Extracting Valuable Data from Classroom Trading Pits

1
Extracting Valuable Data from Classroom Trading
Pits
Ted Bergstrom Eugene Kwok University of
California, Santa Barbara
2
The Origin of Experimental Economics

• The first scientific experiments in economics
  were classroom market experiments
• by Edward Chamberlin at Harvard in 1940s.

3
Chamberlins experiments

• Assigned Buyer Values and Seller Costs.
• Let students mill around and trade.
• Recorded prices.
• Remarked on difference from competitive
  equilibrium outcome.
• Observed excess trading.

4
Revival at Purdue

• Chamberlins experiments went almost unnoticed
  until
• Vernon Smith revisited them in his classroom at
  Purdue.

5
Smiths experiments

• Gave competition a better chance.
• Two main differences from Chamberlin.
• Double oral auction, not pit trading
• Ran 3-5 rounds, repeating same setup
• Found outcomes very close to competitive
  equilibrium

6
Founders of Experimental Economics

• Edward Chamberlin

• Vernon Smith

7
Our Data

• Classroom experiments from Experiments with
  Economic Principles, a principles text by
  Bergstrom and Miller
• Experiments conducted in 31 classrooms, 10
  universities.

8
The Apple Market

• Students assigned roles as apple suppliers or
  apple demanders.
• Suppliers supply at most 1 bushel.
• Demanders demand at most 1 bushel.

9
Buyer Values and Seller Costs

• Two types of demanders
• High ValueBuyer Value is 40
• Low ValueBuyer Value is 20
• Two types of suppliers
• High CostSeller Cost is 30
• Low CostSeller Cost is 10

10
Session 1

• 2/3 of Sellers have low cost, 1/3 high.
• 2/3 of Demanders have low value, 1/3 high.

11
Demand and Supply in Session 1
12
Session 2

• 2/3 of Sellers have high cost, 1/3 low.
• 2/3 of Demanders have high value, 1/3 low.

13
Demand and Supply in Session 2
14

• Session 1 Distribution of Average Prices

15

• Session 2 Distribution of Average Prices

16

• Session 1 Distribution of Quantity Deviations

17

• Session 2 Distribution of Quantity Deviations

18
Enough to convince crudulous students, maybe

• But does the evidence show that competitive
  theory is empirically useful?

19
An alternative hypothesis Profit Splitting

• Demanders meet suppliers chosen at random.
• If mutually profitable trade is available they
  trade, splitting the profits.
• Demander with value 40 and supplier with cost
  30 trade at 35, etc.
• There is trading at 15, 25, and 40.
• If high cost seller meets low value demander, no
  trade.
• .

20
Average Prices are predicted better by
Profit-Splitting

Session 1 Session 2
Competitive 20 30
Profit-Split 20.7 29.3
Actual, Rd 1 21.2 27.0
Actual Rd 2 21.2 28.5
21
Detailed predictions

• Competitive theory and profit splitting theory
  both make detailed predictions beyond average
  price and total quantity.
• Distribution of prices
• Competition implies uniform price.
• Splitting implies trading at 15, 25, and 40.
• Both theories predict who trades with whom as
  well as total number of trades.

22
Demand and Supply in Session 1
23
Session 1 Detailed Price Predictions
Competitive vs Profit-splitting
Price Range 14-16 24-26 34-36 19-21
Competitive 0 0 0 100
Profit-splitting 57 29 14 0
Actual shares, Rd 1 24 18 6 20
Actual shares, Rd 2 16 19 2 30
24

• Session 1 Distribution of All Prices

25
Demand and Supply in Session 2
26
Session 2 Detailed Price Predictions
Competitive vs Profit-splitting
Price Range 14-16 24-26 34-36 29-31
Competitive 0 0 0 100
Profit-splitting 14 29 57 0
Actual shares, Rd 1 7 20 8 32
Actual shares, Rd 2 2 24 8 42
27

• Session 2 Distribution of All Prices

28
Session 1 Detailed Quantity Predictions
Competitive vs Profit-Splitting
Buyer Value Seller Cost LowLow Low High High Low High High Total Trades
Competitive Prediction 197 0 241 0 438
Profit-Splitting Prediction 290 0 145 73 508
Actual, Round 1 221 9 207 34 471
Actual Round 2 218 0 209 38 465
29
Session 2 Detailed Quantity Predictions
Competitive vs Profit-Splitting
Buyer Value Seller Cost LowLow Low High High Low High High Total Trades
Competitive Prediction 0 0 241 201 442
Profit-Splitting Prediction 74 0 148 296 518
Actual, Round 1 26 6 218 211 461
Actual Round 2 18 2 218 213 451
30
Remarks

• Sometimes trading environment is like Smiths,
  much repetition with same environments and public
  trading.
• Sometimes more like Chamberlins or like ours.
• Seems worth understanding what happens in
  environments with intermediate levels of
  information.

31
Mining Classroom Trading Pits

• Data is cheap and abundant.
• Design is less flexible.
• But worth saving and studying.
• Remember where experimental economics started.

32
Thats all for now

• Mine tailings