0.0055
-0.0130
-0.0047
(0.0016)
(0.0056)
(0.0079)
(0.0064)
Offshore and onshore experience
0.0115
0.0076
0.0316⁺
0.0083
(0.0126)
(0.0118)
(0.0155)
(0.0176)
Days on assignmenta
0.0009
-0.0122
0.0038
-0.0024
(0.0124)
(0.0075)
(0.0134)
(0.0062)
Seniority level
-0.0012
-0.0004
-0.0030
-0.0089⁺
(0.0024)
(0.0029)
(0.0044)
(0.0046)
Seniority potential
0.0009
-0.0037
-0.0061⁺
-0.0061
(0.0021)
(0.0040)
(0.0034)
(0.0056)
Total time in firma
0.0043
0.0046
0.0156*
0.0122*
(0.0045)
(0.0047)
(0.0071)
(0.0056)
Prior experience outside of firm
-0.0049⁺
-0.0041
0.0112**
0.0081
(0.0027)
(0.0066)
(0.0034)
(0.0090)
Size (# of wells drilled)a
-0.7758***
-0.7235***
-0.9094***
-0.8527***
(0.0778)
(0.0331)
(0.0395)
(0.0257)
Top international drilling contractor
-0.0021
-0.0044⁺
0.0129**
0.0129***
(0.0014)
(0.0027)
(0.0035)
(0.0021)
Total well costa
0.7355***
0.6922***
0.8049***
0.8148***
(0.0842)
(0.0218)
(0.0480)
(0.0188)
Exploration well
-0.0130
-0.0436***
0.0144⁺
0.0244***
(0.0096)
(0.0117)
(0.0073)
(0.0062)
Offshore rig
0.0063***
0.0073***
-0.0060**
-0.0079***
(0.0011)
(0.0013)
(0.0016)
(0.0014)
Total non-productive daysa
-0.0327
-0.0233*
-0.0516
-0.0356**
(0.0351)
(0.0105)
(0.0302)
(0.0123)
Oil price in start year
0.0015***
0.0017***
-0.0010
-0.0005
(0.0003)
(0.0003)
(0.0007)
(0.0004)
Job specification: completion and
-0.0011
0.0061
0.0056
-0.0021
integration
(0.0037)
(0.0059)
(0.0064)
(0.0098)
Job specification: commercial
0.0260*
0.0241⁺
0.0087
0.0540⁺
(0.0119)
(0.0146)
(0.0199)
(0.0309)
Job specification: other engineering
-0.0082
-0.0089
-0.0055
0.0252
(0.0083)
(0.0138)
(0.0082)
(0.0232)
Co-worker high level of innov tech exper
0.0298*
0.0400***
-0.1319***
-0.0769**
(0.0119)
(0.0116)
(0.0304)
(0.0245)
Unit innovative technology use
-0.0196⁺
0.1208**
-0.0590**
-0.3056***
(0.0093)
(0.0457)
(0.0197)
(0.0790)
Engineer innov technology experience
-0.0143⁺
0.0306
-0.0064
-0.0228**
(0.0070)
(0.0197)
(0.0038)
(0.0071)
Unit technology x engineer technology
0.0062**
-0.0182*
0.0035⁺
0.0218***
(0.0020)
(0.0088)
(0.0019)
(0.0064)
Constant
6.6467***
6.9268***
(0.2652)
(0.1365)
Year Dummies
YES
YES
YES
YES
Unit (Country) Fixed Effect
YES
YES
YES
YES
Observations
1524
1523
1684
1684
R2
0.929
0.890
0.850
0.779
Adjusted R2
0.928
0.887
0.847
0.773
Standard errors in parentheses; ⁺ p<0.10, * p<0.05, ** p<0.01, *** p<0.001 a variable in log form
... We find that neither humans nor AI and complement humans' cognitive capabilities (Raisch & Krakowski, 2021). Prior RBV research has documented the co-occurrence of substitution and complementation when organizations use both familiar and new technologies (Stadler, Helfat, & Verona, 2021). In the case of AI, however, the technology and its novelty remain constant but substitution and complementation are still expected to co-occur. ...
... For example, Shrestha et al. (2019) warn that combining human and machine capabilities in an AI context could also lead to negative complementarities. Second, prior research studied substitution and complementation in different contexts (Hess & Rothaermel, 2011;Stadler et al., 2021), providing little insight into how they are interrelated. In the case of AI, scholars described these resource dynamics as occurring in the same context (Brynjolfsson & McAfee, 2014, p. 134f;Raisch & Krakowski, 2021). ...
... The ways sources of competitive advantage change over time have received substantial research attention (Argyres & Zenger, 2012;Levinthal & Wu, 2010;Polidoro & Toh, 2011). This literature has tended to treat a new resource as either a substitute or a complement, informing a contingency view (Hess & Rothaermel, 2011;Stadler et al., 2021): The resource is either related, thus providing the potential to substitute rather than complement, or it is unrelated, therefore enabling complementation rather than substitution. In contrast, our findings show substitution and complementation as interrelated effects. ...
Article
Full-text available
We apply a resource-based view to investigate how the adoption of Artificial Intelligence (AI) affects competitive capabilities and performance. Following prior work on using chess as a controlled setting for studying competitive interactions, we compare the same players’ capabilities and performance across conventional, centaur, and engine chess tournaments. Our analysis shows that AI adoption triggers interrelated substitution and complementation dynamics, which make humans’ traditional competitive capabilities obsolete, while creating new sources of persistent heterogeneity when humans interact with chess engines. These novel human-machine capabilities are unrelated, or even negatively related, to traditional capabilities. We contribute an integrated view of substitution and complementation, which identifies AI as the driver of these dynamics and explains how they jointly shift the sources of competitive advantage.
... On the other hand, the latter authors conclude that, instead of being complementary, "a firm's intellectual human capital (proxied by its total scientists) and a firm's R&D capability are substitutes for one another… Star scientists and R&D capability also substitute for one another." Similarly, Stadler, Helfat, and Verona (2021) investigate whether individual skills and organizational practices substitute or complement each other, and find that the answer depends upon the familiarity of the work, with a substitution effect for familiar work and a complementarity effect for unfamiliar work. ...
... In further opening this black box to examine how other organizational dimensions influence resource-based advantages, there is a significant opportunity to refine RBV. In this regard, microfoundations theory (Felin et al., 2012;Felin et al., 2015;Nyberg et al., 2014) has led the way, inspiring a recent stream of cross-level empirical research to investigate these contingency factors (Crocker & Eckardt, 2014;Ethiraj & Garg, 2011;Lacetera et al., 2004;Rothaermel & Hess, 2007;Stadler et al., 2021). Our study contributes to this emerging literature. ...
Article
An under-researched and under-theorized factor in the resource-based view's VRIO acronym is the “O”: organizational deployment. Exploiting a rule change that exogenously shifted the workflow interdependence of NBA basketball teams, we examine how the benefit provided by a resource depends upon the way it is deployed in an organization's workflow. We also examine how workflow interdependence influences the relative effectiveness of two capabilities for obtaining resources—external acquisition versus internal development. We find that losing star players is more damaging under conditions of greater pooled interdependence and that external acquisition capabilities are more effective for mitigating this damage. Conversely, we find that losing non-star players is more damaging under conditions of greater reciprocal interdependence and that internal development capabilities are more effective for mitigating this damage. This article is protected by copyright. All rights reserved.
Article
Purpose The Brittelstand are innovative, family-owned firms that offer national and international opportunities in the United Kingdom (UK). These fast-growing businesses are customer-oriented and proud of family ownership and embeddedness of the businesses within communities. While Brittelstand firms are as likely to deploy open innovation models as non-Brittelstand firms, these firms' engagement with customers in regional and national markets and the ability to benefit from this collaboration contrasts with these firms' willingness to engage in open innovation. Design/methodology/approach Using longitudinal data and regression analysis on 13,876 firms with 24,286 observations over 2004–2020, the authors develop and test a theoretical framework of open innovation in the Brittelstand. The authors' model explains the willingness and ability of the Brittelstand firms to engage in open innovation and benefit from it. Findings The authors' results show that Brittelstand firms are less willing than non-Brittelstand firms to collaborate with customers and universities, contrasting prior research on family firms, and distinguishing the innovation model of the Brittelstand from a family business model. The Brittelstand firms who are able to engage in collaboration with customers in domestic markets will outperform the firms' non-Brittelstand counterparts in innovation outputs. Research limitations/implications In line with other studies, this study is associated with several limitations that open opportunities for further research that replicate and/or extends this study. First, this study is unbalanced panel data and the fact that some firms appear in the model only once from 2004–2020. The longitudinal study will allow to enforce causality of the relationship and examines the dynamics of open innovation in the Brittelstand. Second, the indicator on the extent and mechanisms of collaboration with customers could be better explained and measured, for example, using a scale indicator instead of a binary variable for knowledge collaboration across different types of partners and four geographical dimensions. Practical implications First, Brittelstand firms who are less likely to employ open innovation models nationally and with customers. However, those Brittelstand firms who decide to collaborate with customers nationally are more likely to increase the innovation sales compared to those firms that do not engage in such collaborations? This is an interesting and unexpected finding, which means that low willingness of cross-country and cross-regional collaboration for Brittelstand firms is not optimal and engagement in collaboration with customers in domestic markets is beneficial for innovation. Managers and policymakers may use this finding to design and re-design open innovation strategies managers and policymakers with customers within and across regions in the UK. Second, managers may benefit from the integrated view on the two drivers of firm innovation – collaboration with customers and the local embeddedness of such collaboration. Social implications The authors' results show that Brittelstand firms outperform the firms' non-Brittelstand counterparts by adopting an open model of innovation with customers in domestic markets. This means that the most dynamic and fast growing Brittelstand firms are those who collaborate with customers for new ideas and innovation. Originality/value This study describes the phenomenon of the Brittelstand and investigates the link between open knowledge sourcing across different geographical proximities and partners and innovation outputs. First, the authors contribute to open innovation and resource-based view (RBV) literature in family firms by theorizing and empirically testing the open innovation model for the Brittelstand firms. The authors also debate that the Brittelstand firms should overcome this inertia of willingness to collaborate across heterogeneous external partners and convert regional/national embeddedness of the firms with customers into strengths for greater product innovation. Second, the authors contribute to family business literature by explaining how and why the Brittelstand firms can achieve greater innovation outputs. In doing so, the authors draw on the concept of familiness and local embeddedness.
Article
This study focuses on the sectoral and geographical differences in open knowledge collaboration for the innovation performance of UK firms. Drawing on transaction cost theory and appropriability of knowledge, we adopt a resource-based view to focus on limits to open innovation. We use a generalized multi-level mixed model to test our hypotheses, controlling for time, regional and firm unobserved characteristics. Our sample includes 19,510 observations and 17,859 firms, with a small panel element of 1651 firms mainly from the UK Innovation Survey and Business Registry. Our results demonstrate that limits to open innovation differ across knowledge-intense sectors and at different geographic dimensions, with creative sectors experiencing the greatest limits to knowledge collaboration in both national and international markets. We discuss why and how this occurs, and conclude with several theoretical and policy implications.
Preprint
We apply a resource-based view to investigate how the adoption of Artificial Intelligence (AI) affects competitive capabilities and performance. Following prior work on using chess as a controlled setting for studying competitive interactions, we compare the same players’ capabilities and performance across conventional, centaur, and engine chess tournaments. Our analysis shows that AI triggers interrelated substitution and complementation dynamics, which make humans’ traditional competitive capabilities obsolete, while creating new sources of persistent heterogeneity when humans interact with chess engines. These novel human-machine capabilities are unrelated, or even negatively related, to traditional capabilities. We contribute by providing an integrated view of substitution and complementation, including the description of AI as a driver of the two, eventually explaining substantial shifts in the sources of competitive advantage.
Article
Full-text available
This paper introduces the volume on Resource Redeployment and Corporate Strategy, which is devoted to exploring a relatively new justifi-cation for how multi-business firms create value-having flexibility to internally redistribute non-financial resources across their businesses. We clarify how a theory around resource flexibility differs from other theories of how multi-business firms create value. We then synthesize the collection of papers in this volume and describe how they contribute to this line of inquiry. Finally, we offer our own views on opportunities for elaboration of this theory. © 2016 by Emerald Group Publishing Limited All rights of reproduction in any form reserved.
Article
The use of machine learning (ML) for productivity in the knowledge economy requires considerations of important biases that may arise from ML predictions. We define a new source of bias related to incompleteness in real time inputs, which may result from strategic behavior by agents. We theorize that domain expertise of users can complement ML by mitigating this bias. Our observational and experimental analyses in the patent examination context support this conjecture. In the face of “input incompleteness,” we find ML is biased towards finding prior art textually similar to focal claims and domain expertise is needed to find the most relevant prior art. We also document the importance of vintage‐specific skills, and discuss the implications for artificial intelligence and strategic management of human capital. Managerial Summary Unleashing the productivity benefits of machine learning technologies in the future of work requires managers to pay careful attention to mitigating potential biases from its use. One such bias occurs when there is input incompleteness to the ML tool, potentially because agents strategically provide information that may benefit them. We demonstrate that in such circumstances, ML tools can make worse predictions than the prior technology vintages. To ensure productivity benefits of ML in light of potentially strategic inputs, our research suggests that managers need to consider two attributes of human capital—domain expertise and vintage‐specific skills. Domain expertise complements ML by correcting for the (strategic) incompleteness of the input to the ML tool, while vintage‐specific skills ensure the ability to properly operate the technology. This article is protected by copyright. All rights reserved.
Article
New information pertinent to organizational decision making, even when publicly available, may not diffuse rapidly in the form of adoption and transformation of organizational practices. In this study, we examine how different markers of expertise, each representative of human capital at both individual and organizational levels, moderates the speed of response to new information. We do so in the context of medical device utilization, viz. stents, for the treatment of stable coronary arterial disease by physicians practicing in hospitals. Results show that physicians possessing specialized expertise developed through deliberate practice adopt new guidelines significantly faster, as compared with physicians endowedwith general expertise reflected in elite schooling or tenure. Furthermore, we observe significant spillovers within organizations from expertise gained through deliberate practice, indicating that physicians with expertisemarkers associated with deliberate practice are able to act as influential agents and help diffuse new practiceswithin the organization.Our study thus extends the literature on both information diffusion and expertise by providing quantitative and qualitative evidence of the mechanisms at play in the adoption of new best practices.