Friday, February 10, 2012

Experimental Research with FTS

Experimental research is a big part of our history.  In fact, the first version of the FTS Interactive Markets was developed in 1989 and 1990 for conducting experiments on information aggregation; the resulting paper came out in the Journal of Finance in 1991.  I even have a picture of the old text-based interface, on an original PS2 computer:
 
Since then, the FTS Interactive Markets have been used in many other experiments by different researchers.  Even today, we are helping several people with their new experiments; this includes setting up the trading parameters as well as software modifications to accommodate their needs. 
 
Beyond market experiments, though, we also have a “generic” experimental platform for conducting all sorts of experiments easily; these include behavioral experiments, auctions, and so on.   Examples of how to easily design eperiments are described in this document.

You can also use it to add non-market dimensions to a market; one example is adding a cheap-talk phase before a trading phase in a market, as in http://server1.tepper.cmu.edu/Seminars/docs/JobMarketPaper_0107_HongQu_CMU.pdf 
 
The experimental system is fairly straightforward. I’ll use an auction as an example.  You set up the experiment in an Excel workbook. One worksheet contains common information, such as instructions, sent to all subjects. Another worksheet contains information to be sent to individually, such as a private value or signal for the object being sold in an auction. This area also specifies what input is required from the subjects, such as a bid for the object being auctioned.
 
• You run the experimental server on your computer and connect it to your workbook.
• The subjects all connect to your computer using the client program.
• You send the common information
• You send the private information
• The subjects enter their responses (e.g. bids) during some period of time that you choose.
• When the time is up, you “grab” all their responses, calculate the outcomes in your
workbook (e.g. the price paid and who is allocated the object) and send this information
back to everyone.
 
The information flow described above can also be made continuous; in fact, you can mix and match
information that is updated constantly and that is updated manually. For example, if you are running a sealed bid auction, you only need to retrieve bids at the end of each round. If you are running an English auction, you may want to display the highest bid (or all bids) continuously.
 
The experimenter’s screen looks like this:
In the spreadsheet, you can set the colors of each cell, the text alignment, and you can see all the features in the screen shot.  We have a standard template that you can modify to create a variety of different types of experiments quickly and without a lot of effort.  If you have questions, please contact us, we love to talk about experimental research.