The Network‐Performance Relationship in Knowledge‐Intensive Contexts—A Meta‐Analysis and Cross‐Level Comparison

• DOI: http://doi.org/10.1002/hrm.21823
• Date: 01 January 2018
• Published date: 01 January 2018

Human Resource Management, January–February 2018, Vol. 57, No. 1. Pp. 11–36

© 2017 Wiley Periodicals, Inc.

Published online in Wiley Online Library (wileyonlinelibrary.com).

DOI:10.1002/hrm.21823

Correspondence to: Julia Brennecke, Department of Organisation and Management, University of Liverpool

Management School Chatham Street, Liverpool L69 7ZH, UK, Ph: 0044 7880 835274, julia.brennecke@liverpool.ac.uk

Both authors contributed equally to this article.

THE NETWORK-PERFORMANCE

RELATIONSHIP IN

KNOWLEDGE-INTENSIVE

CONTEXTS—A META-ANALYSIS

AND CROSS-LEVEL COMPARISON

JULIA BRENNECKE AND NATALIE STOEMMER

This study examines the generalizability of the network-performance relationship

across individual and group levels, focusing on knowledge-intensive contexts.

Drawing on a meta-analytical approach, we synthesize the results of 102 empir-

ical studies to test whether network characteristics such as centrality, broker-

age, and tie strength similarly infl uence the job performance of individuals and

groups. Results show that while there are no differences in the direction of the

network-performance relationship across levels, there are substantial differences

in magnitude. Individual performance profi ts more strongly from a high num-

ber of direct connections, whereas groups reap higher benefi ts from brokerage

positions. Additional analyses reveal that the network measurement method, tie

content, and performance criteria function as moderators of the network perfor-

mance relationship, but their infl uence is consistent neither across network char-

acteristics nor across levels. By meta-analytically comparing and contrasting the

network-performance relationship for individuals and groups, we contribute to

multilevel research on networks and organizations. Particularly, we move toward

the development of a multilevel homology theory of networks. Implications for

theory, practice, and future research are discussed. ©2017 Wiley Periodicals, Inc.

Keywords: group network, individual network, job performance, meta- analysis,

multilevel

Researchers and practitioners alike have

long recognized the beneficial influence

of network embeddedness on perfor-

mance at different organizational levels,

particularly in knowledge-intensive con-

texts (e.g., Burt, 2004; Cross, Kaše, Kilduff, &

King, 2013; Keller, 2001). At the individual level,

recent meta-analytic summaries of several decades

of research have shown that a central network

position, brokerage, and strong ties are crucial

determinants of job performance and innova-

tiveness (Baer, Evans, Oldham, & Boasso, 2015;

Fang et al., 2015). With the rise of teamwork in

organizations, scholars increasingly investigate

12 HUMAN RESOURCE MANAGEMENT, JANUARY–FEBRUARY 2018

Human Resource Management DOI: 10.1002/hrm

Examining whether

different network

characteristics vary

in their influence

on job performance

depending on the

level of theorizing,

we are able to refine

the scope of the

network-performance

relationship and

improve our

understanding

of cross-level

differences in

organizations.

relationships that are not homologous across indi-

vidual and group levels (Moliterno & Mahony,

2011). Our study focuses on research conducted in

knowledge-intensive contexts where networks are

especially important for individuals and groups to

succeed due to high degrees of task interdepen-

dence and complexity (e.g., Reagans & McEvily,

2003). We investigate centrality, brokerage, and

tie strength as key characteristics of instrumen-

tal as well as expressive networks that the major-

ity of past research on the network-performance

relationship at either level has analyzed. We

examine the influence of these network charac-

teristics on job performance of individuals and

groups, defined as success in completing tasks and

responsibilities in a given role (Fang et al., 2015).

To investigate the generalizability of our findings

not only across levels, we additionally analyze the

influence of meaningful moderators on the net-

work-performance relationship.

Exploring whether the network-performance

relationship is generalizable across levels, we con-

tribute to multilevel research on networks and

organizations. Particularly, we add to the devel-

opment of a multilevel homology theory of net-

works in knowledge-intensive contexts. While

there have been calls in the literature to analyze

the cross-level generalizability of relationships for

some years now (Kozlowski & Klein, 2000; Payne,

Moore, Griffis, & Autry, 2011; Phelps, Heidl,

& Wadhwa, 2012), there are, to the best of our

knowledge, no studies taking on this task with

respect to organizational networks. Examining

whether different network characteristics vary in

their influence on job performance depending on

the level of theorizing, we are able to refine the

scope of the network-performance relationship

and improve our understanding of cross-level

differences in organizations (G. Chen, Bliese, &

Mathieu, 2005). Moreover, our study contributes

to existing research by meta-analytically syn-

thesizing the empirical literature on the impact

of individual-level and group-level networks on

performance with a particular focus on knowl-

edge-intensive contexts. In a research area such

as organizational networks, which has produced

a large body of publications in the past 30 years,

meta-analysis provides the means to consolidate

the often conceptually different studies and dis-

cover common patterns within and across levels.

Theory and Hypotheses

Homology: The Network-Performance

Relationship across Organizational Levels

In multilevel research, the term homology

describes the notion that “constructs and the

the network-performance relationship also at the

level of the group, focusing on groups’ external

networks to other groups (e.g., Kratzer, Leenders,

& van Engelen, 2010; Tsai, 2001). Their basic

assumption is that in today’s complex environ-

ment single groups—just like single individu-

als—are no longer able to possess all the relevant

resources needed to succeed and can reap perfor-

mance benefits from networks (e.g., Ancona &

Caldwell, 1992; Marrone, 2010).

A review and comparison of research conducted

at the individual level and group level highlights

that studies at both levels largely apply the same

network theoretic constructs and their empirical

associations to predict performance

(Borgatti & Foster, 2003; Moliterno

& Mahony, 2011). In other words, a

certain network position of employ-

ees in a coworker network allegedly

has the same effect on employee

job performance as the respective

position of groups in an intergroup

network has on group performance.

More than that, it seems to be com-

mon practice to refer to studies con-

ducted at the individual level when

the actual object of analysis is the

group and vice versa (e.g., Lechner,

Frankenberger, & Floyd, 2010; Zaheer

& Soda, 2009), often without explic-

itly pointing out the cross-level infer-

ence made. In short, research on the

network-performance relationship in

organizational contexts seems to be

guided by the implicit assumption

that the influence of network embed-

dedness on performance is general-

izable—or homologous (G. Chen,

Bliese, & Mathieu, 2005)—across

levels. However, we have yet to test

the homology assumption empiri-

cally. Neglecting to do so and merely

assuming generalizability, scholars

risk oversimplifying their theoretical model and

committing a cross-level fallacy (Rousseau, 1985),

and practitioners may draw flawed conclusions

about which network structures to foster at each

level within their organization.

The purpose of this study is to test the gen-

eralizability of the network-performance relation-

ship for individuals and groups. To this end, we

conduct a meta-analysis and examine whether

the level of theorizing moderates the influence

of network embeddedness on job performance.

This approach not only enables us to validate

within-level findings across a large set of studies,

but also allows us to identify network-theoretic