Investing For Impact Article

Investing for Impact
Bhagwan Chowdhry
University of California, Los Angeles
Shaun William Davies
Downloaded from https://academic.oup.com/rfs/article-abstract/32/3/864/5039908 by Adam Ellsworth, Adam Ellsworth on 07 February 2019

University of Colorado, Boulder
Brian Waters
University of Colorado, Boulder
We study joint financing between profit-motivated and socially motivated (impact) investors
and derive conditions under which impact investments improve social outcomes. When
project owners cannot commit to social objectives, impact investors hold financial claims to
counterbalance owners’ tendencies to overemphasize profits. Impact investors’ ownership
stakes are higher when the value of social output is higher, and pure nonprofit status may
be optimal for the highest valued social projects. We provide guidance about the design
of contingent social contracts, such as social impact bonds and social impact guarantees.
(JEL M14, O35, D86)
Received May 23, 2016; editorial decision April 28, 2018 by Editor Francesca Cornelli.
Authors have furnished an Internet Appendix, which is available on the Oxford University
Press Web site next to the link to the final published paper online.
Organizations in the for-profit sector increasingly rely on joint financing by

profit-motivated and socially motivated agents. For example, some publicly
traded companies are held by both traditional and socially responsible mutual
funds. Within these organizations, profit-motivated and socially motivated
agents have differing missions; some investors value profit maximization above
all else while others have a social goal in mind. To the extent that both types of
investors lay claim to the same organizations, we wonder how is joint financing
congruent with both missions, and given that the missions may be at odds,
how might contracting be used to align incentives among otherwise diverse
investors? These questions are the subject of our paper.
We would like to thank Maitreesh Ghatak, Robert Gertner, Jesse Davis, Archishman Chakraborty, Ivo Welch,
Edward Van Wesep, Robert Dam, Kyle Matoba, and Francesca Cornelli; two anonymous referees; and seminar
participants at CU-Boulder Leeds School of Business, UCLA Anderson School of Management, the HKUST
Conference on the Impact of Responsible and Sustainable Investing, the Emerging Markets Finance Conference,
the Geneva Summit on Sustainable Finance, The University of Chicago Economics of Social Sector Organizations
Workshop, and the Midwest Finance Association annual meeting for their helpful insights and suggestions.
Supplementary data can be found on The Review of Financial Studies Web site. Send correspondence to Bhagwan
Chowdhry, Anderson School of Management, University of California, Los Angeles, 110 Westwood Plaza Suite
C-411, Los Angeles, CA 90095; telephone: (310)-825-5883. E-mail:bhagwan@anderson.ucla.edu.
© The Author(s) 2018. Published by Oxford University Press on behalf of The Society for Financial Studies.
All rights reserved. For permissions, please e-mail: journals.permissions@oup.com.
doi:10.1093/rfs/hhy068 Advance Access publication July 4, 2018
[16:29 4/2/2019 RFS-OP-REVF180069.tex] Page: 864 864–904

We model a project that produces both a monetary payoff and a social benefit
and consider settings in which there is a trade-off regarding which output to
emphasize. While the for-profit project owner is agnostic toward the project’s
social goals, a social impact investor, who values (and is willing to pay for)
the project’s expected social output, also contributes to the project’s cost of
investment. In doing so, the impact investor provides the for-profit owner a
subsidy for directing scarce resources and attention to the production of the

social benefit. Joint financing between the for-profit owner and the impact
investor is therefore mutually beneficial: the for-profit owner receives the
subsidy, effectively monetizing the impact investor’s social preferences, and
the impact investor enjoys increased social output. We show that when the for-
profit owner cannot commit to social objectives, the impact investor must hold a
large enough financial claim in order to incentivize the for-profit owner to pursue
the project’s social goals. Our model therefore provides a theory of impact
investing and shows that socially minded investors should hold large ownership
stakes in socially valuable firms. We further characterize stakeholders’ financial
claims and describe conditions under which they own equity, social impact
bonds, or new securities we propose called social impact guarantees.
The mechanism in our model is closely related to work examining not-for-
profit status as a means to overcome limited commitment. Specifically, Glaeser
and Shleifer (2001) consider a multitask principal-agent problem (Holmstrom
and Milgrom 1991) in which an entrepreneur cannot commit to produce
high-quality goods because doing so is at odds with profit maximization.
The commitment problem is remedied by organizing the firm as a pure
nonprofit as opposed to a pure for-profit firm. Pure nonprofit status weakens
the entrepreneur’s ex post incentive to maximize profits and allows him to
commit ex ante to higher product quality. Instead, in our framework, we derive
an intermediate option between the pure nonprofit and pure for-profit firm;
the for-profit owner receives some cash flow, but a fraction of the cash flow
is pledged to the impact investor in order to weaken the owner’s incentive to
maximize profits. We show that, when this intermediate option is available, pure
nonprofit status is not optimal. Our analysis therefore reconciles the recent trend
among individuals and organizations in coupling their investment activities and
their philanthropic giving.
In the model, the underlying project is characterized by two production
technologies and one unobservable scarce resource, for example, attention.
One production technology produces a monetary payoff and the other produces
a nonrival social benefit. While the two outputs are positively related at the
extensive margin (undertaking the project positively influences the production
of both), there is a trade-off regarding which output to emphasize at the intensive
margin.1 The project is owned by a profit-motivated investor and is funded as
1 For example, the underlying project may be an initiative to improve employee health at a firm. Employee health
is improved through more expansive health insurance and greater emphasis on preventive care and healthy living.
865

The Review of Financial Studies / v 32 n 3 2019
well by a social impact investor. Because funds available for social good are
relatively scarce and socially minded investors face additional constraints in
their portfolio selection problem, the impact investor has a higher opportunity
cost of using capital than does the for-profit owner. In addition, the for-profit
owner employs a manager who allocates unobservable scarce attention between
the two production technologies. While the for-profit owner is agnostic toward
social output, the social impact investor and manager are, to some extent,

socially motivated (like in Besley and Ghatak 2005).
Joint investment allows the for-profit owner to internalize (via a subsidy)
the impact investor’s value of social output. However, since the opportunity
cost of using the impact investor’s scarce capital exceeds that of the for-profit
owner, future payments to the impact investor are inefficient. When the for-profit
owner can commit to pursue social goals, joint investment is accomplished via
donations. Specifically, the owner receives an upfront payment from the impact
investor, which is equal to his expected value of the project’s social benefit. As
a result, the for-profit owner internalizes the impact of greater social output
via a larger upfront donation, and in response, the for-profit owner reduces the
manager’s monetary incentive wage in order to elicit greater attention to the
project’s social goals.
If the for-profit owner is unable to commit to the project’s social objective,
donations are ineffective. While the impact investor is willing to provide a larger
subsidy if he expects greater attention to social goals, he knows that the for-profit
owner has an incentive to renege on any promised increase in social attention
after receiving the upfront donation. To overcome this commitment problem,
the for-profit owner sells a portion of the firm’s cash flow to the impact investor.
When the for-profit owner retains a smaller cash flow claim, his profit motives
are softened, allowing him to commit to greater social attention ex ante and
thus allowing him to secure a larger subsidy from the impact investor. However,
selling financial claims to the impact investor is costly, since the opportunity
cost of using the impact investor’s scarce capital is high. As such, the for-profit
owner chooses joint financing only if the benefits from committing to higher
social output outweigh the cost of selling a cash flow claim to the impact
investor. This is true only when the firm’s potential social value is sufficiently
large.
Our model therefore provides an explanation for exclusionary investment
among impact investors. Only when a firm’s potential social value is high should
social investors purchase financial claims in order to increase attention to social
goals. As such, our results justify the recent growth in impact funds which
The two outputs from the project are (1) a monetary payoff (cost savings) to the firm via lower turnover and a
reduction in sick leave and (2) a social benefit that stems from improvements in long-term health outcomes. The
firm may stress greater sick patient care in order to reduce employee leave, but do little to affect overall health
outcomes or life expectancy. Conversely, the firm could provide nutrition counseling and subsidized gym access
to reduce the likelihood of diabetes and heart disease but do little to reduce employee sick days due to common
illnesses.
866

purchase equity in socially focused firms. Impact funds, which have an explicit
mandate to invest in sustainable and responsible opportunities, increased their
investments in public equity from 276 million dollars in 2013 to nearly 1.5
billion dollars in 2015 (GII 2016). We show, as well, that impact investors
overpay for financial claims (to subsidize for-profit owners) and, thus, earn
lower expected financial returns. This result is consistent with recent studies
that document that social impact investors pay a premium for holding socially

conscious portfolios.2
We consider two extensions of our base model. First, while pure nonprofit
status is never optimal in the base model, we show that nonprofit status (i.e.,
when the profit-motivated project owner does not retain a cash flow claim) may
be optimal in the presence of corporate taxes. If the impact investor places a
large value on the social benefit and corporate taxes are a concern, it is optimal
for the project to be wholly claimed by the impact investor, as opposed to paying
cash out as taxable profit. The analysis, in combination with our base model
results, provides a relation between optimal firm organization and the impact
investor’s value of social output: (1) if the impact investor places little value
on the social benefit, the firm should be organized as a purely commercial firm
with the impact investor holding no financial claim, (2) if the impact investor
places an intermediate value on the social benefit, the firm should exhibit joint
financing via impact investing, that is, the impact investor holds a nonzero
financial claim, and (3) if the impact investor places a large value on the social
benefit, the firm should be organized as a nonprofit with the impact investor
being the lone residual claimant.
Finally, we consider contracts contingent on realized social output and
show that incentive alignment is best achieved when the most intrinsically
profit-oriented agent (the for-profit owner in our model) holds a pay-for-
social-success contract that provides a larger payment when social goals
are achieved. Our model is therefore consistent with the design of social
impact bonds (an example of a pay-for-social-success contract), which are
commonly used when profit-motivated investors partner with local governments
to fund public sector projects. For example, in 2014, the Commonwealth of
Massachusetts announced one of the largest social impact bonds in history.
The contract raised $12 million from private for-profit investors to provide
life skills and employment training to young, at-risk men in the Boston area
with the objectives of reducing recidivism and generating taxpayer savings.
Contingent on the project leading to a successful reduction in reconviction rates,
the contract requires up to $16 million in success payments to investors. To this
end, our model suggests that incentive alignment between for-profit investors
and socially motivated governments is best achieved when the investor is paid
more as a result of social success.
2 See Geczy, Stambaugh, and Levin (2005), Hong and Kacperczyk (2009), Chava (2014), and Hartzmark and
Sussman (2017).
867

We show, as well, that the social impact investor should hold the opposite
contract; a guarantee-against-social-failure contract which guarantees a large
financial payment when the project’s social value is small. We refer to this
security as a social impact guarantee. While financial contracting on social
outcomes in the for-profit sector is rare to date, our model shows that for-
profit owners in socially valuable firms should issue social impact guarantees to
impact investors so that the owners’ own residual claims provide the necessary

incentives to direct attention to social outcomes (similar to the pay-for-social-
success feature of social impact bonds). The emergence of hybrid corporate
structures, such as benefit corporations, which explicitly target social impact
and corporate profits, are a first step in this direction. However, our analysis
demonstrates that the potential to fine-tune the pursuit of the double bottom
line is much larger with appropriately designed financial claims.
Our paper contributes to two bodies of literature. First, we add to a growing
literature examining the private provision of public goods. Besley and Ghatak
(2007) show that the private sector is often the more natural provider of public
goods when social output is inherently bundled with the production of a private
good. This is precisely the setting we have in mind for our model; while there
is a trade-off regarding which output to emphasize, the project has the potential
to produce both a monetary payoff and a social benefit.3 Furthermore, the
trade-off between the monetary payoff and the social benefit gives rise to an
agency conflict regarding which output to emphasize. Consequently, our impact
investing model is directly related to the multitask principal-agent problem of
Holmstrom and Milgrom (1991), in which the two dimensions of output in
our framework are the monetary payoff and a social benefit. Several papers
have examined organizational choice as a means to mitigate multitask agency
conflicts when profit and social purpose are at odds. Glaeser and Shleifer (2001)
prescribe pure nonprofit status as a credible means to dull an agent’s profit
motivation and overcome issues with limited commitment.4 As previously
discussed, the mechanism in our model provides an intermediate solution
between pure nonprofit and pure for-profit status.
In a similar spirit to our intermediate solution, Besley and Ghatak (2017)
suggest a role for “social enterprises” (such as benefit corporations and certified
B corporations) which serve as hybrid organizations between for-profit and
nonprofit organizations.5 Social enterprises elicit more unobservable effort and
provide better alignment with social objectives. In our analysis, we rely on
financial contracting as a means to achieve a hybrid solution, and we show that
this financial contracting aligns incentives among heterogeneously motivated
3 See Baron (2001), Bagnoli and Watts (2003), and Kotchen (2006) for settings in which profitability and social
good production are complements.
4 See also Hansmann (1996).
5 See also Tirole (2001) for an analysis of whether managerial incentives and control structures should be used to
promote stakeholder rights.
868

investors. Hart and Zingales (Forthcoming) prescribe corporate governance and

shareholder activism as a means to balance profitability against social harm.6
While shareholder activism can lead profit-oriented firms to consider social
value, we show that passive investment strategies also lead to greater social
impact if some investors condition stock prices on social preferences.
Second, our paper contributes to an emerging literature which studies the
effects of sustainable, responsible, and impact (SRI) investments on firm

outcomes. Heinkel, Kraus, and Zechner (2001) argue that, by lowering the
cost of capital for socially responsible firms, SRI investment strategies lead to
an increase in the number of social projects that may be undertaken. In our
model, this effect is captured by the subsidy provided by the impact investor,
reflecting his private value of the project’s expected social good. In line with
this feature of our model, there is suggestive evidence that social investors
pay a premium for holding SRI-compliant portfolios: Hong and Kacperczyk
(2009) examine the financial cost to investors from avoiding stocks that are not
SRI acceptable, that is, “sin stocks.” Sin stocks have lower price-to-book ratios
and higher risk-adjusted returns than comparable stocks. Chava (2014) finds
similar evidence in firms’ implied costs of capital. Firms that are excluded
by environmental screens (e.g., firms with substantial carbon emissions) are
characterized by higher expected returns on firm equity.7 Moreover, we show
that SRI investments lead to an improvement in social outcomes by increasing
attention to social goals even amongst firms owned by profit-motivated owners.
1. The Model
We consider a project which may produce both a monetary payoff and a social
benefit — a “social venture.” The project requires an upfront investment equal
to I > 0 supplied by an investor(s) and a single unit of labor supplied by a
manager. If the project is undertaken, the manager allocates an unobservable
scarce resource between two production technologies, one which aids in the
production of the social benefit and another that aids in the production of the
monetary payoff. For ease of exposition, we refer to the unobservable resource
as attention, and we denote by ab ≥ 0 the fraction of attention allocated to
the social technology and by ax ≥ 0 the fraction of attention allocated to the
traditional for-profit technology (where ab +ax = 1).8
The project may either “succeed” or “fail” in the production of the monetary
payoff. When the project succeeds, it generates a payoff x = X > 0, whereas
6 See also Baron (2008) for a setting in which corporate governance is used to further social goals.
7 Geczy, Stambaugh, and Levin (2005) provide similar evidence; investors that adhere to socially responsible
investments add an additional constraint to their portfolio selection problem and underperform relative to an
unconstrained portfolio by as much as 30 basis points per month.
8 The scarce resource also could be interpreted in the context of managerial effort or the distribution of resources
within firms’ internal capital markets.
869

when the project fails, its payoff is x = 0. The probability of a successful outcome
depends on the fraction of attention allocated to the for-profit technology and
is given by
Pr(x = X| ax ) = f (ax ), (1)
where f (·) is a twice continuously differentiable function with 1 ≥ f (ax ) ≥ 0,

f (ax ) > 0, and f (ax ) < 0 for all ax ∈ [0,1]. In addition, the project may succeed

(b = 1) or fail (b = 0) in its production of the social benefit b. The probability of
a successful social outcome is given by
P r (b = 1| ab ) = g(ab ), (2)
where g(·) is, as well, a twice continuously differentiable function with

1 ≥ g(ab ) ≥ 0, g (ab ) > 0, and g (ab ) < 0 for all ab ∈ [0,1].9 The distinguishing
feature between x and b is that b is nonrival and nonexcludable. For simplicity
of notation, we drop the subscript b on ab and denote ax as 1−a hereafter. To
ensure that there is an interior solution for social attention, we assume,
Assumption 1.
lim f (1−a) = ∞ and lim g (a) = ∞.

a→1 a→0
There are three agents: a profit-motivated investor (denoted by p), a social

impact investor (denoted by s), and a project manager (denoted by m). We
assume that the profit-motivated investor owns the investment project and refer
to him as the for-profit owner.10 All agents are risk neutral and have additively
separable utility for cash and social output. If the project is successful in the
production of social output, agent i ∈ {p,s,m} receives nonpecuniary payoff
ψi . While the for-profit owner places no value on a successful social outcome
(ψp = 0), the impact investor and manager receive a strictly positive payoff when
b = 1 (ψs > 0 and ψm > 0).11
The agents differ, as well, in their discount rates across time. Specifically,
the for-profit owner discounts utility at gross rate ρp = 1, while the impact
investor discounts utility at gross rate ρs = 1+δ for δ > 0. The difference in
discount rates may reflect (a) limited capital among socially minded investors,
(b) additional constraints in the impact investor’s portfolio choice problem,
and/or, (c) unmodeled efficiencies coming from for-profit capital providers,
9 Our setup therefore assumes that conditional on the chosen levels of attention, the success or failure of the social
technology is independent of the success or failure of the for-profit technology.
10 In our Online Appendix, we consider the alternative case in which the impact investor owns the project.
11 Equivalently, the for-profit owner is unwilling to pay for social output, whereas the impact investor and manager
are willing to pay ψs and ψm for a successful social outcome.
870

such as through improved monitoring.12 While profit-oriented agents benefit

only from investments with sufficient financial returns, social impact investors
benefit from the combination of financial and social returns. Given this setup,
the for-profit owner and impact investor maximize their discounted payoffs,
xi +ψi b
ki + , (3)
ρi

where ki is the net-payment made to/from agent i at date 0 prior to undertaking
the project, xi is agent i’s monetary payoff at date 1 which depends on the
success or failure of the project’s for-profit technology, and ψi is agent i’s
nonpecuniary payoff from a successful social outcome.
While the for-profit owner and impact investor have sufficient wealth and do
not face a binding wealth constraint, we assume that the manager has no wealth
and cannot contribute to the project’s upfront investment cost. Furthermore,
we restrict the manager from receiving an upfront transfer as part of her
compensation.13 As such, the manager’s payoff can be summarized simply
by her date 1 utility,
xm +ψm b. (4)
The manager’s reservation payoff is γ ≥ 0, and we restrict attention to the range

of γ for which the project generates positive surplus.14
Attention is unobservable and is therefore noncontractible. While the
monetary payoff x is observable and contractible, we assume in our base model
that social output b is noncontractible (we study contractible social output in
Section 3). At date 0, the for-profit owner simultaneously makes take-it-or-
leave-it offers to the manager, a wage contract, and to the impact investor, a
cash flow claim and upfront transfer ks .
We assume standard limited liability constraints – no agent can receive a
negative date 1 cash flow and the total payment across all three agents cannot
exceed the project’s monetary payoff.15 These assumptions imply that the wage
and cash flow sharing rule can be summarized by the vector Ω = (w,y,z), in
which w ≥ 0 is the manager’s equity share, y ≥ 0 is the impact investor’s equity
12 Regarding (a), despite recent growth in sustainable, responsible, and impact investments, the pool of dollars
directed toward social projects is still limited in comparison to the amount of funds commercially available
for investments that generate market returns (GII 2017). Regarding (b), in subsequent analysis, we show that
the impact investor must maintain a concentrated stake in a limited number of social ventures in order to
align incentives and increase social investment. As such, impact investors may be underdiversified relative to
profit-motivated investors.
13 This assumption is consistent with a contactual incompleteness. If the manager can walk away from the project
after accepting the upfront transfer, such transfers cannot be optimal.
14 We provide the explicit conditions on γ in Assumption 2 in the appendix.
15 Our contracting environment is consistent with settings featuring moral hazard and limited liability. See Innes
(1990).
871

share, and z ≥ 0 is the equity share retained by the for-profit owner (and in
which w +y +z ≤ 1).16
If the manager accepts the wage contract, she chooses a ∈ [0,1] to maximize
her expected payoff:

a ∈ arg max E xw +ψm b|a . (5)
a ∈[0,1]

Lemma 1. There exists a one-to-one continuous and differentiable function
mapping any wage contract w to the unique solution to the manager’s problem
in (5),
a(w).
The function a(w) is strictly decreasing in w for a ∈ (0,1).
For a given wage contract w, the manager allocates her attention across the
for-profit and social technologies in order to maximize her expected payoff.
When the wage contract is higher, the manager benefits relatively more from
the production of the monetary payoff and therefore shifts attention away from
social production in favor of greater attention toward financial returns.17
Since the function a(w) is differentiable and strictly decreasing, the inverse
function w(a) = a−1 (w), which maps social attention to the manager’s wage
contract, exists and is differentiable and strictly decreasing.
Lemma 2. The inverse function,
g (a)ψm
w(a) = a−1 (w) = , (6)
f (1−a)X
is a one-to-one continuous and differentiable function. The function w(a) is
strictly decreasing in a.
Following Holmstrom (1984), the first-order approach is valid for the manager’s
attention choice problem. The inverse function which maps social attention to
wages therefore follows from the manager’s first-order condition for optimal
attention. While the stated assumptions for the production functions g(·) and
f (·) guarantee a unique solution to the manager’s attention allocation problem
for a given w, we assume as well that the for-profit owner’s expected payoff is
concave in attention, and thus that the second-order condition for optimality is
satisfied.
16 Since x = 0 when the project fails to produce the monetary payoff, the manager’s wage and cash flow sharing rule
can be equivalently represented as debt contracts.
17 For example, what we have in mind is increasing managerial wages to mitigate the agency cost associated with
corporate social responsibility (CSR). This agency cost is due to firm managers shifting resources away from
profitable uses and toward nonpecuniary CSR activities that provide social benefit. See Cheng, Hong, and Shue
(2016) and Krüger (2015) for empirical studies examining CSR as an agency conflict.
872


Figure 1
Model timing
Contracts are offered at date 0. After agreeing to the contract, capital is

invested, and the manager chooses her attention allocation. Outputs are realized
at date 1. Each agent receives his or her claim to the project’s monetary payoff
and privately enjoys the project’s produced social good. Figure 1 summarizes
the model’s timing.
1.1 Preliminary analysis

We begin by outlining a benchmark value for the level of social attention a.
We consider the equilibrium attention allocation in a commercial firm; that is,
when the impact investor does not participate in the project. When the impact
investor does not participate in the project and the project is therefore funded
solely by investors who do not value social output, social output can be viewed
as a private benefit for the manager. As such, this benchmark reduces to the
typical principal-agent problem (like in Ross (1973) or Jensen and Meckling
(1976)) with unobservable effort. Since the for-profit owner cannot contract on
attention explicitly, he chooses a wage contract which incentivizes the manager
to shift attention away from social output and into the for-profit production
technology. Formally, the for-profit owner solves the following optimization in
order to maximize his expected profit,
maxf (1−a)Xz−I (7)

{w,a }
s.t. w = w(a), (7.1)
0 ≤ f (1−a)Xw +g(a)ψm −γ , (7.2)

0 ≤ w +z ≤ 1, and w ≥ 0, z ≥ 0, (7.3)
where (7.1) is the manager’s incentive compatibility constraint, (7.2) is the

manager’s participation constraint, and where (7.3) are the limited liability
constraints. Let {wo ,ao } be the optimal contract which solves the commercial
benchmark without the participation of the impact investor, and let γo be the
value of the manager’s reservation payoff, such that for γ ≥ γo , the manager’s
873

participation constraint (7.2) binds. Then, for γ < γo , the optimal level of social
attention ao is defined implicitly by,18

dw
−f (1−ao )X(1−w(ao ))−f (1−ao )X = 0, (8)
da a=ao
whereas for γ ≥ γo , the optimal level of social attention ao is pinned down by

the manager’s participation constraint,
f (1−ao )Xw(ao )+g(ao )ψm −γ = 0. (9)
In either case, the optimal wage contract must satisfy the manager’s incentive
compatibility constraint (7.1) and is therefore given by wo = w(ao ). Moreover,
ao is the lower bound on a; any optimal contract featuring the involvement of
the impact investor achieves an attention allocation at least as large as ao .
In our subsequent analysis of impact investing, we compare equilibrium
levels of social attention to the commercial benchmark ao . If equilibrium
attention exceeds the commercial benchmark, impact investing leads to greater
emphasis of social goals. Otherwise, impact investing is ineffective. While
the first-best level of social attention, which maximizes joint surplus among
the for-profit owner, the project manager, and the social impact investor, is
straightforward,

ψm ψb
−f (1−aF B )X +g (aF B ) + = 0, (10)
ρm ρb
our model does not provide clear predictions comparing equilibrium levels of
social attention to the first-best allocation aF B . To see this, consider the simplest
case in which the manager’s participation constraint is slack. Since attention is
not observable, the owner must provide the manager a cash flow claim in order
to reduce attention to social goals. Since paying the manager a higher wage
is costly for the for-profit owner, the owner prefers a level of social attention
above the first-best allocation in order to lower the manager’s wage, all else
equal. On the other hand, a high level of social attention is also costly for the
for-profit owner, since it requires the impact investor to hold a larger financial
claim (see Section 2). In this case, the owner prefers a level of social attention
below the first-best allocation, all else equal. Ultimately, it is unclear which
effect dominates.
Since the relationship between equilibrium levels of social attention and
the first-best allocation are unclear, we consider, as well, a setting in which
the for-profit owner can commit not to alter the manager’s equity wage after
18 To focus on projects for which there is a tension between the production of the monetary payoff and social benefit,
we assume that lim f (1−a) dw da is finite valued. The assumption ensures that the commercial benchmark level
a→1
of social attention is strictly less than 1.
874

the transfer ks is sunk. In the next section, commitment is limited, and the
optimal contract must be robust to renegotiation between the for-profit owner
and manager. While outside the scope of the current model, one can envision
the full-commitment setting as one in which the for-profit owner engages
in repeated interactions with social investors and must therefore maintain a
reputation for achieving sufficient social value. For the purposes of this paper,
the full-commitment model highlights the basic contracting frictions among the

for-profit owner, manager, and impact investor without the added complexity of
limited commitment. In this case, the for-profit owner chooses a contract triplet
Ω = (w,y,z), a corresponding level of social attention a, and a security price
ks in order to maximize his expected profit. Given the contract y and security
price ks , the impact investor’s discounted expected payoff is
f (1−a)Xy +g(a)ψs
−ks + ,
1+δ
and the optimal security price ks sets the impact investor’s expected payoff
equal to zero,
f (1−a)Xy g(a)ψs
ks = + . (11)
1+δ 1+δ
The first term in the security’s price is the discounted expected value of the
impact investor’s equity claim y, whereas the second term captures the impact
investor’s utility from the social benefit. By paying for his social value upfront,
the impact investor provides a subsidy to the for-profit owner which depends
on the project’s expected social output.
When the impact investor contributes to the project’s upfront investment cost,
the for-profit owner’s optimal contracting problem is therefore given by,
f (1−a)Xy g(a)ψs
maxf (1−a)Xz+ + −I (12)
{Ω,a} 1+δ 1+δ
s.t. w = w(a), (12.1)
0 ≤ f (1−a)Xw +g(a)ψm −γ , (12.2)
0 ≤ w +y +z ≤ 1, and Ω ≥ {0,0,0}, (12.3)
where we have substituted for the security’s price ks from (11) into the for-
profit owner’s expected profit in (12). The following lemma characterizes the
profit-maximizing contract when the for-profit owner can credibly promise not
to renegotiate the manager’s wage contract (where we use the subscript v to
denote his “vow” to not renegotiate the contract).
Lemma 3. There exists a unique contract {(wv ,yv ,zv ),av } which maximizes
the for-profit owner’s expected payoff. The contract is characterized by wv =
w(av ), yv = 0, and zv = 1−wv .
875


Figure 2
Model timing with renegotiation
1. For γ < γo , the level of social attention exceeds that in the commercial
benchmark, av > ao .
2. For γ ≥ γo , social attention equals the commercial benchmark, av = ao .
Lemma 3 shows that, when the for-profit owner can commit not to alter the
manager’s wage contract (perhaps in order to maintain his reputation as ethical
among potential future social investors), the profit-maximizing contract sets
the social impact investor’s equity claim to zero. Since the impact investor
discounts future payments at a higher rate than the for-profit owner, any
monetary payment promised to the impact investor at date 1 is inefficient.
Hence, the for-profit owner chooses y to be as small as possible, setting yv = 0.
While the impact investor holds none of the project’s equity, he still pays for
his expected value of social good through the upfront transfer ks = g(a)ψ
1+δ
s
, and it
is through this subsidy that the impact investor convinces the for-profit owner
to implement a higher level of social attention. We therefore interpret the
optimal contract in this setting with full commitment as a grant or a donation
provided by the impact investor to the for-profit owner. We now transition to
the more natural setting with limited commitment and show that grants and
donations are ineffective when commitment is limited. Instead, the impact
investor must hold a nonzero cash flow claim in order to affect the level of
attention directed toward social goals. Thus, Section 2 provides a theory of
impact investments in for-profit firms.
2. The Basic Model of Impact Investing

We now consider the base model with limited commitment. In particular, we
allow the for-profit owner to renegotiate the manager’s wage contract and
therefore influence the level of social attention after receiving the payment ks
from the impact investor. As a result, the impact investor correctly anticipates
the possibility the contract will be renegotiated and adjusts the payment ks
according to the true, rather than promised, level of social attention. Figure 2
summarizes the model’s timing with renegotiation.
Formally, the contract {(w,y,z),a} will be renegotiated by the for-profit
owner and manager if there exists an alternative contract {(w ,y,z ),a } which
876

satisfies limited liability (w +z ≤ 1−y) and which makes both the for-profit
owner and manager better off (with at least one strictly better off):

E xz a = a w ≥ E xz a = a(w) ,

E xw +ψm b a = a w ≥ E xw +ψm b a = a(w) .
We refer to a contract which cannot be renegotiated by the for-profit owner and

manager as “renegotiation-proof.” A direct consequence of our definition for
renegotiation is that cash burning, w +y +z < 1, cannot satisfy renegotiation-
proofness. Moving forward, we restrict attention to the set of full-repayment
contracts with w +y +z = 1. We now provide two lemmas which are useful for
understanding the impact of bilateral renegotiation within social ventures.
Lemma 4. Renegotiation is only possible by increasing the manager’s wage.
To see this, observe that renegotiation increases the manager’s expected payoff
only if the new wage w is larger than the initial wage w. Under the initial
contract, the manager chooses a = a(w) to maximize her expected payoff:
f (1−a(w))Xw +g(a(w))ψm .
An increase in the wage contract therefore leads to an increase in the manager’s
payoff,
da
f (1−a(w))X + −f (1−a(w))Xw +g (a(w))ψm > 0,
dw
Equals zero
by envelope condition
while a decrease in the wage contract reduces the manager’s payoff. As such,
renegotiation is feasible only if the new contract increases the manager’s wage
and subsequently reduces the level of social attention.
Lemma 5. There is a unique threshold y(a) given by,

f (1−a) dw
y(a) = 1−w(a)+ , (13)
f (1−a) da
such that the contract {(w(a),y,z),a} is renegotiation-proof if and only if
y ≥ y(a).
1. The threshold y(a) is increasing in a.

2. When social attention is at the benchmark level a = ao , then y(ao ) = 0.
Lemma 5 illustrates a crucial point. When the desired level of social

attention exceeds the benchmark level under commercial ownership, impact
investors must hold sufficiently large equity stakes in order for contracts to be
877

renegotiation-proof.19 In this case, increasing y and subsequently lowering z

reduces the for-profit owner’s preference for the monetary payoff after receiving
the initial transfer ks . By softening the incentives for renegotiation, equity sales
allow the for-profit owner to commit to a greater level of social attention ex
ante. The greater the desired level of ex ante social attention, the greater the
motivation to renegotiate the initial contract, and the larger must be the impact
investor’s equity stake.

With limited commitment, the for-profit owner’s optimal contracting
problem is given by
f (1−a)Xy g(a)ψs
maxf (1−a)Xz+ + −I (14)
{Ω,a} 1+δ 1+δ
s.t. w = w(a), (14.1)
0 ≤ f (1−a)Xw +g(a)ψm −γ , (14.2)
1 = w +y +z, and Ω ≥ (0,0,0), (14.3)
y ≥ y(a), (14.4)
where we have again substituted for the upfront transfer ks from (11) into the
for-profit owner’s expected profit in (14) and where constraint (14.4) ensures
that the impact investor’s equity claim is large enough so that the contract is
renegotiation-proof.
Proposition 1. There exists a unique optimal contract {(wr ,yr ,zr ),ar }
which maximizes the for-profit owner’s expected payoff and which is
renegation-proof. The contract is characterized by wr = w(ar ), yr = y(ar ), and
zr = 1−wr −yr .
1. For ψs > ψs (and γ < γo ), the social impact investor holds a nonzero
equity claim, yr > 0, and social attention exceeds the benchmark level,
ar > ao . The unique threshold ψs is given by

dy
δf (1−ao )X da
a=ao
ψs = . (15)
g (ao )
2. Otherwise, yr = 0 and ar = ao .
19 Although we cannot distinguish between debt and equity in our model (since x = 0 when the project fails to
produce the monetary payoff), we conjecture that, in a more general model with outcomes XH > XL > 0, an
equity claim sold to the impact investor more strongly dulls the for-profit owner’s incentive to focus on monetary
payoffs than does an equally sized debt claim. As such, our model is consistent with the rapid growth in socially
targeted equity investments in recent years (GIIN 2016).
878


Figure 3
y regions
For ψs > ψs and γ < γo , yr > 0, and outside this region, yr = 0.
The first part of the proposition shows that the impact investor’s equity stake
is set just high enough to satisfy renegotiation-proofness: yr = y(ar ). Increasing
the impact investor’s claim y is costly for the for-profit owner, since the impact
investor has a higher cost of capital. To maximize his expected profit, the
for-profit owner therefore sets y as small as possible without violating
constraint (14.4).
In choosing the optimal level of social attention, the manager therefore faces
a trade-off between increasing the upfront subsidy from the impact investor
(higher y and higher a) and limiting the inefficiency that arises from the impact
investor’s equity stake (lower y and lower a). When the impact investor places
a sufficiently high value on a successful social outcome, that is, ψs > ψs , the
benefit from increasing the subsidy is large enough to offset the cost of a nonzero
equity claim. As such, the for-profit owner chooses ar > ao and yr > 0. However,
when the impact investor’s value for social output is limited, the effect on the
subsidy from an increase in a is small relative to the loss from selling equity to
the impact investor. Thus, for ψs ≤ ψs , the for-profit owner chooses yr = 0 and
settles for the commercial benchmark level of social attention ar = ao .
For the impact investor to hold a nonzero equity stake, not only must the
impact investor have a large enough value of social output ψs > ψs but also the
manager’s reservation payoff must be small enough γ < γo . Figure 3 illustrates
this point. When the manager’s reservation payoff is large, the manager requires
a large wage contract in order to participate in the social venture. As such, there
is no flexibility to reduce the manager’s wage and increase the level of attention
allocated to social goals. Importantly, the manager’s value for social output ψm
is positively related to the threshold γo . That is, all else equal, a larger value
of ψm increases the manager’s expected value of employment and raises the
879

reservation value at which the manager’s participation constraint just binds.

This suggests that impact investments are more likely when the manager’s
value of social output is not too small.20
Corollary 1. If the manager’s participation constraint is slack and ψs > ψs ,
dyr dyr

>0 and < 0.
dψs dδ
Proposition 1 shows that impact investors hold cash flow claims only in
those firms with significant social value. In this sense, our model provides a
theory of directed or exclusionary investment among socially minded investors.
While exclusionary investment campaigns are sometimes framed in the context
of protesting or screening certain companies with low social value, our model
suggests exclusionary investment also may arise as impact investors concentrate
equity positions only in those firms for which joint funding optimally improves
social outcomes. In addition, Corollary 1 shows that impact investors hold
larger equity stakes in more socially valuable firms. Since the size of the upfront
subsidy depends on the impact investor’s expected social value, an increase in
ψs raises the marginal benefit of an increase in social attention to the for-profit
owner. In response, the for-profit owner prefers a higher level of social output
which requires a larger equity sale in order to make the contract renegotiation-
proof. As such, the model predicts that impact investors hold a larger fraction
of firm equity in more socially valuable firms and that social output be higher
and firm profitability lower in such firms (since a is higher).
In addition, the impact investor’s equity stake is decreasing in the impact
investor’s added cost of capital δ. Recent studies suggest that the pool of socially
conscious capital has increased dramatically in recent years (Barber, Morse,
and Yasuda (2017)) and that the millennial generation is more pre-disposed to
investing in firms that focus on social impact. As the opportunity cost of socially
conscious capital declines, the model suggest impact investors will hold larger
ownership stakes in for-profit companies in the years to come.
We now complement the earlier analysis with a numerical example (in Figure
4) in order to highlight the relationship between the size of the monetary payoff,
X, and the threshold values ψs and γo . Panel 4(a) shows that ψs is increasing in
X. When firm profitability is high, it is optimal for the impact investor to hold
a nonzero equity claim and increase social attention above ao only for those
firms with immense social value. The analysis therefore suggests that we are
less likely to obverse impact investments in firms with very high profitability,
as measured for example using return-on-assets or return-on-equity. However,
20 In our Online Appendix, we further analyze the match between managers and projects along the dimension of
ψm .
880

(a) (b)

Figure 4
Comparative statics of ψs and γo with respect to X
The regions for which ar > ao are shaded.
conditional on investment by impact investors, firms with very high profitability

must also have very high social value in order to justify the impact investment.
While panel 4(a) suggests that impact investing is limited among high
profitability firms, Panel 4(b) shows that impact investing is also limited among
low profitability firms (since γo is increasing in X). When firm profitability is
low, managers require a large equity stake in order to satisfy their participation
constraint. If the manager’s wage is large, the ability for impact investing to
affect attention to social goals is restricted. Taken together, panels 4(a) and 4(b)
suggest that impact investing is optimal only for those projects for which the
monetary payoff and social benefit are truly coupled. If the monetary payoff is
too large relative to the project’s social value, impact investing is suboptimal
(since selling claims to impact investors is prohibitively costly). Likewise, if
the monetary payoff is too small relative to the project’s social value, impact
investing will not increase social returns (since social attention is depressed by
the manager’s wage).
While we focus on the optimal design of cash flow rights between profit-
oriented and socially conscious investors in impact investing, a similar increase
in attention to social goals could be achieved by allocating control rights to the
impact investor. For example, suppose that the owner’s decision to renegotiate
the manager’s wage contract and reduce the promised level of social attention
is observable but not contractible. Consider a contract which grants the impact
investor the right to purchase the firm at a pre-specified price (i.e., a call option)
prior to the realization of the monetary and social payoffs. If the owner chooses
to reduce attention to social goals in order to raise expected profits, the social
investor’s purchase option is in the money, and the social investor purchases
the firm from the for-profit owner. While exercising the option is (ex ante)
suboptimal, the option need not be exercised in practice, since the threat alone
881

is sufficient to allow the owner to commit to a higher (i.e., the full-commitment

benchmark) level of social attention.
2.1 Comparing impact investments to nonprofit status

Section 2 shows that impact investments increase attention to social outcomes
in social ventures by softening the for-profit owner’s desire to renegotiate the
manager’s wage contract. The mechanism in this paper is therefore closely

related to that in Glaeser and Shleifer (2001), who show that, by reducing the
value of firm ownership, nonprofit status allows entrepreneurs to commit to
actions that run counter to ex post profit maximization.
Nonprofits, by definition, are barred from distributing any earned profits
to persons who exercise control over the firm (see Hansmann 1996). In the
context of our model, we therefore define a nonprofit firm as a firm in which
the for-profit owner retains none of the project’s residual cash flow.
Definition 1. In a nonprofit firm, the for-profit owner’s equity stake is set to
zero: z = 0.
While Glaesar and Shleifer (2001) model a nonprofit as a firm in which profits
are distributed through the enjoyment of “perquisites” that are less valuable than
cash, an alternative interpretation is that firm profits are held by someone who
both has no control over the firm and values the cash flows less than the for-profit
owner, that is, the impact investor in our model.21
Corollary 2. Nonprofit status is never optimal.
From Proposition 1, the impact investor’s equity stake is given by

r ) dw
yr = 1−w(ar )+ ff(1−a
(1−ar ) da a=ar
which is the smallest financial claim needed
to dissuade the for-profit owner from reneging on the promised level of
social attention ar . Since dwda
< 0 for all a ∈ (0,1), the impact investor’s equity
stake is strictly less than the entire residual cash flow claim, and the for-
r ) dw
profit owner therefore retains a positive equity stake, z = − ff(1−a
(1−ar ) da a=ar
> 0.
While Glaesar and Shleifer (2001) show that entrepreneurs prefer nonprofit
status relative to for-profit status when the value of noncontractible quality is
sufficiently large, our model shows that pure nonprofit status is never optimal
if the for-profit owner can sell only a fraction of the firm’s cash flows to the
impact investor (rather than choosing simply between pure for-profit status and
pure nonprofit status).
To understand this result, recall that in order to commit to the desired level of
social attention, the for-profit owner must not find it desirable to renegotiate the
manager’s wage after receiving the upfront payment from the impact investor.
That is, the change in the for-profit owner’s expected date 1 payoff from a
21 What we have in mind is that the impact investor plows-back produced cash into additional social opportunities
rather than realizing a distribution of profit.
882

decrease in the level of social attention (and a corresponding increase in the

manager’s wage) cannot be positive:

dw
f (1−a)X(1−w(a)−y) +f (1−a)X ≤ 0. (16)
da

+ −
The first term in (16) captures the benefit of shifting attention away from social

production and toward the production of the monetary payoff which increases
the probability that the project successfully produces the payoff X of which a
fraction 1−w(a)−y is paid to the for-profit owner. The second term captures
the cost of increasing the manager’s wage. An increase is needed to achieve the
desired shift in attention from the social technology to the for-profit technology.
When the for-profit owner retains only a small fraction of the firm’s cash
flows, the benefit from reducing social attention and increasing the probability
of a successful monetary outcome is small. In the limit as the for-profit owner’s
cash flow claim goes to zero (like in a pure nonprofit), the benefit of decreasing
the level of social attention goes to zero, while the cost of increasing the
manager’s wage is still strictly positive. As a result, the for-profit owner can
always commit to the same level of social attention by selling only a fraction of
the residual cash flow to the impact investor as he could under a pure nonprofit
firm. Since selling a greater portion of the firm to the impact investor is costly,
pure nonprofit status is therefore never optimal.
In practice, however, nonprofit firms not only exist but they are also abundant.
A key assumption in the model is that the project is capable of generating both a
social benefit and a monetary payoff. If the project instead operates at a financial
loss, involving the for-profit owner is no longer valuable, and it is optimal for the
firm to be wholly owned by the impact investor. Furthermore nonprofit status
may be optimal even among firms that generate positive operating profits if
there exist substantial frictions between for-profit and nonprofit firms, such as
differential tax treatment. In the following subsection, we show formally that
nonprofit status may be optimal if corporate taxes are a concern.
2.2 Corporate taxes

While pure nonprofit status cannot be optimal in our base model, the gain
from impact investing relative to nonprofit status is smallest when the impact
investor’s equity stake is large. This is the case when the desired level of social
attention is largest. Thus, if there are additional benefits to nonprofit status,
such as tax advantages, nonprofit firms may be optimal when the desired level
of social attention is sufficiently high.
Suppose the corporate tax rate for for-profit firms is given by τ and that
nonprofit firms are untaxed. With some abuse of attention, let {(wr ,yr ,zr ),ar }
denote, as above, the unique optimal contract which maximizes the for-
profit owner’s expected after-tax payoff and which is renegation-proof. It is
straightforward to show that corporate taxes affect only the optimal level of
883

social attention ar and not the manager’s wage or impact investor’s equity
claim given ar . When the for-profit owner chooses to operate as a for-profit
firm and sells a portion of the firm’s cash flows to the impact investor, the for-
profit owner’s expected payoff is equal to the value of his expected after-tax
equity claim plus the value of the upfront payment from the impact investor
(which reflects the tax paid on the impact investor’s equity claim) minus the
cost of investment,

f (1−ar )X 1−w(ar )−y(ar ) (1−τ )
(17)
f (1−ar )Xy(ar )(1−τ )+g(ar )ψs
+ −I.
1+δ
In contrast, if the for-profit owner chooses nonprofit status, the entire after-
wage cash flow, X(1−w(ar )), is assigned to the impact investor, and these
profits are not subject to corporate taxation. For simplicity, assume that the
firm implements the same level of social attention under nonprofit status as
is optimal with impact investing. Then, the for-profit owner’s expected payoff
when choosing nonprofit status is given by22
f (1−ar )X(1−w(ar ))+g(ar )ψs
−I. (18)
1+δ
A direct comparison of the expressions in (17) and (18) shows that nonprofit
status delivers the for-profit owner a higher expected payoff if,

τ
y(ar ) ≥ (1−w(ar )) 1− . (19)
δ(1−τ )
Thus for y(ar ) large enough, nonprofit status is optimal, since the tax saving
from nonprofit status outweighs the benefit of retaining a larger cash flow claim
as a for-profit firm.
Furthermore from Corollary 1, y(ar ) is increasing in the impact investor’s
value of social output ψs . As a result, there exists a threshold value ψs such
that for ψs ≥ ψs , the impact investor’s required financial claim is large enough
that nonprofit status is optimal.23 Figure 5 summarizes this point. When the
project’s potential social value is small ψs ≤ ψs , the project’s residual cash flow
is owned entirely by the for-profit owner (as shown in Proposition 1). When the
value of social output is intermediate, ψs < ψs < ψs , the project continues to be
operated as a for-profit firm, but the project’s residual cash flow is jointly owned
by the for-profit owner and impact investor. Over this range, the for-profit owner
22 In actuality, the firm owner would implement a different level of social attention when operating as a nonprofit
firm, which would further increase the value of nonprofit status. Thus, the condition we establish in (19) for
nonprofit status to be optimal is merely sufficient (but not necessary).

23 This directly follows from (1−w(a )) 1− τ
r δ(1−τ ) < 1 and lim ψs →∞ y(ar (ψs )) = 1.
884

(a) (b)

Figure 5
A comparison of impact investing to nonprofit status
The first panel depicts the difference in the for-profit owner’s profit under impact investing (Pr ) and under
nonprofit status (Pnp ) as a function of ψs . The second panel depicts the impact investor’s cash flow claims under
impact investing (yr ) and under nonprofit status (ynp ) as a function of ψs .
wishes to commit to a higher level of social attention than is possible under

commercial ownership, and he therefore sells a portion of the project’s residual
cash flow to the impact investor. Finally, for ψs ≥ ψs , the desired level of social
attention and the impact investor’s required financial claim are large enough
that the project is operated as a nonprofit, and the impact investor receives all
of the project’s residual cash flow.
3. Contractible Social Output

Our base model results in Section 2 are largely positive; for example, the
model shows that impact investors purchase financial claims to incentivize
owners to pursue social goals, which is consistent with the common SRI
practice of targeted financial investments in socially valuable firms. In this
section, we provide normative recommendations for impact investing when a
project’s social output is both measurable and contractible. To date, financial
contracting on social outcomes in the for-profit sector is rare. However, it is
not uncommon for executive compensation to be partly based on predefined
environmental, social and governance (ESG) goals.24 In addition, the public
sector has begun raising external finances via social-outcome-contingent
securities, for example, social impact bonds. These securities have been utilized
to fund childhood literacy programs, recidivism reduction interventions, and
homelessness initiatives. While our focus is on the design of contingent
social contracts in for-profit firms, the intuition in our model is instructive
for understanding the use of such contracts in these alternative settings.
24 For example, the United Nations provides guidelines on integrating ESG benchmarks into executive compensation
contracts (PRI 2012).
885

Furthermore, we expect contracts contingent on social output to be more

commonplace in the for-profit sector in the years to come as the pool of
socially conscious capital increases (Barber, Morse, and Yasuda 2017) and
as well-defined, measurable metrics of social value continue to improve.25
For this section, we assume the manager’s reservation payoff γ is sufficiently
small so that her participation constraint is slack. Consistent with Proposition
1, assuming a slack participation constraint enables us to focus on settings in

which involving the impact investor leads to greater social attention. Like in the
base model, the for-profit owner makes a take-it-or-leave-it offer to the manager.
However, in this case, the optimal wage contract depends on both x and b, since
both realized outputs provide independent signals of the manager’s attention
choice (Holmstrom (1979)). We denote the wage contract as a function of the
social benefit, w(b). If the manager accepts the wage contract, she chooses
a ∈ [0,1] to maximize her expected payoff:
a ∈ arg max E[xw(b)+ψm b|a ], (20)
a ∈[0,1]
which can be written explicitly as

a ∈ arg max = f (1−a )X g(a )w(1)+(1−g(a ))w(0) +g(a )ψm .
a ∈[0,1]
The analysis requires an additional regularity assumption: we assume that f (1−
a)(1−g(a)) is concave in a so that the manager’s expected payoff in (20) is
concave in a under all possible wage contracts.
Lemma 6. There exists a two-to-one continuous function mapping any wage
duplet {w(0),w(1)} to the unique solution to the manager’s problem in (20),
a(w(0),w(1)). (21)
We first outline the commercial firm benchmark as we did in Section 1.

When the impact investor does not participate in the project and social output
is contractible, there exists a wage contract wco (b) (with the for-profit owner
retaining zco (b) = 1−wco (b)) and attention allocation aco that maximizes the for-
profit owner’s expected payoff.26 Notably, the commercial benchmark provides
a lower bound on the level of social attention aco , which we will compare to
the level of social attention under impact investing.
To highlight the contracting tensions when social output is contractible, we
now consider a setting in which the for-profit owner commits not to renegotiate
the manager’s wage contract after receiving the initial transfer from the impact
investor.27 With full commitment, the for-profit owner’s optimal contracting
25 For example, Impact Reporting & Investment Standards (IRIS) (http://iris.thegiin.org/) and GIIRS Ratings and
Analytics for Impact Investing (http://b-analytics.net/giirs-funds) provide independent evaluations of social value
similar to the role credit rating agencies play in providing default information on corporate bonds.
26 For the sake of brevity, the supporting analytic results are provided in the appendix.
27 Although also outside the scope of the current model, one can imagine the full-commitment setting in the context
of a repeated game, in which the for-profit owner uses the prospects of future impact investments in order to
commit today to the promised level of social attention.
886

problem is given by

y(1) y(0)
f (1−a)X g(a) z(1)+ +(1−g(a)) z(0)+
1+δ 1+δ
max (22)
{a,Ω(b)}
g(a)ψs
+ −I
1+δ

s.t. a = a(w(0),w(1)), (22.1)
0 ≤ f (1−a)X[g(a)w(1)+(1−g(a))w(0)]+g(a)ψm −γ , (22.2)
0 ≤ w(b)+y(b)+z(b) ≤ 1, and Ω(b) ≥ {0,0,0}, (22.3)

where we have substituted for the upfront transfer (which sets the social impact
investor’s expected payoff equal to zero) into the for-profit owner’s expected
profit in (22).
Lemma 7. When the for-profit owner commits not to renegotiate the

manager’s wage, there exists a contract, {(wc (b),yc (b),zc (b)),ac }b∈{0,1} , which
maximizes the for-profit owner’s expected payoff. For ψs sufficiently large and
γ sufficiently small, the contract has the following features:
1. The manager’s wage is larger when social output is low, wc (0) > wc (1).
2. The impact investor receives no cash flow claim, yc (b) = 0 for b ∈ {0,1}.
3. The for-profit owner retains any cash flows not paid to the manager
zc (b) = 1−wc (b) for b ∈ {0,1}.
When social output is contractible, the manager’s optimal wage contract is

a “guarantee-against-social-failure” contract featuring wc (0) > wc (1).28 Since
the manager’s participation constraint is slack (as assumed throughout this
section), the for-profit owner wishes to pay the manager as small a wage
as possible while still providing the manager adequate incentives to pursue
the project’s monetary payoff. That is, if the manager’s wage contract is too
small, the manager enjoys greater utility from a successful social outcome
than from a successful monetary payoff, and the manager therefore allocates
too much attention to the project’s social technology. To reduce the level of
social attention and increase the probability of a successful monetary payoff,
the for-profit owner therefore provides the manager a guarantee-against-social-
failure wage contract which makes a greater financial payment when the project
fails to produce the social benefit (and which therefore provides the greatest
28 When the manager’s outside option is instead large, the manager may over- or underemphasize social attention
depending on the size of the manager’s required wage contract. When the manager overemphasizes social goals,
she is compensated with a guarantee-against-social-failure contract, whereas when the manager underemphasizes
social goals, she is compensated with a “pay-for-social-success” contract featuring wc (0) < wc (1).
887

disincentive for social attention). In addition, with full-commitment, the impact

investor, like in our base model, provides an upfront donation to the for-profit
owner in order to increase attention to social outcomes. Since the impact
investor retains none of the project’s produced cash flow, the for-profit owner
is simply paid the residual of manager’s wage contract.
Lemma 7 shows that, when contracting with the manager, the for-profit
owner would like to most cheaply reduce attention to social goals, and this

is accomplished with a guarantee-against-social-failure contract. In the next
section, we show that a similar tension dictates the impact investor’s required
financial claim. When commitment is limited, the for-profit owner would like to
commit to a high level of social attention after the impact investor’s payment is
sunk, and this is most cheaply accomplished by reducing the owner’s financial
claim (and thus by selling claims to the impact investor) when the project fails
to produce the social benefit.
3.1 Impact investing with contractible social output

We now consider contracts contingent on both realized cash flows and social
output when the for-profit owner suffers from limited commitment. Similar
to Section 2, the contract {(w(b),y(b),z(b)),a}b∈{0,1} can be renegotiated if
there exists an alternative contract {(w (b),y(b),z (b)),a }b∈{0,1} which satisfies
limited liability (w (b)+z (b) ≤ 1−y(b)) and which makes both the for-profit
owner and manager better off (with at least one strictly better off):

E xz (b) a = a w (0),w (1) ≥ E xz(b) a = a(w(0),w(1)) ,

E xw (b)+ψm b a = a w (0),w (1) ≥ E xw(b)+ψm b a = a(w(0),w(1)) .
A contract is renegotiation-proof if it cannot be renegotiated by the for-profit
owner and manager. Let SRP denote the set of renegotiation-proof contracts.
Then, the for-profit owner’s optimal contracting problem is identical to the
problem in (22) with the additional constraint,
{Ω(b),a}b∈{0,1} ∈ SRP . (22.4)
Proposition 2. There exists an optimal contract {(wcr (b),ycr (b),

zcr (b)),acr }b∈{0,1} which maximizes the for-profit owner’s expected payoff and
which is renegotiation-proof. For ψs sufficiently large and γ sufficiently small,
social attention exceeds the benchmark level acr > aco , and the impact investor
holds a nonzero financial claim, ycr (0) > 0 and/or ycr (1) > 0. Furthermore (like
in Corollary 2), nonprofit status is never optimal, ycr (0) < 1−wcr (0) and/or
ycr (1) < 1−wcr (1).
The results in Proposition 2 mirror those in Proposition 1 — impact investing

with contractible social output leads to greater social attention than the
commercial benchmark when the impact investor places a large value on the
social benefit (large ψs ) and when the project manager’s participation constraint
888

is slack (small γ ). To ensure that the for-profit owner does not renegotiate
the manager’s wage contract after receiving the upfront payment from the
impact investor, the impact investor must hold a nonzero financial claim to the
project’s produced cash. While selling financial claims to the impact investor is
costly, the benefit from increasing social attention and obtaining a larger upfront
payment is valuable when ψs is sufficiently large. In addition, Proposition 2
shows that the for-profit owner always retains some residual cash flow and thus

that nonprofit status is never optimal. Like in our base model, the benefit of
decreasing attention to social goals goes to zero in the limit as the for-profit
owner’s cash flow claim goes to zero (like in a pure nonprofit), while the cost
of increasing the manager’s wage is strictly positive. This implies that the for-
profit owner can always commit to the same level of social attention by selling
only a fraction of the residual cash flow to the impact investor as he could under
a pure nonprofit firm.
The following corollary elaborates on the impact investor’s cash flow claim.
Corollary 3. The impact investor’s financial claim satisfies,29
1. ycr (0) > 0 if and only if wcr (0) < 1,

2. ycr (1) > 0 only if ycr (0) = 1−wcr (0).
Corollary 3 shows that the impact investor is first compensated when the project
fails to produce the social benefit. By reducing the for-profit owner’s residual
claim when social output is low, this contract most cheaply mitigates the owner’s
preference to reduce attention to social goals after receiving the upfront payment
from the impact investor, and therefore allows the for-profit owner to (most
cheaply) commit to the desired level of social attention ex ante. When the for-
profit owner’s incentive for renegotiation is large enough, the owner may be
unable to commit to the desired level of social attention even when retaining
none of the project’s cash flow when social output is low. In these cases, the
impact investor is also compensated when social output is high.
We provide a numerical example in Figure 6 to further illustrate optimal
contracts with contractible social output. Figure 6(a) depicts the manager’s
compensation contract wcr (b) and her equilibrium attention choice acr as
functions of ψs (when γ is sufficiently small that her participation constraint is
slack), while Figure 6(b) depicts the impact investor’s claim ycr (b). When ψs
is small, a higher level of social attention a increases the subsidy paid by the
impact investor only a small amount. This small increase is more than offset
by the loss due to selling a financial claim to the impact investor. As such, the
for-profit owner offers the wage contract that compels the manager to choose
the commercial benchmark level of social attention aco , and the impact investor
29 Like in Proposition 2, the corollary assumes ψ is sufficiently large and γ is sufficiently small.
s
889

(a) (b)

Figure 6
The equilibrium wage contract, attention allocation, and impact investor’s financial claim as functions
of ψs
receives no financial claim. When ψs is sufficiently large, the for-profit owner

benefits from a higher level of a, since the increase in the subsidy from the
impact investor now outweighs the loss from selling him a financial claim.
Thus when ψs is large enough, the equilibrium level of social attention exceeds
the commercial benchmark, and this is achieved by decreasing the manager’s
wage contract, wcr (0). In the limit as ψs goes to infinity, the for-profit owner
sets wcr (0) = wcr (1) = 0 and the manager chooses acr = 1.
In our numerical example, the impact investor is compensated more heavily
when the project fails to produce the social benefit, ycr (0) > ycr (1). We refer to
this security as a guarantee-against-social-failure contract or more succinctly
as a “social impact guarantee,” since it provides the impact investor a greater
financial return when social returns are not successful. This design alleviates
the conflict of interest between the for-profit owner and the impact investor. In
particular, if the for-profit owner were to renegotiate the initial contract to a
lower level of social attention, the probability of a successful social outcome
would fall and the probability of a successful monetary payoff would rise.
While this lowers the impact investor’s expected social value, it increases the
impact investor’s expect payment when ycr (0) > ycr (1). Moreover, since the
impact investor owns a social impact guarantee, the for-profit owner’s residual
claim is a pay-for-social-success security zcr (1) > zcr (0). In the public sector,
pay-for-social-success financial contracts are commonly referred to as “social
impact bonds.” This, too, mitigates the difference in preferences between the
for-profit owner and impact investor by tying the owner’s expected payment to
a successful social outcome. As such, when the for-profit owner is compensated
with a pay-for-social-success security, he is able to commit ex ante to a greater
level of social attention.
The results of this section provide guidance in designing securities that are
contingent on social outcomes. When projects are financed by both profit-
motivated and socially motivated agents, intrinsically profit-motivated agents
890

(in our model, the for-profit owner) should be more heavily compensated when
social output is high, whereas intrinsically socially motivated agents (in our
model, the impact investor) should be more heavily compensated when social
output is low. The economic tensions at play are therefore consistent with the
design of social impact bonds, which make greater repayments to private (for-
profit) investors in public works projects when these projects are successful
in obtaining their desired social goals. Moreover, our model suggests that for-

profit owners signing cofinancing deals with socially motivated investors should
issue social impact guarantees so that the owners’ own residual claims reflect
the pay-for-social-success feature of social impact bonds.
4. Concluding Discussion
We provide a model of joint financing by profit-motivated investors and social
impact investors in projects that produce both corporate profits and social good,
and we show that impact investors must hold financial claims in order to
incentivize profit-motivated owners to pursue social goals. When a project’s
potential social value is large enough, joint financing is mutually beneficial,
leading to an increase in the owner’s expected profit and an increase in the level
of social output enjoyed by impact investors.
Since it is costly for socially minded investors to hold financial claims, impact
investors invest only in those firms with significant social value. In addition,
we show that impact investments increase attention to social goals only when
the manager’s participation constraint is slack, which is most likely for firms
with sufficiently socially motivated managers. We argue that these results are
consistent with the common practice of targeting investments in companies that
are most likely to provide societal and/or environmental benefits and excluding
investments in those companies that do not. Moreover, we show that social
investors hold a larger fraction of firm equity in the most socially valuable firms.
When the model is extended to include corporate taxes, impact investments in
for-profit firms arise as an intermediate solution between pure for-profit status
and pure nonprofit status.
In practice, governments, in addition to socially minded investors, also
subsidize social activities within for-profit firms. To the extent that private social
capital is limited, the role for governments to aid in the achievement of social
good is vast. However, unlike private investors, the government need not hold
financial claims in order to incentivize profit-motivated agents to pursue social
goals, since the requisite reduction in owners’ profit motives can be obtained
simply through taxation.
Finally, we provide guidance into the design of contingent social contracts,
for example, social impact bonds and social impact guarantees. Our results
suggest that profit-motivated agents be compensated more heavily when
projects succeed in achieving predefined social objectives (like in social impact
bonds), while socially motivated agents be compensated more heavily when
891

projects fail in achieving these objectives (like in social impact guarantees).

This intuition is consistent with the use of social impact bonds as a method for
raising capital from profit-oriented investors to fund social works programs,
such as those designed to reduce recidivism among nonviolent offenders or
provide greater access to high-quality preschools in high need areas.

Appendix
Proof of Lemma 1
For a given wage w, the manager’s problem in (5) is explicitly given by
max f (1−a)Xw +g(a)ψm . (A1)

a∈[0,1]
Because f (·) and g(·) are twice-differentiable, continuous, and strictly concave functions, the
manager’s first-order condition is necessary and sufficient for determining the payoff-maximizing
attention allocation choice ã. For any w > 0, the manager’s payoff-maximizing choice ã is implicitly
defined by
0 = −f (1− ã)Xw +g (ã)ψm . (A2)
Define a(w) = ã as the equilibrium attention choice as a function of w. Because f (·) and g(·) are
twice-differentiable and continuous, a(w) is continuous and differentiable. By the implicit function
theorem, a(w) is strictly decreasing in w:

da ∂
−f (1−a)Xw +g (a)ψm
= − ∂w (A3)
dw ∂
(−f (1−a)Xw +g (a)ψm )
∂a a=ã
f (1− ã)X
= , (A4)
f (1− ã)Xw +g (ã)ψm
which is strictly negative. Therefore, a(w) is a one-to-one mapping between ã and w.
Proof of Lemma 2
By Lemma 1, a(w) : w → ã is one-to-one, continuously differentiable, and strictly decreasing
in w. By the inverse function theorem, there exists a function w(a) = a−1 (w) : ã → w that is one-
to-one, continuously differentiable, and strictly decreasing in ã. The explicit form of w(a) comes
from a rearrangement of (A2):
g (ã)ψm
w(ã) = = w. (A5)
f (1− ã)X
Furthermore, w(ã) is increasing in ψm .
Before proceeding with our the proof of Lemma 3, we provide the explicit assumptions that
guarantee that the set of contracts that satisfy the manager’s participation constraint and raise at
least I from investors is not empty.
Assumption 2. The manager’s outside opportunity is sufficiently small:
γ < max f (1−a)X +g(a)ψm . (A6)

a∈[0,1]
Assumption 2 guarantees that the manager’s participation constraint can be satisfied, that is, if the
manager is given the project’s entire cash flow, w = 1, it is individually rational for her to participate.
The requirement on γ is a necessary condition for the existence of a nontrivial contract derived in
Lemma 3, but the condition is not enough to guarantee existence.
892

Assumption 3.
g(ã)ψs
I≤ , (A7)
1+δ
where ã ∈ arg max f (1−a)X +g(a)ψm .
a∈[0,1]
Assumption 3 guarantees that the project will be funded in equilibrium; that is, the project generates
sufficient enough social good so that the impact investor would be willing and able to cover the

project’s upfront cost I . Together, Assumptions 2 and 3 guarantee that the set of contracts that
satisfies the constraints in the for-profit owner’s problem in (12) is not empty.
Proof of Lemma 3
Before constructing a proof of Lemma 3, we provide two preliminary results.
Lemma 8. If the for-profit owner can commit to not renegotiate contracts, any optimal contract
never exhibits cash burning:
1 = wv +yv +zv . (A8)
Proof of Lemma 8
The proof is constructed as a proof by contradiction: There exists an optimal contract
{(w ,y ,z ),a } that exhibits cash burning, that is, 1 > w +y +z . Now consider an alternative
contract {(w ,y , ẑ), â} with 1 = w +y + ẑ. By Lemma 1, the alternative contract induces the same
level of equilibrium attention as the original contract,
â = a(w ) = a . (A9)
However, the alternative contract yields a strictly higher payoff for the for-profit owner:

y g(a )ψs y g(a )ψs
f (1−a )X ẑ+ + −I > f (1−a )X z + + −I. (A10)
1+δ 1+δ 1+δ 1+δ
Therefore, an optimal contract never exhibits cash burning.
Lemma 9. If the for-profit owner can commit to not renegotiate contracts, an optimal contract
features no cash repayment to the social impact investor:
yv = 0. (A11)
Proof of Lemma 9
The proof is constructed as a proof by contradiction: There exists an optimal contract
{(w ,y ,z ),a } for which the social impact investor receives a strictly positive fraction of the
project’s cash flow, that is, y > 0. Now consider an alternative contract {(w , ŷ, ẑ), â} with ŷ = 0.
By Lemma 8, the for-profit owner’s fractions of the cash flow under each contract are given by:
z = 1−w −y , (A12)
ẑ = 1−w , (A13)
implying ẑ > z . By Lemma 1, the alternative contract induces the same level of equilibrium attention
as the original contract:
â = a(w ) = a . (A14)
However, the alternative contract yields a strictly higher payoff for the for-profit owner:

y
f (1−a )X (1−y −w )+
g(a )ψs 1+δ
f (1−a )X(1−w )+ −I > , (A15)
1+δ g(a )ψs
+ −I
1+δ
893

because δ > 0. Therefore, if the for-profit owner can commit to not renegotiate contracts, an optimal
contract features no cash repayment to the impact investor.
We now proceed with the proof of Lemma 3. Since f (·) and g(·) are continuous functions, the
for-profit owner’s expected payoff is continuous in w, y, z, and a. Constraints 12.1–12.3 guarantee
the set of feasible contracts is closed and bounded. Furthermore, by Assumptions 2 and 3, the
contract {(w,y,z),a} = {(1,0,0),a(1)} satisfies constraints 12.1–12.3 and guarantees the for-profit
owner a nonnegative expected payoff. Thus, the set of feasible contracts is nonempty, closed, and

bounded, and, by the extreme value theorem, a maximum exists.
There are two cases to consider in proving the optimal contract is unique: (1) the manager’s
participation constraint in (12.2) is slack and (2) the manager’s participation constraint in (12.2)
binds. First, consider the former, that is, (12.2) is slack. Using Lemma 8 and Lemma 9, the for-profit
owner’s problem may be rewritten as
g(a)ψs
max f (1−a)X(1−w(a))+ −I. (A16)
a∈[0,1] 1+δ
Because the for-profit owner’s expected payoff in (A16) is concave by assumption, the for-profit
owner’s first-order condition is necessary and sufficient for determining the unique payoff-
maximizing attention allocation choice av . The optimal contract is given by {(w(av ),0,1−
w(av )),av }.
Now, suppose the manager’s participation constraint in (12.2) binds. The optimal level of social
attention is implicitly defined by

g (av )ψm
0 = f (1−av )X +g(av )ψm −γ , (A17)
f (1−av )X

(av )ψm
where we have made the substitution w(av ) = fg (1−a . Because f (·) and g(·) are twice-
v )X
differentiable, continuous, and strictly concave functions, there is a unique value av ∈ (0,1) that
satisfies (A17). The optimal contract is given by {(w(av ),0,1−w(av )),av }.
To complete the proof, we show that av > ao if γ < γo and av = ao if γ ≥ γo . Before proceeding,
we provide a preliminary result.
Lemma 10. The manager’s expected payoff is strictly increasing in w.
Proof of Lemma 10
The manager’s expected payoff with wage contract w is given by
f (1−a(w))Xw +g(a(w))ψm . (A18)
The change in the manager’s payoff with respect to a change in w is given by
da
d −f (1−a(w))Xw +g (a(w))ψm
(f (1−a(w))Xw +g(a(w))ψm ) = dw . (A19)
dw
+f (1−a(w))X
By an envelope condition, the expression simplifies to
= f (1−a(w))X, (A20)
> 0. (A21)
Therefore, the manager’s expected payoff is strictly increasing in w.
894

We now proceed with completing the proof of Lemma 3. First, ignoring the manager’s
participation constraint momentarily, define the level of attention that maximizes the for-profit
owner’s payoff as â where â ∈ (0,1) is implicitly defined by

g (â)ψs dw
0 = −f (1− â)X(1−w(â))+ −f (1− â)X , (A22)
1+δ da a=â
where â > ao because g (·) > 0. If γ < γo there exists some equity wage contract w̃ < wo that satisfies

the manager’s participation constraint and leads the manager to choose attention a(w̃) > ao . The for-
profit owner decreases the equity wage contract from w̃ to wv which occurs when either a(wv ) = â
or the manager’s participation constraint binds. Therefore, av > ao if γ < γo .
If γ ≥ γo , it is not possible to attain any level of attention greater than ao . It is not possible because
a(·) is decreasing in w and, by Lemma 10, the manager’s payoff is increasing in w. This implies
that if γ ≥ γo and the manager’s participation constraint binds at w(ao ), it is not possible to lower
the wage and attain av > ao without violating the manager’s participation constraint. Therefore,
av = ao if γ ≥ γo .
Proof of Lemma 4
Denote the initial equity wage contract, before investment is sunk, as w. The manager’s expected
date 1 utility with wage contract w is given by
f (1− ã)Xw +g(ã)ψm , (A23)
where ã = a(w). By Lemma 10, the manager’s expected date 1 utility is strictly increasing in w.
Therefore, renegotiation is not possible if it involves decreasing the manager’s wage, as the manager
would be worse off. Furthermore, by Lemma 1, the for-profit owner can only affect the manager’s
attention allocation choice by changing w. Thus, renegotiation is only possible if it involves strictly
increasing the manager’s wage.
Proof of Lemma 5
By assumption, the for-profit owner’s discounted expected date 1 payoff Up (w(a),y,z,a) is
twice continuously differentiable and concave in a. The contract {(w,y,z),a} will therefore be
renegotiated by the for-profit owner if and only if it is profitable for the for-profit owner to decrease
attention to social output by raising the manager’s wage (see Lemma 4),

∂Up dw ∂Up
+ < 0.
∂w da ∂a {(w,y,z),a}
Thus, the contract is renegotiation-proof if and only if

∂Up dw ∂Up dw
+ = −f (1−a)X −f (1−a)X(1−w(a)−y) ≥ 0
∂w da ∂a da
A rearrangement of the preceding inequality yields

f (1−a) dw
y ≥ y(a) ≡ 1−w(a)+ . (A24)
f (1−a) da
Finally,

dy dw f (1−a) f (1−a) d 2 w
= −2 − (1−w(a)−y(a))+ > 0. (A25)
da da f (1−a) f (1−a) da 2

Proof of Proposition 1
From Lemma 4, renegotiation is possible only if it involves strictly increasing the manager’s
wage. Since the contract {(w,y,z),a} = {(1,0,0),a(1)} gives the manager all of the project’s equity, it
cannot be renegotiated, and thus satisfies constraint 14.4. As such, the existence of a unique optimal
895

renegotiation-proof contract directly follows from the proof of Lemma 3, which establishes the
existence of a unique optimal contract without renegotiation.
Constraint 14.1 guarantees that wr = w(ar ), and constraint 14.4 guarantees yr ≥ y(ar ). Since
increasing y is costly for δ > 0, the optimal renegotiation-proof contract sets y as small as possible,
yr = y(ar ). We now verify that yr = y(ar ) satisfies our limited liability constraints. Since dw
da < 0,
dy
we have y ≤ 1−w for all a. By construction y(ao ) = 0, and from Lemma 5, da > 0. Together, these
imply that y(a) ∈ [0,1−w] for all a ∈ [ao ,1]. A similar argument like in Lemma 3 establishes that
ar ≥ ao , and thus yr satisfies limited liability. Finally, Lemma 8 establishes that zr = 1−wr −yr .

As discussed above, the optimal contract involves w = w(a) and y = y(a). As a result, the for-profit
owner’s expected payoff is given (as a function of a) by
f (1−a)Xy(a) g(a)ψs
P(a) = f (1−a)X(1−w(a)−y(a))+ + . (A26)
1+δ 1+δ
Differentiating with respect to a yields

dw δ dy
−f (1−a)X(1−w(a)−y(a))−f (1−a)X +
dP da 1+δ da
= (A27)
da f (1−a)Xy(a) g (a)ψs
− + ,
1+δ 1+δ

δ dy f (1−a)Xy(a) g (a)ψs
= −f (1−a)X − + , (A28)
1+δ da 1+δ 1+δ
where we have used the definition of y(a) from (13).
When γ < γo , the level of social attention exceeds the benchmark level, a > ao , if and only if,

dP δ dy f (1−ao )Xy(ao ) g (ao )ψs
= −f (1−ao )X − + > 0.

da a=ao 1+δ da a=ao 1+δ 1+δ
A rearrangement of the preceding inequality yields

dy
δf (1−ao )X da a=a
o
ψs > ψs = , (A29)
g (ao )
where we have made the substitution y(ao ) = 0. Thus, for ψs > ψs , the optimal level of social
dy(a)
attention ar exceeds the benchmark level: ar > ao . Since da > 0, ar > ao implies that yr > 0.
When γ ≥ γo , a similar argument as used in the proof of Lemma 3 gives ar = ao . And by
construction, y(ao ) = 0.
Proof of Corollary 1
From Proposition 1, ar > ao for ψs > ψs . Thus, if the manager’s participation constraint is slack,
the optimal level of social attention ar is given by the first-order condition for the for-profit owner’s
expected payoff:

δ dy
−f (1−ar )X
1+δ da a=ar
dP
= = 0. (A30)
da a=ar
f (1−ar )Xy(ar ) g (ar )ψs
− +
1+δ 1+δ
dyr
We begin with the comparative static of dψs : Taking the total derivative of (A30) with respect to
ψs and rearranging terms gives
rg (a )
dar 1+δ
= > 0, (A31)
dψs − d 2 P
2 da a=ar
896

which is strictly positive since P(a) is strictly concave in a by assumption. Thus, increasing ψs
raises the equilibrium level of social attention. Furthermore, since yr depends on ψs only through
dy
ar and da > 0,
dyr
> 0, (A32)
dψs
as well.
dyr
Next, we consider the comparative static of dδ : Taking the total derivative of (A30) with respect

to δ and rearranging terms yields

dy
−f (1−ar )X 1
da a=a
dar 1+δ
= r
> 0, (A33)
dψs 2P
−d
da 2 a=a r
dy
which is strictly negative since P(a) is strictly concave in a and da is positive valued for all a. Thus,
increasing δ decreases the equilibrium level of social attention. Furthermore, since yr depends on
dy
δ only through ar and da > 0,
dyr
< 0, (A34)
dδ
as well.
From Proposition
1, the impact investor’s equity stake is given by yr = 1−w(ar )+
f (1−ar ) dw
f (1−a ) da a=a
which is the smallest financial claim needed to dissuade the for-profit owner
r r
from reneging on the promised level of social attention ar . Since dw da < 0 for all a ∈ (0,1), the
impact investor’s equity stake is strictly less than the entire residual cash flow claim, and the
for-profit owner therefore retains a positive equity stake, z = − ff(1−a r ) dw > 0.
(1−a ) da a=a r r
Proof of Lemma 6
For a given wage w(b) = {w(0),w(1)}, the manager’s problem in (20) is explicitly given by,
max f (1−a)X(g(a)w(1)+(1−g(a))w(0))+g(a)ψm . (A35)

ã∈[0,1]
Because f (1−a)(1−g(a)) is continuously differentiable and concave in a by assumption, and both

f (·) and g(·) are twice-differentiable, continuous, and strictly concave functions, the manager’s
first-order condition is necessary and sufficient for determining the unique payoff-maximizing
attention allocation choice ã. If
0 > −f (1)X(g(0)w(1)+(1−g(0))w(0))+f (1)X g (0)w(1)−g (0)w(0) +g (0)ψm ,
then ã = 0. In addition, if
0 < −f (0)X(g(1)w(1)+(1−g(1))w(0))+f (0)X g (1)w(1)−g (1)w(0) +g (1)ψm ,
then ã = 1. Finally, for any noncorner solution, the manager’s unique payoff-maximizing choice ã
is implicitly defined by
−f (1− ã)X(g(ã)w(1)+(1−g(ã))w(0))
0= (A36)
+f (1− ã)X g (ã)w(1)−g (ã)w(0) +g (ã)ψm .
Define a(w(0),w(1)) = ã as the equilibrium attention choice as a function of {w(0),w(1)}. Because

f (·) and g(·) are twice-differentiable and continuous, a(w(0),w(1)) is continuous and differentiable
in each argument. Therefore, a(w(0),w(1)) is a two-to-one mapping between {w(0),w(1)}
and a.
897

Proof of Lemma 7
Before constructing a proof of Lemma 7, we rely on Lemma 8 and Lemma 9, located in the
proof of Lemma 3. By Lemma 8, the optimal contract features,
1 = wc (b)+yc (b)+zc (b). (A37)
By Lemma 9, the optimal contract, when renegotiation is not possible, features,
yc (b) = 0. (A38)

We now proceed with the proof of Lemma 7. Since f (·) and g(·) are continuous functions, the
for-profit owner’s expected payoff is continuous in w(b), y(b), z(b), and a. Constraints 22.1-
22.3 guarantee the set of feasible contracts is closed and bounded. Furthermore, by Assumption
2, Assumption 3, and Lemma 6, the contract {(w(b),y(b),z(b)),a} = {(1,0,0),a(1,1)} satisfies
constraints 22.1–22.3 and guarantees the for-profit owner a nonnegative expected payoff. Thus,
the set of feasible contracts is nonempty, closed, and bounded, and by the extreme value theorem,
a maximum exists.
Before proceeding with the proof of Lemma 7, we provide a preliminary result.
Lemma 11. The marginal social-attention-to-cost ratio for w(0),

∂a/∂w(0)
, (A39)
∂P/∂w(0)
is strictly positive and strictly greater than the marginal social-attention-to-cost ratio for w(1),
∂a/∂w(1)
. (A40)
∂P/∂w(1)
Proof of Lemma 11
We begin by considering the direct effects of both w(0) and w(1) on the for-profit owner’s
expected payoff. When renegotiation is not possible, the for-profit owner’s expected net payoff
with contractible social output is given by
g(a)ψs
P = f (1−a)X(1−g(a)w(1)−(1−g(a))w(0))+ −I, (A41)
1+δ
and the direct effect of changing w(0) on the net payoff is given by
∂P
= −f (1−a)X(1−g(a)), (A42)
∂w(0)
and the direct effect of changing w(1) on the net payoff is given by
∂P
= −f (1−a)Xg(a). (A43)
∂w(1)
∂a
The manager’s first-order condition in (A36) is used to calculate ∂w(b) ,
∂a
= −f (1−a)X(1−g(a))−f (1−a)Xg (a), (A44)
∂w(0)
∂a
= −f (1−a)Xg(a)+f (1−a)Xg (a). (A45)
∂w(1)
Using the preceding analysis, the marginal social-attention-to-cost ratios are,
∂a
∂w(0) −f (1−a)X(1−g(a))−f (1−a)Xg (a)
= , (A46)
∂P(a) −f (1−a)X(1−g(a))
∂w(0)
898

f (1−a) g (a)
= + > 0, (A47)
f (1−a) (1−g(a))
∂a
∂w(1) −f (1−a)Xg(a)+f (1−a)Xg (a)
= , (A48)
∂P(a) −f (1−a)Xg(a)
∂w(1)
f (1−a) g (a)
= − . (A49)
f (1−a) g(a)

g (a) (a)
Because (1−g(a)) > 0 and − gg(a) < 0,
∂a/∂w(0) ∂a/∂w(1)
> . (A50)
∂P/∂w(0) ∂P/∂w(1)

Lemma 11 shows, roughly speaking, that for a marginal change in the manager’s social attention,
the cheapest means to do so (i.e, the smallest impact on the for-profit owner’s expected payoff) is
via w(0).
We complete the proof of Lemma 7 by using a proof by contradiction; the optimal wage contract
features wc (0) > wc (1). Suppose not: If γ is sufficiently small such that the manager’s participation
constraint is slack, there exists an optimal wage contract featuring w (0) < w (1). Using Lemma
11, there must exists an alternative wage contract with ŵ(0) > ŵ(1) that induces the same level of
effort and yields a higher expected payoff for the for-profit owner. Therefore, the wage contract
{w (0),w (1)} cannot be optimal. Thus, when γ is sufficiently small and the manager’s participation
constraint is slack, the optimal contract features wc (0) > wc (1).
Derivation of Commercial Firm Benchmark with Contractible Social Output

The for-profit owner’s problem is given by
max f (1−a)X[g(a)z(1)+(1−g(a))z(0)]−I (A51)

{z(b),w(b),a}
s.t. a = a(w(0),w(1)) (A51.1)
0 ≤ f (1−a)X[g(a)w(1)+(1−g(a))w(0)]+g(a)ψm −γ (A51.2)
0 ≤ w(b)+z(b) ≤ 1, and w(b) ≥ 0, z(b) ≥ 0. (A51.3)
To begin, we rely on Lemma 8 located in the proof of Lemma 3. By Lemma 8, an optimal contract
features,
1 = w(b)+z(b). (A52)
Moreover, since f (·) and g(·) are continuous functions, the for-profit owner’s expected payoff is
continuous in w(b), z(b), and a. Constraints A51.1-A51.3 guarantee the set of feasible contracts is
closed and bounded. Furthermore, by Assumption 2, Assumption 3, and Lemma 6, the cash flow
sharing rule w(b) = 1 (and equilibrium attention a(1,1)) satisfies constraints A51.1–A51.3. Thus,
the set of feasible contracts is nonempty, closed, and bounded, and by the extreme value theorem,
a maximum exists.
Finally, it is straightforward to show, like in the proof of Lemma 7, that a unique contract exists
and the contract elicits an attention allocation aco and the contract is characterized by wco (0) >
wco (1). Suppose not: If γ is sufficiently small such that the manager’s participation constraint
is slack, there exists an optimal wage contract featuring w (0) < w (1). Using Lemma 11, there
must exists an alternative wage contract with ŵ(0) > ŵ(1) that induces the same level of effort and
yields a higher expected payoff for the for-profit owner. Therefore, the wage contract {w (0),w (1)}
899

cannot be optimal. Thus, when γ is sufficiently small and the manager’s participation constraint is
slack, the optimal contract features wco (0) > wco (1) and aco < 1.30
Proof of Proposition 2
Before constructing a proof of Proposition 2, we rely on Lemma 8, located in the proof of
Lemma 3. By Lemma 8, an optimal contract features,
1 = wcr (b)+ycr (b)+zcr (b). (A53)

We now proceed with the proof of Proposition 2. Since f (·) and g(·) are continuous functions,
the for-profit owner’s expected payoff is continuous in w(b), y(b), z(b), and a. Constraints 22.1-
22.3 guarantee the set of feasible contracts is closed and bounded. Furthermore, by Assumption
2, Assumption 3, and Lemma 6, the contract {(w(b),y(b),z(b)),a} = {(1,0,0),a(1,1)} satisfies
constraints 22.1–22.3, and, since the contract {(w(b),y(b),z(b)),a} = {(1,0,0),a(1,1)} gives the
manager all of the project’s produced cash, it cannot be renegotiated, and thus satisfies constraint
22.4. Therefore, the set of feasible contracts is also nonempty and, by the extreme value theorem,
an optimal contract exists.
We now proceed to showing that for a sufficiently large ψs and sufficiently small γ , an optimal
renegotiation-proof contract elicits an attention level acr that is strictly larger than aco . Suppose
not: For any ψs , the optimal renegotiation-proof contract elicits a level of social attention equal to
aco . The cheapest renegotiation-proof way to elicit aco is by offering the commercial benchmark
contract: the employee receives the wage contract wco (b), the for-profit project owner receives the
entire residual 1−wco (b) and the impact investor receives no repayment. The for-profit project
owner’s expected profit is given by
g(aco )ψs
f (1−aco )X(g(aco )(1−wco (1))+(1−g(aco ))(1−wco (0)))+ . (A54)
1+δ
Recall, the commercial benchmark level of attention, aco , is strictly less than 1. Furthermore,
wco (0) > 0. For γ sufficiently small, the manager’s participation constraint is slack and wco (0)
may be lowered, holding fixed wco (1), without violating the manager’s participation constraint.
As such, consider an alternative wage contract in which ŵ(0) = wco (0)− for some small > 0
and ŵ(1) = wco (1). Relying on Lemma 11, the alternative wage contract elicits a level of attention
â > aco . Furthermore, to ensure that the contract is renegotiation-proof set ŷ(0) = 1− ŵ(0) and
ŷ(1) = 1− ŵ(1) so that the for-profit owner receives no repayment. Under the alternative contract,
the for-profit owner’s profit is given by
f (1− â)X g(â)(1− ŵ(1))+(1−g(â))(1− ŵ(0)) +g(â)ψs

. (A55)
1+δ
The difference between the for-profit owner’s profit in (A55) and in (A54) is given by
f (1− â)X g(â)(1− ŵ(1))+(1−g(â))(1− ŵ(0))

1+δ
(1+δ)f (1−aco )X(g(aco )(1−wco (1))+(1−g(aco ))(1−wco (0))) ψs g(â)−g(aco )

− + , (A56)
1+δ 1+δ
which is increasing in ψs because g(â)−g(aco ) is strictly positive. Consequently, the for-profit
owner’s profit is strictly higher under the alternative contract for some finite, large ψs . Thus, it is
not true that for any ψs , the optimal renegotiation-proof contract elicits a level of social attention
30 To ensure that the commercial benchmark level of social attention is strictly less than 1, we require an assumption

dw(0)
similar to that used in Section 1: lim f (1−a) da(w(0),w(1)) is finite valued.
a→1 w(1)=0
900

equal to aco . Instead, for a sufficiently large ψs , the optimal level of social attention strictly exceeds
aco . Moreover, the impact investor, like in Proposition 1, must hold a nonzero financial claim if
aco > aco . If not, the for-profit owner will renegotiate the manager’s wage contract to the commercial
benchmark wage contract after the impact investor’s upfront payment is sunk.
To conclude the proof, we show that nonprofit status is never optimal, that is, ycr (0) < 1−wcr (0)
and/or ycr (1) < 1−wcr (1). The proof requires showing that the equilibrium a is continuous in y(b).

Lemma 12. The equilibrium level of social attention is continuous in y(b).
Proof of Lemma 12
Like in the proof of Proposition 1, denote the for-profit owner’s expected payoff, before the
impact investor’s transfer is sunk, as P(a,w(b),y(b),z(b)). Furthermore, denote the initial contract
consisting of the manager’s wage, impact investor cash flow claim, the for-profit owner’s cash flow
claim, and equilibrium attention choice as {(w(b),y(b),z(b)),a}b∈{0,1} . Now, like in the proof of
Lemma 5, denote the for-profit owner’s expected payoff, after the impact investor’s transfer is sunk
(and y(b) is fixed), as Up (a ,w (b),z (b);y(b)). The for-profit owner’s problem after the impact
investor’s transfer is sunk is given by
max Up (a ,w (b),z (b);y(b)) (A57)

{w (b),a }
s.t. a = a(w (0),w (1)), (A57.1)

f (1−a )X g(a)w (1)+(1−g(a))w (0) +g(a )ψm
≥ f (1−a)X(g(a)w(1)+(1−g(a))w(0))+g(a)ψm , (A57.2)

0 ≤ w (b)+z (b) ≤ 1−y(b), and w (b) ≥ 0, z (b) ≥ 0, (A57.3)
where constraint A57.1 maps the wage w (b) to the manager’s optimal social attention choice a ,
constraint A57.2 is the manager’s individual rationality constraint in accepting wage w (b) in lieu
of the initial wage w(b), and constraint A57.3 represents the limited liability constraint.31
The for-profit owner’s problem may be generalized as follows: define χ ≡ {a ,w (b),z (b)}b∈{0,1}
and θ ≡ {y(b)}b∈{0,1} , then the for-profit owner’s objective function may be rewritten as Up (χ ;θ )
which is continuous in χ and θ . Furthermore, define C(θ ) : [0,1]2 [0,1]5 as a compact, vector-
valued, correspondence that maps the impact investor’s initial contract θ to constraints A57.1–
A57.3 and yields the constrained set of feasible contracts to choose from. Therefore, the for-profit
owner’s problem may be restated as,
argmaxχ ∈C(θ ) Up (χ ;θ ), (A58)
and define the solution to the for-profit owner’s problem as,
C ∗ (θ ) ∈ argmaxχ ∈C(θ ) Up (χ ;θ ). (A59)
Note, y(b) only affects limited liability in Constraint A57.3 and the constraint is linear in
y(b). Furthermore, a(w(0),w(1)) is continuous in w(b), and the manager’s individual rationality
constraint is linear in w(b) and continuous in a. As such, C(θ ) is both upper and lower
hemicontinuous. Consequently, by the theory of the maximum, the solution C ∗ (θ ) is continuous
31 Renegotiation-proofness requires that the solution to the for-profit owner’s problem outlined above is the same
as the initial contract P(a,w(b),y(b),z(b)).
901

in θ . Therefore, av (θ ), which is an element of C ∗ (θ ), is continuous in θ . Or, said differently, the

equilibrium level of social attention is continuous in y(b).
Continuing with our proof showing that nonprofit status is never optimal, we use Lemma 12 to
show that, after the impact investor’s upfront payment is sunk, the renegotiation tension is in
the direction of less social attention if a > aco . As above, denote the for-profit owner’s expected
payoff, before the impact investor’s transfer is sunk, as P(a,w(b),y(b),z(b)). Furthermore, denote
the initial contract consisting of the manager’s wage, impact investor cash flow claim, the for-profit
owner’s cash flow claim, and equilibrium attention choice as {(w(b),y(b),z(b)),a}b∈{0,1} . Now,

denote the for-profit owner’s expected payoff, after the impact investor’s transfer is sunk (and y(b)
is fixed), as Up (a ,w (b),z (b);y(b)), where {(w (b),y(b),z (b)),a }b∈{0,1} is the solution to the
for-profit owner’s problem. We now show that if a > a > aco there exists a profitable deviation to
the original contract {(w(b),y(b),z(b)),a}b∈{0,1} which is renegotiation-proof. aco is achieved by
setting y(b) = 0. Because a is continuous in y(b) (by Lemma 12) and by the intermediate value
theorem, there exists an alternative cash flow claim ỹ(b) < y(b) for the impact investor that elicits
ã = a. To determine ỹ(b) first, we begin by lowering y(1) until ã = a or until y(1) = 0. If y(1) = 0
before ã = a, then lower y(0) until ã = a. As such, ỹ(b) < y(b) implying there exists a profitable
deviation to the original contract {(w(b),y(b),z(b)),a}b∈{0,1} which is renegotiation-proof.
The above argument shows that, like in the base model (see Lemma 5), y(b) must be large
enough so that it is not profitable for the for-profit owner to renegotiate the initial contract to
a lower level of social attention. We now show that pure nonprofit status is never optimal. We
proceed by contradiction. Suppose that an optimal contract {(w (b),y (b),z (b)),a }b∈{0,1} invokes
nonprofit status, z (b) = 0. Now consider an alternative contract {(w (b), ŷ(b), ẑ(b)),a }b∈{0,1} in
which ŷ(0) = y (0) and ŷ(1) = y (1)− for > 0 and ẑ(0) = 0 and ẑ(1) = . For the alternative contract
to be renegotiation-proof, it must not be profitable for the for-profit owner to renegotiate the
manager’s wage contract in order to induce a lower level of social attention. Since ŷ(0) = 1−w (0)
and ẑ(0) = 0, the for-profit owner cannot increase the manager’s wage contract when the project
fails to produce the social benefit. Furthermore, this rules out as well the possibility for the owner
to reduce the manager’s wage contract when the contract successfully produces the social benefit,
since the manager’s total expected repayment cannot be reduced. The available deviations therefore
include increasing w (1) and/or decreasing w (0). There are two cases to consider: (a) suppose first
that the owner increases w (1) and decreases w (0) such that the manager’s total expected payoff
remains unchanged. Recall from the proof of Lemma 7 that w (0) provides a greater disincentive for
social attention than does w (1). This deviation therefore leads to a higher level of social attention,
and (as shown above) can therefore not be profitable for the owner. Now (b) suppose that the
owner increases w (1) and (weakly) decreases w (0) such that manager’s total expected payoff
increases. In the limit as → 0, the owner’s residual cash flow claim goes to zero and the benefit of
decreasing the level of social attention also goes to zero, while the cost of increasing the manager’s
wage is still strictly positive. Thus, there exists an > 0 small enough such that this deviation is also
not profitable for the for-profit owner. Then the alternative contract {(w (b), ŷ(b), ẑ(b)),a }b∈{0,1}
is renegotiation-proof and requires a smaller repayment to the impact investor than the original
contract {(w (b),y (b),z (b)),a }b∈{0,1} . This contradicts the optimality of the original contract, and
thus nonprofit status is never optimal.
In an optimal renegotiation-proof contract, the impact investor’s claim {ycr (0),ycr (1)} satisfies
(a) ycr (0) > 0 if and only if wcr (0) < 1 and (b) ycr (1) > 0 only if ycr (0) = 1−wcr (0). We first prove
(b). Suppose not: There exists an optimal renegotiation-proof contract with y (0) < 1−w (0) and
y (1) > 0. Using Lemma 11, there exists an alternative cash flow claim {ŷ(0), ŷ(1)} with ŷ(0) > y (0)
and ŷ(1) < y (1) that is also renegotiation-proof and requires a lower repayment to the impact
investor. Therefore, the cash flow claim {y (0),y (1)} cannot be optimal. Thus, ycr (1) > 0 only if
ycr (0) = 1−wcr (0).
We now prove part (i) by contradiction. Suppose not: There exists an optimal renegotiation-
proof contract with y (0) = 0 and w (0) < 1. There are two cases to consider. First, consider y (1) = 0.
902

However, by Proposition 2, {y (0),y (1)} = {0,0} cannot satisfy an optimal contract if acr > aco .
Next, consider y (1) > y (0). However, as shown above, y (1) > y (0) cannot be optimal. Thus,
ycr (0) > 0 if and only if wcr (0) < 1.
References
Bagnoli, M., and S. G. Watts. 2003. Selling to socially responsible consumers: Competition and the private

provision of public goods. Journal of Economics & Management Strategy 12:419–45.
Barber, B. M., A. Morse, and A. Yasuda. 2017. Impact investing. Working Paper, University of California,
Berkeley.
Baron, D. P. 2001. Private politics, corporate social responsibility, and integrated strategy. Journal of Economic
& Management Strategy 10:7–45.
———. 2008. Managerial contracting and corporate social responsibility. Journal of Public Economics
92:268–88.
Besley, T., and M. Ghatak. 2005. Competition and incentives with motivated agents. American Economic
Review 95:616–36.
———. 2007. Retailing public goods: The economics of corporate social responsibility. Journal of Public
Economics 91:1645–63.
———. 2017. Profit with purpose? A theory of social enterprise. American Economic Journal: Economic
Policy 9:19–58.
Chava, S. 2014. Environmental externalities and cost of capital. Management Science 60:2223–47.
Cheng, I.-H., H. G. Hong, and K. Shue. 2016. Do managers do good with other peoples’ money? Working Paper,
Dartmouth College.
Glaeser, E. L., and A. Shleifer. 2001. Not-for-profit entrepreneurs. Journal of Public Economics 81:99–115.
Geczy, C. C., R. F. Stambaugh, and D. Levin. 2005. Investing in socially responsible mutual funds. Working
Paper, University of Pennsylvania, Wharton.
GIIN. 2016. Impact investing trends. New York: Global Impact Investing Network.
———. 2017. Annual impact investor survey 2017. New York: Global Impact Investing Network.
Hansmann, H. 1996. The ownership of enterprise. Cambridge: Harvard University Press.
Hart, O., and L. Zingales. Forthcoming. Companies should maximize shareholder welfare not market value.
Journal of Law, Finance, and Accounting.
Hartzmark, S. M., and A. B. Sussman. 2017. Do investors value sustainability? A natural experiment examining
ranking and fund flows. Working Paper, University of Chicago.
Heinkel, R., A. Kraus, and J. Zechner. 2001. The effect of green investment on corporate behavior. Journal of
Financial and Quantitative Analysis 36:431–49.
Holmstrom, B. 1979. Moral hazard and observability. Bell Journal of Economics 10:74–91.
———. 1984. A note on the first-order approach in principal agent models. Mimeo.
Holmstrom, B., and P. Milgrom. 1991. Multitask principal-agent analyses: Incentive contracts, asset ownership,
and job design. Journal of Law, Economics, & Organization 7:24–52.
Hong, H., and M. Kacperczyk. 2009. The price of sin: The effects of social norms on markets. Journal of Financial
Economics 93:15–36.
Innes, R. D. 1990. Limited liability and incentive contracting with ex-ante action choices. Journal of Economic
Theory 52:45–67.
903

Jensen, M. C., and W. H. Meckling. 1976. Theory of the firm: Managerial behavior, agency costs and ownership
structure. Journal of Financial Economics 3:305–60.
Krüger, P. 2015. Corporate goodness and shareholder wealth. Journal of Financial Economics 115:304–29.
Kotchen, Matthew J. 2006. Green Markets and Private Provision of Public Goods. Journal of Political
Economy 114:816–834.
PRI Principles for Responsible Investment. 2012. Integrating ESG issues into executive pay. The United Nations
Principles for Responsible Investment.

Ross, S. A. 1973. The economic theory of agency: The principal’s problem. American Economic Review
63:134–39.
Tirole, J. 2001. Corporate governance. Econometrica 69:1–35.
904

Copyright of Review of Financial Studies is the property of Oxford University Press / USA
and its content may not be copied or emailed to multiple sites or posted to a listserv without
the copyright holder's express written permission. However, users may print, download, or
email articles for individual use.

Investing For Impact Article

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Investing For Impact Article

Uploaded by

Copyright:

Available Formats

Investing for Impact

Shaun William Davies

Downloaded from https://academic.oup.com/rfs/article-abstract/32/3/864/5039908 by Adam Ellsworth, Adam Ellsworth on 07 February 2019

Organizations in the for-profit sector increasingly rely on joint financing by

[16:29 4/2/2019 RFS-OP-REVF180069.tex] Page: 864 864–904

Downloaded from https://academic.oup.com/rfs/article-abstract/32/3/864/5039908 by Adam Ellsworth, Adam Ellsworth on 07 February 2019

[16:29 4/2/2019 RFS-OP-REVF180069.tex] Page: 865 864–904

Downloaded from https://academic.oup.com/rfs/article-abstract/32/3/864/5039908 by Adam Ellsworth, Adam Ellsworth on 07 February 2019

[16:29 4/2/2019 RFS-OP-REVF180069.tex] Page: 866 864–904

Downloaded from https://academic.oup.com/rfs/article-abstract/32/3/864/5039908 by Adam Ellsworth, Adam Ellsworth on 07 February 2019

[16:29 4/2/2019 RFS-OP-REVF180069.tex] Page: 867 864–904

Downloaded from https://academic.oup.com/rfs/article-abstract/32/3/864/5039908 by Adam Ellsworth, Adam Ellsworth on 07 February 2019

[16:29 4/2/2019 RFS-OP-REVF180069.tex] Page: 868 864–904

investors. Hart and Zingales (Forthcoming) prescribe corporate governance and

Downloaded from https://academic.oup.com/rfs/article-abstract/32/3/864/5039908 by Adam Ellsworth, Adam Ellsworth on 07 February 2019

[16:29 4/2/2019 RFS-OP-REVF180069.tex] Page: 869 864–904

where f (·) is a twice continuously differentiable function with 1 ≥ f (ax ) ≥ 0,

Downloaded from https://academic.oup.com/rfs/article-abstract/32/3/864/5039908 by Adam Ellsworth, Adam Ellsworth on 07 February 2019

where g(·) is, as well, a twice continuously differentiable function with

lim f (1−a) = ∞ and lim g (a) = ∞.

There are three agents: a profit-motivated investor (denoted by p), a social

[16:29 4/2/2019 RFS-OP-REVF180069.tex] Page: 870 864–904

such as through improved monitoring.12 While profit-oriented agents benefit

Downloaded from https://academic.oup.com/rfs/article-abstract/32/3/864/5039908 by Adam Ellsworth, Adam Ellsworth on 07 February 2019

The manager’s reservation payoff is γ ≥ 0, and we restrict attention to the range

[16:29 4/2/2019 RFS-OP-REVF180069.tex] Page: 871 864–904

Downloaded from https://academic.oup.com/rfs/article-abstract/32/3/864/5039908 by Adam Ellsworth, Adam Ellsworth on 07 February 2019

[16:29 4/2/2019 RFS-OP-REVF180069.tex] Page: 872 864–904

Downloaded from https://academic.oup.com/rfs/article-abstract/32/3/864/5039908 by Adam Ellsworth, Adam Ellsworth on 07 February 2019

Contracts are offered at date 0. After agreeing to the contract, capital is

1.1 Preliminary analysis

maxf (1−a)Xz−I (7)

s.t. w = w(a), (7.1)

0 ≤ f (1−a)Xw +g(a)ψm −γ , (7.2)

where (7.1) is the manager’s incentive compatibility constraint, (7.2) is the

[16:29 4/2/2019 RFS-OP-REVF180069.tex] Page: 873 864–904

whereas for γ ≥ γo , the optimal level of social attention ao is pinned down by

Downloaded from https://academic.oup.com/rfs/article-abstract/32/3/864/5039908 by Adam Ellsworth, Adam Ellsworth on 07 February 2019

f (1−ao )Xw(ao )+g(ao )ψm −γ = 0. (9)

[16:29 4/2/2019 RFS-OP-REVF180069.tex] Page: 874 864–904

Downloaded from https://academic.oup.com/rfs/article-abstract/32/3/864/5039908 by Adam Ellsworth, Adam Ellsworth on 07 February 2019

0 ≤ w +y +z ≤ 1, and Ω ≥ {0,0,0}, (12.3)

[16:29 4/2/2019 RFS-OP-REVF180069.tex] Page: 875 864–904

Downloaded from https://academic.oup.com/rfs/article-abstract/32/3/864/5039908 by Adam Ellsworth, Adam Ellsworth on 07 February 2019

2. The Basic Model of Impact Investing

[16:29 4/2/2019 RFS-OP-REVF180069.tex] Page: 876 864–904

Downloaded from https://academic.oup.com/rfs/article-abstract/32/3/864/5039908 by Adam Ellsworth, Adam Ellsworth on 07 February 2019

Lemma 4. Renegotiation is only possible by increasing the manager’s wage.

Lemma 5. There is a unique threshold y(a) given by,

1. The threshold y(a) is increasing in a.

Lemma 5 illustrates a crucial point. When the desired level of social

[16:29 4/2/2019 RFS-OP-REVF180069.tex] Page: 877 864–904

renegotiation-proof.19 In this case, increasing y and subsequently lowering z

Downloaded from https://academic.oup.com/rfs/article-abstract/32/3/864/5039908 by Adam Ellsworth, Adam Ellsworth on 07 February 2019

0 ≤ f (1−a)Xw +g(a)ψm −γ , (14.2)

1 = w +y +z, and Ω ≥ (0,0,0), (14.3)

[16:29 4/2/2019 RFS-OP-REVF180069.tex] Page: 878 864–904

Downloaded from https://academic.oup.com/rfs/article-abstract/32/3/864/5039908 by Adam Ellsworth, Adam Ellsworth on 07 February 2019

[16:29 4/2/2019 RFS-OP-REVF180069.tex] Page: 879 864–904

reservation value at which the manager’s participation constraint just binds.

Corollary 1. If the manager’s participation constraint is slack and ψs > ψs ,

Downloaded from https://academic.oup.com/rfs/article-abstract/32/3/864/5039908 by Adam Ellsworth, Adam Ellsworth on 07 February 2019

[16:29 4/2/2019 RFS-OP-REVF180069.tex] Page: 880 864–904

From Proposition 1, the impact investor’s equity stake is given by