The Organization of Social Enterprises:

Transacting versus Giving

Ofer Eldar∗

Preliminary and incomplete; Please do not cite or circulate.

Abstract

Traditional models of organizational choice focus on the distinction between the for-profitand non-profit forms. These models ignore a critical distinction between organizations thattransfer subsidies on behalf of donors to beneficiaries, and social enterprises, such as micro-finance institutions and fair trade firms, that commit to transacting with different classes ofdisadvantaged groups. I present a model where entrepreneurs receive subsidies and have achoice between (1) giving to the beneficiaries, and (2) forming a social enterprise that transactswith them, as well as choose to incorporate as a for-profit or a non-profit. I show that socialenterprises have financial and reputational incentives to measure their beneficiaries’ abilities andtailor subsidies to their particular needs. Their effectiveness in utilizing subsidies explains whysuch enterprises have been particularly effective in addressing social missions, such as increasingaccess to capital or improving employment opportunities. Entrepreneurs choose to form socialenterprises when the variance in beneficiaries’ abilities is higher, and therefore the benefits ofmeasurement are greater. Moreover, when the variance in abilities is very high, entrepreneurswill choose to form a for-profit social enterprise, because when the incentives to measure aresufficiently strong, the non-profit form as a commitment device to use subsidies effectively isredundant. The analysis further considers the effects of different subsidy policies and reformproposals for regulating social enterprises. In particular, it shows that conditional subsidies tosocial entrepreneurs, but not tax exemptions, are likely to increase social welfare.

∗Duke University School of Law. 210 Science Drive, PO Box 90360, Durham, NC 27708–0360. Email:eldar@law.duke.edu. I thank David Baron, Jean de Bettignies, John de Figueiredo, Shmuel Leshem, LorenzoMagnolfi, Anup Malani, David Robinson, Larry Samuelson, Aron Tobias, and participants at the Conferenceon the Economics of the Social Sector at Chicago Booth Business School, and the annual meeting of 17thAnnual Strategy & the Business Environment Conference, Duke Fuqua School of Business, for valuablecomments and suggestions. [Additional Acknowledgments to be added].

1 Introduction

In recent years, many forms of hybrid organizations that combine profit and social missionshave emerged to address a variety of social problems. In particular, social enterprises, such asmicrofinance institutions and fair trade social enterprises, are increasingly influential in addressinga variety of development missions, such as facilitating access to capital, improving employing op-portunities and productivity and enhancing consumer welfare (see Dees, 2008; Eldar, 2017). Theseorganizations signal a shift from the traditional model of pursuing development through donativefunding to more entrepreneurial organizations that transact directly with their beneficiaries. Forexample, microfinance institutions lend to poor people, fair trade social enterprises source productsfrom low-income farmers and work-integration firms employ low-income people.

The evolution of these firms presents a puzzle for theories of the firm. First, it is not clear whyaltruistic entrepreneurs would opt to form as for-profits, as many social enterprises do. Traditionaltheories claim that the non-profit form is essential for entrepreneurs to commit to using donativefunding to benefit third parties, in circumstances where monitoring how such funds are used iscostly (Hansmann, 1980; Glaeser & Shleifer, 2001). By subjecting the entrepreneur to the non-distribution constraint, the non-profit form mitigates her profit incentives to expropriate donativefunds, and hence makes it more likely that the entrepreneur will deliver the donation to the ultimatebeneficiaries. A classic example is Oxfam that receives donations and distributes them to the poor.

While these models correctly identify the critical function of the non-distribution constraint inpreventing expropriation, they do not address the issue of subsidy-effectiveness. The entrepreneursthat form donative organizations may be altruistic, but they are not always able to alleviate povertyor fix other social problems simply by transferring subsidies to the relevant beneficiaries. Theexperience of many donative organizations, particularly NGOs and development agencies, in theirquest to alleviate poverty and pursue development missions suggests that the effectiveness of theirprograms is often limited (Easterly, 2001). This problem is particularly acute in the context ofcomplex missions, such as increasing access to capital or improving employment opportunities(Eldar, 2017).

Second, since many social enterprises are formed as non-profits (e.g., non-profit microfinance in-stitutions), the emergence of social enterprises calls into question the traditional distinction betweenfor-profits and non-profits. If microfinance institutions or fair trade firms can form as for-profitsas well as non-profits, and both serve a similar role, then arguably this distinction is of limitedrelevance. Many of the most influential social enterprises are non-profits, for example, ASA andBRAC, two non-profit microfinance institutions with millions of customers. But many similarlylarge and influential microfinance institutions, such as Compartamos and BancoSol, are for-profits,

1

and they have had a similar impact on increasing access to capital in developing countries.In this paper, I offer a theory of social enterprises that focuses on the transactional relationship

with the social enterprises’ beneficiaries, rather than their organizational form. Social enterprises, Iargue, address information asymmetries with respect to their beneficiaries’ attributes (e.g., produc-tivity or creditworthiness). Social enterprises that transact with their beneficiaries have the abilityand incentives to measure their beneficiaries’ attributes. This information enables them to tailorthe subsidies they allocate to their beneficiaries effectively. For example, the training programsof fair trade and work-integration firms are geared towards improving the productivity of theirlow-income workers. The measurement function thus makes social enterprises relatively effective inaddressing development missions, such as increasing employment and access to capital. In contrast,donative organizations that do not transact with their beneficiaries have limited information abouttheir needs and attributes, and as a result their programs tend to be ineffective. This informationis particularly important in the context of complex development goals where subsidies’ effectivenesscan be critical to the success of the program. For example, an organization that wishes to increaseaccess to credit or improve employment rates may need to tailor price discounts and business or em-ployment training to the specific needs of different individuals. In these circumstances, transactingwith beneficiaries works better than simply giving to them.

The measurement function of social enterprise is similar in spirit to the metering function ofcorporations discussed in the well-known paper of Alchian & Demsetz (1972). Alchian & Demsetzargue that for-profit corporations allocate capital efficiently because they continuously “meter” or“measure” their input productivity. In other words, corporations transact in competitive marketswith creditors, employees, suppliers, etc. On the other hand, Alchian & Demsetz argue that non-profits shirk on quality because the entrepreneurs who control the organization have no financialincentives to “meter” or measure input productivity, for example, the quality of a training program.

My theory modifies this analysis in several ways. First, measurement is primarily related tothe transactional relationship between firms and their counter-parties rather than the legal form.Non-profit social enterprises, similar to for-profit ones, also perform a measurement function whenthey transact with their beneficiaries because they have financial (as well as altruistic) incentivesto allocate capital, including donative capital, efficiently. Thus both for-profit and non-profit socialenterprises that transact with their beneficiaries can be effective in pursuing development missions.

Second, the problem with donative organizations is not their lack of goodwill or incentives toimprove quality, but rather an informational problem associated with any organization, includingfor-profits. Following Hansmann (1980), Glaeser & Shleifer (2000) show that non-profits may ac-tually produce higher quality than for-profits, because the non-distribution constraint mitigatesentrepreneurs’ incentives to cut costs by shirking on quality. The major problem with this model

2

is that they assume that entrepreneurs can simply increase the quality of their services, includingcomplex tasks, such as alleviating poverty, at will. However, entrepreneurs may have limited in-formation on how to accomplish such complex tasks. To address this issue, quality in my model ismeasured in terms of the effectiveness of the subsidy, which depends on tailoring it to the needs ofthe beneficiaries. When the organization’s task is to improve employment rates and productivity,it is difficult to know what general policy or training would be effective in accomplishing this objec-tive, and specific policies need to be tailored to the particular attributes of different beneficiaries.Donative organizations have no market platform through which they can observe or measure theirbeneficiaries’ attributes. They are thus more likely to make mistakes in allocating subsidies tobeneficiaries by providing too much, too little, or even the wrong type of subsidy altogether.

This type of informational problem derives not from the non-profit form (as Alchian & Demsetz(1972) seem to suggest), but because donative organizations engage in giving. Therefore, thisproblem applies also to for-profits that engage in corporate charity without transacting with theirbeneficiaries. For example, for-profit corporations, such as Google, have large segments that engagein philanthropy. Despite enthusiasm for corporate philanthropy, it has, like donative organizations,a relatively poor record in pursuing complex social goals (Eldar, 2017). In this respect, I makea distinction between for-profit social enterprises that transact with their beneficiaries, and thosethat simply transfer subsidies to third parties through their corporate charity or corporate socialresponsibility initiatives. In this way, I hope to offer a broader distinction between organizationsthat pursue social goals by transacting with their beneficiaries and those that engage in giving.

Third, although my theory focuses on the relationship between the firm and its beneficiaries, italso explains entrepreneurs’ choice between the for-profit and non-profit forms, and the implicationsof such choice on the effectiveness of subsidies. In particular, it explains why under some conditionseven socially motivated entrepreneurs will choose to incorporate as for-profits. On one hand, whenthe variance in beneficiaries’ abilities is relatively low, the non-profit form may actually induce firmsto transact with their beneficiaries (and tailor subsidies to their needs) rather than just transferringsubsidies to them. The intuition is that consistent with Shleifer & Glaeser (2000), the non-profitentrepreneur produces higher quality, which in my model is measured in terms of the effectiveuse of subsidies. However, when the variance in abilities is high, there is a higher risk that theentrepreneur will make mistakes in allocating subsidies to beneficiaries. Therefore, her incentives(both financial and altruistic) to measure beneficiaries’ abilities and needs are strong enough tomake the entrepreneur use all available donations or subsidies to transact with disadvantagedgroups (rather than just giving them the subsidies). In these circumstances, the non-profit formas a commitment device to use subsidies effectively becomes redundant, and does not improve theincentives of the entrepreneur to pursue social missions effectively.

3

I emphasize that my analysis admits a role for each type of organization, whether it is acorporation that pursues corporate charity, a donative organization or a social enterprise, eitherfor-profit or non-profit. For example, donative organizations and corporations that engage incorporate charity can be effective if there is little need for measurement, that is, when the variancein beneficiaries’ abilities is low. In fact, when we examine the evolution of many social enterprises,a certain pattern emerges. Many of these organizations start out as donative organizations thatcater to the neediest of society. Donative organizations operate in an environment where thebeneficiaries have low variance and low mean in abilities. As the abilities of some of them increase,so does the variance in their abilities, in which case there is a greater need to tailor the subsidiesto their individual attributes. At this stage, a non-profit social enterprise that makes loans tobeneficiaries or employs them becomes an efficient form for pursuing their development. Thatbusiness ultimately converts to a for-profit social enterprise, and over time, when the beneficiariesbecome fully competitive, the firm becomes largely equivalent to a standard corporation. At thatstage, it may still engage in corporate charity as most corporations do.

This paper is related to literature on corporate social responsibility (“CSR”), corporate charity,and social entrepreneurship. Whereas most accounts of CSR do not differentiate between socialenterprise and CSR, I draw an analytical distinction between social enterprises that transactswith their beneficiaries, and CSR or corporate charity that generally involves giving. The mainreason is that most accounts of CSR (e.g., Baron, 2007; Benabou & Tirole, 2010) do not take intoconsideration the effectiveness of subsidies or giving, but simply assume that entrepreneurs, donorsand investors can create public benefits at will by spending money or exerting efforts. De Bettignies& Robinson (2015) demonstrate that CSR may be harmful to social welfare when it is coupled withlobbying to pass suboptimal level of regulation. However, they also assume that the CSR policiesthemselves increase social welfare or public benefits, and therefore their model does not apply tocircumstances where it is costly to measure the public benefits of specific actions. Recently, Besley& Ghatak (2017) acknowledge the difficulty for founders in measuring public benefits, but theynonetheless assume that this task can be delegated to socially-motivated managers who presumablyknow how to increase the quality of such benefits. In practice, however, even socially motivatedentrepreneurs often fail in pursuing social missions due to lack of knowledge and information. Thecontribution of this article is to identify an organizational form that gives the entrepreneurs boththe ability to acquire such information, and incentives to pursue social missions effectively.

This article is also related to a vast literature on development economics, particularly studies ofthe effectiveness of aid programs (for reviews, see Easterly, 2007; Easterly, 2009; Banerjee & Duflo,2009). Despite evidence that giving in the development context is often wasteful, some experimentalstudies have suggested that due to behavioral biases or lack of knowledge, beneficiaries are more

4

likely to use goods (such as fertilizers or bed-nets) when they are given for free rather than sold forvery low prices. These studies, however, fail to compare giving to true social enterprises that havefinancial and altruistic incentives to measure the needs of particular beneficiaries as well as educatethe beneficiaries about their products. Nor do these studies distinguish between circumstanceswhen there is a high variance in the beneficiaries’ abilities (in which case social enterprise mightbe more effective) and low variance in these abilities (in which case a donative organization mightbe preferable). Accordingly, the measurement theory may inform empirical studies that evaluatedifferent institutional mechanisms for pursuing development missions.

This article is structured as follows. Section 2 defines the different forms of organizations andlays out the basic set-up of the model. Sections 3 to 6 each deals with a different type of organizationand the different factors that affect the entrepreneur’s choice of the organizational form. I start theanalysis with corporations that engage in corporate charity mainly because it is the simplest formof organization that engages in transferring or giving subsidies to beneficiaries. I then comparesuch corporations to for-profit social enterprises. In the following sections, I consider the effectof the non-distribution constraint by analyzing donative organizations (which are invariably non-profits) and non-profit social enterprises. Section 7 considers the entrepreneur’s choice among theseorganizational forms. Sections 8 and 9 discuss different regulatory measures to address potentialabuses of social enterprise organizations, and the effect of tax exemptions and subsidy policy.Section 10 considers several extensions. Section 11 concludes.

2 A Taxonomy of Organizations and Basic Set-up

The analysis focuses on organizations that receive a subsidy or a donation, and are expectedto deliver or utilize it for the benefit of third party beneficiaries.1 As shown in Table 1, suchorganizations may adopt one of four models which differ along two main parameters: (1) giving tobeneficiaries versus transacting with them, and (2) for-profits versus non-profits.

1Note that as a result, this theory has no relevance for analyzing organizational choice that does notinvolve such a subsidy. Following Hansmann (1980), Glaeser & Shleifer (2000) and Choi (2015) apply theirmodel of organizational choice to both circumstances where (1) organizations receive a subsidy from donors(who are viewed as the firm’s customers) and deliver it to third party beneficiaries, and (2) when consumershave difficulty in observing the quality of the services (such as educational or health services) provided tothem. My paper deals only with the first case, whereas the second case does not involve a firm that receivesa subsidy.

5

Table 1: Taxonomy of OrganizationsGiving Transacting

For-profits Corporate Charity For-profit Social EnterprisesNon-profits Donative Organizations Non-profit Social Enterprises

Corporations that engage in corporate charity are for-profits that distribute subsidies to thebeneficiaries. Donative organizations are non-profits that distribute subsidies to the beneficiaries.Social enterprises, whether for-profits or non-profits, transact with their beneficiaries as patronsand also distribute a subsidy to them.

The basic set-up of the model is as follows. The firm has a single entrepreneur (representingmanagers and shareholders). The entrepreneur runs a business, for example a bakery, a restaurant,or an agricultural farm. The entrepreneur also faces reputational pressure to help the community,for example by providing training and other assistance to low-income workers with no employmenttrack record. The entrepreneur may decide to help the disadvantaged by (1) giving them some ofher earnings as a cash transfer or by organizing a training program for them, and/or (2) decidingto employ them and train them in her business. In the main model, I will focus on beneficiarieswho suffer from systemic unemployment and need professional training, but the analysis appliesto other disadvantaged groups, such as low-income borrowers or consumers. The entrepreneur alsohas a choice to form as either (1) a for-profit or (2) a non-profit.2

I will consider the entrepreneur’s utility from forming each of the four types of organization.The entrepreneur chooses to assist n beneficiaries, where n ≥ 0, and delivers a subsidy to eachbeneficiary i in the amount of d(i), where i is defined over a continuum for convenience.3 In allthese cases, the firm receives a subsidy or donation from some source to fund the costs of thesedisbursals. For the purposes of this paper, I am indifferent as to the source of the subsidy. Forexample, consumers may pay a premium over the price that reflects the market price of products ofequivalent quality. Alternatively, donors may fund the costs of these disbursals.4 I will assume thatthe subsidy amount D is an increasing continuous function in n, D(n) is concave, and D(0) = 0.For simplicity, I assume that the donation function is exogenous, but this can be understood aspart of a decision process by a donor who can only contract for the number of beneficiaries theorganization will assist, but does not observe how effectively the donation is used.5

2Note that in principle donative organizations also form businesses or hold stakes in businesses to generaterevenues while focusing their charitable activities on giving to the beneficiaries. An example is a privatefoundation that has revenues from its investment activities.

3[Cite papers that use continuum]4The donation may even flow from the entrepreneur herself to the firm.5The donor’s utility function may be given by Y − D(n) + F (n), where F (n) is an increasing concave

6

In addition, I assume that standard workers who are not disadvantaged have the same abilities,ā. The standard workers’ abilities (i.e., ā) are observable by firms. By contrast, each disadvantagedworker i has a lower level of ability, a(i) ∈ [0, a) ∀i ∈ R+. Disadvantaged workers may receivetraining which increases their abilities up to ā. To reach ā, they must receive the appropriate amountof training, which I quantify at ā − a(i). Any training above this amount is a waste. Of course,they must also receive the appropriate type of training (and not just the appropriate quantity),but for simplicity, I focus on quantity only. A critical assumption is that the true abilities of thedisadvantaged cannot be observed by the entrepreneur. For simplicity, I further assume that thesize of the firm proxied by the number of employees (w) and the returns on human capital (r) arefixed by market processes that are outside the scope of my inquiry. Firm revenues equal w× r× a,where r×a is the revenue per worker the firm employs, and w is a fixed number of workers employedby the firm. Thus, the firm’s revenues depend linearly on the workers’ abilities (a), and the revenuesof a “standard” commercial firm that employs only standard workers are equal to w × r × ā, i.e.,a constant. I further assume r ≥ 1 to limit the inquiry to viable productive firms that generatenon-negative returns on capital.

I start the analysis with the simplest form of hybrid organization, namely a for-profit thatgives assistance to some beneficiaries through its corporate charity arm. It is the simplest formin the sense that the entrepreneur is not subject to the non-distribution constraint, and does nottransact with the beneficiaries. I then consider for-profit social enterprises that transact with thebeneficiaries. Finally, I consider donative organizations and non-profit social enterprises, whichsubject entrepreneurs to a constraint on their ability to distribute earnings to themselves.

3 Corporate Charity

Suppose the entrepreneur chooses to help the disadvantaged through corporate charity. Thebasic idea of corporate charity is that the firm makes a disbursal or transfer to the beneficiaries,which may take the form of a training program. The firm only employs standard workers and doesnot employ the disadvantaged. The entrepreneur makes a decision about the number of beneficiariesto whom she will allocate a subsidy to (n) and the amount of subsidies (d(i)) she allocates to eachbeneficiary i:

maxn,d(i)

w × r × ā+D(n)−∫ n

0d(i)di− γ × E

[∫ n0

(d(i)− (ā− a(i)))2 di]

(1)

function that reflects the donor’s utility from making charitable contribution. The first order conditionsimply that F ′(n) = D′(n), and therefore, F ′′(n) = D′′(n) < 0.

7

where the expression,∫ n0 (d(i)− (ā− a(i)))

2 di, is a measure of the effectiveness of the alloca-tions to the beneficiaries, and E denotes the expectation over the abilities of the beneficiaries.

If the allocations to beneficiaries (i.e., d(i)) are too high or too low (i.e., d(i) 6= ā − a(i)),the entrepreneur’s utility is decreased. γ is a parameter that measures the extent to which theentrepreneur dislikes ineffectiveness. This dislike may reflect her own altruism or reputational costsfor failing to help the community. In contrast, the traditional models of non-profits, primarilyGlaeser & Shleifer (2000), assume that subject to the costs of making allocations of higher quantityto beneficiaries, the entrepreneur can increase her utility infinitely by increasing the amount ofallocations to the beneficiaries, without regard to whether or not the beneficiaries actually needthese subsidies.6

In my model, the entrepreneur clearly has no altruistic (or other) incentive to increase d(i)beyond the amount that would help the beneficiary to reach the competitive level. This is consistentwith the increasing concerns in recent years that donative funds should be used effectively tohelp disadvantaged groups gain competitive abilities, rather than spent without regard to theireffectiveness. The second difference from traditional models is that the entrepreneur also does notknow what allocations the beneficiaries need because she cannot observe the beneficiaries’ abilities.But first, I will consider the entrepreneur’s choice without information asymmetries.

A. No Information Asymmetries (Abilities Are Observable)

If the entrepreneur observes the abilities of the beneficiaries, she will choose to allocate to eachbeneficiary an amount that takes the beneficiary’s needs into account. For any given n, we can fixi = i′ and find the optimal allocation d(i′). The solution for every i is the same,7 i.e.,

d∗(i) = ā− a(i)− 12γ ∀i ∈ [0, n]. (2)

The last term 12γ reflects the entrepreneur’s dis-utility from paying, and this dis-utility vanishes asγ →∞. Using equations (1) and (2), the ex ante utility of the entrepreneur is then

6The entrepreneur in Glaeser & Shleifer (2000) chooses q, the quantity of the subsidy distributed to thebeneficiaries, and her utility is essentially given by P−c(q)+γ×q, where P is a constant which is independentof q because quality is not observed when the consumer buys the products, c(q) is a convex cost function,and γ is a parameter that measures the entrepreneur’s altruism.

7To see why equation (2) holds, we can assume that the entrepreneur decides to set d∗(i) = ā−a(i)− 12γ−εfor j workers, and the amount in equation (2) for all other beneficiaries, where ε > 0. The utility of theentrepreneur in equation (3) would then be increased by j × ε (because of the savings in disbursal to thebeneficiaries), but it would be decreased by j

(ε+ ε2

).

8

r × w × ā+D(n∗)− n∗ (ā− E(a)) + n∗

4γ , (3)

The choice of n depends on the marginal utility of increasing n. Accordingly,

D′(n∗) = d∗(n∗) + γ × E (d∗(n∗)− (ā− a(n∗)))2 , (4)

If we plug in d∗, this condition becomes:

D′(n∗) = ā− E(a)− 14γ . (5)

B. With Information Asymmetries (Abilities Are Not Observable)

For any given n, we can fix i = i′ and take the derivative with respect to d(i′). The solution forevery i′ is the same,

d∗(i) = ā− E(a)− 12γ ∀i ∈ [0, n] (6)

The difference from equation (2) is that the entrepreneur does not know the true a. If we plug ind∗ in equation (4) this condition now becomes:

D′(n∗) = ā− E(a) + γV ar(a)− 14γ , (7)

The ex ante utility of the entrepreneur is,

r × w × ā+D(n∗)− n∗ (ā− E(a))− n∗(γV ar(a)− 14γ

), (8)

where n∗ is defined by equation (7). In comparing equation (8) with equation (3), two maindifferences emerge. First, the key point is that the entrepreneur suffers the costs of uncertainty orineffectiveness, i.e., n∗ × γ × V ar(a). These costs reflect the risk that the entrepreneur underpaysdisadvantaged workers whose abilities are lower than E(a), and overpays those with abilities higherthan E(a). These costs may reflect the entrepreneur’s own dissatisfaction from failing to provideadequate training to low-income people or pressures from donors (and people at large) who areunhappy with the impact of the charity program. Note that the larger γ is, the greater the costs.

9

Second, without information asymmetries the entrepreneur would choose to give assistance toa larger number of beneficiaries because the right-hand-side in equation (5) is smaller than theright-hand-side in (7) and D(n) is concave. Intuitively, the entrepreneur will be deterred fromhelping more beneficiaries when she does not know if the assistance will be effective or not. Thissuggests that charity is not necessarily constrained by lack of altruism, but rather by a concern thatdonative funding will not be utilized effectively. This is summarized in the following proposition:

Proposition 1. When there are information asymmetries with respect to beneficiaries’ abilities,the utility of entrepreneurs that care about the effectiveness of their charitable program (γ is large)will be lower and they will choose to assist fewer beneficiaries (n is small).

4 For-Profit Social Enterprise

To address the costs of uncertainty, the entrepreneur may instead form a social enterprise. Socialenterprises employ disadvantaged workers and allocate subsidies to them through the transactionalrelationship rather than by allocating a subsidy to the workers as external beneficiaries. Theadvantage of a social enterprise is that when the entrepreneur interacts with the beneficiaries,she is capable of measuring their true abilities, and as I show below, also has incentives to do sounder certain conditions. The firm may expend a cost C(m) on measuring the abilities of the mdisadvantaged workers it employs. C(m) is an increasing convex function, and C(0) = 0. If theentrepreneur expends C(m) at t = 0, she can observe the true abilities a(i) for those m workers.Note that setting d(i) such that ā − a(i) eliminates reputational costs and increases the revenuesfor the firm.

The entrepreneur does not necessarily have to measure the attributes of all the beneficiaries itemploys. Thus, it may also employ k beneficiaries, but decide not to expend the costs of measure-ment; therefore with respect to k workers the entrepreneur does not observe their true abilities. Thefirm will employ m+ k disadvantaged workers and w−m− k standard workers, where m+ k ≤ w.The firm may still make a disbursal of subsidy to the other g disadvantaged workers that it doesnot employ through a corporate charity program, such that m + k + g = n, the total number ofbeneficiaries.

At t = 0, the entrepreneur chooses m, k, and g. At t = 1, the entrepreneur finds out the trueabilities of m workers, and sets the allocations dm(i), dk(i) and dg(i), which respectively refer to

• the allocations to m beneficiaries she employs after measuring and observing their abilities,

• the allocation to k beneficiaries she employs but whose abilities she does not measure orobserve, and

10

• the allocations to g beneficiaries through the firm’s corporate charity program, where again,the entrepreneur does not observe the beneficiaries’ abilities.

Note that D(n) = D(m+ k+ g) so that the firm faces a choice as to how to allocate each marginaldonation it receives. The social entrepreneur’s problem is:

maxm,k,g,d(i)

r(w −m− k)ā+ r × E∫ m

0(a(i) + dm(i)) di+ r × E

∫ k0

(a(i) + dk(i)

)di+ (9)

D(n)−∫ m

0dm(i)di−

∫ k0dk(i)di−

∫ g0dg(i)di−

C(m)− γ × E[∫ m

0(dm(i)− (ā− a(i)))2 di

]−

γ × E[∫ k

0

(dk(i)− (ā− a(i))

)2di

]− γ × E

[∫ g0

(dg(i)− (ā− a(i)))2 di]s.t.

a(i) + d(i) ≤ a ∀i , and n = m+ k + g.

A. No Information Asymmetries (Abilities Are Observable)

Proposition 2. Without information asymmetries (i.e., when a(i) is observable for every i), theentrepreneur will never form a social enterprise by transacting with the beneficiaries (i.e., k∗ = 0and m∗ = 0).

The proof is in the Appendix, but the intuition is simple. When there are no information asym-metries, there are no benefits to measurement. Social enterprises evolve mainly in the developmentcontext where disadvantaged groups include individuals with varying abilities, for example, withrespect to their productivity or creditworthiness. In the context of other social problems, such asassistance to homeless people, information asymmetries are not particularly potent, and thereforethe need for tailoring is less acute. When the beneficiaries have relatively homogeneous levels ofabilities and needs, there is no need for measurement. In this case, we are more likely to ob-serve policies, such as training programs or credit subsidies, rather than work-integration firms ormicrofinance institutions.

11

B. With Information Asymmetries (Abilities Are Not Observable)

In this sub-section, I show that when there are information asymmetries, an entrepreneur thatcommits to transacting with the beneficiaries has both financial and altruistic incentives to (a)tailor subsidies to their needs, and (b) measure or gather information on the beneficiaries’ abilities,if the costs of measurement are not too high.

Proposition 3. If the social entrepreneur expends C(m) on measurement at t=0, she has financialand altruistic incentives to tailor the subsidies to the needs of these w beneficiaries at t=1, i.e.,dm(i) = ā− a(i) ∀i ∈ [0,m].

Proof. Suppose the entrepreneur decides to expend C(m) and sets the allocations such that dm′(j) =ā − a(j) − ε for the j’th worker, and dm(i) = ā − a(i), ∀i 6= j, where ε > 0. It is clear that shecan increase her utility by setting dm′(j) = ā − a(j). The reason is that increasing dm′(j) by εincreases her utility by (r − 1) ε + b × ε2, the additional profit from setting the subsidy to theright amount, and the reputational benefits from additional effectiveness. Alternatively, settingdm′(j) = ā − a(j) + ε is clearly disadvantageous because it doesn’t increase profits, but will only

result in reputational costs from ineffectiveness.

The importance of Proposition 3 is that the entrepreneur has profit incentives to tailor thesubsidies to the needs of the beneficiaries. If she deviates from dm(i) = ā− a(i), the entrepreneurloses revenues because the beneficiary would perform poorly.

Proposition 4. The social entrepreneur that commits to transacting with the beneficiaries hasfinancial and altruistic incentives to measure the abilities of the beneficiaries, as long as the marginalcosts of measurement (C(m)) are lower than the marginal revenues from increasing the productivityof the beneficiaries and reputational gains from tailoring the subsidies to their needs.

The proof to Proposition 4 is laid out in the Appendix. If the entrepreneur has already com-mitted to employing the beneficiaries, then her revenue depends on their productivity. Withoutmeasuring their abilities, the entrepreneur does not know what subsidy to allocate to them, andtherefore she will give too little assistance to at least some of them, and too much assistance toothers. This will result in loss of revenues. Moreover, because the allocations will not be tailored tothe needs of each beneficiary, the entrepreneur will also suffer reputational loss due to the mismatchbetween what the beneficiaries need and what they get. The financial and altruistic incentives tomeasure are stronger when (a) many beneficiaries’ abilities are below the mean abilities, (b) thevariance of beneficiaries’ abilities is high and she cares greatly about effectiveness (large b), and (c)the business project yields high returns. Proposition 4 shows that the entrepreneur has a financial

12

motive to measure. In practice, as summarized in Lemma 1, if measurement is too costly, the en-trepreneur will always prefer to help the beneficiaries through giving rather than transacting withthem.

Lemma 1. The entrepreneur will always prefer to distribute subsidies through its corporate charityprogram rather than transact with the beneficiaries without measuring their attributes. Thus, k∗ = 0.

Corollary 1. If the entrepreneur chooses to employ its beneficiaries she will expend the measure-ment costs and tailor subsidies to its beneficiaries’ abilities.

When the entrepreneur does not measure, she is likely to set the disbursals in a way thatdoes not help the beneficiary reach the capabilities of standard workers, and hence she will loserevenues. She will also suffer the reputational costs of ineffectiveness. If she is going to suffer thereputational costs of ineffectiveness anyway, she might as well employ standard workers and earnmore revenues, and confine her altruistic endeavors to giving only. In the following discussion, Iturn to the entrepreneur’s choice between transacting and giving.

(i) The Entrepreneur’s Choice Between Social Enterprise and Corporate Charity

The entrepreneur still faces a choice between a social enterprise that measures the beneficia-ries and corporate charity, i.e., choosing between increasing m or g. Whenever the entrepreneurincreases n by one unit, she has a choice between employing the beneficiary (and measuring) anddisbursing a subsidy to the beneficiary. Using Lemma 1, and letting n = n′,m = m′, and g = g′,we can write the ex ante utility of the entrepreneur as follows:

r × w × a+D(n′)− C(m′)− n′ (ā− E(a))− γ × g′V ar(a) + g′

4γ . (10)

If we increase n by increasing m, the marginal change in expected utility is D′(n′) − (ā −E(a)) − C ′(m′), whereas if the entrepreneur increases n by increasing g, the marginal change inutility is D′(n′)− (ā− E(a))− γV ar(a) + 14γ . Thus, the entrepreneur will increase n by increasingm only if (1) C ′(m′) < γV ar(a)− 14γ , and (2) C

′(m′) < D′(n)− (ā− E(a)), i.e., where the costs ofmeasurement are lower than the reputational costs of ineffectiveness, and lower than the marginaldonation received minus the expected disbursals to the beneficiaries. Otherwise, she would preferto increase n by increasing g. Note that D(n) is concave, and C(m) is convex, and that increasingm means also increasing n. At the optimum, there can be two results. The first result is one thatsatisfies what I will call the “effectiveness constraint”, beyond that point the measurement costs ofemploying more beneficiaries exceed the gains in effectiveness

(i.e., γV ar(a)− 14γ

), even if she had

more donations available.

13

C ′(m∗1) = γV ar(a)−1

4γ < D′(m∗1)− (ā− E(a)) . (11)

The second result satisfies the “measurement compensation constraint,” which means that theentrepreneur demands to keep some of the donations minus the disbursal transferred to the bene-ficiaries, as compensation for the costs she incurs for performing the measurement function.

C ′(m∗2) = D′(m∗2)− (ā− E(a)) < γV ar(a)−1

4γ . (12)

.These conditions show that social enterprise (corporate charity) is preferable when V ar(a) is

large (small), and hence the informational advantages of social enterprise due to its measurementfunction are relatively high (low), and when γ is high (low). This leads to the following proposition.

Proposition 5. An entrepreneur will prefer to transact with the beneficiaries if the variance inabilities is high and she cares about effectiveness (γ is high). Otherwise, the entrepreneur will preferto help the beneficiaries through its corporate charity program.

As depicted in figure 1, three main results may emerge depending on V ar(a) and γ:

Case 1. Pure Giving : most corporations engage only in giving by forming a corporate charityarm. A firm will choose to engage only in giving if the effectiveness constraint is notsatisfied for any m, i.e., C ′(0) > γ × V ar(a) − 14γ . The firm will then choose n

∗, suchthat equation (7) is satisfied. This means that if the variance in abilities is extremelylow or γ is small, the firm will prefer to focus only on corporate charity.

Case 2. Mixed Giving and Transacting: this result arises when at the optimum level m∗ =m∗1 as per equation (11). Note that when n = m∗1, the entrepreneur will continue toincrease n until n satisfies equation (7). Thus, the entrepreneur will mix between em-ploying the beneficiaries and giving to them through its corporate charity program. Thiscase is likely to occur where V ar(a) and γ are relatively low, such that the advantagesof measuring are relatively low, and the entrepreneur will continue to use available do-nations beyond the point at which measurement is effective.

Case 3. Pure Transacting: this result arises when at the optimum level m∗ = m∗2, per equation(12), i.e., the measurement compensation constraint is satisfied. This scenario is morelikely when V ar(a) and γ are large. As D(n) is concave, the marginal income fromdonations decreases with n, and therefore at high level of γV ar(a), the measurementcompensation constraint will bind.

14

Figure 1

5 A Donative Organization

A donative organization is a non-profit that only engages in giving or disbursing subsidies tothe beneficiaries. The difference from corporate charity is that the entrepreneur is subject to thenon-distribution constraint. This means that the entrepreneur can only distribute a fraction (V )of the organization’s earning to herself. V is an increasing concave function (following Glaeser &Shleifer, 2001; Choi, 2014), and V (x) < x, ∀x, V ′′(∞) = 0. For simplicity we assume that V = v,and 0 < v < 1. Note that a donative organization may have earnings. A donative organization thatengages primarily in distributing subsidies (such as, training disadvantaged groups) may in principleform a business or invest its capital to make profits, and thereby increase its donative capital as wellas the ability of the entrepreneur to make profit.8 Thus, the entrepreneur’s maximization problemis given by,

maxn,d(i)

v

(w × r × ā+D(n)−

∫ n0d(i)di

)− γ × E

[∫ n0

(d(i)− (ā− a(i)))2 di]. (13)

The solution in this case is:

d∗(i) = ā− E(a)− v2γ ∀i ∈ [0, n]. (14)

Note that since 0 < v < 1, d∗np, the disbursal by a non-profit donative organization in equation(14) is larger and closer to ā − E(a) than d∗fp, the disbursal made by a for-profit in equation (6).The reason is that the non-profit entrepreneur cares less about monetary incentives and more about

8An example is Housing Works which is an organization dedicated to helping homeless people who sufferfrom AIDS, but also operates coffee shops and second hand clothing stores in New York City.

15

effectiveness. Note however, that g∗np < g∗fp, because D(n) is concave, and

D′(n∗) = ā− E(a) + γvV ar(a)− v4γ , (15)

which is larger than the expression in equation (7) (as 0 < v < 1). Accordingly, a donativeorganization prefers to make fewer disbursals than a for-profit. This is summarized in the followingLemma.

Lemma 2. A donative organization will provide each beneficiary a larger disbursal than a for-profitthat engages in corporate charity (d∗np > d∗fp), but will make fewer disbursals to fewer beneficiaries(g∗np < g∗fp).

We may think that donative organizations have greater incentives to dispose of their moneybecause they are non-profits. But many non-profits are relatively slow to give out charity anddisburse donations. For example, large foundations have a great deal of untapped capital that isoften just invested in profitable projects.9 One reason may be information problems with respectto beneficiaries’ attributes, which creates uncertainty as to which charitable programs might beeffective. This suggests that in order to encourage foundations to utilize more of their funds forprojects with a social mission, there is a need to provide them with greater assurance about theeffectiveness of the program.10

6 A Non-Profit Social Enterprise

As above, I now consider the entrepreneur’s decision to employ the beneficiaries. The majordifference from the entrepreneur’s problem in equation (9) is that the entrepreneur is subject tothe non-distribution constraint. Lemma 1 generally applies in a similar way to the non-profit en-trepreneur with minor modifications. The basic idea is that the entrepreneur would prefer to choose

9[Cite]10The entrepreneur’s choice between a donative organization and corporate charity depends on her util-

ity. By comparing the ex ante utilities of the entrepreneur that forms a donative organization and onethat engages in corporate charity, it can be shown that the entrepreneur will choose a donative organiza-

tion over corporate charity if the following condition holds:(v − 1) r × w × ā︸ ︷︷ ︸ + v ×D (n∗np)−D (g∗fp)︸ ︷︷ ︸+

< 0 < 0(g∗fp − v × g∗np

)︸ ︷︷ ︸(ā− E(a)− 14γ

)+(g∗fp − g∗np

)γV ar(a)︸ ︷︷ ︸ > 0

> 0 > 0. Note that this expression does not nec-

essarily increase or decrease in V ar(a) and E(a) (as they also affect n). Naturally, the choice dependson several salient factors, such as the profit opportunities in the relevant industry (r), the availability ofdonations as determined by D(n), and the restrictiveness of v.

16

k∗ = 0, and hence the only meaningful choice is between employing the beneficiaries or transfer-ring subsidies to them through a donative organization. Accordingly, the simplified entrepreneur’sproblem (with k∗ = 0) is:

maxm,g,d(i)

v ×[r(w −m)ā+ r × E

∫ m0

(a(i) + dm(i)) di+D(n)−∫ m

0dmi di−

∫ g0dg(i)di− C(m)

](16)

−γ × E[∫ m

0(dm(i)− (ā− a(i)))2 di

]− γ × E

[∫ g0

(dg(i)− (ā− a(i)))2 di]s.t.

a(i) + d(i) ≤ a ∀i, and n = m+ g.

We again assume that n = n′,m = m′, and g = g′, and the ex ante utility of the entrepreneuris thus:

v ×(r × w × a+D(n′)− C(m′)− n′ (ā− E(a))

)− γ × g′V ar(a) + v

2g′

4γ . (17)

As in the case of for-profits, the entrepreneur will choose m∗, such that either the new effec-tiveness constraint or the measurement compensation constraint is satisfied, that is

C ′(m∗3) =γ

vV ar(a)− v4γ < D

′(m∗3)− (ā− E(a)) , or (18)

C ′(m∗4) = D′(m∗4)− (ā− E(a)) <γ

vV ar(a)− v4γ . (19)

As in the case of a for-profit social enterprise, the higher V ar(a) and γ, the more likely that theentrepreneur will employ the beneficiaries and measure their abilities. Note that γvV ar(a) −

v4γ >

γ × V ar(a) − 14γ , i.e., the expression in the right-hand-side of equations (11) and (12). This willaffect the choice between a for-profit and non-profit entity, which I discuss below. But first, Idiscuss the choice between a non-profit social enterprise and a donative organization. This choiceis determined in a similar way to the choice between corporate charity and a for-profit socialenterprise. Accordingly, the same three cases depicted in figure 1 may emerge as in the case of afor-profit social enterprise: (a) a firm that engages only in giving, which in this context we call,a “donative organization”, (b) a firm that mixes between giving and transacting (i.e., the resultrepresented by equations (18) and (15)), and (c) a firm that only employs the beneficiaries anddoesn’t engage in giving (see equation (19)).

17

7 The Entrepreneur’s Choice

In considering the choices of the entrepreneur, we need to distinguish between three main casesagain based on the term γ × V ar(a).

Case 1. γ×V ar(a) is low. Suppose C ′(0) > γv×V ar(a)−v4γ , such that the non-profit entrepreneur

chooses to form a donative organization. In this case, the for-profit entrepreneur wouldalso focus exclusively on giving because γv × V ar(a) −

v4γ > γV ar(a) −

14γ and C(m) is

convex.

Case 2. γ × V ar(a) is medium. Suppose C ′(mfp) = γV ar(a)− 14γ (see equation (11)). At m =m∗fp, the non-profit entrepreneur will continue to increase the number of beneficiariesshe will employ because C(m) is convex, such that mnp ≥ mfp. Note however that thenon-profit entrepreneur will increase n up to the level at which D′(n∗np)− (ā− E(a)) =γvV ar(a) −

v4γ > γV ar(a) −

14γ = D

′(n∗np) − (ā− E(a)). Therefore, n∗np < n∗fp becauseD(n) is concave. It follows then that g∗fp > g∗np ≥ 0 because n = m+ g.

Case 3. γ × V ar(a) is high. Suppose that C ′(mfp) = D′(mfp)− (ā−E(a)) (see equation (12)).This means that g∗fp = 0, as the for-profit entrepreneur chooses to engage only in trans-acting. This is however also the choice of the non-profit entrepreneur, since equation(12) is essentially identical to equation (19), γvV ar(a) −

v4γ > γV ar(a) −

14γ , and D(n)

is concave. This means that at mfp, the measurement compensation constraint for thenon-profit entrepreneur will also bind. Therefore, mfp = mnp, and g∗np = 0.

Figure 2, summarizes these cases, which again fall into the three business models: (a) pure giving,(2) mixed giving and transacting and (3) pure transacting.11

11Note that these cases are not exhaustive, because in case 2 it is possible that the for-profit will mixbetween transacting and giving, while the non-profit will engage exclusively in transacting.

18

Figure 2

To determine which form the entrepreneur will actually choose, we need to compare the utilitiesshe derives from her decisions. As a preliminary step, I will compare these utilities on the assump-tion that the entrepreneur mixes between social enterprise and giving. The ex ante utility of thenon-profit entrepreneur is given by equation (18), except that n′ and m′ should be replaced with n∗

and m∗. The utility of the for-profit entrepreneur is the same except that we set v = 1. Accordingly,the social entrepreneur will choose the non-profit form if utility from forming a non-profit is higherthan the utility from forming a for-profit, i.e.,

(v − 1) r × w × ā︸ ︷︷ ︸ + v ×D (n∗np)−D (n∗fp)︸ ︷︷ ︸< 0 < 0

(20)

+(n∗fp − v × n∗np

)(ā− E(a))︸ ︷︷ ︸ +

(g∗fp − g∗np

)γV ar(a)︸ ︷︷ ︸ + 14γ

(v2 × g∗np − g∗fp

)︸ ︷︷ ︸

> 0 > 0 < 0

+C(mfp)− v × C(mnp)︸ ︷︷ ︸?

> 0.

In evaluating the entrepreneur’s choice, note that in Case 1, the last two terms equal 0, n∗fp =g∗fp, and n∗np = g∗np. In Case 2, we know that mnp > mfp > 0 and g∗fp > g∗np ≥ 0. That a non-profit would actually employ more of the beneficiaries is consistent with the view that non-profitentrepreneurs produce higher quality (Hansmann, 1980; Shleifer & Glaeser, 2000), which in mymodel is measured in terms of the effective use of subsidies.

19

Note that in both Cases 1 and 2, we cannot tell if the condition in equation (20) would increase ordecrease in V ar(a) and E(a). This makes it hard to predict the circumstances in which entrepreneurswill choose between for-profit and non-profit social enterprises as well as between corporate charityand donative organizations. In fact, this might actually explain the difficulty of many commentatorsin articulating whether a for-profit or a non-profit would be a better vehicle for channeling subsidies,and why many have argued that the distinction between the two forms appears to be increasinglyillusory. This is however subject to one important caveat.

In Case 3, the non-profit entrepreneur’s utility is:

Unp = v ×(r × w × a+D(m∗np)−mnp (ā− E(a))− C(m∗np)

)(21)

The utility of the for-profit entrepreneur is:

Ufp = r × w × a+D(m∗fp)−m∗fp (ā− E(a))− C(m∗fp). (22)

Assuming the entrepreneur’s profits are positive, it is clear that the entrepreneur will choose thefor-profit form because UFP > UNP , as mfp = mnp. This finding leads to the following proposition.

Proposition 6. The entrepreneur will form a for-profit social enterprise that engages exclusivelyin transacting with the beneficiaries if the variance of abilities (Var(a)) is high and the entrepreneuris very altruistic (γ is large).

There are many examples of dedicated for-profit social enterprises that operate in environmentswhere information asymmetries are likely to be severe due to a high variance in abilities. Forexample, fair trade firms that buy products from many disaggregated farmers that have varyingabilities tend to be for-profit firms, and rarely incorporate as non-profits. These firms also tendto have a relatively thin corporate social responsibility programs, and their assistance is channeledprimarily through the transactional relationship with their farmers. Proposition 6 suggests thatthe reason why the for-profit form may be the preferred legal form in the context of certain socialproblems is the severity of information asymmetries. This partly explains the puzzle why in somecontexts we observe for-profits as the dominant form for organizations dedicated to pursuing socialmissions.

8 Social Enterprises and Social Welfare

The choices of the entreprenuer can have an effect on social welfare. To consider this effect, it isuseful to define a social welfare function (S). The social welfare function incorporates (1) the utility

20

of the entreprenuer (Uent), (2) the utility of donor (Ud), and (3) the utility of the beneficiaries whoare employed (Ub). The utility of the entreprenuer is given above. The utility of the donors derivefrom making donations is assumed to be equal to the value of their donations to the firm (D(n)).Thus, by assumption, Ud is normalized to zero because they give just as much as their receive frommaking their donation. The benefit to the workers is quantified at the amount of disbursals theyreceive, but only to the extent that such disbursals are effective. Thus, if worker a(i) receives d(i),such that a(i) + d(i) ≥ ā, the worker’s utility increases by ā − a(i). But if a(i) + d(i) < ā, theworker’s utility increases only by d(i) < ā− a(i). Using equation (6) and (14), the probability thata(i) + d(i) < ā, is the same as the probability that a < E(a) + 12γ for a profit and a < E(a) +

v2γ

for a nonprofit. Let Pfp and Pnp be these respective probabilities for a for-profit and a nonprofit.Accordingly,

S = Ud + Uent + Ub (23)

= 0 + v × (r × w × a+D(n∗)− n∗ (ā− E(a))− C(m∗))− g∗ × γV ar(a) + v2g∗

4γ +m∗ (ā− E(a))

+ Pnpg∗d(i)∗ + (1− Pnp) g∗(ā− E[a|a > E(a) + v2γ ]

)= v × r × w × a+ v ×D(n∗)− v × C(m∗) + n∗ (1− v) (ā− E(a))− g∗ ×

(γV ar(a)− v

2

4γ

)

− g∗ ×(Pnp

v

2γ + (1− Pnp)(E[a|a > E(a) + v2γ ]− E(a)

)),

which in the case of a for-profit entrepreneur (v = 1), turns into:

S = r × w × a+D(n∗)− C(m∗)− g∗ ×(γV ar(a)− 14γ

). (24)

= −g∗ ×(Pfp

12γ + (1− Pfp)

(E[a|a > E(a) + 12γ ]− E(a)

)).

Note that the last term denotes the reputational costs to the entreprenuer from making dis-bursals that are too high or too low, and the second-to-last term is costs to the beneficiaries,first by the entreprenuer underpaying them by 12γ when it engages in giving (which bites onlywhen a < E(a) + 12γ ), and second, the waste from overpaying those who have higher abilities(a < E(a) + 12γ ). It is apparent from this expression that there is a tradeoff between the costsof measurement and the costs of disbursals that don’t meet the needs of the beneficiaries. WhenV ar(a) is high, such that the entrepreneur sets g∗ = 0, the social welfare function further simplifies

21

and the last two terms disappear.It follows from this that the entrepreneur does not assist the beneficiaries in a socially op-

timal way. The reason is that her incentives are always to transact or give less than is so-cially optimal. Thus, the entrepreneur will increase m as long as C ′(m) ≤ D′(n) − (ā− E(a)),but it is socially optimal for her to do so until C ′(m) ≤ D′(n). Similarly, the entrepreneurwill increase n by increasing g as long as ā − E(a) +

(γV ar(a)− 14γ

)≤ D′(n), but because(

Pfp1

2γ + (1− Pfp)(E[a|a > E(a) + 12γ ]− E(a)

))≤ ā − E(a),12 it is socially optimal for her to

do so until(Pfp

12γ + (1− Pfp)

(E[a|a > E(a) + 12γ ]− E(a)

))+(γV ar(a)− 14γ

)= D′(n). How-

ever, the bigger problem is that in choosing between transacting and giving, the entrepreneur’schoices will not be optimal. Thus, when the entrepreneur faces a choice between increasing mand g it will choose former only if C ′(m) < γV ar(a) − 14γ . However, from a social welfareperspective, it would be best if she chose to increase m as long as C ′(m) < γV ar(a) − 14γ +(Pfp

12γ + (1− Pfp)

(E[a|a > E(a) + 12γ ]− E(a)

)). This suggests that policies to promote social

enterprises might be conducive to social welfare.

Proposition 7. The optimal number of beneficiaries that transact with the entrepreneur (mopt)is defined by D′(mopt) = C ′(mopt). The entrepreneur transacts with fewer beneficiaries than issocially optimal (m∗ < mopt).

9 Subsidy and Tax Policy

The foregoing analysis ignored tax considerations and government subsidies in the choice oforganizational form. In the following discussion, I discuss the effect of such policies. First, Istart with identifying the policy that maximizes social welfare. Second, I show how standard taxbenefits, such as the tax deduction or income tax exemptions, can actually decrease social welfare,particularly if allocated to for-profit social enterprises.

9.1 Conditional Subsidies

Consider a government that seeks to maximize social welfare. The government represents theinterests of the citizens and can choose to make transfers. Transfers to the entrepreneur have noeffect on social welfare, since they entail a loss to the government’s utility and a correspondingbenefit to the entrepreneur.13 But the government can condition its grant on the firm taking

12This is easy to show using the constraint that d(i) > 0, and therefore γ > 12(ā−E(a)) .13It should also be emphasized further that within this model, it is also not desirable for government to

try to replace the entrepreneur. Pure government spending in this framework will actually generate waste

22

certain actions, in particular, employing the beneficiaries. The government can therefore subsidizethe employment of the beneficiaries up to the optimal level of in which D′(n) = C ′(m). Recall thatwhen γV ar(a) is high, the entrepreneur will increase n until C ′(m∗) = D′(m∗) − (ā− E(a)) . Letmopt be the optimal m such that C ′(m∗) + (ā− E(a)) = C ′(mopt) = D′(mopt). The governmentcan incentivize the entrepreneur to employ mopt beneficiaries by providing the firm with (mopt −m∗) (ā− E(a)) conditional on the firm employing mopt beneficiaries. Note that if the governmentcan fully observe the donation function and the costs of measurements per each beneficiary, itmight be able to provide a lower subsidy that reflects the marginal increase measurement costs anddecrease in donation, i.e.,

∫moptm∗ D

′(m)−C ′(m). Note further that it is not desirable to require theentrepreneur to employ more than mopt because D(m) < C(m) for m > mopt.

Remark 1. A subsidy to the entrepreneur in the amount of at least∫moptm∗ D(m)−C(m) conditional

on transacting with additional beneficiaries will cause the entrepreneur to employ the number ofbeneficiaries (mopt) that maximizes social welfare.

Note that at this point, the entrepreneur will not engage in any giving. The reason is that for giv-ing to take place, it is necessary that D′(g) > ā−E(a)+γV ar(a)− 14γ , but D

′(mopt) < ā−E(a). Wealso know thatD′(mopt) < C ′(m∗) < γV ar(a)− 14γ <

Pfp2γ +(1− Pfp)

(E[a|a > E(a) + 12γ ]− E(a)

)+(

γV ar(a)− 14γ). Therefore, it is also not desirable for the entrepreneur to increase n above mopt

by increasing g.Finally, it’s worth mentioning that in some cases firms might actually refrain from transacting

with the beneficiaries because the costs of measurement are too high (i.e., C ′(0) > γV ar(a)− 14γ ).This is likely to be the case when the costs of measurement consist of a high fixed cost component(i.e., C(m) = Cf + Cv(m)), where the fixed component accounts for the costs of developing themeasurement technology, such as a business model to monitor employees or to acquire informationon borrowers in rural communities. In this case, the desirable subsidy will also include a lump-sumthat accounts for the fixed costs Cf conditional on the entrepreneur employing m∗.

9.2 Other Tax Benefits

9.2.1 Tax deductions

Tax deductions are typically given to donors to non-profit firms. In this analysis, this meansthat more donations are available to non-profit firms. This can be implemented by multiplyingbecause the government faces informational problems. Suppose the government gives ā − E(a) to eachbeneficiary. The government loses g (ā− E(a)) but the corresponding benefits to the beneficiaries are lower,and the ensuing loss is (1− Pfp) (E[a|a > E(a)]− E(a)) , i.e., the waste from over-paying those who havehigher abilities than the average abilities.

23

D(n) by a scalar (λ) in equations (13) and (16). The effect of λ for both donative organizationsand non-profit social enterprises is to increase n, the number of beneficiaries. For non-profit socialenterprises, the tax deduction relaxes the measurement compensation constraint. Thus, as longas C ′(m) < γvV ar(a) −

v4γ , the tax deduction encourages the firm to employ more beneficiaries,

thereby increasing subsidy effectiveness (see equation (19)). But at the point at which C ′(m) =γvV ar(a)−

v4γ , to the extent that C

′(m) < λD′(m)−(ā−E(a)), the effect is to increase the corporatecharity activities of the firm, and potentially reduce the effectiveness of subsidies.

Moreover, tax deductions naturally bias the entrepreneur’s choices towards the non-profit form.When V ar(a) is medium many of the elements in equation (20) could be reversed because a non-profit would assist a greater number of beneficiaries and receive larger donations, though again,it is not possible to make concrete predictions as to which form the entrepreneur will choose.More importantly, when V ar(a) is large Proposition 6 does not necessarily hold with a tax de-duction, and will depend on multiple parameters, including λ and v. This can potentially leadfor-profit social enterprises that without the tax deductions would engage in purely transactingto incorporate as nonprofits that mix between transacting and giving.14 If we measure quality assubsidy-effectiveness, this analysis thus casts doubt on the claim by Glaeser & Shleifer (2001) thatquality rises with the level of tax advantages.

Moreover, from a social welfare perspective, it is difficult to tell whether or not such a resultwould increase social welfare. Recall that a government a tax deduction is essentially a governmenttransfer, but the transfer of (λ− 1) vD(n∗) to the entreprenuer does not affect welfare itself. Ratherthe transfer increases the g∗ chosen by the entreprenuer by 4g∗. If the tax deduction increasesthe corporate charity, using equation (23), social welfare changes by v× (D(n∗ +4n∗)−D(n∗)) +4n∗ (1− v) (ā− E(a))−4g∗×

(γV ar(a)− v24γ

)−4g∗×

(Pnp

v2γ + (1− Pnp)

(E[a|a > E(a) + v2γ ]− E(a)

)),

and it is not possible to determine ex ante whether or not this change is positive or negative.More interestingly, if tax deductions extended to for-profits have the effect of increasing g but not

m, then social welfare actually decreases. Using equation (24), the change to social welfare is givenby (D(n∗ +4n∗)−D(n∗))−4g∗×

(γV ar(a)− 14γ

)−4g∗×

(Pfp

12γ + (1− Pfp)

(E[a|a > E(a) + 12γ ]− E(a)

)).

This expression is negative because for every n > n∗, D(n) < γV ar(a)− 14γ .[DIFFERENCE FROM NONPROFITS IS THAT THE FOR-PROFIT ENTREPRENEUR

DOESN’T INTERNALIZE THE INEFFICIENCY OF THE GIVING BECUASE OF THE NON-DISTRUBTION CONSTRAINT - GIVING IS CHEAPER FOR IT - SO INCREASING N IS NOTSO BAD]

Remark 2. If the effect of tax deductions is only to increase giving, quality measured as subsidy-14One condition for this to happen is that without a tax deduction, C ′(m∗) = D′(m∗)− (ā− E(a)) as per

equations (12) and(19), but with a tax deduction, γvV ar(a)−v4γ < λD

′(m∗)− (ā− E(a)).

24

effectiveness decreases, and social welfare will decrease in the case of for-profits and may decreasein the case of nonprofits.

Remark 2 may partly explain why the Internal Revenue Service (IRS) is cautious in recognizingsocial enterprises as tax exempt organizations under section 501(c)(3) on the basis that they arecommercial in nature even if they are subject to the non-distribution constraint, and they committo employing the poor or providing them with low-cost services. It further explains why the federalgovernment and tax authorities have resisted recent proposals (see Malani & Posner, 2007) toextend tax deductions to social enterprises that are not subject to the non-distribution constraint.

9.2.2 Income Tax Exemption

The effect of income tax exemption for non-profits is therefore fairly simple. The effect isessentially to mitigate the impact of the non-distribution constraint on the entrepreneur. Thatis equivalent to increasing v, presumably getting it closer to 1. In the case of a nonprofit firm,it is noteworthy that income tax exemption (i.e., a higher v) doesn’t change the choice of theentrepreneur as to legal form. That said, it does reduce the incentives of the entrepreneur totransact and measure because γvV ar(a)−

v4γ is lower.

Tax exemptions may in theory be extended to for-profits as some have suggested (e.g., Malani& Posner, 2007). Within this model this is essentially the same as imposing v > 1. It is apparentfrom the analysis of nonprofits that this will have the effect of reducing quality when V ar(a) islow or medium, because it will increase g and n, but decrease m. Moreover, it is also easy toshow that such a policy will decrease welfare. Note again that the income tax exemption itself,which is a transfer from the government to the entrepreneur (amounting to (v − 1)), does not byitself change welfare; rather, the change in S is due to changes in the choices of the entrepreneur.Thus, the change in S amounts to (D(n∗ +4n∗)−D(n∗)) − (C(m∗ −4m∗)− C(m∗)) − 4g∗ ×(γV ar(a)− 14γ

)− 4g∗ ×

(Pfp

12γ + (1− Pfp)

(E[a|a > E(a) + 12γ ]− E(a)

)). This expression is

invariably negative for the following reason. We know that for m < m∗, C(m′) < γV ar(a) − 14γ ,and therefore the effect of substituting m for g (holding n constant) will clearly be to reduce socialwelfare. Likewise, the effect of increasing n by increasing g is again negative, because we know thatfor n > n∗, D′(n) < γV ar(a)− 14γ .

Remark 3. Income tax exemptions reduce quality measured as subsidy-effectiveness. As a result,it decreases social welfare in the case of for-profits and may decrease it in the case of nonprofits.

25

10 Regulation of Social Enterprises

10.1 Mission-Drift

A potential concern with social enterprises is that they may focus primarily on transactingwith the beneficiaries who, albeit disadvantaged (in the sense that they belong to groups that areunable to transact with commercial firms), have the highest abilities among the disadvantaged.This concern is voiced against many social enterprises, particularly microfinance institutions whichlend only to marginally vulnerable borrowers, or fair trade firms that source products from highlyskilled farmers from rural communities, and work-integration firms that avoid employing workersthat require substantial training.15 The analysis above can accommodate these concerns.

Suppose that the abilities of the beneficiaries (a) are uniformly distributed, and (b) that asocial entrepreneur can choose the workers she will employ after she expends the measurementcosts. Naturally, she will then choose to employ those who have the highest abilities, i.e., thosewith a ∈ [ā −m∗, ā), assuming m∗ < ā. The reason is clear. By transacting with those with thehighest abilities, the entrepreneur can make the least expensive disbursals, and therefore retain asprofits any excess donations she has received. For simplicity, I restrict the problem to for-profitsocial enterprises that engage exclusively in transacting. The excess donations can be quantifiedas D(m∗) − m∗22 , given the distributional assumptions with respect to a and Proposition 3. Bycomparison, if the entrepreneur was required to employ the m∗ workers around the mean abilities(i.e., ā2 ), she would retain a smaller amount of D(m

∗) − m∗ā2 , or if required to employ those withthe lowest abilities, she would retain only D(m∗)−m∗ā+ m∗2 .

One potential solution for this problem is for the donor to require that the beneficiaries bedrawn from communities with individuals whose average abilities are lower (i.e., lower than ā2 ) orthat the maximum abilities are lower (i.e., lower than ā). To be sure, the donor in this case doesnot observe a, but she knows the distribution of a. For example, the donor will require that thefirm employs workers from an area where income is lower and unemployment rates are higher. Thisway even if the social entrepreneur chooses to employ those beneficiaries with the highest abilities,those beneficiaries will not be only the marginally vulnerable, but individuals who are substantiallydisadvantaged.

This means that the donation function will also depend on beneficiaries’ mean abilities. AssumethatD = D(n, ä), where ä is the average abilities of the beneficiaries that actually receive assistance,and D(n, ä) is convex and decreasing in ä ∈ [0, ā2 ]. The entrepreneur utility is then r × w × ā +D(m∗, ä) − m∗ (ā− ä), where m∗ is the number of worker-beneficiaries at the optimum defined

15[Add citations]

26

by equation (12). Note that at this point D′(m∗, ä) = −m∗, which means that the entrepreneurcan choose to reduce ä by one unit to increase the donation she receives by m∗ (i.e., one unitof donation per one worker-beneficiary). Because of the convexity of the donation function in ä,D′(m∗, ä − ε) > D′(m∗, ä) for ε > 0 ∀ε. Therefore, the entrepreneur can increase the amount ofdonations she receives by reducing ä without changing m∗, and she will ultimately choose to focuson the most vulnerable beneficiaries, i.e., ä = m∗2 .

Remark 4. Even when the abilities a(i) of each beneficiary i is not observable, conditional donationsbased on the average abilities of those that transact with it (ä) can mitigate the risk of mission-drift.

This type of conditional donation to social enterprises seems to occur in practice. For exam-ple, some consumers buy coffee from, say, Starbucks, which focuses on small farmers with higherabilities, and others buy coffee from more dedicated fair trade firms (such as Equal Exchange orCafedirect) that source products from less skilled farmers (Eldar, 2017). Accordingly, legal policiesshould focus on developing certification mechanisms for verifying the class of beneficiaries thatsocial enterprises commit to transacting with and their average abilities, and communicating thisinformation to the donors.

10.2 The Risk of Exploitation

One issue not modeled in this article is the risk that a social entrepreneur will exploit thebeneficiaries with whom she transacts. This is a probable scenario in circumstances where thesocial enterprise is a monopolist (which is likely since social enterprises arise when commercialfirms ignore certain groups), or appropriate regulation is lacking (as is often the case in developingcountries). Such exploitation can take the form of predatory pricing practices, exploitative wagesor extreme working conditions. A comprehensive assessment of the mechanisms for addressingsuch exploitation is outside the scope of this article, but naturally, any business, whatever itsorganizational form needs to be regulated.16 Two main policies are often proposed to address therisk of exploitation: (a) caps on distribution and (b) price regulation.

Caps on distribution are supposed to mitigate the profit incentives to exploit the beneficiaries.Paradoxically, such caps can actually eliminate the incentives of the entrepreneur to measure the

16Alternatively, in circumstances when regulation is lacking, one possible mechanism for mitigating therisk of exploitation is for the entrepreneur to adopt the non-profit form (see Glaeser & Shleifer, 2001; Bubb& Kaufman, 2013). In this context, the quality of the service is fair dealings with the beneficiaries, and it ismeasured largely as quantity as in Glaeser & Shleifer (2001). Note that there is in theory a possibility thatbeneficiaries will sanction for-profit firms by refusing to transact with those firms that were treated themunfairly in the past (see Choi, 2015). However, in practice, such sanctions do not seem to be a practicalstrategy for disadvantaged individuals who face difficulties transacting with commercial firms.

27

abilities of the beneficiaries. Suppose we cap the total revenues that can be distributed to theentrepreneur at r×w× ā, the revenues of a commercial for-profit with no regard for disadvantagedpeople (i.e., γ = 0). Again, for simplicity, I restrict the discussion to for-profits. The socialentrepreneur will effectively be required to relinquish the amount ofD(m∗)−C(m∗)−m∗ (ā− E(a)),to the extent that such amount is larger than 0. In these circumstances, the cap on distributionmay eliminate the monetary incentives to measure if the donation amount exceeds the monetaryloss from not measuring; that is if D(k) > k ×

(Pa 0 there willbe some reputational incentives to measure, but if γ is small, such soft incentives may be trivial.Furthermore, as firms that engage in corporate charity would presumably not be subject to caps,the incentive to engage in corporate charity rather than social enterprise will dominate (i.e., theentrepreneur’s utility will be larger than r × w × ā) as long as corporate charity attracts sufficientdonative funding (i.e., if D(n) > n(ā− E(a)) + γ × n× V ar(a)− n4γ > 0). This will clearly resultin a decrease to social welfare as it will reduce the utility of the entrepreneur and the beneficiaries.

Price regulation is another mechanism to mitigate exploitation by essentially setting the amountof subsidies allocated to the beneficiaries to amounts larger than those the entrepreneur has incen-tives to disburse. This means that d(i) is fixed such that d(i) = df , where m× df > m(ā−E(a)).For example, the amount of wages, prices and interest rates can be fixed through regulation, sothat they appear to be fairer to the beneficiaries and distribute more wealth to them. This policy issimilar to capping distribution of revenues to the entrepreneur because it requires the entrepreneurto forego her profit, and in this way reduces her incentives to measure the beneficiaries’ abilities.To take a simple example, suppose df = ā. In this case, the social entrepreneur has no financialincentive to measure since she does not benefit financially from tailoring subsidies to the needsof the beneficiaries. Moreover, there will be substantial reputational costs, γ × n × E(a2), as thedisbursals will be wasteful. Again, this will have the effect of discouraging social entrepreneurshipin favor of corporate charity which is more beneficial financially (as the disbursals will be smallerthan ā), and in this example also more effective and less costly for the entrepreneur’s reputation(as V ar(a) < E(a2)). It will clearly also reduce social welfare due to its effect on the entrepreneurand the beneficiaries. 18

17For derivation of the right-hand side, see the proof of Proposition 4 in the Appendix, particularly equation(27).

18It is easy to show that similar results follow when d(i) is subject to a minimum cap, say ā− E(a), butthe entrepreneur has a choice to increase it up to ā, although clearly the decrease in subsidy-effectivenessand social welfare will be less severe.

28

Remark 5. Caps on distribution and price regulation can distort the entrepreneur’s monetary in-centive to measure and dissuade entrepreneurs from forming social enterprises, thereby reducingthe effectiveness of subsidies and social welfare.

11 Extensions

The model presented in this article is deliberately simple, and does not address in detail otherconsiderations that no doubt affect the formation of social enterprises.

11.1 Firm Size and Scale

Throughout the analysis, the size of the firm (w) remains constant. The common wisdom is thatsocial enterprises tend to be small businesses, and face constraints in scaling their business. Thereason for this is arguably the lack of sufficient subsidies (i.e., D(n) is small and highly concave),which are necessary for the entrepreneur to employ the beneficiaries. But this explanation is notsatisfactory since many large firms engage in extensive corporate social responsibility activitiesthat do not involve transacting with the beneficiaries, and are presumably funded either by theowner (which in my model is represented by the entrepreneur), consumers or the government. Analternative reason for the small size of social enterprises is that the costs of developing measurementtechnology are higher for large commercial firms. For example, a firm like Walmart that employsthousands of workers might be ill-equipped to gather information on the abilities of its employeesin the same manner as a small or mid-sized bakery that operates in one location. Thus, it’s possiblethat under some circumstances C(m) also depends on w, and that increasing w also increases themarginal costs of measurement (i.e., C = C(m,w) and Cmw(m,w) > 0).

On the other hand, many social enterprises, such as microfinance institutions and some fair tradefirms, serve millions of beneficiaries, and seem to be as large as many standard commercial firms. Apossible reason for this is that these firms have developed very efficient measurement technology thatcan be scalable across many locations. In other cases, a donative organization (e.g., Technoserve)may partner with multiple large corporations (e.g., Starbucks), and assist them in developinga measurement technology on the ground. The donative organization essentially channels thetraining subsidies through the corporations (which structurally qualify as social enterprises; seeEldar, 2017) in order to benefit from the informational advantages of transacting, but also fromaggregating knowledge acquired by working with different firms. This knowledge may in turn helpit in reducing the costs of measurement.

29

11.2 Market Differentiation and Donors’ Preferences

In this simple set-up, we observe essentially one model of social enterprise. But social enterprisesdiffer on multiple dimensions, including, as discussed above, the scale of the social enterprise, andthe level of abilities of the beneficiaries the entrepreneur chooses to transact with. Thus, socialenterprises may offer differentiated “products” to donors, some of which focus on the marginallyvulnerable, while others on the least endowed. The extent to which social entrepreneurs adoptdifferent models depends on both supply and demand factors.

On the supply side, as discussed above, the measurement technology available to the en-trepreneur may determine both the scale of the social enterprise (as discussed above), and probablyalso the group of beneficiaries she chooses to assist. On the demand side, the preferences of thedonors could affect the decisions of the entrepreneur. To be sure, the donors in this model wererepresented by a simple donation function, D(n) based on the the assumption that donors cannotcontract for quality ex ante. This is a reasonable assumption since donors rarely examine the im-pact of their donations on the intended beneficiaries who may be located far away from the donors(Hansmann, 1980), and it may take years before the impact of any subsidies (e.g., training) mate-rializes. Nonetheless, under some circumstances, short of measuring the social impact of subsidies,donors could contract with entrepreneurs for certain qualitative elements, for example, to addressthe risk of mission-drift and exploitation of the beneficiaries. Some donors are large foundationsthat control the social enterprise firm and can ensure that the social enterprise transacts with ben-eficiaries that belong to vulnerable communities where the average level of abilities is well belowā, and that the firm does not exploit them. Similarly, some government agencies may conditionsubsidies in part on the level of poverty of the intended beneficiaries, as well as the type of sub-sidies (interest rate discounts and training) that would be provided to them. The CDFI Fund forexample provides subsidies to Community Development Financial Institutions (“CDFIs”) based onthe level of need in the relevant target area and the terms of the financial services they promise toprovide in that area. Accordingly, donors with heterogeneous preferences may be able to affect theentrepreneur’s decisions to a greater extent than is suggested by this model.

Note though that not all donors are well positioned to do this. Despite the increasing availabilityof various certification mechanisms, such as the Fair Trade certification, consumers as donors maybe less able to exercise such control over the social enterprise firm. Accordingly, policies that conveygreater information to consumers may improve the ability of donors to ensure that subsidies theyprovide to organizations are used for their intended purpose (Eldar, 2018).

30

11.3 Subsidy Policies

The model in this article does not provide a practical mechanism for donors - including gov-ernments - for setting the amount of donations. The model suggests that the entrepreneur willreceive donations that exceed the amounts necessary to induce the entrepreneur to transact withthe beneficiaries. The entrepreneur thus gets to keep some excess donations, which allow her toearn even more than an entrepreneur that does not pursue any social goals at all. Due to informa-tion problems, it is unlikely that donors would be able to set the amount of donations such that noexcess is provided to the entrepreneur. However, it may be possible to contemplate a competitiveprocess whereby entrepreneurs bid for subsidies from donors. The donors may offer a fixed amountof subsidies, and the entrepreneur will compete for such fixed sums by promising to transact withmore beneficiaries, with beneficiaries with lower average abilities, and/or at better terms. The com-petitive process will likely reduce the excess donations to the extent that the donors can monitorthe fulfillment of such promises, because transacting with more beneficiaries with lower abilitiesat better terms requires more subsidies. In fact, the CDFI Fund conducts such process wherebycertified CDFIs apply for subsidies in a competitive process in which each CDFI offers to providefinancial services to a specific under-served area at pre-specified terms.

11.4 [Competition]

12 Conclusion

Recent literature on the growth of various forms of corporate philanthropy and the fadingdistinction between for-profits and non-profits has aptly identified the altruistic motives of bothinvestors and entrepreneurs that led to this development. However, it has mostly failed to identifythe key informational asymmetry that social enterprises are designed to address. The analysis inthis article fills this void by explaining how committing to transacting with disadvantaged groupsgives entrepreneurs the incentives to measure or gather information on their beneficiaries’ attributesand needs. The main innovation of social enterprises is not the mixed motives of their entrepreneurs(which exist even in the traditional non-profits), but their unique structure that is based on trans-acting with their beneficiaries. This structure gives the entrepreneurs both the ability and incentivesto gather information on their beneficiaries’ abilities, and to design effective subsidies to help theirbeneficiaries develop. When the informational asymmetries are severe, these incentives are so strongthat the non-profit form may lose its value as a mechanism for ensuring that entrepreneurs can betrusted to pursue social missions.

The analysis further sheds light on various policy issues concerning social enterprises. In par-

31

ticular, it suggests that if we measure the quality of social entrepreneurship by examining theeffectiveness of subsidies in helping disadvantaged groups develop competitive abilities, then in-creases in the amount of subsidies through tax exemptions or direct subsidies will have limitedimpact, and can in fact be counterproductive. Furthermore, caps on distribution coupled withprice regulation can likewise reduce or even eliminate the incentives of social enterprises to pursuesocial missions effectively. It is vital to social enterprises, like any business, to prevent exploitationthrough practices such as predatory pricing or unfair working conditions. However, regulation ofsocial enterprises should be cautious so as not to defeat the role of social enterprises by destroyingentrepreneurs’ incentives to design effective subsidies.

32

References

[1] Alchian, Armen A., and Harold Demsetz. "Production, information costs, and economic orga-nization." The American economic review 62, no. 5 (1972): 777-795.

[2] Baron, David P. "Corporate social responsibility and social entrepreneurship." Journal of Eco-nomics & Management Strategy 16, no. 3 (2007): 683-717.

[3] Banerjee, Abhijit V., and Esther Duflo.