Have any questions?+1 (929) 416 5389 Coupon:

Organizations, both for-profit and not-for-profit, have embraced the notion
of having amission.Mission statements are seen as a way to clarify ðand
perhaps advertiseÞ the role the organization sees itself fulfilling in society.
However, they may also be a way to attract workers whose enthusiasm for
the mission translates into higher job satisfaction and greater effort. The
Gallup Corporation polled workers in the United States and found that
We thank Sam Bowles, Nancy Folbre, John Maluccio, Peter Matthews, Jonathan
Meer, Caitlin Myers, and seminar participants at the World Bank, Middlebury
College, and the University of Massachusetts, Amherst, for their suggestions.
Contact the corresponding author, Erick Gong, at Information
concerning access to the data used in this article is available as supplementary
material online.
[ Journal of Labor Economics, 2016, vol. 34, no. 1, pt. 1]
© 2015 by The University of Chicago. All rights reserved. 0734-306X/2016/3401-0004$10.00
Submitted January 13, 2014; Accepted July 3, 2014; Electronically published October 27, 2015
employees who felt their job was important due to their employer’s mission
were more engaged and productive ðEllingwood 2001; Harter et al. 2006Þ.
While this result comes as no surprise to many, given the underlying selection
process, it is actually quite hard to identify the pure effect of a “mission
match.” Controlling for sorting, exactly how much more productive are
workers who believe in the organization’s mission?
Although well-matched workers may be a boon for any organization,
during tight labor markets, it is unlikely that all workers sort perfectly into
organizations based on their mission preferences. Workers without alternatives
may join organizations with little regard to the mission. In addition,
an organization’s mission may change over time, leading to a misalignment
with workers’ preferences.1 As the Gallup poll also found, workers who are
unmotivated by the organization’s mission are more than four times less
likely to believe that they are paid fairly. In the case of these mismatches, can
increases in traditional financial incentives make up for the motivational
deficit? If pay is linked to performance, will even “mismatched” employees
work hard?
Neither of these questions has been completely addressed in the empirical
literature, but the idea thatworker preferences for an organization’smission
can lead to sorting, screening, and higher effort has been addressed conceptually.
Francois ð2000Þ was an early theoretical contribution that showed
that “public service motivation” could lead workers to provide more effort,
even in organizations with low-powered financial incentives. Akerlof and
Kranton ð2005Þ focus on the motivational importance of “identity” and
conclude that employers who can align worker identities with the mission
goals of the organization can benefit from hard-working employees, again
even with “low-powered” incentives that do not link pay to performance.
The contribution that directly motivates our study is Besley and Ghatak
ð2005Þ. Here, focusing on “mission-oriented” workers, the authors show
that when workers are matched to organizations with which they share a
common mission, financial and mission preferences are substitutes. While
matchedworkers will work hard for the mission often despite little financial
incentive to do so, to elicit effort from mismatched workers ðor those who
are just uninspired by the missionÞ, organizations must rely on “highpowered”
incentives ðe.g., piece rates or commissionsÞ. While the existing
literature has focused, almost exclusively, on howmission preferences can be
used to screen and sort workers ðsee Banuri and Keefer 2012; Dur and
Zoutenbier 2014Þ, we focus on estimating the motivational effect ofmission
1 For example, the priorities and goals of government agencies change as new
administrations are voted in, and this can easily affect the motivation of tenured
bureaucrats. See also Greg Smith’s editorial in the New York Times ðSmith 2012Þ
discussing Goldman Sachs’s increasing focus on profits over clients’ interest as his
reason for leaving the firm.
212 Carpenter/Gong
match on effort, an effect that does not have to be filtered through the process
of employment selection.2
We conduct a real-effort lab experiment to identify the effects of missions
on worker productivity and to examine the extent to which missions
and financial incentives are substitutes when motivating workers. We
conduct a large survey of college students to measure individual mission
preferences. From this pool, we recruit workers to participate in a realeffort
lab experiment. In the experiment, workers are randomly assigned
to one of two organizations; each organization has a well-known mission,
and the mission of each organization is diametrically opposed to that of
the other.3 Workers, based on their assignment to each organization, are
classified as mission matched ðmismatchedÞ—when the worker’s mission
preference aligns ðmisalignsÞ with the organization. All workers perform
a task that generates an amount of output, our measure of worker productivity.
In addition, we randomly assign compensation, where workers are
given either a fixed wage or a fixedwage plus high-powered incentives ðpiece
Our first two results are that missions and incentives have strong effects
on worker productivity; both results are consistent with contemporary
personnel economics. Workers who are mission matched produce 72%
more output than mismatched workers. This result is driven by changes
in both the extensive and intensive margins; mission matching leads to a
33 percentage point increase in choosing to work and a 19% increase in
output ðconditional on workingÞ. While it is not surprising that missions
matter, the magnitude of the effect is large; onematchedworker is equivalent
to about 1.7 mismatched workers. We also find that workers offered incentives
increased their output by 35% on average.
Our third result is perhaps the most interesting. We find that incentives
have differential effects depending on whether a worker is well matched or
not. For those who are matched, piece rates increased productivity by a
modest 13%,while for thosewho aremismatched, piece rates induce an 86%
increase in output. In fact, these high-powered incentives can make up more
than two-thirds of the productivity lost due to a poor match. These results
suggest that high-powered incentives can substitute ðalbeit imperfectlyÞ for
mission matching between workers and employers.
2 The sorting process involves workers deciding which organizations to apply to
for employment, choosing whether to work or not if an offer of employment is
made, and if a worker chooses not to work, deciding whether to seek employment
at another organization. In our study, workers are limited to choosing whether to
work or not for an organization that is randomly selected for them.
3 Examples of this are found in gun control ðNational Rifle Association vs.
Educational Fund to Stop Gun ViolenceÞ, abortion ðAmericans United for Life
vs. NARAL Pro-Choice America FoundationÞ, and marriage ðFreedom to Marry
vs. Institute for Marriage and Public PolicyÞ.
Motivating Agents 213
Our study contributes to the nascent literature estimating the implications
of organizations having clear missions. In a chosen-effort lab environment,
Fehrler and Kosfeld ð2014Þ find an increase in effort among
mission-motivated workers only once firms can screen workers by offering
less financially attractive contracts. If screening is not possible, participants
still chose higher efforts when doing so ultimately benefits a charity,
but the increment is not significant. This finding resembles that of Tonin
and Vlassopoulos ð2010, 2015Þ, who conduct real-effort experiments online
and find small mission effort premia, which are driven to a large extent by
female participants. In another chosen-effort experiment, Gerhards ð2012Þ
finds that mission-motivated agents work harder but that principals fail
to realize that they can reduce their monetary incentives once they have
attracted mission-motivated workers. A key difference between these studies
and ours is that our study utilizes a real-effort task that is directly tied
to the success of the organization. Instead of asking participants to donate
a portion of their earnings to a charity like previous experiments, our design
is distinctive because participants were asked to work directly for the mission.
Our workers produce tangible output that unambiguously helps
advance the mission of the organization.4
Our results have implications for human resource policies at many organizations,
those with clear missions and those that seek to clarify their
missions.5 First, we determine that although matches help average productivity
considerably, it is mismatches that may have the largest implications
for the performance of an organization. Using our measure of a
worker’s intensity of preferences, we generate a group that is relatively unmotivated
by missions and find that matching increases productivity 23%
compared to this “no-match” group, while mismatched workers produced
43% less than the no-matches. As a result, the productivity gap between
matched and mismatched workers is large, and the lesson is clear—while
agents sorting on the mission can help an organization substantially, not
screening to prevent mismatches might hurt productivity even more. Second,
for those organizations that might have a hard time screening workers
based on their mission preferences, there is some good news, at least for
4 Besley and Ghatak ð2005, 617Þ note that “donating our income earned in the
market to an organization that pursues a mission that we care about is likely to be an
imperfect substitute for joining and working in it,” a hypothesis recently confirmed
in Brown,Meer, and Williams ð2013Þ.
5 Missions are not limited to organizations that provide public goods. Many forprofit
organizations also have mission statements. Whether the principals of these
organizations derive any nonpecuniary benefit from these missions is unknown.
Some for-profit organizations may choose missions to minimize labor costs, while
others may have founders who do sincerely believe in the mission. Regardless of
the underlying motivation of why organizations choose their mission, it is still the
case that the workers may be motivated to increase effort if they agree with the
214 Carpenter/Gong
those organizations that have relatively deep pockets. It does appear that
workers respond to high-powered financial incentives, especially those who
are mismatched. Third, we find no evidence that incentives crowd out
intrinsic motivation due to mission matching ða` la Bowles and Polania-
Reyes ½2012Þ; while incentives have a more muted effect for those who are
matched, there is still some incremental increase in productivity when
matched workers receive incentives.
We proceed in Section II by presenting a conceptual framework that
makes explicit our hypotheses about how mission match, the intensity of
one’s mission preferences, and financial incentives should affect participation
and effort. This simplemodel also dictated our study design and the data
that we collected. In Section III, we discuss our study design, including our
design choices, themethodofmissionpreference elicitation thatweused, and
our real-effort experiment. In Section IV,we present our results and provide
our empirical analysis.We offer a few concluding remarks in Section V.
II. Conceptual Framework
We develop a simple framework to examine the effects that both mission
matching and pecuniary incentives have on a worker’s productivity.Weuse
the standard principal-agent paradigm, where the principal offers a wage
contract and the agent ðworkerÞ decides how much effort, e, to exert. A
formal exposition can be found in appendix A.1; appendix A is available
online. Below we summarize the key intuition and predictions.
The agent receives both a fixed wage ðwÞ and a performance incentive
ðpÞ, so assuming effort and output are synonymous, the agent’s standard
linear contract entitles her tow1pe.As is also standard, agent effort is both
costly ðCðeÞ; Ce > 0Þ and increasingly so ðCee > 0Þ.
Motivated by Besley and Ghatak ð2005Þ, agents with “mission preferences”
receive utility increments from two sources if they work for the
right principal ði.e., one engaged in the agent’s preferredmissionÞ. First, just
exerting effort for one’s preferredmission increases utilityMðeÞ andMe > 0,
but the process exhibits diminishing returns Mee < 0. Second, we allow for
differences in the intensity of agent mission preferences, denoted g. In this
case, agents who believe more strongly in the mission supported by the
principal will also receive more utility ðMðgÞ; Mg > 0Þ, but this effect diminishes
too ði.e., Mgg < 0Þ. We assume that our framework subsumes the
standard principal-agent model when g 5 0. That is, Mðe, gÞ 5 0 if g 5 0.
Finally, we assume some interaction between these two forces: diminishing
returns to effort set in later for those agents with more intense mission
preferences ði.e., Meg > 0Þ.
Because not all principal-agent dyads will be characterized by a common
mission, we use v ∈ {21,1} to separate matches, v 5 1, when an
agent’s mission preferences align with the principal, from mismatches, and
Motivating Agents 215
v521, when an agent has mission preferences that are at odds with the
principal’s. In other words, while our matched workers enjoy providing
effort to somedegree, our mismatches findwork evenmore onerous because
they are working for the wrong cause.We see matching andmismatching as
proximate causes for changes in motivation. The ultimate drivers of these
changes could work through various channels, including moral conviction
ðSkitka,Baumanm, andSagis 2005Þ, intrinsicmotivation for the output of the
work ðas discussed in Besley and Ghatak ½2005Þ, warm glow ðAndreoni
1989Þ, and cognitive consistency/dissonance theory ðFestinger 1962Þ.6
Putting this all together, our agent’s mission utility is summarized as
vMðe, gÞ. If helpful, one can think of vMðe, gÞ as either attenuating the cost
of effort when the agent is well matched with the principal’s mission or
adding to it when the agent is mismatched. The agent, therefore, chooses
an effort level e* that maximizes the utility function
UðeÞ 5 ðw 1 peÞ 1 vMðe; gÞ 2 CðeÞ; ð1Þ
and the first-order condition, p 1 vMe 5 Ce, highlights that both the piece
rate ð pÞ and mission matching ðvÞ will determine an agent’s level of effort.
Our agents also have outside options that generate reservation utilities,
U. We assume a continuous distribution of reservation utilities and that
principals do not know an agent’s reservation utility before making a wage
offer. Since our focus is on the agent, we allow principals to make a single
contract offer, and the agent will work if Uðe*Þ >U, that is, when the familiar
participation constraint is satisfied.
We first examine the effects of mission matching. The impact on effort
is clear from ð1Þ. For any offered contract ðw, pÞ, an agent exerts greater
effort when matched compared to when mismatched e* v51 > e* v521. Because
matching results in greater utility U e* v51
ð Þ>U e* v521
ð Þ, it must also be the
case that matches are more likely to satisfy the participation constraint.
Matching should thus affect both the intensivemargin ðincreasing e*Þ and the
extensive margin ðincreasing the probability that Uðe*Þ >UÞ.
Turning to incentives, in our formulation, financial incentives will lead
to increases in an agent’s effort ðde*=dp > 0Þ, both for matches and mismatches.
7 Since, ceteris paribus, incentives increase utility, they must also
increase the probability that the contract offer satisfies the participation
6 Workers may also be motivated by the task itself ðDelgaauw and Dur 2008Þ;
however, this is unlikely in our case. See Sec. III for details of the task.
7 We see our mission utility as potentially a component of intrinsic motivation.
As a result, financial incentives could theoretically crowd out this utility and result
in lower effort levels with a different set-up ðBenabou and Tirole 2003Þ. However,
this is not the case in our model, and, despite the mounting evidence of crowding
recently reviewed by Bowles and Polania-Reyes ð2012Þ, this is not what we find in
the data.
216 Carpenter/Gong
As with matching, incentives will lead to increases on both the intensive
and extensive margins. What is notable is that our framework predicts that
incentiveswill have a stronger effect on both the effort of mismatched agents
ði.e., de*ðv521Þ=dp > de*ðv 51Þ=dpÞ and on whether the participation
constraint is satisfied ðsee appendix A.1 online for the detailsÞ. Part of the
intuition for the differential effect of incentive pay is that, given that the
utility function is additively separable, we can think of agents as having a
common cost of effort function but different benefits from working, ones
that depend on thematch. Because they accrue more benefits fromworking,
matched workers find themselves further along the cost curve than mismatched
workers and therefore must consider larger marginal costs when
deciding whether to contribute a bit more effort as the piece rate increases.
Finally, we expect that the intensity of mission preferences will affect effort
levels. For matches, greater intensity leads to greater effort de*ðv 5 1Þ=
dg > 0, while the opposite is the case for mismatches de*ðv521Þ=dg < 0.
The intuition is simple: those who really care about a mission will work
harder when matched, while those who care but are mismatched will
strenuously resist working. As before, we expect the intensity of preferences
to affect the extensive margin too; greater intensity will lead to an
increased likelihood of working for matches and a decreased likelihood
for mismatches.
To summarize, we expect both matching and incentives to increase an
agent’s effort level and the probability that the agent engages in work at
all. While these predicted effects are mostly straightforward, our formulation
also yields a few more subtle insights. In particular, high-powered
financial incentives should have stronger effects on mismatched agents,
and matching should have greater effects on agents with strong preferences.
III. Study Design
A. Overview
Our study was designed with three primary goals in mind. First, given
that the focus of the previous work in this area has been on the effect of
mission matches ðGerhards 2012; Fehrler and Kosfeld 2014Þ, we thought
it would be useful to create a situation closer to our conceptual framework
in which we could assess the impact of both matches and mismatches.
Although some workers accrue benefits from being in a job with which
they match well, others may suffer through a similar job because they do
not believe in the organization’s mission. We are interested in how much
effort is enhanced or dispirited depending on the quality of the match.
Second, we wanted to assess the potential impact of financial incentives on
the efforts of matched and mismatched workers—the obvious question
here is whether incentive pay could overcome any blunting of motivation
Motivating Agents 217
caused by a poor match. Third, we decided to focus on the pure causal
effects of mission match and financial incentives on effort, not the effects
that might have to be filtered through the ðpotentially interestingÞ
process of job or worker selection. Together, these three criteria lead to
the random assignment ofworkers to a simple two-by-two factorial design:
mission match versus mismatch crossed with weak versus strong financial
We decided to use the real-effort paradigm because it would generate
conservative estimates of any treatment effects. Considering the tradeoffs,
Van Dijk, Sonnemans, and Winden ð2001Þ were among the earliest to
notice that treatment effects discovered in chosen-effort experiments
tended to be considerably smaller in real-effort experiments, consistent
with the idea that intrinsically motivated participants work hard, often
regardless of the incentives. Their results suggest that if one uses similar
chosen-effort experiments as a baseline, team members shirk much less
than expected and tournaments extract less additional effort compared to
piece rates than one might have expected. Similarly, Gneezy and List ð2006Þ
find muted effects of reciprocity in a real-effort gift exchange setting, and
Hennig-Schmidt, Sadrieh, and Rockenbach ð2010Þ find no effect on effort
of substantial wage increases in a data entry field study.
Similarly, we thought we could improve the external validity of our
results by utilizing naturally occurring preferences for a mission and by
making the real-effort task similar to what many workers in our chosen
sector actually do. Not only is it more natural, and perhaps less invasive, to
measure the home-grown preferences for a particular mission than it is to
try to create them, it might also be difficult to reliably construct mission
preferences. Eckel and Grossman ð2005Þ, for example, illustrate how hard it
is to prime teammates with a common mission. As for the task itself, we
settled on clerical work for an organization, the type of work often associated
with entry-level positions. In this way our task is more natural and
representative than other related experiments in which participants earn
money that they can then donate to an organization. Further, the task yields
an unambiguousmeasure of output,which, given the simplicity of the work,
we take as a proxy for effort.
All of our design choices culminated in the following experiment,
which piggybacked on the 2012 presidential election in the United States.
Our study design is illustrated in figure 1. Right after the students returned
to campus in early September, we sent out a web survey to record
their political preferences and the intensity of those preferences. We used
the preference data to type our potential participants. Approximately
2 weeks later ðso as to minimize any links between the survey and our
experimentÞ, we began bringing students to the lab to do a letter-stuffing
experiment. In a first round, all participants stuffed fund-raising letters
into envelopes for the college to assess their ability. In a second round,
218 Carpenter/Gong
participants were randomly assigned to stuff campaign letters for one of
the two major party candidates ðObama or RomneyÞ. The letters were
addressed to actual independent voters in Ohio ðan important swing state
in the electionÞ. Because the student preferences skewed Democratic at the
college, to assure balance, prior to their arrival in the lab, we used a unique
but anonymous identifier ðpostal box numbers on campusÞ to stratify our
participants and assure equal numbers of matches and mismatches for the
two preference groups. Overlaying the matching, we also randomly assigned
financial incentives. After working for a candidate in one of the
four cells of our design, the experiment ended for the participants with
another brief survey and payments. We now expand on each aspect of the
B. Agent Mission Preferences
In our study, the political preferences of respondents represent the
agent’s mission preferences. To measure political preferences, a baseline
survey was conducted where respondents were asked for their voter registration
status ðif registered and party affiliationÞ, which major party
candidate ðObama vs. RomneyÞ they would vote for, and whether they
had a preference for who would win the election. In addition, using the
same question that is on the American National Election Studies survey
ðANES 2012Þ, we asked respondents where they would place themselves
on the following 7-point Likert scale:8
Democrat Democrat
Democrat Independent
Republican Republican
ð1Þ ð2Þ ð3Þ ð4Þ ð5Þ ð6Þ ð7Þ
FIG. 1.—The study design
8 In appendix A.3 online, we present the distribution of political preferences for
the entire sample of survey respondents.
Motivating Agents 219
Based on the response to this question, we classified participants as Democrats
if they describe themselves as a strong Democrat, a Democrat, or
someone who leans Democrat, and we classified participants as Republicans
if they describe themselves as a strong Republican, a Republican, or
someone who leans Republican. To test whether these classifications represent
actual policy choices, we examine the correlation between being
classified as a Democrat and agreement with a range of policy statements
developed by Pew Research ðPEW 2012Þ. We find that our party classification
is strongly correlated with the role that government should take on
issues such as taxation, environmental regulation, social safety nets, immigration,
health care, abortion, and labor unions ðsee online appendix
table A.2Þ.
As suggested in Section II, we hypothesize that in addition to having
clear campaign preferences, participants will vary in the intensity of these
preferences. For example, although two individuals might both report being
a Democrat, one might be more passionate about the party’s mission
than the other. To measure the intensity of our respondent’s political preferences,
we ran a version of the dictator game similar to that used in Carpenter
and Myers ð2010Þ. Respondents were given an endowment of $100
that they could split with the campaign of a major party candidate. They
were told that once the survey was finished we would randomly pick five
people and enact their choices. Respondents first picked between “Obama,”
“Romney,” and “I do notwant to allocate any money and will keep the $100
for myself.” Conditional on picking either candidate, they then input the
amount that should be sent to the candidate’s campaign and, implicitly, the
amount they wished to keep. It was understood that the amounts kept
would be sent anonymously to the campus mailboxes of the randomly
selected respondents. We use the amounts sent to either candidate as a
measure of the intensity of mission preferences.9
C. Principal Missions
In previous studies, the organization or principal had a mission that
resulted in the direct provision of a collective good. In our study, the
organization ðpolitical partyÞ will campaign to elect public officials that
will be charged with not only providing collective goods but also deciding
whether they should be provided in the first place. For instance, Democrats
support the provision of health insurance by the government, while
the Republicans believe that the government should have a minimal role
in providing health insurance coverage ðGrassley 2009Þ. There are other
stark differences between the two parties, which broadly generalize on the
level of taxation and expenditures. Several studies have demonstrated relationships
between party control of government and taxes, health care
9 The results of the political dictator game are summarized in online appendixA.3.
220 Carpenter/Gong
spending, family assistance, worker’s compensation, and overall expenditures
ðGrogan 1994; Besley and Case 2003; Reed 2006Þ.

Click here to request for this assignment help

"Looking for a Similar Assignment? Get Expert Help at an Amazing Discount!"