In this page you will find the supplementary material for the paper by Paolo Crosetto and Antonio Fillipin ‘The Bomb Risk Elicitation Task’, [Journal of Risk and Uncertainty, August 2013, Volume 47, Issue 1, pp 31-65 — here]
* * *
The experiment was carried out in German at the Lab of the Max Planck Institute for Economics in Jena. The original English instructions were translated into English by our research assistants Claudia Zellmann and Florian Sturm. The Baseline Dynamic version can be used, with slight parameter modifications, for the treatments Fast, High Stake, 5×5, 20×20, Random.
* * *
It is possible to run the BRET without any special software – just with paper and pencil, or else implementing a very simple computerised interface. In its simplest, static version (as described on the paper), the BRET amounts to asking the subjects for a number in a given support (e.g., 0-100) and then drawing from a uniform distribution (e.g., a urn) another number in the same support. Then, the subjects win an amount of money proportional to the number chosen, if the drawn number (the bomb) is higher than the chosen number; he gets zero otherwise.
The dynamic, visual BRET relies on a visual representation of probabilities; in the paper-and-pencil version the probabilities are not transparently and easily communicated to subjects. It is hence important that subjects read carefully the instructions and understand the underlying random process. We devised a detailed set of instructions and a set of control questions that allow to administer the paper-and-pencil BRET ensuring comprehension.
The experiment was carried out in python, using wxpython as a GUI library. A z-Tree version was also developed to allow for a wider portability of the BRET. Both versions come with their READMEs that include system requirements, instructions to properly set up the lab, guide to the meaning of the data collected.
Possible treatments: Baseline, 5×5, 20×20, Fast, High Stakes, Random [and all combinations of the above]
Script to collect data: Included in the above BRET_Python.zip
The Python version requires the installation of the free and open source Python interpreter and of the wxPython graphical user interface library on all the machines on which you wish to run the BRET. Moreover, it needs a shared network folder between clients and server, with read ad write access for both. Apart from that, the software does not rely on a server-client architecture and it is at all effect a local client application – the shared folder is used only to store data on a unique location and to notify the experimenter of progress of each single client. The drawback of this architecture is that data are saved on client-specific text files and a script needs to be run in order to create a .csv data file for the session. The advantage, though, is the full flexibility of python and the full scalability of the experiment: with the python BRET an unlimited number of clients can be used at one time with no fear of network congestion or breakdown. Moeover, a very fast or very computation-intensive version of the BRET can also seamlessly be run on an unlimited number of machines.
* * *
Requirements: zTree (by Urs Fischbacher)
Possible treatments: Baseline, 5×5, 20×20, High Stakes, Random, Explosion [and all combinations thereof]
The zTree version requires a working zTree – zLeaf installation on the Lab. The version is limited as it allows the experimenter to carry out only a subset of the possible variations on the BRET allowed by the Python version. Furthermore, the fast time intervals of client updates puts the network structure of zTree under stress. Our tests show that it is safe to run the baseline dynamic version of the BRET with up to ~20 clients. With more clients connected, the network could be affected by delays or could hang; it is then recommended not to run fast versions of the BRET, as this saturates the network with a lower number of clients. For the same reasons, do NOT run a FAST treatment on zTree.The advantage of the zTree version is that data is automatically collected on the server in a .xls file, at the end of the session.
* * *
oTree version [Felix Holzmeister and Armin Pfurtscheller]
Application sources: on Felix’s website here