Documentation for experiment,
data and statistical work to accompany “Traders’ Expectations in Asset Markets:
Experimental Evidence.”
Experimental Software.
The program for this experiment is included here in the file “call_market.ztt.
To run this program, one would need to obtain the zTree software, which can be obtained from Urs Fischbacher (http://www.iew.unizh.ch/home/fischbacher/).
Experimental Instructions:
instructions.doc
Statistical Computations
(Tables 2-7)
The bubble measure computations in Tables 2 and 3 in are shown in Tables2-3.xls.
The regressions (Tables 4-7) were run in SAS.
All regressions were replicated (for
verification purposes-- you can never be too careful) in Matlab
The SAS and Matlab programs for these regressions are included (SAS programs above). Each SAS program is named
after the table it corresponds to. Detailed documentation for the SAS programs
is included as comments within each program.
All SAS programs use the same data set TraderFull.xls, which is
documented in detail below. The Matlab regressions are all included in
MatlabCode.m and use the data file beliefs.prn (space delimited text) which was
derived from TraderFull.xls. The datafile beliefs.prn is also included in Excel
format as beliefs.xls.
Tables 4a-4b
Table 5
Table 6
Tables 7a 7b
Data File: TraderFull.xls
This file provides the entire future belief profile for each subject in each period.
Number of rows in the data. There were six sessions. All sessions except one had four markets and all sessions except one had 9 subjects. One session (session 1) had three markets and one session (session 4) had 8 subjects. Each market involved 15 rounds. Hence, the number of rows in this file = 4 sessions x 4 markets x 9 subjects x 15 rounds + 1 session x 3 markets x 9 subjects x 15 rounds + 1 session x 4 markets x 8 subjects x 15 rounds = 3,045 rows of data.
Number of forecasts specified. In each round t, a subject makes 16-t predictions. In total, a subject makes 120 predictions in a market (15 rounds). Hence, the file includes 24,360 forecasts.
Columns: Session, Period, round, Market, Subject, InitialMoney, InitialStock, numBuy, numSell, Bid, Ask, NewStock, Stock, EqmPrice, FV, Dividend, Profit, TotalProfit, b1, b2, …, b14, b15
Session – Indicates the session id (1-6). Session 1 contained only three markets 1-3. All other sessions contained four markets, numbered 1-4. Session 4 contains 8 subjects (1-8). All other sessions contained 9 subjects (1-9).
Period—Each subjects (except in session 1) played 60 periods, consisting of 4 markets * 15 rounds per market. The column ‘period’ list the periods from 1-60.
Round – This column list the rounds from 1-15. Remember that a round can be in one of four markets, so the 7th round in market 2 corresponds to period 22.
Market – Markets are numbered from 1-4. Each market consists of a 15-period-lived asset that yields dividends for 15 rounds and then expires.
Subject - Subjects are numbered 1-9 (except session 1, where they are numbered 1-8).
InitialMoney = This is the initial cash endowment of a subject. 112 for subjects 1,2,3; 292 for subjects 4,5,6; 472 for subjects 7,8,9
InitialStock = This is the initial stock shares endowment of a subject. 3 for subjects 1,2,3; 2 for subjects 4,5,6; 1 for subjects 7,8,9
numBuy/numSell/Bid/Ask – Each subject enters an order which consists of four numbers as shown below: The number of units he wishes to buy (numBuy) and the maximum price to buy it at (bid). The number of units he wishes to sell (numSell) and the lowest price at which to sell them (ask). Note that a subject could enter (but not execute) both buy and sell orders simultaneously.
NewStock – This is the number of additional shares acquired by a subject in the period. This number could be negative, indicating a sale. Note that summing up the NewStock over subjects in a round (in the same market and session) would always give 0, since every share bought by a subject must be sold by another subject.
Stock – This is the number of total shares owned by a subject in a period. This number summed over subject adds up to 18, except for session 4 (with 8 subjects only), where it sums up to 17.
EqmPrice—This is
the price that clears the market (equates demand and supply). All bidders who
indicated a bid >=EqmPrice will have a value for newstock that is >=0.
Likewise, for all bidders where ask<=EqmPrice,
newstock <=0.
FV = (16 – round)*12. This is the fundamental value of the
asset computed as expected future dividend flow.
Dividend – The actual dividend realization for the round. A
dividend takes the value of 0, 4, 14, and 30 with equal probability.
Beliefs: b1, b2, …, b14, b15 – Each subject submits forecast for each future round. Note that in round t, the subject is predicting for round t as well as future rounds. This is because the prediction task happens prior to trading in each round. The vector b1-b15 indicates a subject’s forecast for periods 1-15. Note that in a given round, a subject cannot submit forecasts for past rounds. So, for example, in round 9, a subject would submit beliefs for rounds 9-15. The values in round 9 for b1-b8 will be 0, indicating no forecasts were made for those rounds.