Rough Draft Is face-to-face communication still necessary for forming bonds with others? Copyedited by: ES MANUSCRIPT CATEGORY: Article [17:07 8/3/2021 OP

Click here to Order a Custom answer to this Question from our writers. It’s fast and plagiarism-free.

Rough Draft Is face-to-face communication still necessary for forming bonds with others? Copyedited by: ES MANUSCRIPT CATEGORY: Article

[17:07 8/3/2021 OP-REST200062.tex] RESTUD: The Review of Economic Studies Page: 574 574–609

Review of Economic Studies (2021) 88, 574–609 doi:10.1093/restud/rdaa060
© The Author(s) 2020. Published by Oxford University Press on behalf of The Review of Economic Studies Limited.
Advance access publication 13 October 2020

Face-to-Face Communication
in Organizations

School of Economics, University of Edinburgh and Centre for Economic Performance,

London School of Economics

Department of Management, London School of Economics; Centre for Economic

Performance, London School of Economics and Centre for Economic and Policy Research


Copenhagen Business School and Centre for Economic Performance, London School of


First version received March 2018; Editorial decision July 2020; Accepted September 2020 (Eds.)

Communication is integral to organizations and yet field evidence on the relation between
communication and worker productivity remains scarce. We argue that a core role of communication
is to transmit information that helps co-workers do their job better. We build a simple model in which
workers choose the amount of communication by trading off this benefit against the time cost incurred by
the sender, and use it to derive a set of empirical predictions. We then exploit a natural experiment in an
organization where problems arrive and must be sequentially dealt with by two workers. For exogenous
reasons, the first worker can sometimes communicate face-to-face with their colleague. Consistently with
the predictions of our model, we find that: (1) the second worker works faster (at the cost of the first worker
having less time to deal with incoming problems) when face-to-face communication is possible, (2) this
effect is stronger when the second worker is busier and for homogenous and closely located teams, and
(3) the (career) incentives of workers determine how much they communicate with their colleagues. We
also find that workers partially internalise social outcomes in their communication decisions. Our findings
illustrate how workers in teams adjust the amount of mutual communication to its costs and benefits.

Key words: Teamwork, Face-to-Face communication, Help in organizations, Queueing theory.

JEL Codes: D23, M11.


Most production activities require the collaboration of individuals (Arrow, 1974; Simon, 1979).
For teams and organizations to function effectively, their members must communicate with each
other, for instance, to exchange technical information or to coordinate decisions (Hayek, 1945).
Yet such communication is often imperfect, as it requires time and effort and may be hampered

The editor in charge of this paper was Adam Szeidl.



nloaded from
/restud/article/88/2/574/5922651 by N

ational U
niversity Library user on 30 N

ber 2021

Copyedited by: ES MANUSCRIPT CATEGORY: Article

[17:07 8/3/2021 OP-REST200062.tex] RESTUD: The Review of Economic Studies Page: 575 574–609


by conflicting incentives (Garicano and Rayo, 2016). Therefore, a key decision in organizations
involves choosing how much co-workers communicate with each other. While this issue is at the
core of a large body of theoretical work, empirical evidence on communication in organizations
has largely lagged behind. This article empirically studies the determinants and consequences of
communication between co-workers in teams. We do this by taking advantage of an extremely rich
dataset and a unique natural experiment in a large and complex public sector organization. In our
setting, individuals working in teams are always able to communicate electronically. Some teams,
exogenously chosen by a computerized system allocating tasks to workers, can also communicate
in person. Therefore, our experiment identifies the effects of being able to communicate face-
to-face, in addition to electronically. Our study makes two contributions. Firstly, we provide
the first evidence on the positive causal link between the ability to communicate face-to-face
and team productivity. More importantly, we provide evidence in support of a fundamental
role of communication as a “help” (Itoh, 1991) or “information subsidy” (Hall and Deardorff,
2006) activity. Consistently with this mechanism, we show that communication improves the
productivity of its receiver while generating a time cost on the part of its sender. We use a simple
theoretical model to show that workers’ observed behaviour is consistent with them understanding
this trade-off and reacting to its costs and benefits in their communication decisions. For instance,
we find that face-to-face communication increases when its potential sender has weaker incentives
to maximize their own performance and can therefore afford to help their colleague. Similarly,
we show that communication is also higher when its receiver needs more help, for instance, as a
result of dealing with a more important problem from a social welfare perspective.

1.1. This study

We study the branch in charge of answering emergency calls and allocating officers to incidents
in the Greater Manchester Police (GMP). To understand the role that face-to-face communication
plays in our setting, we must briefly describe the production process. Each incoming call is
answered by a call handler, who gathers the details and describes the incident in the internal
computer system (see Figures 1 and 2). A radio operator then reads the description and allocates
a police officer on the basis of incident characteristics and officer availability. These two workers
have (partially) different objectives. Operators are responsible for minimizing the response time
of their incidents (i.e. the time until an officer reaches the incident location). Handlers’ main
objective is to minimize the time that incoming calls spend waiting in the call queue, and they
are therefore expected to be ready to take new calls as soon as they have finished a previous call.
Reading and understanding the incident’s description takes the operator some time, which slows
the speed of response. This processing time can however decrease if the operator can communicate
face-to-face with the handler who created the incident. Unfortunately the handler’s assistance is
not costless, as communication takes time, during which the handler cannot be ready to receive
new calls (i.e. must be in ‘not ready’ status). Understanding empirically how workers react to the
trade-off between the benefit of communication (i.e. lower operator response time) and the cost
(i.e. higher handler not ready time) is a core objective of the article.

To fix ideas, we develop a queueing theory model where communication is subject to
a trade-off: it helps its receiver but it is costly to its sender. Problems arrive and must be
sequentially processed by two workers. If a worker is busy dealing with a problem, incoming
problems accumulate in their queue. For each problem the first worker learns information that,
if communicated, will allow the second worker to process the problem faster. The cost of
communication is that it occupies the first worker’s attention, thereby preventing them from
dealing with new problems. We assume that the first worker chooses the amount of help provided,
giving positive weight to the queueing and processing delays at the two levels.


nloaded from
/restud/article/88/2/574/5922651 by N

ational U
niversity Library user on 30 N

ber 2021

Copyedited by: ES MANUSCRIPT CATEGORY: Article

[17:07 8/3/2021 OP-REST200062.tex] RESTUD: The Review of Economic Studies Page: 576 574–609


Figure 1

Operational communications branch.

Figure 2


The model generates several predictions regarding the comparison between a setting where
communication is possible and another where it is not. Firstly, the second worker processes
problems more quickly (and the first worker more slowly) when “communication as help” is
available. Secondly, these effects are larger when: (1) the second worker is busier on average,
as a result of being inherently slow or receiving a high inflow of problems; (2) the first worker
is less busy; and (3) the first worker assigns a higher weight to the objective of decreasing the
second worker’s delay. The model further predicts that an improvement in the efficiency of the
communication technology (i.e. in the benefit to the second worker per unit of communication
effort) decreases the second worker’s delay, without necessarily increasing the first worker’s.

We identify the effect of being able to communicate in person with the help of a natural
experiment. In the GMP, handlers and operators are spread across four rooms in separate buildings.
Each room contains the operators responsible for the surrounding neighbourhoods as well as a
subset of the handlers, who can take calls from anywhere in Manchester. Our empirical strategy


nloaded from
/restud/article/88/2/574/5922651 by N

ational U
niversity Library user on 30 N

ber 2021

Copyedited by: ES MANUSCRIPT CATEGORY: Article

[17:07 8/3/2021 OP-REST200062.tex] RESTUD: The Review of Economic Studies Page: 577 574–609


exploits the fact that the computerized queueing system matching incoming calls to newly
available handlers creates exogenous variation in the co-location of handler and operator. As
a result of this system, operators sometimes receive incidents created by handlers located in the
same room. For other (exogenously determined) incidents, the information will instead have been
entered by handlers based in another location.

When co-located, the workers can engage in face-to-face communication, which comprises of
several features. Firstly, it is two-way, allowing for the quick succession of questions and answers
between the interlocutors. Secondly, it is oral. Thirdly, it is visual and in person. Given our setting
we cannot disentangle the separate effect of these features, and it may be that an alternative
technology (e.g. telephone) including only some of these features may achieve a similar result.
Our main focus is studying when the workers in our setting will choose to use this additional
communication channel at their disposal.

1.2. Results

Our baseline results support the existence of a trade-off associated with face-to-face
communication: operator response time is 2% lower and handler not ready time is 2.5% higher
when the two workers are based in the same room. We confirm these findings by exploiting an
organization-wide relocation of workers and finding that the same pair of workers that used to
work in the same room cease to be associated with differential productivities when not co-located.
We also find that this reorganization, which permanently separated handlers and operators, had
the effect of increasing response time by 8% (around one minute, calculated at the median) for
incidents classified as violent crimes.

The heterogeneity of the baseline effects (i.e. lower response time and higher not ready time
under co-location) is broadly consistent with the comparative statics predictions of the theoretical
model. Specifically, the effects are larger when the operator is an intrinsically slow worker or is
suddenly very busy. Conversely, the effects are smaller when the handler is busy, proxied by the
number of recently received incidents. Lastly, the effects are larger when handler and operator
are the same gender, similar age and have a longer history of working together; characteristics
that we would expect generate higher altruism on the handler towards the operator.

To test the remaining prediction on the role of the efficiency of the communication technology,
we use the proximity within the room when handler and operator are co-located. Our expectation
here is that teammates with neighbouring desks require less walking time to engage face-to-face,
so each second spent by the handler should translate into more information transmission and
therefore a higher benefit to the operator. Consistently with this prediction, we find that response
time is faster even when the same pair of workers are located closer together inside the room.

We also find that the career incentives of handlers determine the help that they are willing to
provide to their colleagues. Communication is higher when a handler has just been upgraded in pay
and has weaker incentives to minimize their not ready time. Along the same lines, communication
decreases in the month of a handler’s performance review meeting, when we would expect
handlers to be more concerned about their own performance. We regard these results as particularly
interesting, as they illustrate how seemingly unrelated institutional features of the organization
can affect the amount of help between co-workers. An immediate implication is that handlers do
not fully internalise the welfare of the public ringing the police in their communication decisions.

While handlers may not be fully optimizing social welfare, we show that they are still partially
responsive to (proxies for) it. We find that types of incidents for which victim satisfaction depends
more on response time, are associated with more communication. Similarly, we find that co-
location leads to more face-to-face communication when the handler, upon picking up the phone,
learns that there is a crime ongoing.


nloaded from
/restud/article/88/2/574/5922651 by N

ational U
niversity Library user on 30 N

ber 2021

Copyedited by: ES MANUSCRIPT CATEGORY: Article

[17:07 8/3/2021 OP-REST200062.tex] RESTUD: The Review of Economic Studies Page: 578 574–609


Overall, the pattern of heterogeneity indicates that handlers respond to the (private and social)
costs and benefits of communication when determining how much face-to-face communication
to engage in.

1.3. Related work

As Dewatripont (2006) and Garicano and Prat (2013) argue, theoretical work in economics has
identified two main obstacles to the transmission of information in teams and organizations.
On the one hand, conflicting incentives and incomplete contracting can make communication
difficult or even impossible (Crawford and Sobel, 1982; Prendergast, 1993; Dessein, 2002;
Garicano and Santos, 2004; Alonso et al., 2008; Rantakari, 2008; Friebel and Raith, 2010).
Another body of work instead assumes away incentive problems and posits that communication
is directly costly, for instance because it uses the agents’ valuable time (Radner, 1993;
Bolton and Dewatripont, 1994; Van Zandt, 1999; Garicano, 2000; Dessein and Santos, 2006;
Dessein et al., 2016). While the details of the models in this second literature vary significantly, a
common feature is the trade-off between the benefit in terms of better or faster decision-making,
and the (time) cost incurred to communicate. The mechanism that we highlight in this paper
belongs to this second class of models and, to the best of our knowledge, ours is the first article
to provide empirical support for this broad trade-off. However, we also provide evidence on
the role of incentives, in showing that the career concerns of workers determine the amount of
communication they are willing to engage in.

Field evidence in economics on workplace communication is relatively scant, and often
does not identify effects on productivity (Gant et al., 2002; Palacios-Huerta and Prat, 2012;
Bloom et al., 2014). A recent exception is Menzel (2019), who implements an experiment to
encourage workers to share their knowledge and measures the resulting benefits. However, he
does not explore the cost–benefit trade-off, and workers’ reaction to it, which are the focus of this

Outside economics, the study of organizational communication reflects contributions from
psychology, sociology, and operations research (Jablin et al., 1987; Pace and Faules, 1994;
Harris, 2002). Broadly speaking, two separate paradigms dominate this literature. The earliest
is the “engineering-centric” view (Shannon and Weaver, 1949; Kmetz, 1998), which treats
organizations as information-processing systems, and internal communication as mechanical
information transmission between linked processors. Later, the “transactional” view (Barnlund,
2008) contributed the notion that communication occurs between humans, and therefore is affected
by variables such as its context, the medium used, and the relation between sender and receiver.
Our article combines aspects of these two views: we model workers as information-processing
nodes but empirically acknowledge that the efficiency of their communication may depend on
their incentives, their history together, or the match in their demographics.

The finding that face-to-face communication allows co-workers to help each other links the
paper with the wider literature on teamwork (Gaynor et al., 2004; Chan, 2016) and, more narrowly,
with field studies on employee cooperation (Itoh, 1991; Drago and Garvey, 1998; Hamilton et al.,
2003; Berger et al., 2011). This latter work is mostly concerned with the role of incentives and is
typically silent about the actual mechanism through which this cooperation occurs. We contribute

1. A related literature studies how the patterns of scientific collaboration depend on the ability to communicate
remotely (Agrawal and Goldfarb, 2008; Forman and Zeebroeck, 2012) or in person (Catalini, 2017; Catalini et al., 2018).
The closest paper here is Catalini (2017), who uses the relocation of departments in a French university to analyse how
search costs, monitoring costs, and the associated research productivity vary with physical proximity between academics.


nloaded from
/restud/article/88/2/574/5922651 by N

ational U
niversity Library user on 30 N

ber 2021

Copyedited by: ES MANUSCRIPT CATEGORY: Article

[17:07 8/3/2021 OP-REST200062.tex] RESTUD: The Review of Economic Studies Page: 579 574–609


to this work both by highlighting interpersonal communication as a leading mechanism and by
identifying determinants additional to the incentive structure.

1.4. Plan

The article is organized as follows. Section 2 describes the institutional setting. Section 3 presents
and solves the theoretical model. Section 4 outlines its empirical predictions as adapted to our
institutional setting. In Section 5, we present the data and the main empirical strategy. In Section 6,
we present and interpret the baseline results. In Section 7, we examine the heterogeneity of these
findings. Section 8 discusses implications for social welfare. Section 9 concludes.


2.1. Organization and production

The Operational Communications Branch (OCB) is the unit in charge of answering police 999
calls and allocating officers to the corresponding incidents. We focus on the team consisting of its
two primary workers: call handlers and radio operators. Figures 1 and 2 visualize the production

Incoming calls are allocated to handlers using a first-come-first-served system, matching the
call at the front of the queue with the next handler that becomes available. The handler questions
the caller, chooses the opening code (i.e. the “type” of incident) and the grade (a coarsely defined
degree of urgency), describes the incident in its log and ticks a box to officially create the incident.
This information is recorded by the software GMPICS. The handler then indicates their status as
“not ready” or instead “ready” to receive new calls. If “ready,” a new call can arrive at any point
and must be immediately answered by the handler.

When an incident is created, it immediately appears on the GMPICS screen of the operator
overseeing the subdivision where the incident occurred. The operator processes the information
in the log and allocates a police officer, who attends the incident scene. The allocation of incidents
to operators is deterministic, since at any point in time there is only a single operator in charge of a
specific subdivision. Therefore, handlers do not decide to which operator they assign an incident
(they can observe the operator’s ID number in GMPICS).

2.2. Face-to-face communication as “help” from handler to operator

Reading the log and gathering the information necessary to allocate an officer takes time. This time
can be decreased if the operator is able to interact face-to-face with the handler, for two reasons
(see Supplementary Appendix A for an extended discussion). Firstly, information is sometimes
unclear in the log. Operators have several channels through which they can clarify doubts or gather
additional details, but conversing with the handler is a fast and efficient way to fill information

Secondly, the logs sometimes contain too much information, rather than too little. For obvious
reasons, handlers typically record many more details than are needed for allocation purposes. This
implies that operators have to sift through the log and extract the specific elements guiding the

2. Alternative channels include conducting targeted searches on individuals or addresses in the GMP databases,
and contacting the caller directly. In addition, operators can electronically message handlers. The availability of electronic
messages enabling real-time Q&A implies that our study may be identifying features of face-to-face communication that
are absent in real-time electronic messaging systems, such as oral communication and the presence of visual cues.


nloaded from
/restud/article/88/2/574/5922651 by N

ational U
niversity Library user on 30 N

ber 2021

Copyedited by: ES MANUSCRIPT CATEGORY: Article

[17:07 8/3/2021 OP-REST200062.tex] RESTUD: The Review of Economic Studies Page: 580 574–609


Figure 3

Location and radio operations coverage of OCB rooms.

allocation of officers. This challenge is compounded by the fact that the information is often not
structured optimally (from the operator’s perspective). Operators often take less time asking the
handler to concisely provide the important details, than extracting these details themselves from
the log.

Assisting operators is not costless for handlers, as it implies that they must be ‘not ready’ to
answer incoming calls.3

2.3. The natural experiment

From November 2009 to January 2012, OCB staff were spread across four buildings or “rooms,”
in different parts of Manchester: Claytonbrook, Leigh, Tameside, and Trafford. Every room
accommodated the operators overseeing the surrounding subdivisions (Figure 3 displays the
areas overseen from each location). Handlers were also dispersed across (and uniquely linked to)
the four locations.

As discussed earlier: (1) incoming calls were deterministically matched with the operators in
charge of the subdivisions from which they originated and (2) the first-come-first-served queueing
system exogenously matched incoming calls to available handlers. This meant that, for exogenous
reasons, operators would sometimes be reading the descriptions of incidents created by same room
handlers, while on other occasions the handlers were based in a different part of Manchester. As
we argue in Section 5, this exogenous variation provides the foundation for the main empirical
strategy of the article.

3. A handler can set “not ready” anticipating that there may be questions or switch from “ready” to “not ready”
when approached by an operator. However, handlers cannot interrupt an ongoing call to discuss an earlier case with the


nloaded from
/restud/article/88/2/574/5922651 by N

ational U
niversity Library user on 30 N

ber 2021

Copyedited by: ES MANUSCRIPT CATEGORY: Article

[17:07 8/3/2021 OP-REST200062.tex] RESTUD: The Review of Economic Studies Page: 581 574–609


In January 2012, a major reorganization of the OCB reassigned all handlers to a single location
(Trafford), while radio operators were divided between Claytonbrook and Tameside. This put an
end to the natural experiment that we study here.

2.4. Workers and performance indicators

While there are no educational prerequisites to work at the OCB, a high school diploma is in
practice necessary to be a successful applicant. Once selected, workers undergo intensive training
programs. Salaries for handlers are slightly below the median Manchester salary and turnover is
quite low (around 5% annually).

The GMP has a small number of key performance indicators (Supplementary Appendix
Figure A1 displays them in a recent Annual Report). The most important are: (1) the allocation
time of incidents (the time between their creation by the handler and the allocation of an officer),
(2) the response time (the time between creation and the officer reaching the incident’s scene), and
(3) the call queuing time (the average time that incoming calls spend in the queue before being
answered). These measures are critical to the GMP (see Supplementary Appendix Figure A2),
for two main reasons. Firstly, nation-wide numerical targets were introduced by the UK Home
Office in 2008. For instance, the target for call queuing times was for 90% of calls to be answered
within ten seconds.4 Secondly, these measures are important determinants of public satisfaction.
For instance, UK-wide survey evidence suggests that response time is one of the most important
variables predicting citizens’ satisfaction with the police forces (Dodd and Simmons, 2002/03).5

Operators are held responsible for the allocation and response times of the incidents that they
personally deal with, while handlers are held responsible for the call queueing time. Handlers’
responsibility is a joint one, as they all take calls from a common queue. As is the case with other
public sector organizations (Burgess and Ratto, 2003), there is no performance pay providing
explicit incentives to handlers. However, handlers do have career incentives to contribute to the
reduction of call queueing time. Specifically, handlers can be moved to a higher pay grade (while
continuing to perform the same job) and they can be promoted to other jobs of higher status and
salary (specifically radio operator and handler supervisor). Career progression depends partly on
a handler’s supervisor evaluation, which takes place annually following their performance review
meeting. An important ingredient in this evaluation is a handler’s average “not ready” time, as it
is deemed that handlers being too often “not ready” are not contributing to the group objective
of reducing the call queueing time.

In Table 1, we provide suggestive evidence on the importance of “not ready” time for handlers’
career progression. Specifically, we regress a handler’s promotion or upgrade in a year on a set

4. For Grade 1 incidents, the targets were for a maximum of 2 and 15 minutes for allocation and response time,
respectively. The equivalent targets for Grade 2 (respectively Grade 3) were 20 and 60 minutes (respectively 120 and 240
minutes). While these targets were nominally scrapped in June 2010, police forces continued to regard them as objectives
and to believe that they were being informally evaluated on their basis (Curtis, 2015). Furthermore, information on
response times continued to be discussed in the reports produced by the HMIC (the central body that in the UK regulates
and monitors police forces). For an example, see HMIC (2012). They were also discussed in the reports by the GMP to
the Manchester City Council Citizenship and Inclusion Overview and Scrutiny Committee.

5. While important, these measures do not directly capture the “quality” of the GMP dealing with an incident.
For instance, they do not reflect whether the appropriate officer was allocated, or whether the attending officer was in
possession of all the relevant information prior to arrival. Measuring every dimension of quality is of course difficult.
Nevertheless, in Section 8, we replicate our baseline regressions using additional dependent variables such as whether
the incident escalated to becoming a crime, and, if so, whether the crime was cleared. A superior “quality” measure of
performance is whether the crime victim was satisfied with the police response. Unfortunately, we are unable to use this
measure as additional dependent variable in the main analysis of the article, as the number of survey responses is very
low and it mostly falls outside our baseline sample period.


nloaded from
/restud/article/88/2/574/5922651 by N

ational U
niversity Library user on 30 N

ber 2021

Place your order now for a similar assignment and have exceptional work written by one of our experts, guaranteeing you an A result.

Need an Essay Written?

This sample is available to anyone. If you want a unique paper order it from one of our professional writers.

Get help with your academic paper right away

Quality & Timely Delivery

Free Editing & Plagiarism Check

Security, Privacy & Confidentiality