Initially Contestable Property Rights and Coase - Evidence From The Lab - 2023

Journal of Environmental Economics and Management 120 (2023) 102842
Contents lists available at ScienceDirect
Journal of Environmental Economics and

Management
journal homepage: www.elsevier.com/locate/jeem
Initially contestable property rights and Coase: Evidence from the

lab✩
Lana Friesen ∗, Ian A. MacKenzie, Mai Phuong Nguyen
School of Economics, University of Queensland, Brisbane, 4072, Australia
ARTICLE INFO ABSTRACT
JEL classification: This article investigates how the existence of initially contestable property rights affects
C92 the efficiency of the Coase theorem. We design a two-stage experiment that incorporates a
K13 stage where property rights are initially allocated to participants followed by a stage that
K41
allows bargaining between participants. In stage one, participants endogenously choose rent-
Q52
seeking effort (and thus the probability) to initially obtain property rights before entering an
Keywords: unstructured bargaining game. We find the presence of costly rent seeking to obtain the property
Coasean bargaining rights makes it significantly less likely that the efficient outcome is reached. We introduce
Transaction costs
bargaining costs and find that allowing for symmetric bargaining costs has no impact on the
Experiment
likelihood of the efficient outcome being reached, whereas asymmetric bargaining costs – costs
Property rights
Contest that differ if the player wins or loses the initial property rights – substantially reduces the
Natural resources likelihood of reaching an efficient outcome. Our analysis is applicable to contexts with initially
contestable and tradable nature resource rights.
1. Introduction
The ‘Coase theorem’ states that reaching the efficient level of externality is independent of the initial allocation of property rights.
Regardless of the initial distribution of property rights, parties can negotiate to capture gains from trade and reach an efficient
equilibrium (Coase, 1960). Although this idea has heavily influenced environmental economic debate over the last 60 years, it
is well known that for this result to hold it requires restrictive conditions, which ensure clearly defined, secure, and enforceable
property rights as well as zero transaction costs (e.g., Shogren, 1992; Cherry et al., 2013; Bar-Gill and Engel, 2016; Medema, 2020;
Deryugina et al., 2021).
One often overlooked reality is that natural resource property rights may be initially contestable among parties. For instance,
initial property right contestability could occur because (i) the creation of new secure property rights may have no existing precedent
for the initial assignment of property rights, or (ii) there exists ambiguity and contention over the initial ownership of the property
rights (for an extensive list of environmental applications where property rights are initially contestable see Deryugina et al., 2021,
pp. 85–86). In all scenarios it is likely that the property rights – especially those created for the purpose of trading environmental
externalities – may require effort by parties to determine and secure the initial assignment (Robson and Skaperdas, 2008; Dari-
Mattiacci et al., 2009; MacKenzie and Ohndorf, 2013). Agents may invest in costly (sunk) activities – such as rent seeking, litigation
or conflict – to not only influence the probability of obtaining the initial distribution of property rights but also to secure the property
rights for any future successful bargaining. Indeed this was highlighted early in the debate by Ronald Coase, who commented:
✩ Acknowledgments: We would like to thank the Editor and two anonymous referees for helpful comments, Nick Feltovich for sharing his z-Tree program for
unstructured bargaining, and the University of Queensland for funding. The usual disclaimer applies.
∗ Corresponding author.
E-mail addresses: l.friesen@uq.edu.au (L. Friesen), i.mackenzie@uq.edu.au (I.A. MacKenzie), maiphuong0612@gmail.com (M.P. Nguyen).
https://doi.org/10.1016/j.jeem.2023.102842
Received 10 November 2022
Available online 21 June 2023
0095-0696/© 2023 The Authors. Published by Elsevier Inc. This is an open access article under the CC BY-NC-ND license
(http://creativecommons.org/licenses/by-nc-nd/4.0/).
L. Friesen et al. Journal of Environmental Economics and Management 120 (2023) 102842
[T]he legal delimitation of rights provides the starting point for the rearrangement of rights through market transactions. Such
transactions are not costless, with a result that the initial delimitation of rights may be maintained even though some other would
be efficient. Or, even if the original position is modified, the most efficient delimitation of rights may not be attained. Finally, waste
of resources may occur when the criteria used by the courts to delimit rights result in resources being employed solely to establish a
claim (Coase, 1959, p.27, n. 54).
Following the environmental economics textbook case of the Coase theorem, prima facie, the initial allocation of property rights is
distinct from any bargaining process and it would follow that there is no detrimental impact on efficiency (e.g., MacKenzie and
Ohndorf, 2013). Yet from a behavioral perspective, the initial allocation – and the efforts expended to determine the assignment –
may impact efficiency (Hoffman et al., 1994; Feltovich, 2019). The aim of this article is to investigate the connection between the
initial allocation of natural resource property rights, rent-seeking effort, (asymmetric) bargaining costs, and efficiency.
This article investigates how the existence of initially contestable property rights affects the bargaining efficiency of the Coase
theorem. We create a two-stage framework over the determination of property rights that creates an externality. In stage one, two
players choose how much to invest in rent-seeking effort in order to influence the probability of obtaining the initial endowment of
property rights. Following the relevant theoretical contest literature (Tullock, 1980; Hillman and Riley, 1989; Robson and Skaperdas,
2008; MacKenzie and Ohndorf, 2012), we model the rent-seeking stage as a winner-take-all contest with each player having a
probability of winning (and thus securing) the initial allocation of property rights, which is dependent on their effort relative to
total outlays.1 Once the initial allocation of property rights has been secured via the contest, stage two allows the players to bargain
over the externality level. Bargaining in stage two may be costly. When bargaining is costly we consider two scenarios: where these
costs are symmetric or asymmetric among players. In the asymmetric cost case, we allow a difference in bargaining cost dependent
on whether the player wins or loses the initial property right. Given the framework, we investigate how the existence of costly rent-
seeking impacts the efficiency of Coasean bargaining and understand how ex ante and ex post transactions costs (i.e., rent-seeking
and bargaining costs, respectively) interact to potentially affect the efficiency of the Coase theorem.
We design the experiment with these two main stages: a contest stage reflecting the rent seeking of contestable property rights
where participants submit their bids into a lottery, and a bargaining stage, where parties participate in an unstructured bargaining
game to determine the final externality distribution (out of a possible seven distributions). We have four treatments that differ in
the presence of a contest and bargaining costs. Our control treatment randomly allocates the initially contestable property rights
to players. In this treatment there are no bargaining costs to reach an efficient outcome. In the remaining three treatments we
incorporate a winner-take-all contest (Cason et al., 2020), where participants can invest sunk costs to influence the probability that
they obtain the initial allocation of property rights. These three remaining treatments differ in the presence of bargaining costs by
allowing for zero bargaining costs as well as asymmetric and symmetric bargaining costs. With symmetric bargaining costs an equal
amount is taken away from both parties (regardless of whether they are the initial property-rights holder or not) and for asymmetric
costs the same total amount is removed but the allocation among the parties is different (i.e., each player incurs different bargaining
costs depending on whether they are the initial property-rights holder or not). There is no a priori rationale for assuming bargaining
costs must be symmetric among players. Indeed, one can reasonably consider scenarios where the property-rights holder may have
lower bargaining costs due to, for example, the absence of search costs.
We find that the presence of costly rent-seeking activity to capture the property rights makes it significantly less likely that the
efficient equilibrium is reached relative to a random allocation of property rights. In particular, this is because the two parties are
significantly less likely to reach any agreement in the bargaining stage. When agreement is reached it almost invariably leads to the
most efficient outcome being chosen.2 We find that allowing for symmetric bargaining costs has no impact on the likelihood of the
efficient outcome being reached, whereas when bargaining costs are asymmetric we find it significantly and substantially reduces
the likelihood of reaching the efficient outcome relative to symmetric bargaining costs. We also find that the property-rights holder,
that had to use costly rent-seeking efforts, receives a larger share of the available surplus (in successful bargains) compared to
property-rights holders who received the property rights randomly. This may explain why efficiency is lower when costly rent
seeking is required to establish the property rights.
The implications of our work highlight potential efficiency issues in contexts where natural resource property rights are initially
ambiguous but the determination of these property rights allows negotiation to occur. For example, this is applicable to cases where
land, water or mineral rights are initially contested but eventually traded/settled. Our analysis shows that if rent-seeking activities
are taken into account (such as the use of litigation, lobbying and rent seeking to initial obtain the rights) then this has a detrimental
effect on the efficiency of bargaining. Further, if bargaining costs are asymmetric then this compounds the problem and reduces the
likelihood of reaching the efficient bargaining equilibrium.3
1 In particular, Coasean bargaining has been theoretically investigated in this context when the bargaining space has been exogenously restricted (MacKenzie
and Ohndorf, 2013), in the presence of taxes (MacKenzie and Ohndorf, 2016a), and exchange transfers are limited (MacKenzie and Ohndorf, 2016b).
2 Our main goals in this article are to better understand how the existence of rent-seeking activities distorts the efficiency of bargaining and how the incentive
to invest in rent-seeking activities changes in distinct bargaining environments. As such, we are not interested in the expected aggregate payoff net of rent-seeking
costs. Indeed, it is straightforward to consider that, given our treatments, a random allocation (thus exogenously restricting rent-seeking efforts to zero) and
assuming zero bargaining costs would result in the highest expected aggregate payoffs. For investigations into net payoffs and rent-seeking costs see MacKenzie
and Ohndorf (2013).
3 For example, the determination of Native American water rights ownership dates back to the U.S. Supreme Court ruling Winters v. United States (1908), but
where, from the 1980 onwards Native Americans were able to negotiate the settlement of water rights (Tarlock, 2010). Allowing the exchange and negotiation
2
2. Related literature
It is well documented that property rights can be insecure, contestable, or ambiguously determined (e.g., Demsetz, 1964;
Buchanan, 1975; Grossman, 2001; Skaperdas and Syropoulos, 2002). How players interact when property rights are contestable
has been a main focus of inquiry within the Public Choice literature, where costly efforts are expended to influence the probability
of obtaining the initial allocation of property rights (Tullock, 1980; Hillman and Riley, 1989; MacKenzie and Ohndorf, 2012) via
either litigation (Farmer and Pecorino, 1999; Baye et al., 2005) or political lobbying (Baye et al., 1993; Dickson et al., 2022).
A predominant method to investigate contestable property rights is the use of contest theory. A contest is a situation in which
individuals expend sunk costs – rent-seeking costs – to influence either the probability of obtaining a property right (Tullock, 1980;
Hillman and Riley, 1989) or – in some applications – a share of the property rights (Cason et al., 2020; Dickson et al., 2022). In
the standard Tullock contest (Tullock, 1980) a pure-strategy Nash equilibrium of efforts exists in which efforts are increasing in the
value of the property rights.
In our article we extend the use of contest theory to model initially contestable property rights by incorporating (costly) Coasean
bargaining. We allow players to expend effort to influence the probability of property rights initial allocation but, in contrast to much
of the literature, we then allow property rights exchange via costly Coasean bargaining. Contestable property rights in a Coasean
context has been theoretically studied by Robson and Skaperdas (2008) and MacKenzie and Ohndorf (2013, 2016a,b). The main
focus has been on how limits to the surplus can affect rent-seeking efforts (MacKenzie and Ohndorf, 2013, 2016b) as well as whether
players decide to litigate or not (Robson and Skaperdas, 2008). The aim of this article is different: our focus is on understanding
how the costly efforts to secure property rights can impact the efficiency of bargaining in the presence of (a)symmetric bargaining
costs and how these costs can impact the incentive to invest in rent-seeking efforts.
We test the theory of initially contestable Coasean property rights with experimental laboratory evidence. Our key focus is on
understanding how the system of securing the property rights – the initial allocation process – and the existence of bargaining
costs impacts efficiency. Although a number of experiments study the Coase theorem, there has been, up to now, no systematic
experimental investigation of the Coase theorem with initially contestable rights and – despite the importance of transaction costs for
the Coase theorem – only a few experiments include bargaining costs, none of which consider an analysis of asymmetric bargaining
costs.4
Our first contribution is shedding light on how the initial allocation process impacts bargaining efficiency. We use a novel
experimental design, where we model the initial allocation of property rights as a Tullock contest involving measurable monetary
efforts (Tullock, 1980; Robson and Skaperdas, 2008; MacKenzie and Ohndorf, 2013). This contrasts with existing Coasean
experiments, which typically use simple competitive real-effort tasks to determine who gets the property right (Shogren, 1992;
Shogren and Kask, 1992). As well as corresponding more closely with theoretical models, our approach ensures that we can precisely
measure the relative efforts exerted by both parties, which is important for our investigation of rent-seeking costs. It also introduces
a degree of uncertainty that may be relevant in real-world settings where, even if you incur more rent-seeking costs (exert more
effort) than other players, you are not guaranteed to win the property rights.5
The earliest Coasean experiments created a simple environment where subjects bargain over certain outcomes, there are no
transaction costs, and property rights (and any subsequently agreed contracts) are perfectly enforced (Hoffman and Spitzer, 1982,
1985, 1986). These experiments demonstrated that Coasean bargaining is highly efficient even with large group sizes (Hoffman
and Spitzer, 1986), limited information (Hoffman and Spitzer, 1982, 1986) and even when bargaining over physically harmful
outcomes (Coursey et al., 1987). However, while bargaining generally leads to the efficient outcome, equal splits are common
in these experiments, even when they are irrational (and lead to a lower payoff than the outside option) for the property-rights
holder. Harrison and McKee (1985) show that at least some of these irrational choices result from subjects not understanding the
meaning of ‘‘unilateral’’ property rights, while Shogren (1998) and Rhoads and Shogren (2003) explain using the idea of ‘‘constrained
self-interest’’, that is, the property-rights holder’s ‘‘payoff is a weighted average of pure self-interest and equal splits on the expected
reward’’ (Rhoads and Shogren, 2003, p.75).
In our contest approach, property-rights holders ‘‘earn’’ the right to hold the property rights by expending rent-seeking costs
rather than exerting real effort. Even though theory predicts that the expenditure of sunk rent-seeking costs should not affect
bargaining outcomes, our results suggest otherwise; the presence of rent seeking makes it significantly less likely that an efficient
outcome is reached. Studies have previously demonstrated how social preferences impact players’ willingness to exploit their
strategic advantages. In the context of Coasean bargaining experiments (e.g., Hoffman and Spitzer, 1985; Shogren, 1992), efficiency
of rights is an new aspect of the new water market in Arizona and the newly introduced bill to allow lease water rights in the Colorado River Indian Tribe
Water Resiliency Act of 2021 (S. 3308), with growing calls for the U.S. Congress to pass legislation in order to allow for easier leasing of water rights. Another
example is the U.S. Clean Water Act of 1972 that ensured development within wetlands was blocked (i.e., the property rights of wetlands were protected).
Yet bargaining occurred as the Army Corps of Engineers negotiated the issuance of development grants for developers that created additional restoration and
improvements of wetlands elsewhere (For full details see Deryugina et al., 2021).
4 Croson (2009) and McAdams (2000) provide surveys of the earlier Coasean bargaining experiments, while Prante et al. (2007) conduct a meta-analysis.
5 We are only aware of a limited number of early ultimatum bargaining experiments that use a related approach by using an auction to allocate roles (Güth
and Tietz, 1985, 1986) with results demonstrating how proposers are more willing to exploit their strategic power when they have to pay for it. These auctions
are, of course, deterministic. In addition, our novel bargaining environment is considerably more complex because subjects are not simply negotiating over how
to divide a given pie size but instead negotiating over which option out of seven possible options to choose, as well as the consequent division of the available
surplus.
3
is usually unaffected by the allocation method but the distribution of gains is affected. The latter result is supported in cases
where players are more self-interested and exploit their bargaining power when roles and endowments are earned rather than
allocated (Cherry et al., 2002; Oxoby and Spraggon, 2008; Barber, IV and English, 2019; Feltovich, 2019).
A second contribution of our paper is investigating how bargaining costs—both symmetric and asymmetric—can impact
efficiency and the choice of rent-seeking efforts. Surprisingly—given its importance to the Coase theorem—only a few papers (i.e.,
Rhoads and Shogren, 1999; Cherry and Shogren, 2005) explicitly include bargaining costs in their experimental setting, and only one
varies these costs across treatments. The exception is Rhoads and Shogren (1999) who find that high bargaining costs significantly
reduce the efficiency of bargaining outcomes compared to zero costs but small bargaining costs did not. In contrast we compare the
presence versus the absence of bargaining costs, as well as the composition of bargaining costs (symmetric versus asymmetric).6
A conceptually similar article to the research presented here is by Cherry and Shogren (2005) who investigate the impact of
insecure property rights in the presence of bargaining costs, which are symmetric across parties and held constant across treatments.7
Interestingly, Cherry and Shogren (2005) find that bargaining efficiency and property rights security are inversely related: insecure
property rights result in increased bargaining efficiency. This inverse relationship is illustrated within a two-stage game. Cherry and
Shogren (2005) first allow bargaining for property rights (which determines the likelihood of winning a payoff). These property
rights are either secure or insecure. If the property rights are secure, the property-rights holder can unilaterally opt for the outside
option in the event that a bargaining agreement is not reached. If property rights are insecure, then there is a chance that the
property-rights holder’s choice of the outside option is not enforced, and instead both parties receive their disagreement payoffs. In
such an event, players invest in rent-seeking effort to obtain their own outside option.8 One intuition for their result is that there
exists bargaining in the ‘shadow of conflict’ (Anbarci et al., 2002; Skaperdas and Syropoulos, 2002): the potential costs of securing
their outside option (e.g., through lawsuits, rent seeking) will mean players are more willing to bargain with insecure property
rights, even in the presence of bargaining costs.
While the components of our experiment – the ability to bargain and rent seek – are similar to Cherry and Shogren (2005),
our ‘inverted’ structure creates a unique institutional setting to analyze. By analyzing rent seeking in the first stage, our focus is
on initially contestable property rights that can be secured by players investing in rent-seeking effort (rather than being placed in
an ad hoc environment of (in)secure property rights). Once property rights are secured bargaining can commence. This is realistic
in settings where investment in legal and administrative (institutional) processes is a prerequisite for bargaining to commence.
For example, trading of land and water titles are often granted only after legal proceedings have determined ownership.9 We find
opposing results to Cherry and Shogren (2005): when rent-seeking costs are used to secure property rights, there is a negative impact
on bargaining efficiency. Although rent-seeking costs are sunk at the time of bargaining—the existence of rent-seeking efforts lowers
bargaining efficiency. Rather than bargaining in the shadow of conflict, in our experiment we observe bargaining in the presence
of the sunk cost fallacy. This is compounded by observations that the initial property-rights holder obtains a larger share of the
available surplus when there exist rent-seeking costs. By following our ‘inverted’ approach, we can additionally investigate how the
potential for future (costly) bargaining impacts the incentive to invest in rent seeking. We show having asymmetric bargaining costs
will increase players’ incentives to rent seek relative to cases with either symmetric or no bargaining costs.
By analyzing an environment in which property rights are secured by players’ rent-seeking efforts with the potential for trade, we
are able to shed light on contemporary issues with property rights exchange over many natural resources including, but not limited
to, land, minerals, and water rights. The article is organized as follows. In Section 3 the theoretical model is presented. In Section 4
the experimental design is described and in Section 5 the results are detailed. Concluding remarks are provided in Section 6.
3. The model
Suppose there exists a set of two agents {𝐴, 𝐵}. Agent 𝐵 participates in an activity that generates harm on agent 𝐴. Agent 𝐵 has
the ability to reduce the level of harm by investing in abatement activities, 𝑥 ≥ 0. The investment in abatement occurs at a private
cost to agent 𝐵 denoted by 𝐶(𝑥) with 𝐶 ′ (𝑥) > 0, and 𝐶 ′′ (𝑥) ≥ 0. The damage experienced by agent 𝐴 is denoted by 𝐷(𝑥), where
−𝐷′ (𝑥) > 0, and 𝐷′′ (𝑥) ≥ 0. Each agent has wealth endowment 𝑊 0 .
As a benchmark we allow frictionless exchange to occur between agents over the level of harm (abatement). In particular,
agents can bargain over the level of harm à la (Coase, 1960). Let 𝑥∗ be the (ex post ) efficient Coasean bargaining solution, so that
6 Shogren (1998) and Rhoads and Shogren (2003) examine how delay costs (as distinct from bargaining costs) affect Coasean bargaining, with these delay
costs symmetric across parties and roles. Finally, while Aivazian et al. (2009) investigate bargaining costs in their experiment, they do so by measuring outcomes
(e.g., bargaining time, number of proposals) rather than explicitly imposing costs of these actions in the experiment.
7 Several other papers investigate the impact of ill-defined or insecure property rights. For example, Bar-Gill and Engel (2016) find that Coasean bargaining
is just as efficient with either absolute or relative property rights, Aivazian et al. (2009) examine the impact of potentially ill-defined property rights when the
core may be empty, and Shogren and Kask (1992) investigate the impact of imperfect contract enforcement. However, in contrast to our experiment, property
rights in these experiments are initially clearly defined and allocated, but may be insecure in application afterwards. We instead consider the case of initially
insecure (contestable) property rights but once awarded they become secure.
8 While participants in our experiment take part in a Tullock contest, this does not occur in Cherry and Shogren (2005); instead the disagreement values are
used as experimental outcomes.

9 In the context of investing in rent-seeking activities, it is observed that ex ante bargaining is unlikely to occur because the initial insecurity of property
rights make any ex ante bargaining non-credible. This is clear from the rent-seeking literature, where players attempt to individually capture rents. For example,
this is observed over the initial distribution of tradable pollution property rights, where firms lobby for their own initial allocation that can only be traded once
initial ownership has been formally established (Hanley and MacKenzie, 2010; MacKenzie and Ohndorf, 2012; MacKenzie, 2017; Dickson and MacKenzie, 2022).
4
𝑥∗ = min𝑥 {𝐶(𝑥) + 𝐷(𝑥)}. If agent 𝐵 initially has the ownership of the property rights then exchange can occur to 𝑥∗ , where agent
𝐵 can obtain gains from exchange. Analogously, if agent 𝐴 is endowed with the full ownership of the property rights then 𝐴 can
obtain gains from exchange. Under conventional Coasean argumentation, the agent that does not hold the initial property rights
must compensate the other agent for their cost to reach the efficient level 𝑥∗ as well as compensate to elicit agreement. Note there
exists gains from exchange for both agents, regardless of the initial property-rights owner.10
From this benchmark model, we incorporate transaction costs within the Coasean framework. In particular, we allow two forms
of transaction costs that include (i) a cost to influence the probability of obtaining the initial endowment of the property rights,
what we denote as rent-seeking costs, and (ii) a cost involved in the bargaining process, denoted as bargaining costs. Rent-seeking
costs are those costs associated with activities that are used to establish the initial ownership of the property rights. We thus assume
that property rights are initially insecure and agents have the choice to invest in effort in order to obtain the property rights and
allow for the potential of trade, i.e., to determine the initial endowment of property rights.11 Agents’ efforts can be interpreted as
litigation or legal services used to determine the probability of obtaining property rights ownership, lobbying and rent seeking, as
well as (violent) conflict. Formally, we model agents’ attempts to obtain the initial endowment of property rights using a Tullock
contest with endogenously determined linear costs of (sunk) effort 𝑎, 𝑏 ∈ R+ for agents 𝐴 and 𝐵, respectively. The probability of
agent 𝐵 obtaining the initial endowment of property rights is 𝑝𝐵 (𝑎, 𝑏) where
{ 𝑏
𝑎+𝑏
if max{𝑎, 𝑏} > 0
𝑝𝐵 (𝑎, 𝑏) = 1
(1)
2
otherwise
and 𝑝𝐴 (𝑎, 𝑏) = 1 − 𝑝𝐵 (𝑎, 𝑏) for all 𝑎, 𝑏 ∈ [0, ∞). That is, if agent 𝐵 increases effort, ceteris paribus, then the probability of obtaining the
property rights increases and decreases for agent 𝐴.
Second, we include bargaining costs that resemble the traditional costs discussed within the Coasean environment. In particular,
we suppose that there are costs associated with exchange and present them as reductions in agents’ expected payoffs after voluntary
exchange has been determined. To capture the whole host of potential combinations of costs, we allow both winners and losers to
experience bargaining costs. We denote 𝑇𝑗𝑊 and 𝑇𝑗𝐿 as the bargaining costs when agent 𝑗 ∈ {𝐴, 𝐵} win (𝑊 ) or lose (𝐿) the initial
endowment of property rights. Clearly, for sufficiently large bargaining costs, there may be an incentive for agents not to reach an
efficient outcome. Our focus here is on how transaction costs affect the incentive to invest in rent seeking, therefore we assume that
these bargaining costs are small enough to allow Coasean bargaining. As a consequence of including the two types of transactions
costs, we now have a two-stage Coasean game that we solve as a subgame-perfect Nash equilibrium.
In stage one of the game, agents choose the level of rent seeking to influence the probability of obtaining the initial allocation
of property rights. Once stage one is complete agents can freely bargain (at cost) to determine the agreed level of harm. In the
second stage – as property rights have already been initially allocated – the rational agent will attempt to maximize their gains
from exchange. Within the bargaining game we create a general framework by not placing any structure on the bargaining process
(we can, however, adopt structures such as a Nash bargaining solution, see Appendix A). This aligns to the experimental design in
which we allow for an unstructured bargaining process that mimics the reality of bargaining over property rights. Consequently, we
would expect, as is standard in Coasean argumentation, that the efficiency of the bargaining outcome is independent of the initial
property owner, how the initial allocation of property rights were allocated (and the associated costs), as well as the existence (and
any asymmetry) of bargaining costs. This is presented within Proposition 1.
Proposition 1. Assuming bargaining costs are sufficiently small, the Coase theorem predicts that efficient abatement (𝑥∗ ) should occur
regardless of
(a) which party holds the initial allocation of property rights

(b) the process used to determine the initial property rights allocation
(c) the presence or absence of bargaining costs,
(d) the symmetry or asymmetry of bargaining costs
We now turn to stage one. First, note that the value of the property rights is affected by the composition of bargaining costs. As we
model rent seeking through a contest, it is well known that the players’ equilibrium rent-seeking efforts are normally monotonically
increasing in the value of rent at stake. As can be observed in Appendix A, each agents’ incentive to invest in rent-seeking activities
is now dependent on the relative difference between their bargaining costs when they win or lose 𝑇𝑗𝐿 − 𝑇𝑗𝑊 for 𝑗 ∈ {𝐵, 𝐴}. Second,
note that if there exists symmetric bargaining costs—symmetric in the sense that each agent experiences the same cost regardless
of winning or losing, i.e., 𝑇𝑗𝐿 = 𝑇𝑗𝑊 —then we find an identical effort level to the case when no bargaining costs exist. From this we
can derive a number of predictions that are related to how the endogenously determined rent-seeking costs are influenced by the
conventional bargaining costs.
Proposition 2. Assuming bargaining costs are sufficiently small and holding bargaining power constant, rent-seeking costs chosen by players
are increasing in the magnitude of the difference in transaction costs between losing and winning. This leads to the following predictions:
10 Thus both the ‘‘efficiency’’ and ‘‘invariance’’ versions of the Coase theorem are relevant for our analysis (see Hurwicz, 1995; Robson and Skaperdas, 2008;
Medema, 2020).
11 Thus this diverts from the literature that assumes a fixed cost of entry into the Coasean bargaining game (e.g., Anderlini and Felli, 2001, 2006).
5
Table 1
Treatment design.
Treatment label Contest Bargaining costs Asymmetric Subjects(Groups)
RandomNo NO NO NO 40(6)
ContestNo YES NO NO 38(6)
ContestSym YES YES NO 40(6)
ContestAsym YES YES YES 40(6)
(a) Rent-seeking efforts are unaffected by the presence of symmetric bargaining costs, where symmetric means the same costs when winning
and losing the initial property rights.
(b) Rent-seeking efforts are increasing in the difference between bargaining costs when losing versus winning the property rights.
We now turn to the experimental design.
4. Experimental design
The design of the experiment closely follows the theoretical model, where all treatments consist of two stages. Stage one is
the contest stage and stage two is the bargaining stage. The purpose of the contest stage is to assign the property rights to one
participant in each pair. Consistent with past experiments (e.g., Hoffman and Spitzer, 1982; Shogren, 1992) we use the neutral
framing of ‘‘controller’’ in the experimental instructions to refer to the agent who holds the property rights, i.e., the ‘‘property-rights
holder’’, but use the latter term throughout this article. In each treatment, participants can bargain over their final payoff in the
second stage (which is equivalent to bargaining over the level of the externality).
There are four treatments, which differ in the presence of contests and bargaining costs, as shown in Table 1. In three treatments,
a Tullock contest is used to allocate the property rights but the bargaining costs vary. In the ContestNo treatment, there are no
bargaining costs, while in the ContestSym treatment all participants have to pay the same amount of bargaining costs when they
reach agreements in the bargaining stage. In the ContestAsym treatment, different participants pay different amounts of bargaining
costs depending on their roles and their ownership of the property rights. Since our design differs considerably from other Coase
experiments, we also include a control treatment, RandomNo, where property rights are randomly assigned and there are no costs
associated with bargaining. Our design enables us to separately identify the impact of contests and bargaining costs of different
types. As noted earlier, bargaining costs can be asymmetric in different ways. However, as shown in Appendix A, the value of the
property rights, and therefore bidding behavior in the Contest treatments, is affected only when a person’s bargaining costs differ
when winning and losing, and we chose to study this case.
4.1. Bargaining stage
In the bargaining stage of all treatments, participants are given the list of options shown in Table 2 and each pair is asked to
choose one option from this list. To avoid potential experimenter demand effects, total payoffs are not presented to the participants,
however, since participants are provided with a basic calculator, these can easily be computed. The table shows that there is a
trade-off between the payoffs to both participants of a pair. As a result, asking participants to pick an option for their pair creates
conflict between both participants.
The list of options and payoffs is the same in all treatments and is consistent with the theoretical model outlined in the previous
section, where subjects are given an endowment of $E600 in the bargaining stage so that they bargain over gains rather than losses
as shown in the Table. As shown in Appendix A, this fixed endowment does not alter the marginal incentives of the players nor
the value of the property rights. Choosing an option corresponds to choosing the level of abatement (or externality).12 Option 3
is the efficient choice, regardless of which player holds the property rights. Option 1 is the profit maximizing choice for player B
that causes the most harm, and Option 7 is the zero harm option. The asymmetry in the payoffs between the players (Hoffman and
Spitzer, 1985; Shogren, 1992), avoids creating a focal point at the middle option.
Participants are told that there are two ways to choose an option from the list. First, the property-rights holder can unilaterally
choose an option. Second, the two players can negotiate and reach an agreement regarding the division of the total available
surplus. To negotiate, subjects make an offer, stating the option that they want to choose and the payoffs that they want to keep for
themselves. Their partner then decides whether to accept the offer or to make other offers. Once made, offers cannot be withdrawn.
As shown in Table 2, successful negotiation can result in greater payoffs for both partners than a unilateral choice. For example,
suppose Player B is the property-rights holders. Unilaterally choosing Option 1 yields a payoff of $E600 for Player B and $E100 for
Player A. If instead they agree to choose Option 3, then there is $E900 to be split between them, or an extra $E200. If they choose
an equal split of the gains from trade then player A would receive $E200 and player B $E700. Clearly other allocations are possible.
12 The underlying functions we use are 𝐶(𝑥) = 6𝑥2 and 𝐷(𝑥) = 4𝑥2 − 90𝑥 + 500. Subjects are given a discrete version of this choice problem, with the options
chosen to ensure a reasonable difference between the options. The actual underlying abatement levels range from 0 to 10 with the efficient choice being option
3 with 𝑥 = 4.
6
Table 2
Options and payoffs in the bargaining stage in all treatments.
Options 1 2 3 4 5 6 7
Payoff A ($E) 100 214 396 542 575 590 600
Payoff B ($E) 600 589 504 285 172 87 0
Total payoffs 700 803 900 827 747 677 600
Table 3
Bargaining costs ($E).
Treatment Type A Type B
Property-rights Non-property-rights Property-rights Non-property-rights
holder holder holder holder
RandomNo 0 0 0 0
ContestNo 0 0 0 0
ContestSym 50 50 50 50
ContestAsym 15 95 5 85
In contrast to earlier Coasean experiments (Hoffman and Spitzer, 1982, 1985, 1986; Shogren, 1992; Rhoads and Shogren, 1999;
Cherry and Shogren, 2005), which employ face-to-face bargaining via a mediator, we use anonymous bargaining conducted over
the computer to reflect that Coasean-type negotiations are often done at arm’s length by lawyers and other employed negotiators
who, while not anonymous, are less likely to be swayed by fairness considerations. Another contrast is that we use a more general,
unstructured, bargaining protocol (Anbarci and Feltovich, 2018; Feltovich, 2019; Karagözoğlu and Kocher, 2019) rather than a
structured sequence of offers and counteroffers. An unstructured bargaining protocol ensures that both parties are strategically
equal in the bargaining process, face no constraints on either the timing or sequence of offers, and generates a rich set of data
regarding the bargaining process (Camerer et al., 2019). Subjects are given three minutes to bargain. If no agreement is reached
nor a unilateral choice made by the property-rights holder, the disagreement outcome is implemented.13
In RandomNo and ContestNo there are no bargaining costs when choosing an option by negotiation, while in the ContestSym
and ContestAsym treatments, participants have to pay bargaining costs if they choose options by negotiation. Table 3 shows the
amount of bargaining costs in each treatment. While total bargaining costs are the same in both treatments ($E100), the allocation
of bargaining costs differs. In ContestSym, all participants pay the same bargaining costs regardless of their role and whether they
have the property rights. In contrast in ContestAsym, bargaining costs depend on the role and whether the person initially holds the
property rights with bargaining costs $E80 higher for the non-property-rights holder than the property-rights holder. Our design
maximizes this asymmetry in order to maximize the difference between treatments.
Note that we have chosen to impose bargaining costs only if negotiation is successful in order to encourage negotiation and give
it the greatest chance of success. Such costs could reflect the expense of contract preparation (e.g., legal fees) once an agreement
has provisionally been accepted. This approach also aligns more closely with our unstructured bargaining protocol compared to the
alternative approach of separate per-unit costs of making offers, evaluating offers, and making counteroffers used by Rhoads and
Shogren (1999) and Cherry and Shogren (2005) in their structured bargaining environment. Finally, our design allows us to directly
compare the impact of bargaining costs on the equilibrium rent-seeking bids. Through the subgame-perfect equilibrium, the structure
of bargaining costs impacts the ex ante value of the property rights and thus the equilibrium rent-seeking bids (Proposition 2). If
we were to choose an alternative design then the ex ante value of property rights would be endogenously determined and provide
less clarity as to the relationship between rent-seeking bids and bargaining costs.
We chose bargaining cost values that are large enough to be meaningful but still ensure that bargaining is beneficial. That is,
in both ContestSym and ContestAsym there remains a significant rent available to be shared if negotiation is successful. If Player A
is the property-rights holder with Option 7 as their outside option, then the available rent is either $E300 when bargaining costs
are zero or $E200 when bargaining costs are present. If instead Player B is the property-rights holder, with Option 1 as the outside
option, then the available rent is lower at $E200 when bargaining costs are zero or $E100 when bargaining costs are present. Thus,
regardless of the presence of bargaining costs – and who initially holds the property rights – there is an incentive to reach the
efficient outcome. These theoretical predictions are summarized in Table 4.
4.2. Property rights stage
In the first stage of the experiment, the initial allocation of property rights is determined. RandomNo is the control treatment
where the property rights are allocated randomly using a virtual coin toss so each person has an equal chance of being the
property-rights holder. In the other treatments, property rights are allocated via bidding in a Tullock contest. In the Tullock
contest each subject submits a bid for the property rights where the probability of winning is given by the standard contest success
13 As described, we also conduct a control treatment without contests and bargaining costs, which enables us to study the effect of this change in bargaining
design compared to the existing literature.
7
Table 4
Theoretical predictions.
Scenario Equilibrium bargaining outcome
Initial property owner 𝑥∗ : option 3 (100% efficiency)
(either A or B obtain the rights) obtained in both
The initial allocation process 𝑥∗ : option 3 (100% efficiency)
(random versus the use of rent seeking) obtained in both
The introduction of transaction costs 𝑥∗ : option 3 (100% efficiency)
(bargaining costs versus no bargaining costs) obtained in both
Composition of bargaining costs 𝑥∗ : option 3 (100% efficiency)
(symmetric versus asymmetric bargaining costs) obtained in both
Scenario Equilibrium rent-seeking costs
Symmetric bargaining costs versus {𝑎∗ , 𝑏∗ } identical in both scenarios,
zero bargaining costs 𝑎∗ = 𝑏∗ = 137.5
Asymmetric bargaining costs versus {𝑎∗ , 𝑏∗ } larger under asymmetric
symmetric bargaining costs bargaining costs, 𝑎∗ = 𝑏∗ = 157.5
function (Tullock, 1980), as presented in Eq. (1). The design of the contest follows the standard experimental design (Dechenaux
et al., 2015; Cason et al., 2020). Importantly, at the bidding stage subjects are made fully aware of the bargaining stage that is to
follow. Consistent with theory (see Appendix A), their bids should be based on expectations of the value in the bargaining stage,
which is derived from their expectations of being a property-rights holder and non-holder. Furthermore, anticipated bargaining
power will affect the optimal bid. To derive the predicted bid, we assume that bargaining power is equal (i.e., 0.5), which seems
a reasonable assumption a priori, and aligns with the conventional Nash bargaining solution. Consistent with Proposition 2, the
optimal bid in both ContestNo and ContestSym is $E137.50 but higher in ContestAsym at $E157.50.14 Subjects receive an endowment
from which their bid is deducted. In ContestNo and ContestSym the endowment is $E200, while it is $E220 in ContestAsym to allow
for the higher optimal bid. Earnings from stage one equal the difference between the endowment and the bid. To maintain relative
earnings across treatments, in RandomNo subjects received an endowment of $E63 being equal to the expected earnings in stage
one of the other treatments.15 Subjects were given one minute to enter their bid and were informed that if no entry was made in
the time allowed a bid of zero was assumed. Our predictions are summarized in Table 4.
4.3. Procedure
The experiment proceeded as follows. First, subjects undertake a simple risk elicitation task (Gneezy and Potters, 1997).
Participants were given A$5 and had to decide how much to invest in a risky asset. If the investment succeeds, they win three times
the amount invested, but earn nothing if it fails. The chance of success is 50%. Subjects are only informed of the results at the end of
the session. Second, detailed instructions for task two (the main task) were distributed and read aloud. The experimental instructions
are contained in the Supplementary Appendix. Subjects then answered a series of quiz questions to check their understanding, before
undertaking two practice periods. Eight real periods followed. Subjects were paid their earnings in only one randomly selected
period, in addition to their earnings in the risk task.
At the beginning of the experiment, subjects were randomly assigned the role of A or B, and then assigned to one of three
matching groups per session. In each period they are randomly matched with a subject of the other type in their matching group. Each
matching group contains either six or eight participants. Matching groups and types remain unchanged throughout the experiment.
We conducted two sessions for each treatment. Sessions lasted on average 90–120 min. All payoffs are denoted in experimental
dollars ($E). At the end of the experiment, payoffs are converted to Australian Dollars with the rate E$18 = A$1. Average earnings
were A$31.97 (equivalent to US$23.63 at the time of experiment) and ranged from $5.50 to $59.40. The experiment was fully
computerized using z-Tree (Fischbacher, 2007). To recruit participants, invitations were sent out to the students who have registered
in the experimental database (Greiner, 2015). Participants were students with 75% being undergraduate students, 72% studying
business, economics or commerce, 48% were male, and 66% were international students.
4.4. Experimental hypotheses
In this section, we derive two experimental hypotheses based on our theoretical model and the parameters chosen. We also
add a third hypothesis regarding the distribution of gains. First, consider the efficiency of the bargaining outcome. Proposition 1
14 Since our main interest, as detailed in Hypothesis 2 below, is in the comparative static predictions, this assumption regarding bargaining power is of less
consequence. Indeed for Hypothesis 2 to hold we only require that the distribution of the subject beliefs regarding bargaining power is held constant across
treatments.
15 Average earnings were very similar across the four treatments, being A$31.86 in RandomNo, A$32.87 in ContestNo, A$32.08 in ContestSym, and A$31.10 in
ContestAsym. This design avoids a situation where subjects in different treatments enter the bargaining stage with systematically different endowments, which
might influence bargaining outcomes.
8
Table 5
Mean (S.D.) of main outcome variables.
Treatment Efficient outcome Bid Bid/Endowa PR Holder’s PR Holder’s
reached (% of pairs) BPowerb shareb
RandomNo 0.72 n/a n/a 0.08 0.69
(0.45) (0.42) (0.11)
ContestNo 0.62 134 0.67 0.32 0.75
(0.49) (65) (0.33) (0.39) (0.11)
ContestSym 0.63 127 0.64 0.02 0.76
(0.48) (65) (0.32) (0.78) (0.12)
ContestAsym 0.40 158 0.72 0.29 0.79
(0.49) (73) (0.33) (0.68) (0.10)
All 0.59 140 0.67 0.16 0.74
(0.49) (69) (0.33) (0.59) (0.12)
a
Bid as a proportion of provided endowment in each treatment.
b
Efficient bargains only.
should hold in our experiment. Bargaining is defined as efficient if the total payoff in a pair is maximized; i.e., Option 3 is the
outcome. As described earlier, bargaining costs are large enough to be meaningful but small enough that negotiating to Option 3
generates gains from trade. In our experiment, once property rights are allocated they are well-defined and secure, satisfying the
requirements of the Coase theorem. The property-rights holder can make a unilateral choice that is implemented with certainty. We
translate Proposition 1 into the following testable Hypotheses of the Coase theorem.
Hypothesis 1. Bargaining leads to the efficient option being chosen
(a) Regardless of whether player A or B is the property-rights holder.

(b) Equally often in the RandomNo and ContestNo treatments.
(c) Equally often in the ContestNo and ContestSym treatments.
(d) Equally often in the ContestSym and ContestAsym treatments.
Second, we consider the bids in the contest stage. As described in Proposition 2, optimal bids are only affected when the
bargaining costs differ between winning and losing the property rights. These costs are different only in our ContestAsym treatment.
Thus, this leads to the following testable hypothesis.
Hypothesis 2. Bidding in stage 1:
(a) Stage 1 bids are the same in ContestNo and ContestSym.

(b) Stage 1 bids are higher in ContestAsym than ContestNo and ContestSym.
(c) Stage 1 bids are the same for both Player A and Player B.
Hypothesis 2 relies on the assumption that the distribution of participants’ beliefs regarding their bargaining power is the same
across treatments.
Third, we consider the distribution of the rents (gains from trade) when subjects reach agreement in the bargaining stage. Since
the property-rights holder has the greater outside option (disagreement payoff) we expect them to have greater bargaining power and
therefore receive more than half of the rent available in any agreement (Binmore et al., 1989). However, as discussed in Section 2,
experimental evidence suggests that property-rights holders are less likely to exploit this bargaining power when the property rights
are allocated randomly (Feltovich, 2019). These findings suggest the following hypothesis.
Hypothesis 3. Property-rights holder’s rent:
(a) The property-rights holder has greater bargaining power and therefore appropriates more than half of the rent.
(b) The property-rights holder receives a greater share of the rent in the Contest treatments compared to the RandomNo treatment.
5. Results
In this section, we formally test our three hypotheses beginning with efficiency, followed by bidding behavior in stage one, before
finally considering the distribution of rents. Summary statistics for the measures analyzed in this section are provided in Table 5.
5.1. Efficiency of bargaining outcomes
In this section, we examine whether the efficient outcome is reached. Since the outcome is the same for both parties in a pair, the
analysis is conducted at the pair level. Bargaining outcomes are summarized in Table 6. Recall that bargaining outcomes are reached
in one of three possible ways. Of the 632 pair observations of bargaining outcomes, 62% result in an agreement between the parties,
9
Table 6
Summary of bargaining outcomes.
Method of outcome % of outcomes Option chosen within a method (% of outcomes)
Outside (option 1 or 7) Efficient (option 3) Other option
Agreement 62 0 95 5
Unilateral choice 24 84 1 15
Disagreement outcome 14 100 n/a n/a
Overall 100 34 59 7
24% involve the property-rights holder choosing unilaterally, and in 14% no active choice is made and instead the disagreement
outcome is implemented. Of those pairs reaching an agreement, almost all (95%) choose the most efficient option. Thus, reaching
an agreement and choosing the efficient option are virtually synonymous.16 This is unsurprising since if both parties agree (and
therefore incur any bargaining costs) they should always choose the option with the largest total payoff. This also implies that any
observed treatment differences will result from differences in the frequency of successful negotiations rather than differences in the
option chosen if negotiation is successful.
Of those property-rights holders choosing unilaterally, 84% choose the best option for themselves, either Option 1 or 7 depending
on their type. Option 3 is chosen unilaterally in only 1% of unilateral choices or less than 0.25% of overall choices, reinforcing that
the correlation between successful negotiation and choosing the most efficient option is extremely strong. Options 5 (5% of unilateral
choices) and 6 (7% of unilateral choices) are the most common alternative unilateral choices. The unilateral choice of other options
is more common in ContestNo (28% of unilateral choices) than in either ContestSym (13%) or ContestAsym (12%), suggesting that
avoidance of transaction costs is not the reason for such choices.
We next test if Hypothesis 1 holds by examining the likelihood of the efficient option being chosen in each treatment.
Result 1. Efficiency of bargaining outcomes
(a) The likelihood of the efficient outcome is unaffected by whether player A or B is the property-rights holder.
(b) The efficient outcome is significantly less likely to occur in ContestNo than RandomNo.
(c) There is no significant difference in the likelihood of the efficient outcome in ContestNo compared to ContestSym.
(d) The efficient outcome is significantly less likely to occur in ContestAsym than ContestSym.
Support. Fig. 1 shows the frequencies of outcomes disaggregated by treatment. The figure shows that Option 3, the efficient
option, is the most common outcome in all treatments, followed by the two outside options (1 and 7). Option 3 occurs more often in
RandomNo (72%) than in ContestNo (62%), similarly in ContestNo and ContestSym (63%), and less in ContestAsym (40%). Thus, even
in the absence of any transaction costs (either rent seeking or bargaining costs), the subjects fail to achieve the efficient outcome
more than one-quarter of the time. This contrasts with the theoretical predictions shown in Table 4 of 100% efficiency across all
treatments.
To formally test Hypothesis 1, we employ random effects panel regressions, using outcomes from the 79 pairs in each of the 8
rounds of the experiment. We use the following specification:
𝐸𝑓 𝑓 𝑖𝑐𝑖𝑒𝑛𝑡𝑖𝑡 = 𝛼 + 𝛽 ′ 𝑇𝑖 + 𝛿 ′ 𝐸𝑖𝑡 + 𝛾 ′ 𝑋𝑖 + 𝑢𝑖 + 𝑒𝑖𝑡 (2)
where 𝑖 indexes the property-rights holder and 𝑡 indexes the round. The dependent variable (𝐸𝑓 𝑓 𝑖𝑐𝑖𝑒𝑛𝑡𝑖𝑡 ) is an indicator of
whether or not property-rights holder 𝑖 reached the efficient option in round 𝑡, 𝑇𝑖 is a vector of time invariant treatment indicators
𝐶𝑜𝑛𝑡𝑒𝑠𝑡𝑁𝑜𝑖 , 𝐶𝑜𝑛𝑡𝑒𝑠𝑡𝑆𝑦𝑚𝑖 , 𝐶𝑜𝑛𝑡𝑒𝑠𝑡𝐴𝑠𝑦𝑚𝑖 (with 𝑅𝑎𝑛𝑑𝑜𝑚𝑁𝑜𝑖 as the base category); 𝐸𝑖𝑡 is a vector of additional experimental controls
containing 𝑃 𝑒𝑟𝑖𝑜𝑑𝑡 , that takes on an integer value to control for learning, and 𝑇 𝑦𝑝𝑒𝐴𝑖 is an indicator of the property-rights holders
type; 𝑋𝑖 is a vector of the property-rights holder characteristics, 𝑢𝑖 is the random effect and 𝑒𝑖𝑡 is the idiosyncratic error clustered
at the matching group level.17
Given the binary nature of the dependent variable, we use a probit regression.18 Results are shown in column (1) of Table 7,
with the bottom panel of the table reporting p-values from Wald tests of treatment comparisons.
The results reveal the following. First, the property-rights holder type has no significant effect, consistent with the standard Coase
theorem and Hypothesis 1a. Second, the Contest significantly reduces the probability of reaching the efficient option compared to
Random in all three Contest treatments. Third, while the difference between ContestNo and ContestSym is insignificant (supporting
16 This is true of agreed outcomes in all treatments; specifically, 98% of successful bargainers in RandomNo choose Option 3, compared to 91% in ContestNo,
96% in ContestSym, and 93% in ContestAsym.

17 We include the following property-rights holder characteristics: is male, locally born, an economics major, participating in their first experiment, and their
university year. Note that due to the strong correlation between gender and our measure of risk preference (specifically, the amount invested in the risk task) we
do not include both in the regression. Full regression results including demographics are reported in column (1) of the Supplementary Appendix Table A1 and
show that locally born property-rights holders (𝑝 < 0.01) and male property-rights holders (𝑝 < 0.1) are significantly more likely to reach the efficient outcome
while other demographics are not significant.
18 Our results are qualitatively unaffected if we instead use a linear probability model.
10
Fig. 1. Frequency of outcomes disaggregated by treatment.
Table 7
Random effects regressions of main results.
Dependent variable Efficient Bid Bid/ BPower- Share-
option endowment efficient efficient
Method Probit Tobit Tobit Linear Linear
(1) (2) (3) (4) (5)
Type A −0.1457 −2.9925 −0.0130 −0.0681 0.0007
(0.1383) (8.7298) (0.0424) (0.1107) (0.0204)
Period 0.0756∗∗∗ 6.6849∗∗∗ 0.0325∗∗∗ 0.0300∗∗ 0.0070∗∗
(0.0239) (1.7694) (0.0086) (0.0131) (0.0031)
ContestNo −0.3422∗∗ – – 0.2246∗∗∗ 0.0576∗∗∗
(0.1682) (0.0741) (0.0214)
ContestSym −0.4333∗∗∗ −18.9966 −0.0898 0.0096 0.0760∗∗∗
(0.1485) (24.4689) (0.1196) (0.1222) (0.0211)
ContestAsym −0.9549∗∗∗ 25.3626 0.0564 0.0890 0.0858∗∗∗
(0.1569) (24.6007) (0.1184) (0.1183) (0.0204)
Constant 0.4023 86.7401∗∗∗ 0.4356∗∗∗ −0.2226 0.6173∗∗∗
(0.2998) (22.0987) (0.1087) (0.1742) (0.0312)
PR Holder Demographics YES YES YES
Individual Demographics YES YES
N 632 944 944 372 372
No. Left-censored obs 41 41
No. Right-censored obs 282 282
ContestNo vs. ContestSym 0.6514 n/a n/a 0.0817 0.4519
ContestSym vs. ContestAsym 0.0071 0.0391 0.16 0.6167 0.6976
Coefficient estimates reported with robust standard errors clustered on matching groups in parentheses.
The omitted treatment in (1), (4) and (5) is RandomNo. The omitted treatment in (2) and (3) is ContestNo.
Column (1) includes all outcomes with the level of observation at the PR holder.
Columns (2) and (3) include all bids with the level of observation at the individual.
In (2) the lower limit is at zero and the upper is at the endowment in that treatment.
In (3) the lower limit is zero and the upper limit is at one.
Columns (4) and (5) include all efficient bargaining outcomes with the level of observation the PR holder.
The last two rows report p-values from Wald tests of treatment comparisons.
*𝑝 < 0.10, **𝑝 < 0.05, ***𝑝 < 0.01.
11
Hypothesis 1c), the efficient option is significantly less likely to be reached in ContestAsym than ContestSym. Fourth, the likelihood
of choosing the efficient option increases with experience. These results support Hypotheses 1a and c, but not 1b or d. This implies
that rent-seeking costs reduce the efficiency of outcomes, but bargaining costs do so only if they are asymmetric. These effects are
not only significant but also large in magnitude. The average treatment (marginal) effect in the ContestNo and ContestSym treatments
versus the RandomNo is between 12–15 percentage points (about an 18% reduction), while the effect is even larger in ContestAsym
with a 33 percentage point reduction (nearly a 50% reduction) compared to RandomNo. □
With regard to the overall efficiency of Coasean bargaining, across all of our treatments efficiency is substantially lower than
the 100% level predicted by theory. In fact, even in the absence of any transaction costs (either rent-seeking or bargaining),
bargaining failed nearly 30% of the time, increasing to nearly 40% of the time when rent-seeking costs were added (but still no
bargaining costs). These lower levels of efficiency contrast sharply with outcomes from earlier experiments that found that Coasean
bargaining was highly efficient. For example, Hoffman and Spitzer (1982, 1985, 1986) all report that at least 90% of decisions
in their experiments were Pareto optimal (i.e., maximized joint profits), while Shogren (1992) found that 87% of decisions were
efficient even when bargaining over uncertain outcomes. As discussed earlier, a key design difference was our use of an unstructured
bargaining protocol compared to the structured process conducted face-to-face via a mediator employed in these earlier experiments.
A possible explanation for our lower efficiency results is that unstructured bargaining might impose additional, implicit, costs of
bargaining, which reduce the likelihood of negotiating to the efficient outcome. Of course, since we use the same bargaining protocol
across all of our treatments, such costs are held constant in our treatment comparisons. On the other hand, our results suggest that
the success of Coasean bargaining reported in earlier experiments might be an upper bound.
In the presence of bargaining costs, we found that only 63% of outcomes reached the efficient outcome with symmetric bargaining
costs, falling further to only 40% with asymmetric bargaining costs. These lower efficiency levels correspond more closely with
those in other papers with bargaining costs. For example, Cherry and Shogren (2005) found efficiency of around 45% in the case
of secure property rights with bargaining costs, while Rhoads and Shogren (1999) observed efficiency of around 50% in their high
cost treatment. Compared to these papers we chose to impose bargaining costs only for successful bargains (rather than a design
where subjects pay for engaging in the bargaining process itself - e.g., per-offer made). As described earlier, we chose this design in
order to encourage bargaining, to allow us to derive the equilibrium rent-seeking bids, and because it seemed more natural in an
unstructured bargaining context. Our design is therefore likely to be conservative in terms of understating the true negative effects
of bargaining costs both symmetric and asymmetric.
5.2. Bidding behavior
We now test Hypothesis 2 regarding bidding behavior in stage one of the three contest treatments using individual-level data.
Result 2. Bidding in stage 1
(a) There is no significant difference in bidding between ContestNo and ContestSym.

(b) Absolute bids are significantly higher in ContestAsym than in ContestSym but bids as a proportion of endowment are not significantly
different.
(c) There is no significant difference in bidding between different types of property-rights holders.
Support. Recall that the optimal bid as summarized in Table 4, assuming equal bargaining power, in both ContestNo and
ContestSym is $E137.50 but higher in ContestAsym at $E157.50. Bids are summarized in Table 5 with the distribution shown in
Fig. 2. While average bids are close to the theoretical predictions, there is a large variation in bidding behavior with bids ranging
from 0 to the maximum possible in all treatments. There are no bids exactly at the predicted amounts, instead across all treatments
the modal bid is the maximum amount. However, as predicted, bids are higher in ContestAsym. Recall that to compensate for the
higher predicted bid in ContestAsym and ensure that subjects entered the bargaining stage with similar earnings in all treatments
we provided a higher endowment of $E220 compared with $E200 in ContestNo and ContestSym. Hence, we also compare bids as
a proportion of the endowment. As shown in Table 5, the difference between treatments is less pronounced using this measure
although the mean proportional bid remains higher in ContestAsym (0.72) than in either ContestNo (0.67) or ContestSym (0.64).
To formally test Hypothesis 2, we again employ a random effects panel regression but this time using 158 individual level bids
in each of the 8 rounds of the experiment. The specification is as follows:
𝐵𝑖𝑑𝑗𝑡 = 𝛼 + 𝛽 ′ 𝑇𝑗 + 𝛿 ′ 𝐸𝑗𝑡 + 𝛾 ′ 𝑋𝑗 + 𝑢𝑗 + 𝑒𝑗𝑡 (3)
where 𝑗 indexes the individual and 𝑡 indexes the round. The dependent variable (𝐵𝑖𝑑𝑗𝑡 ) measures how much individual 𝑗 bids in
round 𝑡, 𝑇𝑗 is a vector of time invariant treatment indicators, 𝐶𝑜𝑛𝑡𝑒𝑠𝑡𝑆𝑦𝑚𝑗 , 𝐶𝑜𝑛𝑡𝑒𝑠𝑡𝐴𝑠𝑦𝑚𝑗 (with 𝐶𝑜𝑛𝑡𝑒𝑠𝑡𝑁𝑜𝑗 now the base category)
𝐸𝑗𝑡 is a vector of additional experimental controls, 𝑃 𝑒𝑟𝑖𝑜𝑑𝑡 , that takes on an integer value to control for learning, and 𝑇 𝑦𝑝𝑒𝐴𝑗 is an
indicator of the individual’s type, 𝑋𝑗 is a vector of the individual’s characteristics, 𝑢𝑗 is the random effect and 𝑒𝑗𝑡 is the idiosyncratic
error clustered at the matching group level.19
19 Individual characteristics are the same as those described earlier for property-rights holders in Footnote17 . Full regression results including demographics
are reported in columns (2) and (3) of the Supplementary Appendix Table A1 and show that subjects participating in their first experiment bid less (𝑝 < 0.05),
while those in later university years bid more (𝑝 < 0.05), while other demographics are not significant.
12
Fig. 2. Distribution of bids across treatments.
To account for the bunching of bids at the upper limit (and to a lesser extent at the lower limit of zero), we use Tobit mixed
panel regressions with the upper limit equal to the endowment in the treatment. The results are reported in columns (2) and (3)
of Table 7. We find no significant difference between bidding in ContestNo and ContestSym as predicted by theory and consistent
with Hypothesis 2a. We find a significant difference in bidding between the ContestSym and ContestAsym treatments but not between
ContestAsym and ContestNo, thus partially supporting Hypothesis 2b. However, the difference between bidding in ContestSym and
ContestAsym becomes insignificant once we adjust for the different endowments and compare bidding as a percentage of the provided
endowment. This is consistent with the high frequency of maximal bids in all treatments. While the optimal bid is derived on the
basis that bargaining always succeeds yet this is not the case in the experiment. The high likelihood that bargaining fails makes it
even more valuable to become the property-rights holder, which might explain the high number of maximum bids. Consistent with
Hypothesis 2c, we find no difference in bidding behavior between different types of property-rights holders. Finally, we find strong
evidence that bids increase with experience.20 □
5.3. Distribution of gains
We next test whether Hypothesis 3 holds. Our analysis in this section includes all successful bargains that reach the efficient
outcome.21 As in Section 5.1, our analysis in this section is conducted at the pair level. We construct two different measures of
the distribution of gains received by the property-rights holder, BPower and Share. Our first measure, bargaining power (BPower),
captures the share of the rent (i.e., the gains from trade) obtained by the property-rights holder and corresponds to Eqs. (4) and
(5) in our theoretical model within Appendix A. The rent is defined as the gain in total payoffs that pairs can achieve by moving
from the option that maximizes the property-rights holders’ payoffs (either Option 1 or Option 7) to the efficient option (Option 3),
net of any bargaining costs. In order to allow for social preferences, we also examine the share of the total available payoff (Share),
20 Estimation of a two-step model shows that the likelihood of bidding the maximum amount increases with experience, while interior bids are unchanged
by experience. In addition, the two-step results reveal that increased bids in ContestAsym compared to ContestSym result from increases in both the likelihood of
bidding at the maximum and from increases in the size of interior bids.
21 As noted earlier, only 5% of successful bargains result in an inefficient option being selected (Table 6). Results in columns (1) and (3) of the Supplementary
Appendix Table A2 show that including these outcomes in our regression analysis does not qualitatively change our findings.
13
Fig. 3. Distribution of the property-rights holder’s bargaining power by treatment (Efficient bargains).
net of any bargaining costs, that each party receives in the event of a successful bargaining outcome. Summary statistics for the
property-rights holders’ BPower and Share are shown in Table 5, while Fig. 3 shows the distribution of BPower in each treatment
and Fig. 4 the distribution of Surplus.
Result 3. Property-rights holder’s rent
(a) On average, the property-rights holder receives less than half of the available rent, implying that the non-property-rights holder actually
has the greater bargaining power. In contrast, on average, the property-rights holder receives more than half of the available (net)
surplus.
(b) The property-rights holder has significantly higher BPower in ContestNo compared to RandomNo, but no other treatment differences
are significant. In contrast, the property-rights holder’s Share is significantly higher in all three Contest treatments compared to
RandomNo but there are no differences between the Contest treatments.
(c) The property-rights holder’s type has no impact on either BPower or Share.
Support. As shown in Fig. 3, BPower is quite variable, with the distribution more right-shifted in ContestNo and ContestAsym.
As shown in Table 5, the mean is positive in all treatments, with the property-rights holder having higherBPower in the ContestNo
and ContestAsym treatments and able to extract around one-third of the rent compared to nearly zero BPower in the RandomNo and
ContestSym treatments. Thus, contrary to Hypothesis 3a, property-rights holders often agree to receive less than half of the available
rent. In fact, in 18% of all successful bargains the property-rights holder actually receives less than in their outside option implying
a negative BPower. Negative BPower could arise if the agreed outcome involves a more even sharing of the surplus, for example, an
agreed split of (450, 450) in RandomNo would give equal shares of 0.5 to each party but imply a BPower of −0.5 for the property-rights
holder. Indeed, as shown in Table 5, property-rights holders get more than half of the available net surplus; about three-quarters on
average. This Share is somewhat lower in RandomNo compared to the Contest treatments, suggesting that rent-seeking costs matter.
Fig. 4 also shows that the distribution of Share is more left shifted in RandomNo. Exactly equal shares of 0.5 are also more common
in RandomNo (12%) than in ContestNo (0), ContestSym (5%) and ContestAsym (0).22
22 As shown in Fig. 4 there are a few observations (2%) where the property-rights holder’s agreed Share is less than 50%, which is hard to rationalize via social
preferences and we might consider these mistakes. Similarly, outcomes where BPower > 1 are equally difficult to rationalize as in this case the non-property-rights
holder receives less than in the outside option and instead gives a very large share (typically > 0.9) to the property-rights holder. Such potentially irrational
14
Fig. 4. Distribution of property-rights holder’s share of net surplus by treatment (Efficient bargains).
To formally test Hypothesis 3 we again use a random effects panel regression but at the pair level taking the perspective of
the property-rights holder. The specification is as in Eq. (2) with the differences being that the dependent variable is now either
BPower or Surplus, and that we use a linear regression. Results are reported in column (4) of Table 7. We find significantly higher
BPower in ContestNo compared to RandomNo, but no other significant differences between treatments. Thus we find partial support
for 3b. A possible explanation is that obtaining the property right gives you more entitlement in ContestNo versus RandomNo,
but the introduction of bargaining costs makes this more opaque. We also find no impact of player type but BPower increases
with experience. Regression results with Share as the dependent variable are reported in column (5) of Table 7. We find that the
property-rights holder’s Share is significantly higher in all three of the Contest treatments compared to the RandomNo treatment.
However, there are no differences between the three contest treatments. There is a small but statistically significant trend over time
as property-rights holders obtain a higher Share with experience. The type of the property-rights holder has no impact on the Share
obtained. Thus Hypothesis 3 holds when we use the Share measure, which allows for social preferences.23 □
Our results show that when parties have to incur rent-seeking costs to secure the property-rights, then the efficiency of Coasean
bargaining is significantly reduced. Result 3 suggests this might result from the larger shares that property-rights holders expect to
receive when they have incurred rent-seeking costs, which may make it less likely that an agreement is reached. To further support
this interpretation, we examine if the perceived ‘‘legitimacy’’ of the property-rights holder affects bargaining outcomes in the Contest
treatments. We conjecture that BPower might be proportional to the relative rent-seeking costs expended in stage one. Given our
experimental design involves a probabilistic contest, a subject may become the property-rights holder despite expending less effort
to establish the property right. In such a case, the legitimacy of the property-rights holder might be questioned meaning they are
less able to take advantage of that position. While on average those winning the property right did spend more contesting it in stage
one, due to the random nature of the contest outcome about 21% of the time the property-rights holder actually bid less than their
partner. In an additional 18% of cases, both partners bid an equal amount.24
outcomes, where either Share < 0.5 or BPower > 1, only occur in 3% of successful bargains overall, but are more common in ContestSym (4%) and ContestAsym
(6%), which suggests that the presence of bargaining costs could be the reason. However, due to the small numbers, we cannot detect differences in these
irrational outcomes. In the Supplementary Appendix Table A2 columns (2) and (4) we show that excluding these potentially irrational outcomes from our
analysis does not qualitatively change any of our regression findings.
23 Full regression results including demographics are reported in columns (4) and (5) of the Supplementary Appendix Table A1 and reveal that no demographic
factors significantly affect either measure.

24 As expected, due to randomization, the likelihood of these occurrences is very similar across treatments.
15
Fig. 5. Effect of PR-holder legitimacy on average BPower and surplus.
Result 4. Property-rights holders that are considered less legitimate have significantly and substantially less BPower and extract a
significantly smaller Share of the net surplus.
Support. First, we define a property-rights holder as legitimate if they bid more than their partner. As shown in Fig. 5, legitimate
property-rights holders command about 30% of the available rent on average and nearly 80% of the available surplus, whereas
property-rights holders who obtained the property right despite bidding less (and therefore spending less on rent seeking) actually
have a negative BPower. To test if these differences are significant, we include an indicator for whether the property-rights holder
bid less than their partner in the first stage (Bid Less) in our regressions of Share and BPower. The results reported in columns
(1) and (2) of the Supplementary Appendix Table A3 show that a less legitimate property-right has significantly and substantially
less bargaining power. Since average BPower is only 0.16 (refer to Table 5) the effect is large and implies that BPower is actually
negative for less legitimate property-rights holders. While the impact on Share is also significant the magnitude is small compared
to the average share of 0.74.25 □
6. Concluding remarks
This article considers Coasean bargaining when property rights are initially insecure and contestable. We design a two-stage
experiment where, in stage one, participants invest effort to influence the probability of obtaining the initial allocation of property
rights, which we denote as costly rent-seeking activity. In stage two, once the property rights are secure, players can bargain over
the level of externality. Within stage two, we allow for bargaining costs that can either be symmetric (identical to a player regardless
if they win or lose the initial property rights) or asymmetric (a difference exists if the property rights are either won or lost).
We find that costly rent-seeking activities to initially establish property-rights ownership has a detrimental impact on bargaining
efficiency. Compared to a random initial property-rights allocation, the use of costly rent-seeking effort significantly reduces
bargaining efficiency. The bargaining efficiency is significantly less likely when it is coupled with (either symmetric or asymmetric)
bargaining costs. In terms of how much is endogenously invested to capture the initial property-rights allocation, we find that there
is no significant difference between cases with zero bargaining costs and symmetric costs. This confirms our theoretical predictions.
When bargaining costs are asymmetric – i.e., the individual experiences different costs depending on if the initial property rights
are won or lost – we find investment in rent-seeking costs are larger than in the case with symmetric bargaining costs. This, again,
25 We also ran regressions using an alternative measure of potential legitimacy, the difference in bids between the property-rights holder and non-holder
(BidDiff ). As reported in columns (3) and (4) of the Supplementary Appendix Table A3, while BidDiff does positively affect both the BPower and Share of the
property-rights holder, the effect is only weakly significant (𝑝 < 0.1).
16
confirms our theoretical predictions. Finally, we observe that the property-rights holder extracts a greater share of the available net
surplus when having to contest for the property-rights, which might explain why bargaining is also less successful in this case.
Our analysis is useful in Coasean environments when parties invest in activities to acquire initial property-rights ownership. In
particular, our focus is on cases where property rights are initially contestable, but where use of effort – such as litigation, rent
seeking, and conflict – may result in the property-rights ownership becoming secure, which may lead to property-rights exchange.
Our findings may help in understanding the incentives to invest in costly rent-seeking activities over tradable property rights and
how this can be impacted by the process of initial property-rights allocation and the presence of bargaining costs. Overall, this
brings insight to more realistic settings of the Coase theorem that experience both rent seeking and bargaining costs and provides
evidence of a detrimental impact on the efficiency of bargaining when bargaining costs are asymmetric. Consequently, policy could
be designed to ensure parties experience similar bargaining costs.
Declaration of competing interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared
to influence the work reported in this paper.
Appendix A. Theoretical model
Consider a set of agents {𝐴, 𝐵}. Suppose agent 𝐵 participates in an activity that generates harm on agent 𝐴. Agent 𝐵 has the
ability to reduce the level of harm by investing in abatement activities, denoted as 𝑥 ∈ [0, 𝑥] ̄ ⊂ R+ , where 𝑥̄ eliminates the harm.
The investment in abatement occurs at a private cost to agent 𝐵 denoted by 𝐶(𝑥) with 𝐶(0) = 0 = 𝐶 ′ (0), 𝐶(𝑥) ̄ = 𝐶,̄ 𝐶 ′ (𝑥) > 0, and
′′
𝐶 (𝑥) ≥ 0. The damage experienced by agent 𝐴 is denoted by 𝐷(𝑥) where 𝐷(0) = 𝐷, 𝐷 (𝑥) ̄ ′ ̄ −𝐷 (𝑥) > 0, and 𝐷′′ (𝑥) ≥ 0.
̄ = 0 = 𝐷(𝑥), ′
∗ ∗
Let 𝑥 be the (ex post ) efficient Coasean bargaining solution, so that 𝑥 = min𝑥 {𝐶(𝑥)+𝐷(𝑥)}. If agent 𝐵 initially has the ownership
of the property rights then exchange can occur to 𝑥∗ , where agent 𝐵 can obtain gains from exchange given by
[ [ ]]
𝑅𝐵 ≡ 𝜇𝐵 𝐷̄ − 𝐶(𝑥∗ ) + 𝐷(𝑥∗ ) ≥ 0, (4)
where 𝜇𝐵 ∈ [0, 1] represents agent 𝐵’s relative bargaining power.26 Analogously, if agent 𝐴 is endowed with the full ownership of
the property rights then their gains from exchange are given by:
[ [ ]]
𝑅𝐴 ≡ 𝜇𝐴 𝐶̄ − 𝐶(𝑥∗ ) + 𝐷(𝑥∗ ) ≥ 0, (5)
where 𝜇𝐴 ∈ [0, 1]. We leave the 𝜇𝐴 and 𝜇𝐵 in a general form, but note that if the analysis was restricted to a Nash bargaining
solution then the share of the rent would be split such that 𝜇𝐴 = 𝜇𝐵 = 1∕2.
Under conventional Coasean logic, the agent that does not hold the initial property rights must compensate the other agent for
the cost incurred to reach (accept) the optimal level 𝑥∗ and also bear their own cost as well as compensating their rival to reach the
efficient equilibrium. That is, the agents must incur 𝐶(𝑥∗ ) + 𝐷(𝑥∗ ) for voluntary exchange to occur as well as the rent to the rival,
which is either 𝑅𝐵 or 𝑅𝐴 . Note that in this setting it continues to be optimal for gains from exchange to occur as the alternative
would be to incur either 𝐶̄ or 𝐷, ̄ for agents 𝐵 and 𝐴, respectively, which is a weakly higher expense than participating in exchange.
We suppose that there are costs associated with exchange and present them as reductions in agents’ expected payoffs after
voluntary exchange has been determined. To capture the whole host of potential combinations of costs, we allow both winners and
losers to experience bargaining costs. Formally, agents’ payoffs structures are given by:
( )
𝑈𝐵 (𝑏) = 𝑊 0 + 𝑝𝐵 (𝑎, 𝑏) 𝑅𝐵 − 𝑇𝐵𝑊 − [1 − 𝑝𝐵 (𝑎, 𝑏)][𝐶(𝑥∗ ) + 𝐷(𝑥∗ ) + 𝑅𝐴 + 𝑇𝐵𝐿 ] − 𝑏, (6)
0
( 𝑊
) ∗ ∗ 𝐿
𝑈𝐴 (𝑎) = 𝑊 + 𝑝𝐴 (𝑎, 𝑏) 𝑅𝐴 − 𝑇𝐴 − [1 − 𝑝𝐴 (𝑎, 𝑏)][𝐶(𝑥 ) + 𝐷(𝑥 ) + 𝑅𝐵 + 𝑇𝐴 ] − 𝑎, (7)
where 𝑇𝑗𝑊 and 𝑇𝑗𝐿 are the bargaining costs when agent 𝑗 ∈ {𝐵, 𝐴} win or lose the initial endowment of property rights, and 𝑊 0 is
the initial wealth of each agent. Clearly, for sufficiently large bargaining costs, there may be an incentive for agents not to reach
an efficient outcome. Our focus here is on how bargaining costs affect the incentive to invest in rent-seeking effort, therefore we
assume that these bargaining costs are small enough to allow Coasean bargaining.27 We solve the game as a subgame-perfect Nash
equilibrium, with efforts:
𝑣𝐴 𝑣2𝐵 𝑣2𝐴 𝑣𝐵
𝑏∗ = and 𝑎∗ = , (8)
(𝑣𝐵 + 𝑣𝐴 )2 (𝑣𝐵 + 𝑣𝐴 )2
where 𝑣𝐵 = 𝑅𝐵 + 𝐶(𝑥∗ ) + 𝐷(𝑥∗ ) + 𝑅𝐴 + 𝑇𝐵𝐿 − 𝑇𝐵𝑊 and 𝑣𝐴 = 𝑅𝐴 + 𝐶(𝑥∗ ) + 𝐷(𝑥∗ ) + 𝑅𝐵 + 𝑇𝐴𝐿 − 𝑇𝐴𝑊 . There are a number of distinctions
that exist from our benchmark case. Equilibrium rent-seeking costs are increasing in 𝑇𝑗𝐿 , decreasing in 𝑇𝑗𝑊 , and costs are increasing
in the size of 𝑇𝑗𝐿 − 𝑇𝑗𝑊 . Note also that costs are identical to the case of no transaction costs when 𝑇𝑗𝐿 = 𝑇𝑗𝑊 .
26 The loser, then, is able to capture (1 − 𝜇𝐵 ) of the rent.

27 In particular, we require 𝑅𝐵 + 𝐶(𝑥∗ ) + 𝐷(𝑥∗ ) + 𝑅𝐴 + 𝑇𝐵𝐿 > 𝑇𝐵𝑊 and 𝑣𝐴 = 𝑅𝐴 + 𝐶(𝑥∗ ) + 𝐷(𝑥∗ ) + 𝑅𝐵 + 𝑇𝐴𝐿 > 𝑇𝐴𝑊 for agents 𝐵 and 𝐴, respectively.
17
Appendix B. Supplementary data
Supplementary material related to this article can be found online at https://doi.org/10.1016/j.jeem.2023.102842.
References
Aivazian, V.A., Callen, J.L., McCracken, S., 2009. Experimental tests of core theory and the coase theorem: Inefficiency and cycling. J. Law Econ. 52 (4), 745–759.
Anbarci, N., Feltovich, N., 2018. How fully do people exploit their bargaining position? The effects of bargaining institution and the 50-50 norm. J. Econ. Behav.
Organ. 145, 320–334.
Anbarci, N., Skaperdas, S., Syropoulos, C., 2002. Comparing bargaining solutions in the shadow of conflict: How norms against threats can have real effects. J.
Econom. Theory 106 (1), 1–16.
Anderlini, L., Felli, L., 2001. Costly bargaining and renegotiation. Econometrica 69 (2), 377–411.
Anderlini, L., Felli, L., 2006. Transaction costs and the robustness of the coase theorem. Econ. J. 116 (508), 223–245.
Bar-Gill, O., Engel, C., 2016. Bargaining in the absence of property rights: An experiment. J. Law Econ. 59 (2), 477–495.
Barber, IV, B.S., English, W., 2019. The origin of wealth matters: Equity norms trump equality norms in the ultimatum game with earned endowments. J. Econ.
Behav. Organ. 158, 33–43.
Baye, M.R., Kovenock, D., Vries, C.G.d., 1993. Rigging the lobbying process: An application of the all-pay auction. Am. Econ. Rev. 83 (1), pp. 289–294.
Baye, M.R., Kovenock, D., de Vries, C.G., 2005. Comparative analysis of litigation systems: An auction-theoretic approach. Econom. J. 115, 583–601.
Binmore, K., Shared, A., Sutton, J., 1989. An outside option experiment. Q. J. Econ. 104 (4), 753–770.
Buchanan, J.M., 1975. The Limits of Liberty: Between Anarchy and Leviathan. University of Chicago Press, Chicago.
Camerer, C.F., Nave, G., Smith, A., 2019. Dynamic unstructured bargaining with private information: theory, experiment, and outcome prediction via machine
learning. Manage. Sci. 65 (4), 1867–1890.
Cason, T.N., Masters, W.A., Sheremeta, R.M., 2020. Winner-take-all and proportional-prize contests: Theory and experimental results. J. Econ. Behav. Organ. 175,
314–327.
Cherry, T., Cotten, S., Shogren, J., 2013. Coasean bargaining. In: Shogren, J.F. (Ed.), Encyclopedia of Energy, Natural Resource, and Environmental Economics.
Elsevier, Waltham, pp. 5–9.
Cherry, T.L., Frykblom, P., Shogren, J.F., 2002. Hardnose the dictator. Amer. Econ. Rev. 92 (4), 1218–1221.
Cherry, T.L., Shogren, J.F., 2005. Costly coasean bargaining and property right security. Environ. Resour. Econ. 31 (3), 349–367.
Coase, R.H., 1959. The federal communications commission. J. Law Econ. 2, 1–40.
Coase, R.H., 1960. The problem of social cost. J. Law Econ. 3, 1–44.
Coursey, D.L., Hoffman, E., Spitzer, M.L., 1987. Fear and loathing in the coase theorem: Experimental tests involving physical discomfort. J. Legal Stud. 16 (1),
217–248.
Croson, R., 2009. Experimental law and economics. Annu. Rev. Law Soc. Sci. 5 (1), 25–44.
Dari-Mattiacci, G., Onderstal, S., Parisi, F., 2009. Seeking rents in the shadow of coase. Public Choice 139 (1), 171–196.
Dechenaux, E., Kovenock, D., Sheremeta, R.M., 2015. A survey of experimental research on contests, all-pay auctions and tournaments. Exp. Econ. 18 (4),
609–669.
Demsetz, H., 1964. The exchange and enforcement of property rights. J. Law Econ. 7, 11–26.
Deryugina, T., Moore, F., Tol, R.S., 2021. Environmental applications of the coase theorem. Environ. Sci. Policy 120, 81–88.
Dickson, A., MacKenzie, I.A., 2022. Permit markets with political and market distortions. Environ. Resour. Econ..
Dickson, A., MacKenzie, I.A., Sekeris, P.G., 2022. The role of markets on resource conflicts. Scand. J. Econ. 124 (3), 677–708.
Farmer, A., Pecorino, P., 1999. Legal expenditure as a rent-seeking game. Public Choice 100 (3), 271–288.
Feltovich, N., 2019. Is earned bargaining power more fully exploited? J. Econ. Behav. Organ. 167, 152–180.
Fischbacher, U., 2007. Z-Tree: Zurich toolbox for ready-made economic experiments. Exp. Econ. 10 (2), 171–178.
Gneezy, U., Potters, J., 1997. An experiment on risk taking and evaluation periods⋆. Q. J. Econ. 112 (2), 631–645.
Greiner, B., 2015. Subject pool recruitment procedures: organizing experiments with ORSEE. J. Econ. Sci. Assoc. 1 (1), 114–125.
Grossman, H.I., 2001. The creation of effective property rights. Am. Econ. Rev. 91 (2), 347–352.
Güth, W., Tietz, R., 1985. Strategic Power Versus Distributive Justice: An Experimental Analysis of Ultimatum Bargaining. Professur für Volkswirtschaftslehre,
insbesondere Verhaltensforschung.
Güth, W., Tietz, R., 1986. Auctioning Ultimatum Bargaining Positions. Issues in West German Decision Research, Frankfurt, Lang, pp. 60–73.
Hanley, N., MacKenzie, I.A., 2010. The effects of rent seeking over tradable pollution permits. B.E. J. Econ. Anal. Policy 10 (1), 56.
Harrison, G.W., McKee, M., 1985. Experimental evaluation of the coase theorem. J. Law Econ. 28 (3), 653–670.
Hillman, A.L., Riley, J.G., 1989. Politically contestable rents and transfers. Econ. Polit. 1 (1), 17–39.
Hoffman, E., McCabe, K., Shachat, K., Smith, V., 1994. Preferences, property rights, and anonymity in bargaining games. Games Econom. Behav. 7 (3), 346–380.
Hoffman, E., Spitzer, M.L., 1982. The coase theorem: Some experimental tests. J. Law Econ. 25.
Hoffman, E., Spitzer, M.L., 1985. Entitlements, rights, and fairness: An experimental examination of subjects’ concepts of distributive justice. J. Legal Stud. 14:2.
Hoffman, E., Spitzer, M.L., 1986. Experimental tests of the coase theorem with large bargaining groups. J. Legal Stud. 15.
Hurwicz, L., 1995. What is the coase theorem? Jpn. World Econ. 7 (1), 49–74.
Karagözoğlu, E., Kocher, M.G., 2019. Bargaining under time pressure from deadlines. Exp. Econ. 22 (2), 419–440.
MacKenzie, I.A., 2017. Rent creation and rent seeking in environmental policy. Public Choice 171 (1), 145–166.
MacKenzie, I.A., Ohndorf, M., 2012. Cap-and-trade, taxes, and distributional conflict. J. Environ. Econ. Manage. 63 (1), 51–65.
MacKenzie, I.A., Ohndorf, M., 2013. Restricted Coasean bargaining. J. Public Econ. 97, 296–307.
MacKenzie, I.A., Ohndorf, M., 2016a. Coasean bargaining in the presence of Pigouvian taxation. J. Environ. Econ. Manage. 75, 1–11.
MacKenzie, I.A., Ohndorf, M., 2016b. Caps on coasean transfers. Oxf. Econ. Pap. 68 (2), 566–584.
McAdams, R.H., 2000. Experimental law and economics. Encyclopedia Law Econ. 1, 539–561.
Medema, S.G., 2020. The coase theorem at sixty. J. Econ. Lit. 58 (4), 1045–1128.
Oxoby, R.J., Spraggon, J., 2008. Mine and yours: Property rights in dictator games. J. Econ. Behav. Organ. 65 (3), 703–713.
Prante, T., Thacher, J.A., Berrens, R.P., 2007. Evaluating coasean bargaining experiments with meta-analysis. Econ. Bull. 3 (68), 1–7.
Rhoads, T.A., Shogren, J.F., 1999. On coasean bargaining with transaction costs. Appl. Econ. Lett. 6 (12), 779–783.
Rhoads, T.A., Shogren, J.F., 2003. Regulation through collaboration: final authority and information symmetry in environmental Coasean bargaining. J. Regul.
Econ. 24 (1), 63–89.
Robson, A., Skaperdas, S., 2008. Costly enforcement of property rights and the Coase theorem. Econom. Theory 36, 109–128.
Shogren, J.F., 1992. An experiment on Coasian bargaining over ex ante lotteries and ex post rewards. J. Econ. Behav. Organ. 17.
Shogren, J.F., 1998. Coasean bargaining with symmetric delay costs. Resour. Energy Econ. 20 (4), 309–326.
Shogren, J.F., Kask, S.B., 1992. Exploring the boundaries of the coase theorem: efficiency and rationality given imperfect contract enforcement. Econom. Lett.
39 (2), 155–161.
Skaperdas, S., Syropoulos, C., 2002. Insecure property and the efficiency of exchange. Econ. J. 112 (476), 133–146.
Tarlock, A.D., 2010. Tribal justice and property rights: The evolution of winters v. United states. Nat. Resour. J. 50 (2), 471–499.
Tullock, G., 1980. Efficient rent-seeking. In: Towards a Theory of the Rent-Seeking Society. Texas A&M University Press.
18

Initially Contestable Property Rights and Coase - Evidence From The Lab - 2023

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Initially Contestable Property Rights and Coase - Evidence From The Lab - 2023

Uploaded by

Copyright:

Available Formats

Journal of Environmental Economics and Management 120 (2023) 102842

Contents lists available at ScienceDirect

Journal of Environmental Economics and

Initially contestable property rights and Coase: Evidence from the

ARTICLE INFO ABSTRACT

used as experimental outcomes.

(a) which party holds the initial allocation of property rights

We now turn to the experimental design.

4.1. Bargaining stage

4.2. Property rights stage

design compared to the existing literature.

4.4. Experimental hypotheses

Hypothesis 1. Bargaining leads to the efficient option being chosen

(a) Regardless of whether player A or B is the property-rights holder.

Hypothesis 2. Bidding in stage 1:

(a) Stage 1 bids are the same in ContestNo and ContestSym.

Hypothesis 3. Property-rights holder’s rent:

5.1. Efficiency of bargaining outcomes

Result 1. Efficiency of bargaining outcomes

𝐸𝑓 𝑓 𝑖𝑐𝑖𝑒𝑛𝑡𝑖𝑡 = 𝛼 + 𝛽 ′ 𝑇𝑖 + 𝛿 ′ 𝐸𝑖𝑡 + 𝛾 ′ 𝑋𝑖 + 𝑢𝑖 + 𝑒𝑖𝑡 (2)

96% in ContestSym, and 93% in ContestAsym.

Fig. 1. Frequency of outcomes disaggregated by treatment.

5.2. Bidding behavior

Result 2. Bidding in stage 1

(a) There is no significant difference in bidding between ContestNo and ContestSym.

𝐵𝑖𝑑𝑗𝑡 = 𝛼 + 𝛽 ′ 𝑇𝑗 + 𝛿 ′ 𝐸𝑗𝑡 + 𝛾 ′ 𝑋𝑗 + 𝑢𝑗 + 𝑒𝑗𝑡 (3)

Fig. 2. Distribution of bids across treatments.

5.3. Distribution of gains

Result 3. Property-rights holder’s rent

factors significantly affect either measure.

Fig. 5. Effect of PR-holder legitimacy on average BPower and surplus.

Declaration of competing interest

Appendix A. Theoretical model

26 The loser, then, is able to capture (1 − 𝜇𝐵 ) of the rent.

Appendix B. Supplementary data

Supplementary material related to this article can be found online at https://doi.org/10.1016/j.jeem.2023.102842.

You might also like