ESSAYS ON COMPLIANCE AND COOPERATION

“ESSAYS ON COMPLIANCE AND COOPERATION” Clara Villegas Palacio Uppsats för licentiatexamen vid Institutionen för nationalekonomi med statistik Hande...
Author: Gregory Bruce
1 downloads 2 Views 532KB Size
“ESSAYS ON COMPLIANCE AND COOPERATION”

Clara Villegas Palacio

Uppsats för licentiatexamen vid

Institutionen för nationalekonomi med statistik Handelshögskolan vid Göteborgs universitet

Göteborg Juni 2009

Institutionen för nationalekonomi med statistik Handelshögskolan vid Göteborgs universitet Vasagatan 1, Box 640, SE 405 30 Göteborg 031 786 0000, 031 786 1326 (fax) www.handels.gu.se [email protected]

Taxes, Permits and the Adoption of Abatement Technology under Imperfect Compliance Clara Villegas-Palacioa,b,* , Jessica Coriac. ** Department of Economics. School of Business, Economics and Law. University of Gothenburg. Gothenburg, Sweden. E-mail: [email protected] b Facultad de Minas. Universidad Nacional de Colombia – Sede Medellín. Medellín, Colombia. E-mail. [email protected] cDepartment of Economics. School of Business, Economics and Law. University of Gothenburg. Gothenburg, Sweden. E-mail: [email protected] a

ABSTRACT This paper analyzes the effects of the choice between price-based and quantity-based emission regulations on compliance incentives and social welfare in the presence of incomplete enforcement and technology adoption. We show that in contrast to taxes, the extent of violations under tradable emission permits (TEPs) decreases with the rate of technology adoption. However, in terms of welfare, the ranking of the instruments is not so straightforward: taxes induce lower emission damages while TEPs induce lower abatement, investment, and expected enforcement costs. Thereby, the overall ranking depends on the extent to which these effects offset each other. Key words. Technological adoption, environmental policy, imperfect compliance, enforcement, social welfare. JEL classifications: L51, Q55, K32, K42.

1 * Corresponding author at: Department of Economics. School of Business, Economics and Law. P O Box 640 SE 405 30. Gothenburg, Sweden Phone +46-(0)31- 7862642. Fax +46-(0)31-786 1043 ** We are grateful to Carlos A. Chávez, Katrin Millock, Katarina Nordblom, and Thomas Sterner for valuable comments and suggestions. We also thank seminar participants at University of Gothenburg for valuable comments. Economic support from The Swedish Agency for International Development Cooperation (SIDA) to the capacity building program at the Environmental Economics Unit of the University of Gothenburg is gratefully acknowledged

I.

INTRODUCTION

In the long run, technological change is considered the primary solution for environmental problems (Kneese and Schultze, 1978), and it has long been recognized that environmental policy creates incentives that affect the process of technological development (Jaffe et al., 2002). Many scholars have therefore analyzed how alternative policy instruments affect the rate and direction of technological change. Among market-based policies, the analyses tend to support the use of emission taxes (price-based regulation) over transferable emission permits, or TEPs (quantity-based regulation). 1 The fact that the emission price is fixed under the tax while it decreases under permits creates a wedge between the two instruments and between the rates of adoption they induce. Previous analyses of technology adoption under different policies share a common and implicit assumption: Firms perfectly obey environmental regulations and enforcement of policies is costless. However, reality generally differs from this assumption. In some cases a fraction of firms do not comply with an environmental regulation as a result of incomplete enforcement, and the expected enforcement costs can be quite significant. The interaction between incomplete enforcement and technology adoption can be thought of in two ways: (1) Incomplete enforcement, and therefore the possibility that firms do not comply with a regulation, may influence the profits of firms from technology adoption and the adoption decision, and (2) the existence of a new Downing and Prince (1986), Milliman and Prince (1989), Jung, Krutilla and Boyd (1996), Keohane (1999), Kennedy and Laplante (1999), and Montero (2002). 1

2

technology that reduces the abatement costs may influence a firm’s compliance decisions since the marginal benefit of violations is reduced. Therefore, inclusion of technology adoption considerations into a comparison of different policy instruments, in a world of imperfect compliance, may change the ranking of the instruments since such considerations induce different adoption-compliance behaviors. The purpose of the present paper is to analyze the interaction between incomplete enforcement and technology adoption under price-based and quantitybased policies. We compare emission taxes and TEPs in terms of: (i) how compliance changes with the use of new technologies, (ii) how technology adoption is affected by enforcement parameters such as probability of being monitored and structure of sanctions imposed in case non-compliance is detected, and (iii) how the ranking of price-based and quantity-based policies using a social welfare measure is affected by the adoption-compliance output. To our knowledge, the interaction between technology adoption and imperfect compliance and its effects has not yet been directly addressed. Rouseeau and Proost (2005) include rule making, implementation, monitoring, and enforcement costs for both firms and the government into the cost comparison of policy instruments. While they compare emission taxes, emission standards, and technology standards, they do not compare instruments within the market-based regulations as is the main focus of the present paper. Montero (2002) studies the impact of incomplete enforcement of a regulation on the choice between price and quantity instruments, and shows that both instruments perform equally good as long as the benefit and cost curves are known with certainty. However, if these curves are uncertain to the regulator, the quantity instrument performs relatively better than the price instrument. 2 Macho-Stadler (2006) compares total final emission level 2

The ranking of priced-based versus quantity-based environmental regulation was first studied by Weitzman (1974), who analyzed the choice between these two types of instruments when there is

3

achieved with standards and with market-based instruments when imperfect compliance is present, and finds that taxes are superior to the other instruments. There are two important differences between her analysis and the present paper: First, when comparing market-based instruments, she considers the optimal audit policy to be one that minimizes total emissions subject to an enforcement budget constraint. We, on the other hand, consider that the enforcement authority’s goal is to minimize the extent of violations. Second, she does not consider the effect of an enforcement policy on technology adoption, while we do.

In line with the previous literature on technology adoption (Milliman and Price, 1989; Jung et al., 1996; Requate 1995, 2001, 2003, and 2005), our results suggest that permits do not provide higher adoption incentives than emission taxes. However, under permits, the fall in the permit price produced by technology adoption reduces the benefits of violating the environmental regulation at the margin and ultimately leads both adopters and non-adopters to modify their compliance behavior. Thus, in contrast to taxes, the extent of violations under TEPs decreases with the rate of adoption as well as with the enforcement efforts. In terms of welfare, the ranking of instruments is not straightforward. On one hand, there is less damage from emissions with taxes. On the other hand, abatement costs, investment costs, and expected enforcement costs under taxes are never lower than when using TEPs. Thereby, the overall and final ranking depends on the extent to which these effects offset each other.

uncertainty. After Weitzman (1974), the comparison between priced and quantity-based policies has been further developed (Roberts and Spence, 1976; Yohe, 1978; Finkelshtain and Kislev, 1997; Hoel and Karp, 2002; Montero, 2002; Moledina et al., 2003; Baldursson and von der Fehr, 2004; Quirion, 2004; Stranlund and Ben-Haim, 2008).

4

The paper is organized as follows. Section 2 presents the model of adoption and Section 3 introduces the compliance behavior under emission taxes and TEPs. Section 4 compares these policies under three criteria: (1) the extent of environmental violation, (2) the required enforcement strategy for perfect compliance, and (3) the social welfare achieved under each scheme. Finally, Section 5 offers some concluding remarks. II.

ADOPTION INCENTIVES

We consider a competitive industry of size one consisting of a group of riskneutral firms that are homogeneous in abatement costs. In the absence of environmental regulation, each firm emits one unit of a homogeneous pollutant. The abatement costs of a firm are a function of the firm’s emissions level, e , and are denoted c(e) 3. The abatement cost function is strictly convex and decreasing in emissions: c' (e) < 0; c' ' (e) > 0 . Emissions generate damages represented by, the also strictly convex, function D(e) , with D ' (e) > 0 , D' ' (e) > 0 . 4 Assume there is an environmental authority that sets an environmental target – a maximum level of emissions – and then chooses a policy instrument to reach this target.5 A new technology arrives and firms must decide whether or not to invest. The new technology allows firms to abate emissions at a lower cost, given by kc(e) , where

Henceforth, for the sake of notation we will use parentheses to denote a function and brackets to denote multiplication. 4 Total abatement cost can also be denoted in terms of abatement level, c(a i ) . The function is then 3

strictly convex and increasing in abatement, c '(a ) > 0; c ''(a ) > 0 . A reduction in damage from emissions can also be interpreted as a benefit from abatement. Abatement generates a concave benefit function, B (a i ); B ' (a ) > 0; B ' ' (a ) < 0 . It is not necessary to assume that the targeted emission level is set at the optimum level, i.e., that it satisfies the conditions where the marginal damage from emissions equals the marginal cost of pollution abatement. The authority could have set a standard for aggregate emissions and decided to use a price-based or quantity-based instrument to achieve this level in an efficient way. However, the analysis of policy ranking would be affected by the chosen target emission level as is presented in subsequent sections.

5

5

k ∈ (0,1) is a parameter that represents the drop in abatement cost. 6 Buying and installing the new technology implies a fixed cost that differs among firms. A fraction

α of the firms in the industry, from now on called Group 1, have a lower fixed investment cost, k . Firms in Group 2, which corresponds to a fraction 1- α of the industry, each have to invest a higher amount, represented by k , if they want to adopt the new technology ( k < k ). Firms can be regulated through uniform emission taxes or TEPs. In an emission tax system, firms are required to self-report their emissions. A firm is noncompliant if it attempts to evade some part of its tax responsibilities by reporting an emission level that is lower than its true level. In the case of regulation using permits, a firm should buy one permit for each unit of emissions. A firm that buys fewer permits than its emissions is out of compliance. Let EC A and EC NA be the total expected costs of abatement and compliance for adopters and non-adopters of the new technology, respectively. These costs are composed of abatement costs and expected fines in case a firm is caught violating. Let ∆EC= ECNA − EC A be the expected savings from adoption. Firms will adopt the new technology if adoption implies savings larger than or equal to its fixed investment cost. Let λ denote the fraction of firms adopting the new technology. There are three possible values for λ depending on the extent to which adoption savings offset the adoption costs: (i)

Zero technology adoption. No firm will adopt the technology if adoption savings do not offset the lowest fixed investment cost; i.e., if ∆EC < k , then adoption is not profitable for any firms and λ = 0 .

Since the interval in which k belongs is open, the new technology always reduces the abatement costs but never makes them equal to zero, i.e., there is no perfectly clean technology. 6

6

(ii)

Partial technology adoption. If adoption savings range from k to k , then adoption is profitable only for firms in Group 1; i.e., λ = α if k ≤ ∆EC < k .

(iii)

Universal technology adoption. If adoption savings are larger than the highest investment costs, then all firms in the industry will adopt the new technology; i.e., λ = 1 if ∆EC ≥ k .

Notice that since adoption savings depend on the expected costs of abatement and compliance, the rate of adoption varies with the stringency of the policy, with the choice of policy instrument, and with the monitoring and enforcement design. III.

COMPLIANCE BEHAVIOR AND TECHNOLOGICAL ADOPTION.

In line with Malik (1990), Stranlund and Dhanda (1999), Stranlund and Chávez (2000), and Chávez et al. (2008), we assume that the regulator cannot observe a firm’s emissions unless costly monitoring is undertaken. Let π denote the probability that the regulator audits a firm. We assume that π is known among firms and that once the regulator monitors a firm, it is able to determine the firm’s compliance status perfectly. We assume that the probability of a firm being monitored is exogenous and uniform across firms. 7 If the monitoring reveals that the firm is non-compliant, it faces the penalty F (v) , where v is the extent of the violation. This is a strictly convex function in the level of violation: F ' (v) > 0; F ' ' (0) . 8 For zero violation, the penalty is zero F(0)=0, but the marginal penalty is greater than zero: F’(0)>0. 9

As Sandmo (2002) notes, the assumption of an exogenous probability of being monitored is a simplification. It is more realistic to assume that monitoring probability is a function of regulated firms’ actions. We leave this point for future work. 8 Standlund, Chávez, and Villena (2008) mention some authors who assume that the penalty function is strictly convex: Harford, 1978, 1987; Sandmo, 2002; Cremer and Gahvari, 2002; Macho-Stadler and Perez Castrillo, 2006. Strandlund, Chávez, and Villena assume a linear penalty function in their model, an assumption that is not common in the literature. If the probability of being monitored is exogenous and the marginal penalty is constant, the decision on reporting emissions will be of the type reporting everything or reporting nothing. If the price of emissions is lower than the expected marginal fine 7

7

The game between the regulator and firms is described by the following twostage mechanism: Stage 1. The environmental authority sets the environmental target before the arrival of the new technology and chooses a policy instrument to reach it. We assume that the regulator does not modify the level of the environmental policy in response to the availability of the new technology. 10 The enforcement strategy is exogenously determined and consists of a probability of being monitored and a sanction scheme. The enforcement strategy is set regardless of the regulatory scheme selected by the environmental authority; i.e., firms face the same enforcement policy independent of the regulation mechanism. 11 Stage 2.

Firms make compliance and adoption decisions. The adoption

decision is made based on the comparison of the expected costs of abatement and compliance under the old and the new technology. 12 The next subsections analyze the rate of adoption and the adopters’ and nonadopters’ compliance behavior under both market-based instruments. (constant), a firm will report all of its emissions, while if the price of emissions is higher than the expected marginal fine, it will not report any of its emissions (see Sandmo, 2002, and Heyes, 2000). 9 In order to allow perfect compliance to be a possible choice, we do not rule out the possibility that the marginal penalty when violation is zero is higher than the marginal abatement cost evaluated at the required emission level. 10 Requate and Unold (2003) analyze incentives through environmental-policy instruments to adopt advanced abatement technologies when the regulator anticipates new technologies, and show that taxes and permits are equivalent if the regulator moves just after firms have invested. If, by contrast, the regulator moves prior to the firms’ investment decisions, only permits will succeed in inducing first best outcomes. 11 This assumption does not contradict reality since in many cases the institutional arrangements separate the design of the regulatory instrument from the design of enforcement strategies. However, in a subsequent section of the present paper we will consider the case where the monitoring probability is set to guarantee perfect compliance according to the selected policy instrument. 12 Authors like Lai et al. (2003) argue that a surprisingly large number of firms comply with pollution regulations even though the expected penalties for non-compliance are low. They establish environmental social norm models that consider collective environmental actions among firms. Our model does not include the effect of social norms or non-monetary sanctions in case of noncompliance; instead we consider the expected monetary fines imposed by the enforcement authority to be the only costs of non-compliance.

8

3.1

Uniform Emission Tax Let us assume that firms must pay a uniform tax t per unit of pollutant emitted

and that they self-report their emissions. If a firm makes a truthful report, the total amount of taxes to be paid is te . Since there is incomplete enforcement, the firm could try to evade a fraction of its tax liability by reporting a lower level of emissions. If the firm reports emissions equal to r , where r < e , then the total tax payment is given by tr . In this case, the firm’s violation equals the difference between the actual emissions and reported emissions, vi = ei − ri . If the firm is caught in violation, a penalty is imposed according to the penalty function explained above. Adopters select the emission and report levels that minimize their expected costs of abatement and compliance: 13 (1)

Mine,r kc(e) + tr + π F (e − r )

s.t. e − r ≥ 0 . Notice that the constraint in the optimization problem reflects the fact that there are no economic incentives to over-report emission levels. The Lagrange equation for (1) is ϕ = kc(e) + tr + π F (e − r ) − β [ e − r ]

and the Kuhn-Tucker

conditions, which are necessary and sufficient to determine a firm’s optimal choices of emissions and permits are: (2)

∂ϕ β 0, = kc '(e) + π F '(e − r ) − = ∂e

(3)

∂ϕ = t − πF ' (e − r ) + β = 0 , ∂r

The problem of the firms that do not adopt the new abatement technology is analogous to problem (1); the main difference is that the abatement costs for these kinds of firms are given by c(e) instead of kc(e) . 13

9

(4)

∂ϕ = e − r ≥ 0; β ≥ 0; β [e − r ] = 0 . ∂β Proposition 1: With uniform emission taxes, adopters’ actual and reported

levels of emissions are lower than those of non-adopters. In addition, the actual level of emissions of firms is independent of the enforcement strategy while the reported level of emissions depends on the monitoring probability and on the sanctions structure. Proof 1:

To obtain a firm’s emission level, combine (1) and (2) to get

kc '(e) + t = 0 . Each firm chooses its emission levels such that the marginal abatement

cost equals the tax rate. The emission levels for adopters and non-adopters are, respectively, (5)

e A (t ) = {e kc' (e) + t = 0},

(6)

e NA (t ) = {e c' (e) + t = 0}.

Since there is a uniform tax rate, in equilibrium firms’ marginal abatement costs are equal irrespective of their adoption status: c' (e NA ) = kc' (e A ) . Given that k ∈ (0,1) , it is necessary that c' (e NA ) < c' (e A ) , which is only possible (given the properties of the abatement cost function) if e NA > e A . Note that since the tax is exogenous and not influenced by the enforcement strategy, the actual emissions of firms do not depend on the parameters of the enforcement problem, which is in line with Harford (1978) and the standard in the literature. Let us now look at a firm’s emission report and extent of violation. When the firm is noncompliant, then e − r > 0 , which from (4) implies that β = 0 and ∂ϕ = t − πF ' (e − r ) = 0 . The report levels of adopters and non-adopters firms are, ∂r

respectively,

10

(7)

rA (t , π , F ) = {r t − π F '(eA − rA ) = 0} ,

(8)

rN A (t , π , F ) = {r t − π F '(eNA − rNA ) = 0} .

Equations (7) and (8) state that firms choose to report a level of emissions such that the marginal expected fine equals the marginal benefit of non-compliance, i.e., the tax. Combining both equations, we obtain e A − rA = e NA − rNA . Notice that rNA > rA since e NA > e A . Hence, the emissions reported by adopters are lower than the emissions reported by non-adopters. Proposition 2:

Q.E.D.

With uniform taxes, the extent of violation of firms is

independent of the adoption status and is therefore the same for adopters and nonadopters of the new technology. Proof 2: The size of violation is given by v(t , π , F ) = e(t ) − r (t , π , F ) . From (7) and (8), we obtain that πF ' (e A − rA ) = πF ' (e NA − rNA ) , and since the enforcement strategy is exogenously set and is independent of the adoption status, it is straightforward to observe that e A − rA = e NA − rNA .

Q.E.D.

The intuition behind this result is as follows. On one hand, since the enforcement strategy does not depend on adoption status, the expected marginal cost of evasion does not change with adoption. On the other hand, the marginal benefit of violation does not depend on adoption status either, since it is given by the unit tax rate. Therefore, given that the marginal benefits and expected marginal costs of disobeying the law are the same for all firms, the extent of the violation is the same regardless of adoption status. Then, technological adoption does not provide additional incentives for compliance when emission taxes are used. The expected costs of abatement and compliance for adopters and nonadopters are expressed as:

11

(9)

EC A (t , π , F ) = kc(e A (t )) + trA (t , π , F ) + πF (e A (t ) − rA (t , π , F )) ,

(10)

EC NA (t , π , F ) = c(e NA (t )) + trNA (t , π , F ) + πF (e NA (t ) − rNA (t , π , F )) . Proposition 3: When uniform emission taxes are used, the adoption rate does

not depend on the enforcement strategy but is determined only by the tax rate. Proof 3: Subtracting (9) from (10), we obtain the adoption savings as follows: (11)

∆EC = c(eNA (t )) − kc(e A (t )) + t [rNA (t , π , F ) − rA (t , π , F )] .

Since rNA (t , π , F ) − rA (t , π , F ) = e NA (t ) − e A (t ) , equation (11) can be re-written as: (12) ∆EC = c(eNA (t )) − kc(e A (t )) + t [eNA (t ) − e A (t )] . The first and second terms in (11) give account of the decreasing in abatement costs when the firm adopts the new technology. The third term gives account of the difference in tax payment on reported emissions without and with adoption. Note that adoption savings increase with the level of the tax and the extent of the reduction in abatement costs (i.e., they decrease in k ). Incomplete enforcement does not affect the rate of adoption since neither the emissions level nor the tax rate is a function of monitoring probability or of the sanctions structure.

3.3

Q.E.D

Tradable Emissions Permits A firm regulated by TEPs can abate a fraction of its emissions and buy permits

to compensate the remaining fraction. The equilibrium price of each permit is represented by p , and a firm that emits e should spend pe on buying permits. Assume that the authority issues L emission permits each period and that the possession of a permit gives the legal right to emit one unit of pollutant. 14

14

For the sake of comparison between TEPs and uniform taxes, we say that the quantity of permits initially issued by the authority corresponds to the environmental target used by the authority to set the tax rate.

12

In the presence of imperfect compliance, polluters have an incentive to buy a quantity of permits lower than ei to reduce their expenditure in permits. Let l denote the quantity of permits held by a firm in equilibrium and l 0 be the number of emissions permits, if any, initially allocated to a firm. A firm is noncompliant if after trade it holds a number of permits that is lower than its corresponding units of emissions. The extent of violation is then given by v= e − l . 15 The permit price is endogenously determined by the violation level and technology adoption rate. The larger the extent of violation, the lower the demand for permits and the lower the permit price. On the other hand, the diffusion of the new technology lowers the aggregate marginal abatement costs and therefore lowers the permit price. Since in the model the rate of adoption can take three discrete values, let us denote the price when no firm adopts the new technology pNA , the price with partial adoption pPA , and the price with universal adoption pUA . By construction, pNA coincides with the tax level set by the authority in a scenario of perfect compliance. It holds that pNA > pPA > pUA as will be shown in subsequent paragraphs. Adopters select the emission level and demand for permits that minimize total expected costs: (13)

Mine ,l kc(e) + p[l − l 0 ] + πF (e − l ) ,

s.t. e − l ≥ 0 . The Lagrange equation for (13) is ϕ = kc(e) + p[l − l 0 ] + πF (e − l ) − β [e − l ] and the Kuhn-Tucker conditions, which are necessary and sufficient to determine the firm’s optimal choices of emissions and permits, are:

15

We assume that the enforcement authority keeps perfect track of each firm’s permit holding but can not observe emissions without a costly audit. Assume, for instance, that all transactions performed in the market have to be registered with the authority. Since the authority has information about initial allocation, it is able to have perfect information about each firm’s permit holding at any point in time.

13

(14)

∂ϕ = kc' (e) + πF ' (e − l ) − β = 0 , ∂e

(15)

∂ϕ = p − πF ' (e − l ) + β = 0 , ∂l

(16)

∂ϕ = e − l ≥ 0; β ≥ 0; β [l − e] = 0 . ∂β Proposition 4: With tradable emission permits, the actual emissions and the

quantity of permits that adopters hold in equilibrium are lower than the actual emissions and the quantity of permits that non-adopters hold in equilibrium. In addition, the level of emissions of firms is independent of the enforcement strategy while the quantity of permits that firms hold in equilibrium depends on the permit price and on the monitoring probability and sanctions structure. Proof 4: From the solution to the optimization problem, the level of emissions for adopters and non-adopters are, respectively: (17)

e A ( p) = {e kc' (e) + p = 0},

(18)

eNA ( p ) = {e c' (e) + p = 0}. Equations (17) and (18) state that in equilibrium, each firm chooses its

emissions such that the marginal abatement cost equals the permit price. Since the adopters’ marginal abatement cost is lower, eNA ( p ) > eA ( p ) . Let us now look at the quantity of permits firms hold in equilibrium. When the firm is non-compliant, then e − l > 0 , which from (16) implies that β = 0 and ∂ϕ = p − πF ' (e − l ) = 0 . The number of permits held by adopters and non-adopters ∂l

firms is, respectively: (19)

l A ( p, π , F ) =

{l

p − π F '(eA − l A ) = 0} ,

14

lN A ( p, π , F ) = {l p − π F '(eNA − lNA ) = 0} .

(20)

Equations (19) and (20) show that in equilibrium, firms hold a quantity of permits such that the marginal expected fine equals the marginal benefit of noncompliance, i.e., the permit price. Since the permit prices and the enforcement strategies faced by adopters and non-adopters are the same, from equations (19) and (20) we obtain that eA − l A = eNA − lNA . Given that e NA > e A , it follows that lNA > l A for the equality to hold. Therefore, the quantity of permits held by adopters is lower than the quantity held by non-adopters. Q.E.D. Proposition 5: With tradable emission permits, a firm’s extent of violation is

independent of its adoption status and is therefore the same for adopters and nonadopters of the new technology. However, its extent of violation is decreasing in the rate of adoption. Proof 5:

Equations (19) and (20) state that the extent of violation is

determined by the condition stating that the marginal expected fine equals the permit price. The extent of violation in a scenario of universal adoption is given by

πF ' (v ) = p and in a scenario of partial adoption by πF ' (v ) = p . Given that the UA

UA

pA

pA

permit price is decreasing in the rate of adoption, F ' (vUA ) < F ' (v PA ) . Since the marginal penalty is increasing in the extent of violation, v PA > vUA . Q.E.D. The extent of violation for adopters and non-adopters is given by v( p (π , λ ), π , F ) = e( p (π , λ )) − l ( p (π , λ ), π , F ) . This result is in line with Stranlund and

Dhanda (1999) and Chávez et al. (2008), who find that changes in abatement cost parameters do not affect the extent of violation as long as enforcement and the permit

15

price remain the same. The uniform monitoring effort should be tied only to the observable equilibrium permit prices. 16 The fact that the extent of violation is decreasing in the rate of adoption means that technology adoption does provide incentives to improve compliance when firms are regulated by TEPs. These incentives are directly related to the decrease in the permit price. Since the equilibrium price of permits falls with adoption, there is a decrease in the marginal benefit of violating and consequently the extent of violation is reduced. The permit price that clears the market for each adoption rate is given by the equilibrium between supply and demand for permits. The supply of permits is determined by the total quantity of permits allocated by the environmental authority, and the demand for permits is the sum of the permits all firms decide to hold in equilibrium. The equilibrium permit price is given by:

(22)

k   = = lNA  if λ 0  pNA L ∑ i =1   αn n   p= l A + ∑ lNA  if λ = α.  pPA L = ∑ =i 1 =i α n   n    pUA L ∑ = = l A  if λ 1 i =1  

The permit price is increasing in monitoring probability given that when the monitoring probability increases, the demand for permits increases. Permit price is decreasing in technology adoption: the larger the fraction of firms adopting the new technology, the lower the demand for permits and therefore the larger the reduction in permit price. Hence, pNA > pPA > pUA .

Some authors have studied how targeting enforcement efforts to specific groups of firms can induce greater compliance with regulations (Harrington, 1988; Russell, 1990; Hardford, 1991; Hardford and Harrington, 1991; Livernois and McKenna, 1999; Hentschel and Randal, 2000; Friesen, 2003; Rousseau, 2007). 16

16

The adoption savings are given by: (21) ∆EC = c(eNA ( p (π , λ )) − kc(e A ( p (π , λ ))) + p (π , λ )[l NA ( p (π , λ ), π , F ) − l A ( p (π , λ ), π , F )] . Adoption savings decrease as the new technology is diffused into the industry. The reduced price reduces the adoption savings and prevents high adoption cost firms from overinvesting since they can buy cheaper permits instead of investing. On the other hand, the adoption savings increase with the monitoring probability due to the increase in the permit price. IV. POLICY INSTRUMENTS COMPARISON In this section we compare taxes and TEPs under three criteria: (1) extent of violation in reported emissions and permit holdings of adopters and non-adopters of the new technology , (2) enforcement efforts, and (3) social welfare. 4.1

Extent of violation in reported emissions and permit holdings In previous sections we analyzed the extent of violations in reporting and the

quantity of permits held in equilibrium for the two alternative economic instruments in a context of technology adoption. The compliance incentives are given by the comparison between marginal expected costs of violation and marginal benefits of non-compliance. The marginal expected cost of non-compliance is the marginal expected sanction and is the same for adopters and non-adopters of the new technology since there are no targeted enforcement strategies. Therefore, the extent of violation is determined by the marginal benefits of non-compliance. If the tax rate is higher than the equilibrium permit price, then the extent of violation turns out to be higher under tax regulation. 4.2

Enforcement strategy for perfect compliance So far we have assumed that the regulator sets the monitoring probability π

regardless of the regulatory scheme. However, if the objective of the enforcement

17

authority is to guarantee perfect compliance, and since the instruments differ in the violation they induce, monitoring probabilities must vary between instruments and rates of adoption. 17 Let us assume that the sanctioning structure [ f , g ] is constant and that the enforcement authority only adjusts the monitoring probability in order to guarantee perfect compliance. Table 1 presents the minimum monitoring probabilities required for the extent of violation to be zero under the two alternative instruments and three possible rates of adoption. Table 1. Minimum monitoring probabilities for comparison of extent of violation.

Taxes

Adoption rate

Tradable permits

Zero

F '(0) Permits π min − NA = pNA

Partial

F '(0) Permits π min − PA = P

PA

Universal

π

Taxes min − NA

= F '(0)

t

F '(0) Permits π min −UA = pUA

Notice that under TEPs, but not under taxes, the minimum monitoring probability required for zero violation is a function of the rate of adoption. This result is in contrast to Amacher and Malik (1998), who find that the associated costs of enforcement faced by the regulator in a tax system depend on a firm’s choice of technology. The key difference is driven by the assumption of a regulator engaging in ex-ante regulation. Amacher and Malik (1998) assume a bargaining process where the

Assuming perfect compliance is not a rare assumption in the literature; see e.g., Malik (1990), Malik (1992), Amacher and Malik (1998), Stranlund and Chávez (2000), and Chávez et al. (2008).

17

18

regulator offers the firm a less stringent policy if the firm agrees to employ a more environmentally friendly technology. Thus, since compliance behavior is not affected by the adoption decision under taxes, the monitoring effort required to guarantee perfect compliance is independent of the adoption rate as well. In contrast, the enforcement effort to achieve perfect compliance decreases with the rate of adoption under TEPs.

4.3

Social Welfare In order to compare instruments, we define a social welfare function, W .

Social welfare is given by aggregate damages from emissions, D(e) , the abatement costs, C (e) , the investment costs, I (λ ) , and the expected enforcement costs, E (π ) : 18 (29)

W= − D(e) − C (e) − I (λ ) − E (π ) .

Damages from emissions and abatement costs are a function of emissions level, which in the case of partial technology adoption should be calculated as the weighted average of adopters’ and non-adopters’ emissions: (1 − α )e NA + α e A  . Investment cost, in the case of partial adoption, is given by the expenditure of the lower adoption cost firms, α k . In the case of universal adoption, such investment is equal to α k + (1 − α )k . We assume that the enforcement efforts are intended to achieve perfect compliance. 19 Therefore, the monitoring probabilities are set in order to induce zero Notice that this is not an analysis of the cost-effectiveness of the instruments. Chávez et al. (2008) analyze the cost-effectiveness of a tradable emissions permit system in the presence of costly enforcement, and conclude that a conventional tradable permits program cannot be cost-effective since the individual firms under a transferable emissions permit system do not internalize the monitoring costs required to induce perfect compliance. We do not address this point in our analysis. 19 If we keep the assumptions that the regulator sets the enforcement strategy regardless of the selected policy instrument and that it is not necessarily intended to achieve perfect compliance, the expected enforcement becomes equal across instruments and does not affect the ranking of instruments in terms of social welfare. 18

19

violations and the expected enforcement costs are given by the monitoring costs times the number of audits required to achieve perfect compliance. 20 If we assume that the fixed cost of auditing is equal to x , then total expected enforcement costs are given by E (π ) = xπ mín . Since policy instruments differ in the adoption profits they induce, there are six potential alternative welfare scenarios to analyze, as shown in Table 2. Table 2. Scenarios for welfare comparison.

Scenario

Adoption Rate Taxes

Permits

1

Zero

Zero

2

Partial

Zero

3

Universal

Zero

4

Partial

Partial

5

Universal

Partial

6

Universal

Universal

See Appendix A for details of the components of the welfare function under each scenario of adoption In Scenario 1, there is no adoption under either scheme. Therefore, there is no difference in provided social welfare between the two policy instruments, since they

If we instead assume, like in Sections 1 and 2 of the present paper, that the monitoring probability is exogenous and is set regardless of the regulatory instrument used, we would have the same expected enforcement for both policies in all the cases. 20

20

are both set at the same level and induce the same level of abatement, investment, and extent of violation. 21 In Scenarios 2 and 3, there is no technological adoption under permits. The permit price remains equal to the tax. Hence, both policies imply the same extent of violation and expected enforcement costs. However, since there is more technology adoption under taxes, the investment costs are higher and emissions are lower with taxes than with permits, implying higher abatement costs. Finally, the lower emission level implies less damage from emissions. In Scenarios 4 and 6, both schemes induce partial and universal adoption, respectively. Since the permit price is lower than the tax (with the gap being larger in Scenario 6), violations and expected enforcement costs are lower under permits. Investments costs are the same, while permits are preferred to taxes when it comes to abatement costs. Again, the lower emissions level under taxes induces less damage from emissions, as in the previous cases. Finally, in Scenario 5, taxes lead to universal adoption while permits lead to partial adoption. The permit price is lower than the tax, implying less violation and lower expected enforcement costs. In addition, permits imply lower investment costs and more damage from emissions. In conclusion, since abatement costs, investment costs, and expected enforcement costs are never larger under permits, the critical element determining the ranking between taxes and permits is the damages from emissions function. On one hand, there are lower damages from emissions with taxes. On the other hand, abatement costs, investment costs, and expected enforcement costs with taxes are never lower than with permits. If the reduction in damages generated by a lower level of emissions under taxes is higher than the increase in abatement costs, 21

Remember that under perfect compliance and zero technology adoption, the permit price equals the tax rate and therefore the abatement levels under the two policy schemes coincide.

21

investment costs, and expected enforcement costs, then taxes outperform permits in the social welfare function. The steeper the marginal damage from emissions, the larger the drop in damages produced by a reduction in emissions and therefore the higher the probability that taxes perform better than permits. The emission level targeted by the policy also has an impact on the probability of taxes outperforming permits. If the policy is set such that the required final emission level is higher than the optimal level, then the marginal damages at the target level are higher than the marginal damages at the optimal level, and therefore, the probability for taxes to perform better than permits increases. V.

CONCLUDING REMARKS

The results presented in this paper are important for choosing and designing environmental regulations and their enforcement strategies. We analyze how the choice of policy instruments affects the incentives to comply with environmental regulations and to adopt new technologies in a context of technological change and incomplete enforcement. We have shown three main results: First, compliance incentives are affected by the technology adoption rate under TEP regulation but not under taxes. Indeed, the larger the adoption rate in a TEP system, the lower the permit price and therefore the greater the incentives to comply with the regulation. The fact that the emissions price is fixed by the regulator under taxes while it decreases under TEP creates a wedge between them and between the rates of adoption and compliance they induce. Therefore, the expected enforcement costs necessary to guarantee perfect compliance under TEPs are lower than under taxes. This becomes relevant in a setting where expected enforcement costs are an important component of the regulation costs and when the regulatory agency is budget constrained and has as its main objective to achieve perfect compliance in reported emissions and emission permit holding.

22

Second, the adoption rate under taxes is not influenced by the compliance behavior of firms, while under TEPs it is. In a setting of imperfect compliance, if the main purpose of the regulator is to spread the use of a new abatement technology to achieve a lower level of final emissions, the traditional result that taxes are preferred over TEP regulations is not affected by the presence of weak enforcement and imperfect compliance. Third, social welfare is composed of four elements that vary with the rate of adoption that each policy induces, i.e., (i) damages from emission, (ii) abatement costs (iii) investment cost, and (iv) expected enforcement costs, and we conclude that taxes never perform better than permits in terms of abatement costs, investment costs, and expected enforcement costs. However, the picture is different if we look at damage from emissions, since taxes induce less emission damage than permits. Therefore, the final ranking will depend on the relative weight given to emission damages compared to the other effects. As stated earlier, for our welfare analysis we considered the monitoring probability that ensures perfect compliance. A different result may arise if the only objective of the enforcement agency is to minimize aggregated emissions, as was explored by Macho-Stadler (2006). There are some other aspects that in practice do affect the welfare comparison and that are outside the analysis of the present paper. Differences in distributional consequences and differences in political acceptability of the instruments are some of them. The stringency of the tax and the TEPs system is subject to complicated political economy process for instance. The regulator may know that permit prices will fall during the course of a TEPs program. She may therefore make the TEP scheme tougher than she would have with a Tax scheme – to counteract. One aspect not addressed in this paper is the effect of the rate of adoption on the optimal enforcement strategy. If the regulator wants to minimize enforcement, she/he could modify the parameters of the enforcement strategy in response to the 23

adoption process, varying the probability of monitoring or the sanctions schemes. If firms could foresee this behavior, they could modify their initial adoption decisions, which in turn could affect social welfare and the incentives to comply with the regulation. In this sense, a hold-up problem arises, since the ex-post optimal behavior of the regulator is not consistent with the optimal incentives provided to firms exante. VI.

REFERENCES

Amacher, Gregory S., and Arun S. Malik. 1998. “Instrument choice when regulators and firms bargain”. Journal of Environmental Economics and Management

35:225-241

Baldursson, Fridrik M., Nils-Henrik M. von der Fehr 2004. “Prices vs. quantities: The irrelevance of irreversibility”. Scandinavian Journal of Economics 106 (4). 805-821. Becker, Gary., 1968. “Crime and Punishment: An Economic Approach”. The Journal of Political Economy 76 (2), 169-217. Chavez, Carlos A., Mauricio G. Villena, and John K. Stranlund. 2008. “The Choice of Policy Instruments to Control Pollution under Costly Enforcement and Incomplete Information.” Revised and resubmitted to the Journal of Applied

Economics.

Cohen, MarK.A. 1998. “Monitoring and Enforcement of Environmental Policy”. International Yearbook of Environmental and Resource Economics. Volume III, edited by Tom Tietenberg and Henk Folmer; Edward Elgar publishers (1999). Decker, Christopher S., and Christopher R. Pope. 2005. “Adherence to environmetal law: the strategic complementarities of compliance decisions”. The Quarterly Review of Economics and Finance. 45: 641-661. Finkelshtain, Israel.., Yoav Kislev, 1997. Prices versus quantities: the political perspective. Journal of Political Economy 105, 83–100. Friesen, Lana, 2003. “Targeting enforcement to improve compliance with environmental regulations”. Journal of Environmental Economics and Management 46, 72–85.

24

Harford, Jon. D. 1978. “Firm Behavior under Imperfectly Enforceable Pollution Standards and Taxes”. Journal of Environmental Economics and Management 5(1), 26-43. Harford, Jon.D. 1987. “Self-reporting of Pollution and the Firm’s Behavior under Imperfectly Enforceable Regulations”. Journal of Environmental Economics and Management 14(3), 293-303. Harford, Jon D. 1991. “Measurement error and state-dependent pollution control enforcement”. Journal of Environmental Economics and Management 21, 67– 81. Harford Jon.D., Winston Harrington. 1991. “A reconsideration of enforcement leverage when penalties are restricted”. Journal of Public Economics 45, 391–395. Harrington Wiston. 1988. “Enforcement leverage when penalties are restricted”. Journal of Public Economics 37, 29–53. Hentschel, E. and Randall, A. 2000. “An integrated strategy to reduce monitoring and expected enforcement”. Environmental and Resource Economics 15: 57-74. Hoel, Michael., Larry Karp., 2002. “Taxes versus quotas for a stock pollutant”. Resource and Energy Economics 24 (4), 367–384. Jaffe, Adam., Richard G. Newell, Robert Stavins. 2002. “Environmental Policy and Technological Change”, Environmental and Resource Economics 22, 41-69. Jones, Carol A., Suzanne Scotchmer. 1990. “The social cost of uniform standards in a hierarchical government” Journal of Environmental Economics and Management 19, 61-72. Kneese, Allen., Charles L. Schultze. 1978. “Pollution, Prices, and Public Policy,” The Brookings Institute, Washington, DC . Kaplow, Louis., Steven Shavell, 1994. “Optimal Law Enforcement with SelfReporting of Behavior”. The Journal of Political Economy 102(3), 583-606. Lai Ching-Chong, Yang Chih-Yu, Chang Juin-Jen. 2003. “Environmental Regulations and Social Norms”. International Tax and Public Finance 10, 63-75. Livernois John., McKenna C.J. 1999. “Truth or consequences – enforcing pollution standards with self-reporting”. Journal of Public Economics 71, 415–440.

25

Macho-Stadler, Inés. 2008. “Environmental regulation: choice of instruments under imperfect compliance”. Spanish Economic Review 10 (1), 1-21 Malik, Arun. 1990. “Markets for Pollution Control when Firms are Noncompliant.” Journal of Environmental Economics and Management 18: 97-106. Malik, Arun. 1992. “Enforcement costs and the choice of policy instruments for controlling pollution”. Economic Inquiry, October, 714-721. Malik, Arun. 1993. “Self-Reporting and the Design of Policies for Regulating Stochastic Pollution”. Journal of Environmental Economics and Management 24(3), 241-257. Montero, Juan Pablo. 2002. Prices vs. quantities with incomplete enforcement, Journal of Public Economics Vol. 85, 435-454. Montero, Juan Pablo , Jessica Coria (2008). Instrument choice with costly regulatory adjustments. Working paper. Moledina, Amyaz .A., Jay S. Coggins., Stephen Polasky., Christopher Costello. 2003. Dynamic environmental policy with strategic firms: prices versus quantities. Journal of Environmental Economics and Management 45 (2S), 356–376. Quirion, Philippe., 2004. Prices versus quantities in a second-best setting. Environmental and Resource Economics 29 (3), 337–359. Requate, Till. 1995. “Incentives to adopt new technologies under different pollution control policies”. International tax and Public Finance (2), 295-317 Requate, Till., Wolfram Unold, 2001. “On the incentives created by policy instruments to adopt advanced abatement technology if firms are asymmetric”. Journal of Institutional and Theoretical Economics 157 (2001), 536-554. Requate, Till., Wolfram Unold, 2003. “Environmental policy incentives to adopt advanced abatement technology: Will the true ranking please stand up?”. European Economic Review 47 (2003) 125-146. Requate, Till, 2005. “Dynamic incentives by environmental policy instruments – a survey”. Ecological Economics 54, 175-195 Roberts, Marc J., Michael Spence., 1976. Effluent charges and licenses under uncertainty. Journal of Public Economics 5, 193–208.

26

Rousseau, Sandra., Stef Proost. 2005. “Comparing Environmental Policy Instruments in the Presence of Imperfect compliance – a Case Study” Environmental and Resource Economics. 32: 337-365 Rousseau, Sandra. 2007. “Timing of environmental inspections: survival of the compliant” Journal of regulatory economics. 32: 17-36 Sandmo, Agnar. 2002. “Efficient Environmental Policy with Imperfect Compliance”. Environmental and Resource Economics 23 (1), 85-103. Stranlund, John K., Kanwalroop K. Dhanda. 1999. “Endogenous Monitoring and Enforcement of a Transferable Emissions Permit System.” Journal of Environmental Economics and Management 38(3): 267-282. Stranlund, John K., Carlos A. Chávez. 2000. “Effective Enforcement of a Transferable Emissions Permit System with a Self-Reporting Requirement”, Journal of Regulatory Economics, vol. 18 (2): 113-131, September. Stranlund, John K. 2007. “The Regulatory Choice of Noncompliance in Emissions Trading Programs.” Environmental and Resource Economics 38(1), 99-117. Stranlund, John K., Carlos A. Chávez and Mauricio G. Villena. 2009. “The optimal pricing of pollution when enforcement is costly” Forthcoming in Journal of Environmental Economics and Management doi:10.1016/j.jeem.2008.12.002 Stranlund, John K., Yakov Ben-Haim. 2008. “Price-based vs. quantity-based environmental regulation under Knightian uncertainty: An info-gap robust satisficing perspective.” Journal of Environmental Management 87(3): 443-449. Yohe, Gary.W., 1978. Towards a general comparison of price controls and quantity controls under uncertainty. The Review of Economic Studies 45 (2) , 229– 238.

27

APPENDIX A. Table A. 1. Social welfare comparison by components.

Combination

Damages from abatement

Abatement Costs

Investment Costs

Expected enforcement costs

Taxes = Permits

Taxes =

Taxes =

Permits

Permits

Permits  Taxes

Permits  Taxes

Taxes = Permits

Permits  Taxes

Permits  Taxes

Taxes = Permits

Permits  Taxes

Taxes = Permits

Permits  Taxes

Permits  Taxes

Permits  Taxes

Permits  Taxes

Permits  Taxes

Taxes = Permits

Permits  Taxes

Scenario 1. Taxes λ = 0 . Permits λ = 0 Comparison

Taxes = Permits

(t = PNA ) Scenario 2. Taxes λ = α . Taxes λ = α Comparison

Taxes  Permits

(t = PNA ) Scenario 3. Taxes λ = 1 . Permits λ = 0 Comparison

Taxes  Permits

(t = PNA ) Scenario 4. Taxes λ = α . Permits λ = α Comparison

Taxes  Permits

(t > PPA ) Scenario 5. Taxes λ = 1 . Permits λ = α Comparison

Taxes  Permits

(t > PPA ) Scenario 6. Taxes λ = 1 . Permits λ = 1 Comparison (t > PUA )

Taxes  Permits

28

Conditional cooperation and social groups –

Experimental results from Colombia 1

Peter Martinssona, Clara Villegas-Palaciob,c, Conny Wollbrantd,e a

Department of Economics. School of Business, Economics and Law. University of Gothenburg. Gothenburg, Sweden. E-mail: [email protected] b Department of Economics. School of Business, Economics and Law. University of Gothenburg. Gothenburg, Sweden. E-mail: [email protected] c Facultad de Minas. Universidad Nacional de Colombia – Sede Medellín. Medellín, Colombia. E-mail. [email protected] d Department of Economics. School of Business, Economics and Law. University of Gothenburg. Gothenburg, Sweden. E-mail: [email protected] e Corresponding author at: Department of Economics. School of Business, Economics and Law. P O Box 640 SE 405 30. Gothenburg, Sweden. Phone +46-(0)31- 7862642. Fax +46-(0)31-786 1043

Abstract We explore conditional cooperation in different social groups using a one-shot public goods experiment, and find significant differences in behavior between different social groups.

Keywords: Conditional cooperation, Experiment, Public goods, Social group. JEL classification: C91, H41.

1

Acknowledgments: Financial support from the Swedish Research Council (Vetenskapsrådet) and the Swedish International Development Cooperation Agency (Sida) to the Environmental Economics Unit at the University of Gothenburg is gratefully acknowledged. We are grateful to the School of Engineering (Escuela de Ingeniería de Antioquia), Medellín, Colombia, and to Universidad Nacional de Colombia, Sede Medellín for their support in running the experiment. We are grateful to Antonio Villegas-Rivera for excellent logistic support and to Marta Matute, and Alba Upegui for helpful comments.

1. Introduction Voluntary contribution to public goods is frequently found both in the field and in the laboratory (e.g., Gächter, 2007). Fischbacher et al. (2001) developed a one-shot public goods experiment in which subjects are asked for their: (i) unconditional contribution to a public good, as in standard public goods experiments, and (ii) conditional contribution to the public good given all possible average contributions (rounded to the nearest integer) of other group members. By investigating the profile of conditional contributions, subjects can be grouped into contributor types such as free-riders and conditional cooperators (i.e., their degree of cooperation is conditional on their beliefs about others’ cooperation). Early evidence from type classification following Fischbacher et al.’s approach used university students in Western countries as subjects (see, e.g., Gächter, 2006, for an overview). Generally, conditional cooperators are the dominating type. For example, Fischbacher and Gächter (2006) find that 55% of subjects are classified as conditional cooperators, whereas 23% are free-riders. However, most of the conditional cooperators are not perfect conditional contributors but contribute slightly less than others. Kocher et al. (2008) replicated the experiment by Fischbacher et al. (2001) in three different countries and found differences in both the distribution of types and the share of conditional cooperation. Herrmann and Thöni (2009) conducted the same experiment in two rural and two urban locations in Russia and found that their fractions of conditional cooperators varied from 48% to 60% within location, but that the differences between the locations were insignificant. The evidence from studies testing the effect of cultural background on behavior using a standard multi-period public goods

game has been mixed as well (e.g., Brandts et al., 2004; Burlando and Hey, 1997; Herrmann et al., 2008). When comparing experimental findings between locations, several factors may drive differences: (i) cross-country differences (e.g., religion and social norms); (ii) within-country differences (e.g., rural versus urban areas); and (iii) social group differences (e.g., age, trust, and income). Kocher et al. (2008) varied cross-country factors but did not control for within-country differences, while Herrmann and Thöni (2009) tested for within-country differences along the rural-urban dimension. In both cases, university students were used as subjects. The objective of the present paper is to investigate how social group differences as measured by social class belonging affect cooperative behavior. Using the design of Fischbacher et al. (2001) and controlling for cross-country and within-country differences, we vary along one dimension of the social group – social class belonging. We use university students recruited from two universities in Medellin, Colombia, who differ in social class belonging: (i) socio-economic strata 2 and 3 (i.e., the “medium-low” group) and (ii) socio-economic strata 4, 5, and 6 (i.e., the ”high” group). 2 A substantial part of public goods are local, e.g., teamwork and local environmental public goods governed by common property regimes such as lakes, pastures and irrigation systems. Heterogeneous cooperation preferences among social groups may lead to different provision levels of a public good, suggesting the need for differential policies to achieve a certain level of the public good (see Gächter, 2006, for a 2

There are six social strata in Colombia: 1 (low- low), 2 (low), 3 (medium- low), 4 (medium), 5 (mediumhigh), and 6 (high). Strata 1-3 receive domestic public service subsidies such as provision of water, electricity, and gas; 5-6 pay additional contributions toward the cost of public services while 4 does not, but this group does not receive subsidies either. The strata are indirect indicators of people’s socio-economic conditions.

policy discussion). Thus, it is important to investigate heterogeneous cooperation preferences across social groups.

2. Experimental design and procedure

We employ a standard linear public goods experiment following the same format as Fischbacher et al. (2001), where subject i’s payoff in tokens is given by 4

π i = 20 − ci + 0.4∑ ci ,

(1)

i =1

where 20 is the endowment and c the amount invested in the public good. Each group consists of four randomly matched members. The marginal return from the public good is set to 0.4, ensuring a conflict between the dominant strategy to contribute zero, i.e., to free-ride, and the full contribution Pareto optimum solution. Subjects are asked to indicate how much they would like to contribute both unconditionally and conditionally to the public good. In the case of conditional contributions, subjects are asked how much they would like to contribute conditional on the average contribution of the other members of the group, which ranges from 0 to 20 (i.e., the strategy method). To make each decision incentive compatible, the payoff relevant decision for three randomly selected members was the unconditional contribution. By using their average unconditional contribution, the contribution of the fourth member is given by his or her contribution table. Then, each member’s monetary payoff can be calculated using equation (1). After the experiment, subjects were asked to guess the total contribution of the other three group members, and accuracy of guesses was monetarily rewarded.

Again, the experiments were conducted at one socio-economically “medium-low” (Universidad Nacional de Colombia) and one “high” university (Escuela de Ingeniería de Antioquia), both in Medellín, Colombia. 3 At both places, we ran two sessions with 24 subjects each; students of mathematics, psychology, and economics were excluded. The procedure of the experiment was the same at both places. An experimental session consisted of six stages: (i) the handed-out instructions were read aloud; (ii) examples and individual exercises; (iii) decision making; (iv) beliefs elicitation; (v) socio-economic questionnaire; and (vi) cash payment in private upon departure. Each session lasted approximately 90 minutes and the payoffs were calibrated to reflect opportunity costs. For the medium-low group, each token equaled 750 Colombian Pesos (COP) while the corresponding figure was 1,000 COP for the high group. 4 Average earnings were 25,000 COP for the high group, and 23,000 COP for the medium-low group (both figures include a show-up fee of 5,000 COP).

3. Results We follow the standard approach when defining the four contributor types (see Fischbacher et al., 2001). Conditional contributors submit a contribution table showing a monotonically increasing own contribution for an increasing average

3

At Universidad Nacional de Colombia (the ”medium-low” group), around 80% of the student population belong to strata 2 and 3, 11% to stratum 4, and only 5% to strata 5 and 6 (see Rico, 2005). This is a public university where the cost of a six-month term is around one minimum monthly salary for students of stratum 3. At Escuela de Ingeniería de Antioquia, a private university, students mainly belong to a strata 4, 5, and 6, and the cost is 10 times higher. 4 In cases with samples with different opportunity costs, either the absolute amount or the opportunity cost can be kept constant. Although stake size does not seem to matter in one-shot public goods game (Kocher et al., 2008), we decided on opportunity cost. The exchange rate at the time of the experiment was 1USD= approximately 2,000 COP.

contribution of the other members. 5 Free-riders are characterized by a zero contribution for every possible average of the other members. Unconditional contributors submit the same positive contribution independent of others’ average contribution. Hump-shape contributors (also known as triangle contributors) show monotonically increasing contributions up to a given average level of others’ contributions, after which their contributions decrease. The category referred to as Others constitutes the remaining participants. Table 1 displays the distributions of types by social group. The dominating type is conditional cooperators; 51% and 62% of the high group and the medium-low group, respectively. This is very close to the figures reported by, e.g., Fischbacher et al. (2001) and Fischbacher and Gächter (2006). Interestingly, 25% of the subjects in the high group were classified as free-riders compared to 4% in the medium-low group. We reject the null hypothesis of no differences in distribution of types between groups at the 5% significance level (p=0.03; Chi2-test). This is explained by a rejection of the hypothesis of no differences in share of free-riders between the two groups at the 1% significance level (p=0.004; Chi2-test). Table 1 also presents the average unconditional contribution for each type; the difference between the groups is not statistically significant. >>> Table 1

The relationship between subjects’ own conditional contribution and the average contribution of other group members is shown in Figure 1. When the average contribution of others is zero, subjects in the medium-low group contribute more than 5

We also include those without a monotonically increasing contribution but with a highly significant (at one percent) positive Spearman rank correlation coefficient between own and others’ contributions (see Fischbacher et al., 2001; Fischbacher and Gächter, 2006).

those in the high group, indicating a higher degree of altruism. Also, the difference in slope between the perfect conditional cooperation line and the plotted line, which represents degree of self-serving bias, is significantly larger in the high group. The regression results confirm the results in Figure 1. >>> Figure 1 Using two-sided Mann-Whitney U-tests, we find no significant difference in mean unconditional contribution between groups; it is 7.98 tokens in the medium-low and 7.68 in the high group (p=0.75). These levels of unconditional contributions around 40% of the endowment are in line with earlier findings (e.g., Kocher el at., 2008). We elicited beliefs about others’ contribution in the unconditional case, and find no significant differences in beliefs between the high group (8.83) and the medium-low group (8.23) (p=0.71). Furthermore, regression results reveal that both groups can be classified as imperfect conditional cooperators (Table 2). In addition, the high group displays a significantly higher level of self-serving bias, which is similar to findings from the analysis of the conditional contribution tables; however, while the degree of altruism is positive, it is insignificant. >>> Table 2

4. Conclusion By conducting the experiments in the same location and using university students, we hold cross- and within-country differences constant. Our results suggest that different social groups exhibit differences in terms of both composition of types and extent of

conditional cooperation. How much this explains the differences found in, e.g., the crosscontinent study by Kocher et al. (2008) is an issue for future research. Our results are of policy relevance, showing why general policies within a location may lead to heterogeneous behavioral responses. As a consequence, policy makers may need to consider differential policy schemes. Following Gächter (2006), a social group where most individuals are conditional cooperators needs policies that sustain beliefs for cooperation of its integrants. In contrast, where free-riding dominates, policies involving monitoring and penalties may be required to enhance cooperation.

References Brandts, J., Saijo, T, Schram, A.J.H.C., 2004. A four country comparison of spite and cooperation in public goods games. Public Choice 119: 381-424. Burlando, R.M., Hey, J.D., 1997. Do Anglo-Saxons Free Ride More? Journal of Public Economics 64: 41-60. Fischbacher, U., and Gächter, S.,2006. Heterogeneous Social Preferences and the Dynamics of Free riding in Public Goods. CeDEx Discussion Paper No. 2006-01, University of Nottingham. Fischbacher, U., Gächter, S., Fehr, E., 2001. Are people conditionally cooperative? Evidence from a public goods experiment. Economic Letters 71: 397-404. Gächter, S., 2007. Conditional cooperation: Behavioural regularities from the lab and the field and their policy implications. In: Psychology and economics: A promising new cross-disciplinary field (Cesinfo seminar Series), Eds. B.S. Frey and A. Stuzter, Cambridge, MIT Press, 19-50. Gächter, S., Herrmann, B.,2006. The limits of self-governance in the presence of spite: Experimental evidence from urban and rural Russia. Working Paper 2006-13, University of Nottingham. Herrmann, B., Thöni, C.,2009. Measuring conditional cooperation: A replication study in Russia. Experimental Economics 12, 87-92. Herrmann, B., Thöni, C., Gächter, S.,2008. Antisocial punishment across societies. Science 319: 1362-1367. Kocher, M., Cherry, T., Kroll, S., Netzer, R., Sutter, M.,2008. Conditional cooperation on three continents. Economics Letters 101: 175-178.

Kocher, M. Martinsson, P., Visser, M.,2008. Does stake size matter for cooperation and punishment? Economics Letters 99: 508-511. Rico, D.,2005. Evaluación del costo de las matrículas en la Universidad Nacional de Colombia, Sede Medellín. Colombia.

Oficina de Planeación. Universidad Nacional De

Table 1. Distribution of player types, average unconditional contribution, and guessed contribution. High socio-economic group

Medium-low socio-economic group

Avg. uncond. Avg. guessed Districontrib. contribution bution Distribution Unconditional cooperators 0.00% 0 (0) 0 (0) 4.17% Conditional cooperators 54.17% 9.64 (4.68) 9.88 (4.78) 62.5% Hump-shape contributors 8.33% 6.5 (7.85) 11 (7.53) 8.33% Free-riders 25.00% 3.83 (7.02) 6.5 (7.43) 4.17% Others 12.5% 8 (4.47) 7.67 (4.23) 20.83% Note: Avg. uncond. contrib = average unconditional contributions; standard errors in parentheses Avg. guessed. contrib = average guessed contributions; standard errors in parentheses

Avg. uncond. contrib. 0.5 (0.71) 9.33 (5.12) 8.75 (7.46) 0.5 (0.71) 6.6 (3.95)

Avg. guessed contribution 0 (0) 9.5 (4.93) 8.25 (6.4) 2 (0) 7.3 (4.16)

Table 2. Regression results (dependent variable: unconditional contribution). Coefficients Standard error Constant 0.181 0.984 Beliefs 0.948** 0.102 Beliefs x High socio0.140 -0.312* economic High socio-economic 1.778 1.387 Note. *** denotes significance at the 1% level and * denotes significance at the 5% level. Tobit regression gives similar results.

Average own conditional contribution

Figure 1. Average own contribution.

20 18 16 14 12 10 8 6 4 2 0 0

5

10

15

20

Average contribution of the other three group members Perfect conditional Cooperation

Medium-low Social Group

High Social Group

Do in-group and out-group disclosure affect cooperation?* Peter Martinssona

University of Gothenburg, Sweden Clara Villegas-Palaciob

University of Gothenburg, Sweden Universidad Nacional de Colombia- Sede Medellín, Colombia

Abstract

This paper investigates the effects of in-group and out-group disclosure on contributions to a public good using the one-shot experimental design by Fischbacher et al. (2001). We conduct a between-subjects laboratory experiment with four treatments, and find that disclosure does not significantly increase the contributions to a public good or the proportion of conditional cooperators. Some disclosure schemes affect both the proportion of full contributors to a public good and the heterogeneity in cooperation behavior. Our results are of high policy relevance since in some circumstances, disclosure schemes may have negative effects on cooperation.

JEL classification: C91, H41. Key words: Conditional cooperation, Disclosure, Image motivation, Public Good Experiment.

Acknowledgments: Financial support from the Swedish Research Council (Vetenskapsrådet) and Sida (the Swedish International Development Cooperation Agency) to the Environmental Economics Unit at the University of Gothenburg is gratefully acknowledged. We have received valuable comments from Conny Wollbrant. We are grateful to Lina María Berrouet for excellent research assistance. a Department of Economics, University of Gothenburg, Gothenburg, Sweden; e-mail: [email protected]; Ph +46 31 786 52 55. b Corresponding author; Department of Economics, University of Gothenburg, Gothenburg, Sweden. e-mail: [email protected]; Ph: +46 31 786 26 42.

1. Introduction Public good experiments have consistently found that average contribution levels to public goods are significantly higher than the zero contribution predicted by standard economic theory (e.g., Leyard, 1995; Zelmer, 2003). Ariely et al. (2009) discuss three broad motivations for individuals to behave pro-socially: (i) intrinsic, (ii) extrinsic, and (iii) image motivation. While the intrinsic and extrinsic motivations focus on factors such as altruism and monetary rewards, image motivation focuses on the fact that an individual derives utility from how other people perceive him or her. To achieve this utility, an individual can signal good behavior to others, and the particular type of action fulfilling this is of course related to the prevailing norms in the society (see, e.g., Akerlof, 1980; Ellingsen and Johannesson, 2008). The present paper focuses on image motivations, and in particular analyzes the effect of different disclosure schemes on contributions to a public good. Traditionally, identities and contributions are not revealed to participants in laboratory public goods experiments. However, as discussed in, e.g., Andreoni and Petrie (2004), many situations in daily life do consist of publically revealing both identities and contributions. These situations range from public announcements at fundraising events to official reporting of pollution levels of companies, with the common purpose of using social approval to increase contributions to public goods. Rege and Telle (2004) tested whether social approval affects cooperation in a one-shot public good experiment. In the treatment where the subjects themselves revealed their contributions to the other members of their group, contributions to the public good were significantly higher compared to in a standard setting with no disclosure. Interestingly, while Weimann (1994) and Croson (2001) find in multi-period public good experiments that disclosing

1

information to other group members about contributions without revealing any identities has no significant effect on contribution, Sell and Wilson (1991) find, based on a similar set-up, that disclosure of contribution, but not identity, does have a significant impact on contributions. In a field experiment, Soetevent (2005) investigated church offering used for internal causes, which he defined as a public good, among 30 churches in the Netherlands, and did not find a significant difference related to whether money was collected in open collection baskets or in closed collection bags. The above experiments focused on disclosing information about contributions to group members only, i.e., in-group disclosure. To our knowledge, the effects of disclosing information about contributions only to group members and of disclosing information to all participants in an experiment including non-members of the group have not yet been compared. An exception in the public good experiment literature is Rege and Telle (2003), where each participant’s identity and contribution to the public good is revealed to a third-party but not to other group members. 1 Andreoni and Bernheim (2009) study choices of subjects in a dictator game in which the instructions emphasize that everyone present in the lab will observe the outcome associated with each dictator. In their experiment, the dictator chooses the transfer with probability 1-p, and nature sets it equal to some fixed value (zero or one) with probability p. The probability of intervention of nature and the fixed value are common knowledge, but the receiver cannot observe whether nature intervened. They find that when with probability zero nature decides a transfer equal to zero, 57% of the dictators divided the prize equally, and that when with probability zero nature decides a transfer equal to one, 69% of the dictators divided the prize equally. One of the interpretations of the authors to these findings is that people like others to see them as fair.

1

There is literature on third-party punishment, e.g., Fehr and Fischbacher (2004).

2

Using a one-shot public good game based on the design in Fischbacher et al. (2001), we focus on the effect of three different disclosure schemes: (i) in-group disclosure, where each subject’s identity and contribution is revealed to the group members only, (ii) out-group disclosure, where a subject’s identity and contribution are disclosed to all subjects in the experimental session, but group belonging is not disclosed, and (iii) joint in-group and out-group disclosure, i.e., a subject’s contribution is disclosed to all subjects of the session. We compare the outcome of these disclosure treatments with that of a baseline treatment without disclosure. The outcome of such a comparison can help in designing more efficient information-revealing policies to encourage contributions to a public good. The rest of the paper is organized as follows: Section 2 presents the experimental design and procedures. Section 3 contains the results, which are then discussed in Section 4. Finally, Section 5 offers some concluding remarks.

2. Experimental design Our experiment builds on the experimental design by Fischbacher et al. (2001). In their design, they elicit both unconditional and conditional contributions to a public good. In the unconditional setting, subjects are asked how much they would like to contribute to a public good, which replicates a standard one-shot public good experiment. In the conditional contribution setting, the strategy method is used, i.e., subjects are asked, in what is called the conditional contribution table, how much they would like to contribute to a public good conditional on the average contribution levels (rounded to the nearest integer) of the other group members. Each group consists of four members, but group belonging is not revealed to subjects

3

before or during the experiment. In the disclosure treatments, contributions and identity are revealed. We use a standard linear public good experiment. Each subject is given 20 tokens as endowment and the marginal per capita return from the public good is set to 0.4. Thus, subject i’s payoff in tokens is given by 4

π i = 20 − ci + 0.4∑ ci ,

(1)

i =1

where ci is the amount invested in the public good by individual i. Each token earned in the experiment equaled 750 Colombian Pesos, which corresponded to 0.34 USD at the time of the experiment. In order to make each of the choices incentive compatible, for three of the subjects in each group, the unconditional contribution counts as their contribution to the public good. The contribution from the fourth subject, who was randomly selected, is based on his or her conditional contribution table, where the conditional contribution he or she reported for the average unconditional contributions of the other three members is counted as his or her contribution to the public good. Thus, by adding the three unconditional contributions and the conditional contribution by the fourth member, the total contribution by the group to the public good can be calculated using equation (1). Our experimental design is a 2x2 design. The experimental design is summarized in Table 1. Treatment 1, the baseline treatment, is a standard public goods game setting with complete anonymity regarding both the identity of and the contributions made by the subjects. In Treatment 2 (the out-group disclosure treatment), each subject is asked one at the time, by using the experimental identification numbers, to stand up at the end of the experiment when all participants are gathered, whereby his or her income-relevant decision (the unconditional contribution or conditional contribution) is publically announced by the experimenter to all 4

participants in that session, without any reference to group belonging. In Treatment 3 (the ingroup disclosure treatment), the contributions of the subjects are disclosed to group members only. In this treatment, the four group members come together, one group at a time, in a room next door. Once the four group members are seated, each subject is asked one at a time by using experimental identification numbers to stand up, whereby his or her income-relevant decision (the unconditional contribution or conditional contribution) is publically announced by the experimenter. In Treatment 4 (the joint in-group and out-group disclosure treatment), the four group members are asked to come together, one group at a time, and sit down on four chairs in front of all other participants. Once seated, the income-relevant decision is revealed by the experimenter, using the same procedure as in the other two treatments. Table 1. The 2x2 design of the public goods experiments. Out-group disclosure (Contributions and identity announced to all participants in the session) No Yes Treatment 1 Treatment 2 Standard Public Good game Public good game with No without disclosure only out-group In-group disclosure disclosure (Contributions and identity announced only to group Treatment 4 Treatment 3 members) Public good game with Yes Public good game with both in-group and outonly in-group disclosure group disclosure In addition, we asked the subjects about their beliefs regarding others’ unconditional contribution. As in Gächter and Renner (2006), we monetarily rewarded subjects for accurate guesses. Subjects could earn one additional token by correctly guessing the true contribution. An experimental session consisted of the following stages: At the beginning of a session, participants took the Mach-IV test (Christie and Geis, 1970). According to Vecchio and Sussmann (1991), the resulting test score can be used as a proxy of the degree of an individual’s

5

selfishness. After all participants had completed the Mach-IV test, the experiment instructions were handed out and read aloud to the subjects (the instructions are presented in Appendix 1). Several examples and individual exercises were provided as well. To check for the subjects’ understanding of the experiment, the experimenter publically solved the exercises once all participants had finished answering them. Any additional questions the subjects had were then answered in private. The subjects simultaneously decided how much to contribute unconditionally to the public good, and filled in the contribution table where they indicated their contribution to the group account given the average contribution (rounded to the nearest integer) of the other three members of his or her group, which could range from 0 to 20. After collecting the decision sheets, the participants were asked in writing about their beliefs regarding the total unconditional contribution levels of the other three participants to the public good account. The subjects then completed a socio-economic questionnaire, while at the same time, using the random number generator in EXCEL, the experimenter randomly selected one member in each group for whom the conditional contribution was the income relevant decision and calculated the amount to be paid to each subject. In case of any of the disclosure treatments, the contributionrevealing stage was conducted after collection of the questionnaires. Finally, all subjects were paid privately in cash.

3. Experimental results Our subjects were students at Universidad Nacional de Colombia-Sede Medellín, Colombia. Participants were randomly selected from a list of people who registered in response to an e-mail invitation to participate in the experiment. We ran four treatments (with two sessions per treatment) corresponding to the 2x2 design described in Table 1. In each session, there were

6

24 participants randomly allocated to groups of four. We tested the null hypothesis of homogeneity of participants across treatments. Using the Kruskal-Wallis test we can not reject the null hypothesis in the dimensions of the Mach IV index, income, and number of people in the room that each subject knew from before. Using a chi-square test we can not reject the null hypothesis in the dimension of gender (p-value=0.19). On average, subjects earned 24,000 Colombian Pesos (approximately 10.9 USD) in the 90 minutes that the sessions lasted, including an additional show-up fee of 5,000 Colombian Pesos (approximately 2.27 USD).

3.1

Unconditional Contributions to the Public Good Figure 1 shows the histograms of unconditional contributions per treatment. Compared

to the no disclosure treatment, the main effect of out-group disclosure is a decrease in the proportion of unconditional contributions equal to zero, as well as an increase in the proportion of subjects contributing 50% of the endowment. A similar effect is found for in-group disclosure but additionally, there is an increase in the proportion of unconditional contributions equal to the whole endowment. Interestingly, Treatment 4, which combines in-group and out-group disclosure, results in a uniformly distributed contribution pattern. We performed a KolmogorovSmirnov test of the null hypothesis of equal distributions in treatment pairs, and can not reject the null hypothesis at the 10% significance level in any of the pair-wise tests. We also tested the null hypothesis of normal distribution of the unconditional contribution in each treatment; by using the Skewness-Kurtosis test, we can reject the null hypothesis in Treatment 4 (p-value 0.04) at the 5% significance level, but not in any of the other treatments. Figure 1. Histograms unconditional contributions per treatment.

7

T1. No disclosure

T2. Out-group disclosure

.3

.2

.1

0

T3. In-group disclosure

T4. Joint in-group and out-group disclosure

.3

.2

.1

0 0

5

10

15

20

0

5

10

15

20

Unconditional contribution

Following Rege and Telle (2004), we examined the number of participants who in each treatment contributed everything or nothing (full contributors and non-contributors respectively). The null hypothesis of equal proportions of full contributors in all four treatments is rejected at the 10% significance level (p-value=0.09) using a chi-square test. However, the proportion of non-contributors is not statistically different across the four treatments (p-value=0.75) using the same test. We then conducted pair-wise tests between treatments of the null hypothesis of equal proportions of full contributors and non-contributors, and ended up rejecting only the null hypothesis of equal proportions of full contributors between Treatments 1 and 4 at the 10% significance level (p-value=0.08) and between Treatments 2 and 4 at 5% significance level (pvalue=0.03). Using a chi-square test, we can reject the null hypothesis of equal proportions of subjects contributing 50% of the endowment between Treatments 1 and 2 (p-value=0.08), and in Treatments 2 and 4 (p-value=0.04). The mean unconditional contributions (in percent of endowment) are higher for the treatments that include some degree of disclosure. In the standard public good game, subjects on average contributed 39.9%. Introducing out-group disclosure increases the average contribution

8

to 43.8% whereas introducing in-group disclosure increases the average contribution to 43.2%. Finally, by combining out-group and in-group disclosure, the average contribution increases to 48.1%. We conducted a Wilcoxon-Mann-Whitney test to test the null hypothesis of equal distributions in treatment pairs, and we can not reject the null hypothesis for any of the pairs of treatments at the 10% significance level. 2 Similarly, the null hypothesis of equal distribution of unconditional contributions across the four treatments is not rejected at the 10% level based on a Kruskal-Wallis test (p-value=0.68). Using the Levene test, we also tested the null hypothesis of equal variance in the unconditional contributions in treatment pairs. 3 We can reject the null hypothesis for the pairwise tests involving Treatment 4 and for the Treatments 1 versus 2 test, as shown in Table 2. The highest spread of contributions is in Treatment 4, indicating that more detailed information about individual contributions to a larger audience does not necessarily increase the unconditional contribution to a public good but does increase the heterogeneity in contribution behavior. Table 2. Variance of unconditional contributions in each treatment. Out-group disclosure No

Yes

No

Variance Treatment 1: 29.13

Variance Treatment 2: 19.97

Yes

Variance Treatment 3: 28.10

Variance Treatment 4: 41.60

Treatment 1 vs. Treatment 3

Treatment 2 vs. Treatment 4

In-group disclosure

Ho: Variance is the same across treatments 2

Ho: Variance is the same across treatments Treatment 1 vs. Treatment 2 P(W0)= 0.07

Treatment 3 vs. Treatment 4 P(W0)= 0.03

Treatment 1 vs. Treatment 4

Treatment 1 vs Treatment 2, p-value=0.43; Treatment 1 vs Treatment 3, p-value=0.64; Treatment 1 vs Treatment 4, p-value=0.18; Treatment 2 vs Treatment 4, p-value=0.61; Treatment 3 vs Treatment 4, p-value=0.50. 3 One advantage of Levene’s test (Levene, 1960) is that it is less likely than other tests to be biased for departures from normality.

9

P(W0)= 0.78

P(W0)= 0.009

P(W0)= 0.01

Note: W0 denotes Levene's statistic

3.2

Types of contributors We use the conditional contribution tables to analyze the relationship between a subject’s

own conditional contribution and the average contribution of the other members in his or her group. Following Fischbacher et al. (2001), in Figure 2 we plot the relation between the average own conditional contribution (on the vertical axis) and the other members’ average contribution (on the horizontal axis). The figure shows that the average own conditional contribution is increasing in the average contribution of the other members, which indicates that people on average behave as conditional contributors. This pattern is consistent with, e.g., Fischbacher et al. (2001) and Kocher et al. (2008). As can also be seen, when the average contribution of others is zero, subjects on average contribute more than zero in all treatments, indicating some degree of altruism. The slope represents the degree of self-serving bias; it is less than one, which indicates imperfect conditional cooperation, and not significantly different across treatments.

Figure 2. Average own contribution level for each average contribution level of other group members, by treatment.

10

Average own contribution according to the contribution table

25 20 15 10 5 0 0

5

10

15

20

Average contribution level of other group members

We classify subjects into the four standard categories of contribution behavior defined by Fischbacher et al. (2001): Free-riders, conditional cooperators 4, hump-shape and others. The number of subjects who fall into the different categories is shown in Table 3 together with the average unconditional contribution and the guessed average unconditional contribution of each type of subjects. Comparing our results in the standard public good game setting (Treatment 1) with results in Fischbacher et al. (2001) and Kocher et al. (2008), we find that the proportion of conditional cooperators in our case is higher than the 50% obtained by Fischbacher et al. (2001), but lower than the 80.6% obtained by Kocher et al. (2008) in the US. 5 The number of free-riders in our Treatment 1 is much lower than in the other two studies. The fraction of participants falling into the “others” category is lower in our study as well. 4

We classify subjects as conditional cooperators if their contribution is monotonically increasing with the average contribution of other group members. For subjects with non-monotonically increasing contributions, we count them as conditional cooperators if the Spearman rank correlation coefficient between own and others’ contributions is highly significant (at the 1% level) (as in, e.g., Fischbacher et al., 2001, and Fischbacher and Gächter, 2006). 5 The fraction of conditional contributors in our study falls in between the numbers obtained by Kocher et al (2008) in the US (80.6%) and the figures obtained in Austria (44.4%) and in Japan (41.7%). The proportion of conditional contributors in our study is higher than the figures obtained by Herrmann and Thöni (2009): between 48% and 60% are conditional contributors depending on location in Russia.

11

Table 3. Distribution of contributor types per treatment. Treatment 1 (n=48) Treatment 2 (n=48) Treatment 3 (n=48) Treatment 4 (n=48) Dist. Avg. Guess Dist. Avg. Guess Dist. Avg. Guess Dist. Avg. Guess Avg. Unc. Avg. Unc. Avg. Unc. Avg. Unc. Unc. Cont. Unc. Cont. Unc. Cont. Unc. Cont. Cont. Cont. Cont. Cont. 4.17 0.50 2 4.17 9.00 12 6.25 0.0 2.78 12.50 0.67 3.5 Free-rider % (0.71) (0.47) % (12.7 (2.83) % (0.00) (1,92) % (1.03) (7.13) [2] [2] 3) [3] [6] 9.33 9.48 70.83 8.62 8.57 64.58 9.06 9.62 75.00 11.58 10.17 Conditional 62.50 % (5.12) (4.87) % (3.81) (3.35) % (4.84) (4.95) % (5.48) (5.46) cooperator [30] [34] [31] [36] 4.17 0.50 0.17 0.00 4.17 20.00 0.00 Unconditio % (0.71) (0.24) % % (0.00) % nal [2] [0] [2] [0] cooperator 8.33 8.75 8.25 8.33 5.00 6.75 4.17 3.00 3.33 6.25 6.67 8.22 Hump% (7.45) (6.24) % (3.56) (2.99) % (0.00) (2.83) % (2.89) (3.67) shape [4] [4] [2] [3] 20.83 6.60 7.4 16.67 11.25 11.05 20.83 8.80 9.07 6.25 7.00 3.44 Others % (3.95) (4.11) % (4.65) (5.24) % (2.79) (3.94) % (10.30) (1.64) [10] [8] [10] [3] Dist.: Distribution. Avg. unc. cont.: average unconditional contribution. Guess avg. unc. cont.: guessed average unconditional contribution The number of players in each category is presented in brackets; standard deviations in parentheses.

The distribution of types of contributors is not significantly different across treatments (pvalue=0.40 in a chi-square test). In a more detailed analysis, we test the null hypothesis of equal distributions of player types in treatment pairs using a chi-square test. We can only reject the null hypothesis in the test between Treatments 1 and 4 at the 10% significance level (p-value=0.08), which according to Table 4 is caused by the fact that the proportions of free-riders and conditional cooperators are larger in Treatment 4 while the proportion of others is smaller. We conducted a Wilcoxon-Mann-Whitney test of the null hypothesis of equal distributions of unconditional contributions for each type of subject in treatment pairs. The average unconditional contribution of conditional cooperators under joint in-group and out-group disclosure is significantly different than the average unconditional contribution of conditional cooperators under out-group disclosure (p-value=0.03) and under in-group disclosure (p-value=

12

0.08). 6 The increase in the average unconditional contribution of conditional cooperators in Treatment 4 may indicate that, given that there is a disclosure policy, conditional contributors are more sensitive to changes in the stringency of the policy, and therefore increase their unconditional contribution 7. Based on a Wilcoxon-Mann-Whitney test, we can not reject the null hypothesis of equal distribution of guessed average unconditional contributions in treatment pairs. This indicates that disclosure schemes do not impact beliefs about others’ contributions. 3.3

Intrinsic and image motivations In order to check for differences in the intensity of social disapproval motivations and

intrinsic motivations among different types of subjects, in the final questionnaire we asked the participants to rate on a 1-5 scale anticipated guilt and feelings of shame, respectively, when contributing less than average, with 1 being “no guilt/shame at all” and 5 being “extreme guilt/shame.” We also asked them to rate on a 1-5 scale the expected social disapproval if contributing the lowest compared to others and the importance of social disapproval, respectively, with 1 being “no social disapproval at all/no importance at all”, and 5 “very much disapproval/extremely important.” Table 4 shows that we cannot reject the null hypothesis that free-riders and conditional cooperators expect the same level of feelings of shame when contributing less than average. However, they expect significantly different levels of guilt and of social disapproval if breaking the social norm. This, together with the fact that there are no significant differences in the importance of the expected social disapproval between the two 6

We also found significant differences in average unconditional contributions of hump-shaped contributors between Treatments 3 and 4 at 10% and in average unconditional contributions of other patterns between Treatments 1 and 2 at 10%. 7 We do not rule out the possibility that the difference in average unconditional contribution of conditional cooperators between Treatments 1 and 4 is influenced by of the presence of new conditional cooperators in this treatment. However, since the proportion on conditional cooperators is not significantly different in Treatments 1 and 4, we do not think this is the dominant effect in the increase of the average unconditional contributions

13

groups, supports the idea that image motivation is equally important for different types of players while intrinsic motivation is not. Finally, the magnitude of expected social disapproval (image motivations) is not significantly different across treatments. Table 4. Social and internal sanctions for free-riders and conditional cooperators. Statement Free-riders Conditional cooperators Guilt for contributing less than average 1.27 (0.47) 2.10 (1.00)

P-value 0.09

Shame for contributing less than average

1.36 (0.50)

2.09 (0.99)

0.18

Expected social disapproval

2.09 (0.70)

2.90 (1.07)

0.07

Importance of social disapproval

1.72 (0.90)

2.21 (1.02)

0.47

Note: Standard deviations are presented in parentheses.

4. Discussion and conclusions Using a public good experiment, we studied the effect of social approval on contributions to a public good. To accomplish this, we implemented a between-subject design with three different disclosure treatments: in-group disclosure, out-group disclosure, and joint ingroup and out-group disclosure, in addition to a treatment without any disclosure. We analyzed the effect of the treatments on both unconditional contributions and on distribution of contributor types using the design approach by Fischbacher et al. (2001). The results from the unconditional contributions in the treatment without disclosure are in line with studies using a similar design (Fischbacher and Gächter, 2006). We find that disclosure increases unconditional contributions, although the effect is not statistically significant at conventional levels. However, we find that when implementing joint in-group and out-group disclosure, the proportion of subjects contributing the whole endowment significantly increases

14

compared to in the anonymity treatment, while the proportion of non-contributions does not change significantly. In a more detailed analysis of the distribution of unconditional contributions we find that implementing out-group disclosure results in a higher proportion of subjects contributing 50% of their endowment. Our results of significantly higher variance in unconditional contributions with joint ingroup and out-group disclosure indicate that disclosure policies with larger audiences and more detailed information induce a higher heterogeneity in cooperation behavior and that unconditional cooperation may be moved in various ways. This can be costly for groups where subjects decrease their contribution as a reaction to the disclosure strategy. The direction in which unconditional contribution moves with disclosure may depend on underlying characteristics of subjects such as the importance their assign to social approval, and on the degree of internalization of the social norm for cooperation. In some cases, as Rege and Telle (2004) point out, disclosure strategies may have a negative effect on cooperation. Those circumstances are, according to them, when the effect of social approval depends crucially on the social norms and if the share of people adhering to the social norm for contribution is low. Our results suggest one more circumstance where a disclosure policy may work in the opposite direction in a public good game: When a subject assigns low importance to social approval and he or she has not internalized the cooperation norm, and the expected share of people contributing more to the public good increases, the benefits obtained from taking advantage of the increase in others’ contributions (free-riding) may be higher than the social costs associated with it. For this kind of subjects an identity-revealing policy may work in the opposite direction, decreasing contributions rather than increasing them.

15

Simulations are needed to determine the effects of different disclosure schemes on the stability of cooperation. In that respect, it is possible to predict two counteracting effects of introducing joint in-group and out-group disclosure on the stability of voluntary cooperation. On one hand, as Gächter (2006) mentions, heterogeneity in cooperation preferences may lead to fragile cooperation. In presence of disclosure, Conditional cooperators learn more quickly from other team members’ contributions than in anonymity, and their reactions to the presence of freeriders will most likely be a decrease in contributions. On the other hand, the fact that the share of conditional cooperators increases in the presence of joint disclosure may help maintain high cooperation. Fischbacher and Gächter (2009) show that a decline in contribution over time is caused by imperfect conditional cooperation rather than by changes in beliefs about others’ contributions. The latter effect is then decreasing over time given imperfect conditional cooperation. The total effect depends on the relative sizes of these two effects. Our results are of high policy relevance. A policy maker who intends to apply a disclosure scheme to stimulate cooperation in a public good context must first consider several aspects. We provide empirical evidence that in some cases disclosure schemes do not stimulate cooperation significantly and that in some cases it could even have a negative effect. Finally, and in contrast to Rege and Telle (2004), we do not find any support for the assumption commonly made in economic analysis of social norms that people have preferences for social approval (see, e.g., Akerlof, 1980). This limits the scope for policies relying on a social disapproval mechanism.

16

References Alpizar, F., Carlsson, F., Johansson-Stenman, O. 2008. ”Anonymity, reciprocity, and conformity: Evidence from voluntary contributions to a national part in Costa Rica”. Journal of Public Economics 92, 1047-1060. Akerlof, G.A., 1980. “A theory of social custom, of which unemployment may be one consequence”, The Quarterly Journal of Economics, 94 (4), 749-775. Andreoni, J., Bernheim, B.D., 2009. ” Social image and the 50-50 norm: a theoretical and experimental analysis of audience effects”. Econometrica, forthcoming. Andreoni, J., Petrie, R. 2004. “Public goods experiments without confidentiality: a glimpse into fund raising”. Journal of Public Economics 88, 1605-1623. Ariely, D., Bracha, A. and Meier, S. 2009. “Doing Good or Doing Well? Image Motivation and Monetary Incentives in Behaving Prosocially”. American Economic Review, forthcoming. Christie, R., Geis,F., 1970. Studies in Machiavellianism. Academic Press, New York. Croson, R.T.A. 2001. “Feedback in voluntary contribution mechanisms: an experiment in team production”. Research in Experimental Economics 8, 8-97 Ellingsen, T., M. Johannesson, 2008. “Pride and Prejudice:

The human side of

incentive theory”. American Economic Review 98:3, 990-1008. Fehr, E. And U. Fischbacher, 2004. “Third Party Punishment and Social Norms”. Evolution and Human behavior 25, 63-87. Fischbacher, U., S. Gächter and E. Fehr, 2001, “Are people conditionally cooperative? Evidence from a public goods experiment”. Economic Letters 71, 397-404. Fischbacher, U., Gächter, S. 2006. Heterogeneous social preferences and the dynamics of free riding in public goods. CeDEx Working Paper No. 2006-01, University of Nottingham. Fischbacher, U., Gächter, S. 2009. “Social preferences, beliefs, and the dynamics of free riding in public good experiments”, American Economic Review, forthcoming.

17

Gächter, S., Renner,E. 2006. “Leading by Example in the Presence of Free-Rider Incentives”, Mimeo, CeDEx, University of Nottingham. Herrmann, B., Thöni, C. 2009. “Measuring conditional cooperation: A replication study in Russia”. Experimental Economics 12, 87-92. Kocher, M. 2008. “Robustness of conditional cooperation in public goods experiments”. Discussion paper Kocher, M., Cherry, T., Kroll, S., Netzer, R., Sutter, M. 2008. “Conditional cooperation on three continents”. Economic Letters 101, 175-178. Ledyard, J., 1995, “Public goods: A survey of experimental research” in Kagel, J. and A. Toth (eds.), Handbook of Experimental Economics, Princeton University Press, Princeton, NJ. Levene, H. (1960). In Contributions to Probability and Statistics: Essays in Honor of Harold Hotelling, I. Olkin et al. eds., Stanford University Press, pp. 278-292. Rege, M. and K. Telle, 2003. "Indirect Social Sanctions from Monetarily Unaffected Strangers in a Public Good Game” Discussion Papers 359, Research Department of Statistics Norway. Rege, M.

and K. Telle, 2004. “The impact of Social Approval and Framing on

Cooperation in Public Good Situations”. Journal of Public Economics 88, 1625-1644. Sell, J., Wilson, R. 1991. “Levels of Information and Contributions to Public Goods”. Social Forces, 70, 107-124. Soetevent,A.R 2005. “Anonymity in giving in a natural context: An economic field experiment in thirty churches”. Journal of Public Economics 89 (2005), 2301–2323. Vecchio, R. P. and M. Sussmann, 1991. “Choice of Influence Tactics: Individual and Organizational Determinants” Journal of Organizational Behavior, 12, 73-80. Weimann, J. 1994. “Individual behavior in a free riding experiment”. Journal of Public Economics 54, 185-200. Zelmer, J. 2003. “Linear Public Goods experiments: A meta-analysis”. Experimental Economics 6, 299-310.

18

Appendix 1 INSTRUCTIONS You will be taking part in an experiment on decision-making. The experiment is designed so that your earnings will depend on your decisions. Your earnings in this experiment will be in “tokens.” At the end of the experiment, the tokens will be converted into Colombian Pesos (COP) at an exchange rate of: 1 token = 750 COP. In addition, you will receive a show-up fee of 5,000 COP. Your earnings will be paid in cash at the end of the session in a separate room to preserve the confidentiality of your earnings. The results from this study will only be used for academic purposes. Talking is not allowed throughout the entire session. Any violation of this rule will result in exclusion from the session and not receiving any payment. If you have any questions regarding these instructions, please raise your hand and member of the experimenter team will come to attend to you. We will begin with a detailed description of the experiment followed by some practice exercises before the experiment begins.

1. The basic decision You will now learn how the experiment is conducted. First we will introduce the basic decisionmaking situation. Then we will give you one example and will ask you to answer control questions that will help you gain an understanding of the decision-making situation. You will be a member of a group of four people. The groups are assembled randomly. At the beginning of the experiment you will receive (on paper) a number of tokens, called an “endowment”. Each of the four members of the group has to decide how to divide his or her endowment. You can put all, some, or none of your tokens into the group account. Each token you do not deposit in the group account will automatically be transferred to your private account. Your income from the private account: For each token you put into your private account, you will earn exactly one token. For example, if you have an endowment of 20 tokens and you put zero tokens into the group account (and therefore 20 tokens in the private account), then you will earn exactly 20 tokens from the private account. If instead you put 14 tokens into the group account (and therefore 6 tokens in the private account), then you will receive an income of 6 tokens from the private account. Nobody except you earns tokens from your private account. Your income from the group account: 19

Everybody receives the same income from the group account, which is based on the total number of tokens put into it by the group. Your income from the group account will therefore be earned from the tokens that the other group members put into the group account; not just the tokens that you invest in it yourself. For each group member, the income from the group account will be determined as follows: Income from the group account = sum of all contributions to the group account x 0.4

For example, if the sum of all contributions to the group account is 60 tokens, you and the other group members will get an income of 60 x 0.4 = 24 tokens from the group account. If the four group members deposit a total of 10 tokens in the group account, then you and the others will receive an income of 10x0.4=4 tokens from the group account. Your total income: Your total income is the sum of the income from your private account and the income from the group account: Income from your private account (=your endowment – your contribution to the group account) + Income from the group account (=0.4 x sum of all contributions to the group account) Total income

Before we finish reading the instructions, we will ask you to answer the following control questions. This will help you see whether you have understood everything correctly. If there are any questions or problems, please raise your hand. One of the members of the experimenter team will come to you and answer your questions privately. 2. Control questions Please answer the following control questions. Their purpose is to make you familiar with calculating the various incomes in tokens that you might earn depending on the decisions you will make about endowment allocation. Please answer all questions and write down all calculations. 1.

Assume you have an endowment of 20 tokens. Assume also that all group members (including yourself) put nothing into the group account. a. What is your total income? _____________ b. What are the incomes of the three other group members?_____,____ and ____

20

2.

Assume you have an endowment of 20 tokens the same as the other three group members. Assume also that all group members (including yourself) put their entire endowment into the group account. a. What is your total income? _____________

3.

Assume you have an endowment of 20 tokens. Assume also that the other group members collectively put a total of 30 tokens into the group account. a. What is your total income if you, in addition to the 30 tokens of the other three group members, put 0 tokens into the group account? i. Your income from your private account : _______ ii. Your income from the group account: _______ iii. Your total income __________ b. What is your total income if you, in addition to the 30 tokens of the other three group members, put 8 tokens into the group account? i. Your income from your private account : _______ ii. Your income from the group account: _______ iii. Your total income __________ c. What is your total income if you, in addition to the 30 tokens of the other three group members, put 15 tokens into the group account? i. Your income from your private account : _______ ii. Your income from the group account: _______ iii. Your total income __________

4.

Assume you have an endowment of 20 tokens. You put 8 tokens to the group account. a. What is your total income if the other three group members, in addition to your 8 tokens, put another 7 tokens into the group account? i. Your income from your private account : _______ ii. Your income from the group account: _______ iii. Your total income __________ b. What is your total income, if the other three group members, in addition to your 8 tokens, put another 12 tokens into the group account:

21

i. Your income from your private account : _______ ii. Your income from the group account: _______ iii. Your total income __________ c. What is your total income, if the other three group members, in addition to your 8 tokens, put another 22 tokens into the group account? i. Your income from your private account : _______ ii. Your income from the group account: _______ iii. Your total income __________ If you finish these questions before the others, we advise you to think about additional examples to familiarize yourself further with these types of decision-making situations. 3. The Experimental Procedure The experiment consists of decision-making situations like the one that we have just described. In the following we explain the procedure in detail. As you know, you have an endowment of tokens. You can put these tokens into a group account and the remaining tokens will automatically be deposited into your private account. Each person in the group will have the same endowment. Each group member has to make two types of decision. In the following instructions, we will refer to them as the “unconditional contribution” and the “contribution table decision”. •

With the unconditional contribution to the group account, you decide how many of the tokens in your endowment to put into the group account. Write this amount under “Your unconditional contribution to the group account” on the first page of your decision sheet. You must write down an integer number that cannot be smaller than zero or larger than the total number of tokens that you were given in your endowment. The difference between your endowment and the amount you put into the group account is the amount that will go into your private account automatically.

Your second task is to fill out a contribution table on page 2 of the decision sheet. In the contribution table you have to indicate how many tokens you would like to put into the group account for each possible average contribution of the other three group members (rounded up or down to the nearest integer number). What you actually contribute will depend on what the other group members actually contribute. This will become clear to you when you see the following contribution table example: (Rounded) Average contribution of the other Group members to the Group account. 0

Your contribution to the Group account is:

22

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20

The numbers in the left column are the possible (rounded) average contributions of the other three group members. Assume for this example that the other 3 group members can contribute on average a maximum of 20 tokens. Using the column on the right, you simply have to write down how many tokens you would like to contribute to the group account for each possible average contribution of the others. You must make an entry in each field of the right column. For example, you must write down how many tokens you want to contribute to the group account if the others contribute on average 0 tokens to the group account; how many you want to contribute if others contribute on average 0 tokens to the group account; how many you want to contribute if the others contribute 1 or 2 or 3 tokens, etc. In each field, you must write down an integer, a number that is neither smaller than zero nor larger than the total number of tokens in your endowment. You can of course write down the same number in different fields. After all participants of the experiment have made their unconditional contribution decisions and have filled out their conditional contribution tables, one member of each group will be selected randomly. Each person will receive payment for one decision only and this random selection mechanism is used to determine whether it is the unconditional contribution decision or the contribution table decision that is paid. For the randomly selected group member, only the contribution table will be income relevant. For the 3 group members who are not selected, the unconditional contribution decision will be the income-relevant decision. When you make your unconditional contribution and when you fill out the contribution table, you do not know whether you will be selected randomly. You will therefore have to think carefully about both types of decision because both could affect the amount that you earn. The following two examples should illustrate this: Example 1. Assume that after you handed in your decisions you are selected by the random mechanism. This implies that your income decision will be the contribution table. For the other 3 group members, the unconditional contribution is the relevant decision. Assume they have made

23

unconditional contributions of 0, 2, and 5 tokens. The rounded average contribution is therefore 2 = (7/3=2.33). If you have indicated in your contribution table that you will put 1 token into the group account if the others contribute 2 tokens on average, then the total contribution to the group account is 0+2+5+1=8. Thus all group members earn an income of 0.4*8=3.2 from the group account plus the respective incomes from their private accounts. If you have indicated instead that you will contribute 19 tokens to the group account if the others contribute 2 on average, then the total contribution to the group account is 0+2+5+19=26. All group members earn an income of 0.4*26=10.4 tokens from the group account plus the respective incomes from their private accounts. Example 2. Now assume that you have not been selected by the random mechanism, which means that for you and two other group members the unconditional contribution is the incomerelevant decisions. Assume further that your unconditional contribution to the group account is 16, and that those of the other two group members are 18 and 20. The average unconditional contribution is therefore 18 ((16+18+20)/3). If the randomly selected group member indicates in the contribution table that he or she contributes one token to the group account when the other 3 group members contribute 18 on average, then the total contribution of the group to the group account is 16+18+20+1=55 tokens. All group members will therefore earn 0.4*55=22 tokens from the group account in addition to the respective incomes from their private accounts. If the randomly selected group member instead indicated in the contribution table that he or she will contribute 10 tokens to the group account when the other 3 group members contribute 18 on average, then the total contribution of the group to the group account is 16+18+20+19=73 tokens. All group members will therefore earn 0.4x73=29.2 tokens from the group account in addition to the respective incomes from their private accounts. The random selection of the participants will be implemented using the EXCEL random number generator. Each group member is assigned a letter A, B, C, or D. After all the participants have made their unconditional contribution and contribution table decisions, a letter A, B, C, or D will be randomly drawn using the Excel random number generator. The letter drawn determines the player for whom the contribution table is income relevant. TREATMENT 1 – Standard Public Goods Game 4. How does the experiment work? All your decisions are anonymous. You will be identified only by your “experimental idnumber,” which was given to you when you entered the room, during and after this experiment. During the whole experiment and also after the experiment, the decisions made by the subjects will remain unknown to all others except the academic research team. a. In the beginning of the experiment, the experimenter will give you the “Decision sheet” where you will find your endowment in tokens.

24

b. Using this information, you must make your decisions on the unconditional contribution to the group account. You will write down this decision in the space reserved for this in the first page of the decision sheet. On the second page of the decision sheet you will find the contribution table, which you have to fill out as well. c. After you and everyone else have handed in the decisions, an experimenter will randomly select the participant for whom the contribution table will be income relevant. d. While the instructor is calculating your earnings, you will be asked to answer a questionnaire. We will use the “experimental id-number” to identify you when making the payments, so please hold on to your card. On a desk in another room, the experimenter will place envelopes with the earnings that each of you have made during the experiment. When you go to claim your payment, you should present your “experimental id-number” and pick up the envelope with the same “experimental id-number” as your “experimental id-number.” Please raise your hand if you have any questions and an experimenter will attend to you. TREATMENT 2 – Out-group disclosure 4. How does the experiment work? You will be identified only by your “experimental id-number,” which was given to you when you entered the room, during and after this experiment. a. In the beginning of the experiment, the experimenter will give you the “Decision sheet” where you will find your endowment in tokens. b. Using this information, you must make your decisions on the unconditional contribution to the group account. You will write down this decision in the space reserved for this in the first page of the decision sheet. On the second page of the decision sheet you will find the contribution table, which you have to fill out as well. c. After you and everyone else have handed in the decisions, an experimenter will randomly select the participant for whom the contribution table will be income relevant. d. While the instructor is calculating your earnings, you will be asked to answer a questionnaire. e. All other participants will know your decisions on how much you contributed to the group account. Similarly, you will know each of their decisions on how much they contributed to their group accounts. You do not know and you will not get to know who is in the same group as you. The other participants in your group will not know you are in the same group as them. The procedure for this is as follows: After all participants have made their decisions and completed the final questionnaire, the instructor will announce the contribution of each participant. For instance, the instructor says: “Participant 53 (the participant who has this number must stand up) contributed X (the relevant contribution will be read) tokens to the group account.” For those participants who the contribution table is income relevant, the contribution that he or she indicated in the contribution table will be disclosed. For those participants who the unconditional contribution is the income relevant contribution, this will be the disclosed 25

contribution. The other participants will know whether you contributed and if so how much to the group account, but they will not know to which group you belong. Similarly, you will know what each of the other participants has chosen, but you will not know to which group they belong. We will use the “experimental id-number” to identify you when making the payments, so please hold on to your card. On a desk in another room, the experimenter will place envelopes with the earnings that each of you have made during the experiment. When you go to claim your payment, you should present your “experimental id-number” and pick up the envelope with the same “experimental id-number” as your “experimental id-number.” Please raise your hand if you have any questions and an experimenter will attend to you. TREATMENT 3 4. How does the experiment work? You will be identified only by your “experimental id-number,” which was given to you when you entered the room, during and after this experiment. a. In the beginning of the experiment, the experimenter will give you the “Decision sheet” where you will find your endowment in tokens. b. Using this information, you must make your decisions on the unconditional contribution to the group account. You will write down this decision in the space reserved for this in the first page of the decision sheet. On the second page of the decision sheet you will find the contribution table, which you have to fill out as well. c. After you and everyone else have handed in the decisions, an experimenter will randomly select the participant for whom the contribution table will be income relevant. d. While the instructor is calculating your earnings, you will be asked to answer a questionnaire. e. Only the other members of your group will know your decisions on how much you contributed to the group account. Similarly, you will know each of their decisions on how much they contributed to their group accounts. You do not know and you will not get to know how much others in other groups have contributed. The procedure for this is as follows: After all participants have made their decisions and completed the final questionnaire, the instructor will call each of the members in a group to come together to a room next door. The contribution of each participant will be announced in that room. All groups will be called independently (one by one) to come to a room next door. For instance, the instructor will in the room next door say: “Participant 53 (the participant who has this number must stand up) contributed X (the relevant contribution will be read) tokens to the group account.” For those participants who the contribution table is income relevant, the contribution that he or she indicated in the contribution table will be disclosed. For those participants who the unconditional contribution is the income relevant contribution, this will be the disclosed contribution. The other members of your group will know whether and if so how much you contributed to the group account. Similarly, you will 26

know what each of the other members of your group chose. You will not know how much the participants in the other groups contributed, and participants in other groups will not know how much you contributed. We will use the “experimental id-number” to identify you when making the payments, so please hold on to your card. On a desk in another room, the experimenter will place envelopes with the earnings that each of you have made during the experiment. When you go to claim your payment, you should present your “experimental id-number” and pick up the envelope with the same “experimental id-number” as your “experimental id-number.” Please raise your hand if you have any questions and an experimenter will attend to you. TREATMENT 4 4. How does the experiment work? You will be identified only by your “experimental id-number,” which was given to you when you entered the room, during and after this experiment. a. In the beginning of the experiment, the experimenter will give you the “Decision sheet” where you will find your endowment in tokens. b. Using this information, you must make your decisions on the unconditional contribution to the group account. You will write down this decision in the space reserved for this in the first page of the decision sheet. On the second page of the decision sheet you will find the contribution table, which you have to fill out as well. c. After you and everyone else have handed in the decisions, an experimenter will randomly select the participant for whom the contribution table will be income relevant. d. While the instructor is calculating your earnings, you will be asked to answer a questionnaire. e. All other participants will know your decisions on how much you contributed to the group account. Similarly, you will know their decisions on how much they contributed to their group accounts. Moreover, the other members of your group will know your decisions on how much you contributed to the group account. Similarly, you will know each of their decisions on how much they contributed to their group accounts. The procedure for this is as follows: After all participants have made their decisions and completed the final questionnaire, the instructor will call each member of a group to a stage in the front of the room, where the contribution of each group member will be announced. All groups will be called to come to the stage at the front of the room. For instance, the instructor says: “Participants 53 (the participant who has this number must stand up) contributed X (the relevant contribution will be read) tokens to the group account.” For those participants who the contribution table is income relevant, the contribution that he or she indicated in the contribution table will be disclosed. For those participants who the unconditional contribution is the income relevant contribution, this will be the disclosed contribution. The other participants will know whether and if so how much you contributed to the group account and they will know to which group you belong. 27

Similarly, you will know what each of the other participants chose and to which groups they belong. We will use the “experimental id-number” to identify you when making the payments, so please hold on to your card. On a desk in another room, the experimenter will place envelopes with the earnings that each of you have made during the experiment. When you go to claim your payment, you should present your “experimental id-number” and pick up the envelope with the same “experimental id-number” as your “experimental id-number”. Please raise your hand if you have any questions and an experimenter will attend to you.

28