Saturday, July 10, 2010

Running Economic Experiments Online for Teaching Principles and Intermediate

When I was in graduate school, probably during my second year though maybe it was toward the end of the first year, we spent an afternoon of fun doing tw0-sided auction experiments to do a live simulation addressing the question whether a market comes to equilibrium. I'm not sure the experience had a profound effect on me. Occasionally when I taught intermediate micro I would take part of a class session to do such experiments in an abbreviated format. Students want to know that a market will come to equilibrium. The session was aimed at providing the students with the requisite intuition.

Now many years later I've been thinking about teaching intermediate online and what might be done to engage the students. In the back of my head when I pose a question of this sort to myself, I have a technology in mind to facilitate the teaching. In this case It was the forms tool in Google Spreadsheets. Students could submit their information via a form and the spreadsheet could tabulate all submissions so the class as a whole could monitor that. (Also, the url can otherwise be kept private so the rest of the world wouldn't monitor the activity.)

For about a month or so I've been keeping this thought on the back burner, envisioning that the experiment would be done in rounds delivered asynchronously, perhaps one round a day. But I wasn't satisfied with that for a variety of reasons - students might find it insufficiently compelling so wouldn't be drawn in, the rules were that the first one who accepted an offer had the rights to it but that rule was inconsistent with asynchronous participation, and it just seemed clunky. Because I wasn't satisfied I didn't act on it.

Then a few days ago I asked - what if the auction were held in a live synchronous session, either face to face or online, but where all the the student participants submitted their offers via the forms. All of a sudden this started to make sense to me. I envisioned that if I could get this to work in the simple two-sided auction that I had participated in while a grad student then I could modify the economic environment and do many other experiments as a way for the students to "experience" the economics they'd be studying. So I started to play with this by building forms and seeing what would happen as I submitted various responses with the forms.

Envisioning this as totally online, I thought we'd do a live chat session during the auction. The spreadsheet would record the offers, but there were issues. What if a students made a mistake in submission and sent something in that was nonsense. That would have to be voided. Could it be done automatically or was there a need to do it manually. Likewise a student might want to withdraw an offer, seeing it disadvantageous relative to other recent transactions. How would that work? And when an offer was accepted that would have to be recorded and then then offer would need to be taken off the list of currently active offers. Could all this be automated too? I reasoned that if I'm the "auctioneer" then either I'd make mistakes with manual recording or it would be a time consuming process. So I wanted to automate as much of this as I possibly could.

Alas, I made a discovery yesterday with the Google forms. It is simply not possible to keep running totals of results. Each new submission adds a row of data to the spreadsheet. The newest data comes into row 2. All previous data gets pushed down a row. If you have formulas in other columns than where the data is being collected, those formulas get pushed down two. Not being able to keep running totals, it is not possible to automate the entire process. But I've come up with a reasonably Spartan schema for the manual recording.

You can check out the form and the spreadsheet. Make a submission or two or three to see how it looks. The key to the manual recording is that if a letter (A for Accept, W for Withdraw, V for Void) in front of a price, then from the spreadsheets perspective the entry has been converted to alpha format and so gets ignored in the calculation of maximum bid or minimum ask. The rest of the recording should be just a copy and paste. Still, I don't yet know if this really can work work for the students. They will need three tabs open - one for the spreadsheet, one for the form, and one for the chat. Perhaps each can be put into a frame within one page, and in the not too distant future I'll experiment with that a bit. But there will be a lot of info on the screen then and I'm not sure that will work. So that needs to be tested.

Also, I need to test how long it takes to run one round of such an experiment and how many students can be accommodated during it. Ideally you can do 4 or 5 rounds in an hour with a few minutes left to spare for reflection at the end. But that needs to be tested too.

I should also add that earlier I tested using mail merge in Outlook as a way of giving the students in advance the data they need to have to start the auction going. (Buyers know the value of buying one unit of the good. Likewise sellers know the cost of providing one unit of the good. Both the values and the costs are idiosyncratic. The students don't know their distribution in advance. At the end, the distribution can be shown and if the transaction price settles down it can be compared to the theoretical equilibrium price.) From a sending point of view, that works like a charm. There is quite a different issue whether the students read their email frequently enough to use this as a good way of distributing the information for participating in the auction. That too needs to be tested.

Assuming all those tests get passed, perhaps following substantial modification of how to teach the course, it then becomes incumbent to integrate these experiments with the rest of the class and to figure out how many of these experiments would be fruitful to use over the duration of the semester. In my head I've been able to come up with quite a few different scenarios to e run as experiments. Whether once the novelty has worn off students continue to enjoy participating in them and taking them seriously, I don't know. But I would like to try this.


Greg Eisenmann said...

This is a very interesting idea. I can't tell which came first: the idea itself, or the thought to use google spreadsheets. Either way, I wonder if that would be the best tool to use, especially for a realtime "game" like that. I don't know your webpage design experience, but it seems like you could tune such a page to your specific needs. It seems like it wouldn't be that difficult to setup a page with all of the boxes you had on your survey, then have all of the data sent to a spreadsheet, which, incidentally, wouldn't be bound by the whims of google. (Not that I know anything about designing a webpage, but perhaps you do.)

I don't want to negate any of your ideas; this post merely piqued my curiousity, so I thought I'd maybe try to give some feedback. I guess my biggest question is "are you sticking with google spreadsheets because you really think it will work, or because you really haven't looked at other options?"

If this question is annoying, feel free to ignore it! I'll be interested to see how the idea progresses and develops.


Lanny Arvan said...

Greg - thanks for the comments. I showed an earlier version of this to a colleague who is also an economist. He suggested I get a database programmer. Perhaps eventually it will come to that. But right now I kind of like Google Spreadsheets because it addresses many of the development issues and then I don't have to worry about back end issues at all. So to answer your question, yes, I've consider other alternatives. For now, I think this is the best one. Will have to let use determine whether it is a good choice.