Data Set Description for c1_data.txt:
"On the Origin of Convention: Evidence from
Symmetric Bargaining Games"
International Journal of Game Theory, 24(2), 1995,
187-212.
Draft Date: December 29, 1996
I) Input format for the primary data set
A) Data set name: c1_data.txt
B) Data Format:
Variable Column Definition
SESSION 1-6 Session number - SESS01-SESS14
DESIGN 8-15 xxxxyyzz
xxxx = 1 or 2 populations (1POP) or (2POP)
yy = with labels (WL) or without (NL)
zz = with security (WS) or without (NS)
DATE 17-25 xxxxxx_yz
xxxxxx = Date of session
y = 1 or 2 denotes first or second session run
on that date
z = A or B denotes sessions run at the same
time different rooms
PERIOD 27-28 values 1-45 for sessions 1-11, 1-70 for sessions
12-13 and 1-50 for session 14
SUBJECT 30-34 xxx_y
xxx = the subject's label. For the 2 population
designs, SESS01-SESS02 and SESS07-SESS14
the label is either ROW or COL. For the 1
population design this value is assigned as
SUB.
y = Subject number within a population.
ROW_1-ROW_7 and COL_1-COL_7 in 2
population design and SUB_1-SUB_8 in 1
population design.
SUBNUM 36-37 Unique subject ID for each session, numbered 1-14
in 2 population designs and 1-8 in 1 population designs
CHOICE 39 This value is identical to OWNCH listed next. This is
the variable that is used in ALL analysis that are based
only on row and column totals.
OWNCH 41 Subject's own choice taken from each subject's record
sheet. The record sheets were filled in by the subject,
period-by-period, during the experiment. This value is
identical to CHOICE listed above. This variable is
included to make it easy to look at individual cell
values RiCi and to analyze individual subject data.
OTHCH 43 Subject's record sheet entry for the subject that they
were paired with that period. This data is taken from
each subject's record sheet. This variable is included
to make it easy to look at individual cell values RiCi
and to analyze individual subject data.
C) SAS Input Program
For users familiar with SAS, the following program lines will read in this data file:
DATA INNALL; INFILE '_path_\CONV1AGG.DT2'; INPUT (SESSION DESIGN DATE PERIOD SUBJECT SUBNUM CHOICE OWNCH OTHCH) ($ 1-6 $ 8-15 $ 17-25 27-28 $ 30-34 36-37 39 41 43 ); RUN;
II) Sample of data for 2POP and 1POP sessions taken from c1_data.txt
A) Session SESS01, Period 1, Design 2POPWLNS (14 subjects; 7 row and 7 column; 2 populations)
123456789012345678901234567890123456789012345 SESS01 2POPWLNS 061890_1A 1 ROW_1 1 2 2 2 SESS01 2POPWLNS 061890_1A 1 ROW_2 2 3 3 3 SESS01 2POPWLNS 061890_1A 1 ROW_3 3 2 2 2 SESS01 2POPWLNS 061890_1A 1 ROW_4 4 2 2 1 SESS01 2POPWLNS 061890_1A 1 ROW_5 5 2 2 2 SESS01 2POPWLNS 061890_1A 1 ROW_6 6 2 2 2 SESS01 2POPWLNS 061890_1A 1 ROW_7 7 2 2 2 SESS01 2POPWLNS 061890_1A 1 COL_1 8 2 2 2 SESS01 2POPWLNS 061890_1A 1 COL_2 9 3 3 3 SESS01 2POPWLNS 061890_1A 1 COL_3 10 2 2 2 SESS01 2POPWLNS 061890_1A 1 COL_4 11 1 1 2 SESS01 2POPWLNS 061890_1A 1 COL_5 12 2 2 2 SESS01 2POPWLNS 061890_1A 1 COL_6 13 2 2 2 SESS01 2POPWLNS 061890_1A 1 COL_7 14 2 2 2
B) Session SESS03, Period 45, Design 1POPNLWS (8 subjects; 1 population)
123456789012345678901234567890123456789012345 SESS03 1POPNLWS 041990_1A 45 SUB_1 1 1 1 1 SESS03 1POPNLWS 041990_1A 45 SUB_2 2 1 1 1 SESS03 1POPNLWS 041990_1A 45 SUB_3 3 1 1 1 SESS03 1POPNLWS 041990_1A 45 SUB_4 4 1 1 1 SESS03 1POPNLWS 041990_1A 45 SUB_5 5 3 3 1 SESS03 1POPNLWS 041990_1A 45 SUB_6 6 1 1 3 SESS03 1POPNLWS 041990_1A 45 SUB_7 7 1 1 1 SESS03 1POPNLWS 041990_1A 45 SUB_8 8 1 1 1
III) Recording Errors
A) General discussion
Sessions SESS01-SESS11 were conducted by hand. Each period, each subject entered their choice on their record sheet and circled their choice on a reporting sheet that was collected by the experimenters each period. The experimenters then matched the subjects according to a predetermined random matching protocol and filled out a result sheet that was given to the appropriate paired subject for that period. The paired subject then entered the value from the result sheet for that period on their record sheet and calculated their period earnings and balance. (Sample record sheets, reporting sheets and result sheets are included with the instructions attached to this data appendix.)
Since each subject has a value for their own choice and their pair's choice each period, we have two 'complete' records for each subject pair. If these records are to match: (1) each subject must enter the same value on their record sheet and their reporting sheet, (2) the experimenter must correctly match the reporting sheets, (3) the experimenter must correctly transfer the choice from the reporting sheet to the result sheet, (4) the result sheet must be returned to the correct subject and (5) each subject must correctly transfer the results from the result sheet to their record sheet.
B) Aggregate data considerations
An error at step (1), (3) or (5) by either subject or by the experimenter will, in general, result in the period row and column totals calculated from the row record sheets not being the same as the period row and column totals calculated from the column record sheets. To examine the extent and potential impact of these errors we independently entered both the row and column record sheet data. (The 1POP data was independently entered twice in its entirety.) The two data sets were then compared to see if the period row and column totals differed between the row records and the column records. After all differences were checked for entry errors 4 differences existed.
In two cases subjects entered different values on their reporting sheets than they entered on their record sheets, (1) above.
(i) SESS01, period 2, column subject # 2 entered a value of 3 on the record sheet and reported a value of 1.
(ii) SESS11, period 45, row subject # 4 entered a value of 1 on the record sheet and reported a value of 3.
In two cases subjects entered different values on their record sheets than were reported to them on their reporting sheets, (5) above.
(i) SESS07, period 3, row subject #6 received a value of 2 on the result sheet and entered a value of 3 on the record sheet.
(ii) SESS07, period 3, row subject #7 received a value of 1 on the result sheet and entered a value 3 on the record sheet.
C) Individual cell and subject considerations:
Errors caused by (2) and (4) above do not affect the row and column totals. Error (4) never occurred in our experiments. Error (3), where the experimenters incorrectly implemented the matching protocol did occur. In seven periods some error was made in the matching protocol. (SESS01, periods 2 and 3; SESS07, period 1; SESS09, periods 36 and 37;SESS10, periods 24 and 25.) If the matching error resulted in pairing the wrong subjects it is judged that these effects can be safely ignored.
It is also possible that the experimenter errors resulted in subjects receiving values that precluded the existence of pairs. That is, row subject 1's choice could be sent to subject 3 while row subject 1 received a column subject choice on the result sheet from other than column subject 3. If subject choices are heterogeneous, this type of mismatching can result in errors that will affect the cell values RiCi and result in different cell totals for the row subject record sheet data and the column subject record sheet data.
This happened in three periods.
(i) SESS01, period 3, 2 row and 2 column subjects were mismatched
(ii) SESS07, period 1, 4 row and 4 column subjects were mismatched
(iii) SESS10, period 24, 2 row and 2 column subjects were mismatched
It is important to note that the information conditions of the experiment were such that no subject could have been aware of these errors except in the expected sense that they might expect subjects and/or experimenters to make a few mistakes.
D) Analysis data sets
Although the above discrepancies are few and have made no impact on any of our analysis to date, a decision must be made as to which values to use in our analysis to insure the exact replicability of results across different interested researchers.
We have decided to:
(1) use each subject's own record of their own choice for all aggregate analysis where subject pair information is not needed. This is the variable CHOICE on the enclosed data file.
(2) use the row subject data files for all analysis that are based on an analysis of the detailed cell, RiCi, data. This analysis uses the variables OWNCH and OTHCH for just the ROW subjects.
(3) base all individual subject analysis on each subject's own record sheet data. This analysis uses the variables OWNCH and OTHCH for each ROW and COL subject.
III) Design
SESSION DESIGN DATE METHOD PERIODS
SESS01 2POPWLNS 061890_1A By Hand 1 to 45
SESS02 2POPWLNS 061890_1B By Hand 1 to 45
SESS03 1POPNLWS 041990_1A By Hand 1 to 45
SESS04 1POPNLWS 041990_1B By Hand 1 to 45
SESS05 1POPNLWS 061890_1A By Hand 1 to 45
SESS06 1POPNLWS 061890_1B By Hand 1 to 45
SESS07 2POPWLWS 041290_1A By Hand 1 to 45
SESS08 2POPWLWS 061290_1A By Hand 1 to 45
SESS09 2POPWLWS 061290_2A By Hand 1 to 45
SESS10 2POPWLWS 061390_1A By Hand 1 to 45
SESS11 2POPWLWS 061390_2A By Hand 1 to 45
SESS12 2POPWLWS 111590_1A By Computer 1 to 70
SESS13 2POPWLWS 111590_2A By Computer 1 to 70
SESS14 2POPWLWS 111690_1A By Computer 1 to 50