Job Market Paper

Figure 1: Subject View of Experimental Environment

The Determinants of Territorial Property Rights in a Spatial Commons Experiment

This project, funded via the IFREE Small Grants program, contributes to the burgeoning institutional literature concerning the determinants of private property rights and the mechanisms that promote effective common pool resource (CPR) management. In our paper we show that the presence of costly specialization opportunities within groups can improve overall commons management, and that this improvement corresponds to the establishment of stronger private property rights within our commons environment.

The inspiration behind this project draws from two major sources. Firstly, in his seminal work "Towards a Theory of Property Rights", Harold Demsetz outlines several contrasting modes of group management of common resources. In the absence of state ownership and management of resources, groups may establish either communal or private management regimes. This project investigates whether the presence of costly specialization opportunities may influence which sort of regimes groups establish and how these regimes map onto overall CPR management. Secondly, we take heed of Elinor Ostrom's long line of work on CPR management and the Institutional Analysis and Design framework she and her coauthors have developed in order to identify common features of successful CPR management regimes. In particular we look to the work done by Marco Janssen and Elinor Ostrom in "TURFS in the Lab: Institutional Innovation in Real-Time Dynamic Spatial Commons", which urges experimental researchers to explore social dilemmas in robust spatiotemporal environments that allow these dilemmas to be framed and implemented in a more naturalistic manner.

We chose to use the virtual platform of OpenSim to implement our experiment. A screenshot of our environment can be seen in Figure 1. In our experimental environment each subject controlled an avatar that could be moved around a virtual space in real time. These experiments were run using normal recruiting methods - subjects were brought into a computer lab being monitored by an experimenter, familiarized with the software, and then interacted in our experimental task. OpenSim's built-in tools for creating online spatial environments for avatars to interact in allowed us to implement this environment with relative ease.

The basic harvesting task that our experiment implements has groups of four subjects managing berry bushes whose population grew over time in accordance to a logistic growth model. Managing the commons in this environment involved harvesting the berry bushes to an intermediate extent, but not depleting them entirely. However, since the berry bushes were unowned, the private return of harvesting berry bushes exceeded the social return and a tragedy of the commons could arise. It was up to the members of each group to implement and enforce usage rules that would prevent this outcome. To this end, subjects had access to tools such as the ability to communicate using natural language, engage in costly punishment, transfer resources, and monitor one another within the spatial environment.

Figure 2: The Separated and Mixed field layouts

Our field contains five different colors of berry bushes, spatially-arranged in one of the two layouts in Figure 2. Subjects could choose to specialize in any color other than silver. By specializing in a color they would be able to earn bonus income from collecting berries of that color. In the costly specialization treatment condition, this choice could only be made once per period. In the costless specialization treatment condition, subjects could change their specialization choice at any time, thus making it easy to ensure that the bonus could be attained for any non-silver (RGBY) berry picked.

We observed a wide variation of outcomes in our data. Many groups immediately depleted the 48 berry bushes in our environment, while some were able to establish effective usage rules to prevent overharvesting. The diversity of these rules is apparent in our chat data, as can be seen in Appendix C of the paper. Here's one clear example of a group attempting to informally privatize the commons:

AA: there's four colors
AA: let's each pick one
WH: why? can't you guy tell that way is a dumb approach to this
AA: to what? depleting all the bushes?
WM: i spy a silver! later guys!
AA: it's not a competition
AA: if we work together we all end up with more real american money that we can use in the real world
KO: i would vote for one color each
AA: me too
WH: yeah. i only made $200 when if we all picked a corner and stuck to it before. we would make like $700 (Note: 250 experimental dollars = 1 USD)
KO: ya
AA: i averaged 600 in all four rounds last time picking one color each
AA: its way way better
KO: but no one should enter others territory
WM: everyone pick a color right now and we only stick to that one color
WH: yellow for me
AA: deal
WM: i'll get red
KO: blue
AA: ok green it is
WM: leave at least 2 berries on the silvers
KO: should we have punishment for intruders?
WH: if you want to. costs you money too though
KO: yes
WM: everyone remember their color and no one go in each others territory
KO: so everyone please stick to your color
KO: stick to your colors pleae
KO: please

Figure 3: Territory Hull Definition

Our main result is that we see improved commons management as measured by the total number of berries picked within groups under the costly specialization treatment condition, indicating that this treatment condition is helping groups avoid the tragedy of the commons. To help further understand the mechanism by which this may be working, we devise a measure of territoriality for each group. This measure uses the minimal convex hull approach employed in the ecology literature: For each subject in a group, we define that subject's territory as the minimal convex hull that contains all of the berry bushes that that subject picked berries from. Figure 3 shows one example of this, where the berry bushes circled in orange represent the bushes harvested by one player and the bushes circles in purple represent the bushes harvested by another. These bushes are used to define convex hulls for each player, and then we define an uncontested territoriality measure that is equal to the share of our virtual field that belong to any player's territory that belongs to exactly one player's territory. Using this measure, we show that our costly specialization treatment condition improves not only overall commons management but also the territoriality exhibited in these harvesting patterns, providing evidence that this treatment variable is affecting the propensity of groups to establish private property rights in our commons environment.