I

1.

& EXP&UMENTti ANALYSIS OF ULTIMATUM BARGAINING

W~FIXStit?Kk, Rqlf SCXMI’M’BERCrERand Bernd SCHWARZE*

V&w&&, 2 K&z. fooo Xdln 41, West Germany

Raivcd July 1982, final I e&on received January 1983

Them are many experimental studies of bargaining behavior, but surprisingly enough
nearly no attempt hits been made to inrcstigate the so-called ultimatum bargaining
behavior expMner&Sy. The special property of ultimatum bargaining games is that
on every stage of’ the’ bargaining process only one player has to decide anti that
b&i the IaH stage the set of outcomes is already restricted to only two resuhs. To
make the ultimatum aspect obvious we concentrated on situations with two players
a& two stages, In the ‘easy games’ a given amount c has to be distributed among the
two players, whereas in the ‘complicated games’ the players have to allocate a bundle
of black and white chips witb di&rent values for both players. We performed two
main tqerimentir for easy games as well a3 for complicated games. By a speci22
experiment it was investigated how the demands of subjects as player t are related to
theit aecqtance decisions as pbdyer 2.

1. IqFoiill&M
.A @me in strategic or extensive form, which is played to solve a
distri%utiu~problem, is called a bargaining game, Such a game has perfect
infonmatisa ii all & inr’ormation sets are singletons, i.e., there are no
kr~ultaneous d&sions and every player is always completely h:formed about
alI the I previous i-;lcx;isiorw.Consider a batgaining game with ptrfkct
inlormation whose plays a= al tite. Such s game is called an ultimatum
bar&ning ,ganrei&the last decision of every play is to choose between two
pred@crmined-resuk Oftea a game itself does not satisfy this definition, but
%?nt$ins~srib~ for whioh this is true
. I@ G!qt3mmmlhq#nh&-,ow LusuaUy speak3 of an ultimatum if one *rty
can 2e&rkt the-sat of .$IOssibic.qreemcnts to one single prqosal whkh the
& pqt~Q GUI-.&t&r843x3@,or re@cLSince in an ultimatum bargaining
m I@ s& of ~~ble~.outt;onxes is narrowed down to only two results
b&JNFti l&Stck$lllim is :madk,this %CFk&k3 our terminolqy-
, .
.ii s_ . ,Ti”3 i‘ ,, _ ,:,,,:. I I 1
:w ~~-~ l&.. qo ti&tk~ Rd&rcj Seltpin(univ~&ty of BiehMd) !md two
~raErrrsl#‘f&th&%aMAe ad*.
 t. ’ ’

tion. Thus it is
amcr~rnt c which was to be distributed among the two subjects. All
mes with c43nlpkte infoma~io !I.
‘IIke nmkr kof games ranged from 9 to 12. So the chtinces to meet a
EISplayer 2 were rather low for ali players 1
the same room at desks which were far enou
e verbal comrmmication. Furthermore., players
oppotdte sides of the room. Each pnrscicipant could se? all :he
d had a complete: control thaf the expeiment was performed
to the instruction rules in the appendix. We did not observe
attempts t,o exchange messages during the experiments. Be;wzetl experiments
cornmunicrxtion was not restricted.

2.2. lib3ygmnes
fn an ea:Gygamt the two subjects were first determined to be player 1 and
player 2. The subject chosen to be player I then dec!arss \&ich amount a,
he claims for himself The difference between the amount c c >Q), which can
be distributed, and al. is what player 1 wants to leave for player 2. Given the
decision of player 1 player 2 has to decide whether he accepts player l’s
proposal or not. If 2 accepts, player 1 gets a, and player 2 gets c -aI.
Otherwise both players get zero.
Every subject in the subgroup of pIayers I got a form (table 1) which
informed him about the total amount c to be distributed. Player 1 ha
write down the amount of money a, which he demands for himself. Then the
forms were collected and distributed by I:hance to the subjects in the other
subgroup. Player 2 had to indicate whether he accepts the pro,,losal of player
1 or not. Two tickets were attached to each form, one for player 1 and one
for player 2. On e,zch ticket there was a capital letter, indicating the game,
and the player na~~nber.So, for instance. Xi is on the ticket of the subject
who is player 1 in game X. We called XI the sign of this subject. The
subjects had to show their tickets :to get their payo&.
Table 1
The form given to wbjects engaged in easy games.
__
The amount c to bc distributed is c= DM . . .
Player 1 can dcmnnd *everyamov nt up ts c= DM . . .
--_--------__-----___-__
sign ofplayer l:...l
Decision of player 1: I &max~d DIM . . .
-------.--_-------_---
Sign of player 2:. . .2
I mccpt player l’s demmd:...
I xtiwc pieycr l’s dunand: . , .
~indicatc the dccisim you prcfw by ao ‘X?)
m-p
2.3. Ckm&caed games
The experimenls of complicate mes Were j&ormed in a similar way. In
a compl&ated game player .I first to divide a bundle of 5 ‘&.~k and 9
white chip. In order to do this player 1 determines a vector (m,, m2)
indicating the decition for one bundle (1) with. m, (2; 5) blat?: and mt ( 2<9)
white eltip and t!& complen~entary bundle (II) with (5 -in,j black and
(9 -RQ white chips. After the decision of pfayer 5. player 2 has lo decide
whether he wants to have bundIe (I) or bundle (II). The other bundle is given
to phxyer 1. flayer f got IMA 2 for each chip. Piayer 2 was paid DM 2 for a
Ma& chip az<ldDM t for a white chip. Both players were informed abolut
these VaLUs.
The fo1.mgiven to ‘Se sufrJects engaged in a comp,icated game is shown in
table 2. .&&n several exarnp~es were cafctiated to make ore that every
subject completity understood the rules of the game. Sotllg subjects had
diEcul& to Ie~n how the distribution of chips detttrmincs the money
Y&
iln the ~xxn@cated game the. rational d&tion behavior is not so obtious.
A ratio& pLayer 2 wal alwz~~ choose the buindle which YeIds a higher
payoff for himi. For player 1 it is evident thai he has to deign bundles I and
If such W. rihe bundle, which @ayet 2 will prefer, conta& as few wh;te
Tilb
TheBofxtlgiWItos~~ iJl compibtcd games..
-I_-s--__ e_
!sigaQfplarcr I:...1
toaxaelxtwr~
3 W
9 white chip), or
cm tk tlxnaiegdrip
_-_.-_-._-___-c-__-__-_
‘sign arplayef fk:...z?
IDccisicmof~yu2:
!i~~sC~~~~af~&WsCaa~~...
!I cbotJ8e the tm nyQiipb@... -
bym’mj ,.
.-.-I-
 373

chips as possible. Knowing this some easy calcuJatlons show that the op;.imal
dezision of pJayer 1 is given by im,, n1&==(5,0) or (0,9). This will induce
pliayer 2 to choose I in the first case and tJ in the second case. The
equihbrium payoff for pJayer 1 h DM 18, whereas pla!rer 2 receives Dh4 JO.
ff player 2 would deviate, he WCrid get DM 9 whereas player 1 s payoff
woulld be J>M 10, i,e., a deviation of player 2 would cost player 1 much more
than player % himself.
The complicated game is a we&known distributi(jn procedure [see, for
instance, Kuhn (1978), Steinhaus (1948) Giith (1979)], often zalle,j ‘the
method of divide and choose’. In the economic literature it is most’ry applied
to the problem of cutting cakes fairly. En cur example there are two difTelent
‘cakes’ and two individuals with dii’ferent preferences.
The method of divide and choose yields an euvyfree allocaltion [Pawner and
ScJnnei.dler (1974)] which is even Pareto-optimal in our special case. In
genera& this method determines ;rn allocation which is not Parelro-efKcient
[Giith (1976)]. Observe that a complicated game has other envyfree: and
Pareto-optimal allocations beside the t ?tt.ilibrium ailocation. If pJaJ{er 2
receives the bundle (5, I) of 5 black and 1 &Me chips and ptayer 1 gets the
residual bundfe (0. 8), this allocation is illso envyfree and Pareto-ef,%-lettr.
The same is true if player 2 receives thz bundle (5,2) and player I the
rcsiduai bundle (0,7). All other Pareto.-eff cient allocations arp rot envyfree.
Furthermore, the equilibr’um payoff of playes 1 is his maximal pa.yol‘f an the
set of envyfree allocations. This demonstrates that the method of I&&e and
choose shows player I to exploit the preferences of player 2. Playrr 2 would
prefer to be the one who determines two bundles J and III betwLten which
player 1 has tc choose.

3- Experimental resdts
The subjects were graduate students of economics (University of Cologne)
attending a seminar to get credit for the final exams. J: is alLmost sure that
none of the students .cras familiar with game theor;,. Aft~.xrpilot studies iu the
summer semester of 1978 the main experiments wt;i’” fszrformed at the
beginning of the next. winter semester.

For the -sake of completeness we also show the results of the pilot
experiment with easy games in table 3. The results of one game, qxcified by
a capital letter in column (i), appear in one line. The s<cond co’!umn of table
3 gives the amount c to be distributed. Tht: third one the demanl;l of player
1. A ‘1’ h the. fourth column indicates tSlal player 2 accepted, w’lkerea(sa %’
says that 2 reXusedplayer l’s proposal, Conflict resulted in three (lqmcs C, G
and 11) aif the nine games in table 3.
J.E.B.O.- D
 Naive de&ion bel~avior in eesygames.
-I- --m---^---

c account
to be Demand of
distributed player i Decision of
Game (DM) tDw player 2
-_I-_ .---_
A I0 6.00 1

: 9
8 am
4.00 1i
D 4 2.00 1
E 5 3.50 1
Ei 6 3.00 !
3‘_50 I
7 10
1;: 5.00 1
5.m I
s 9 5.60 1
K 9 5.55 1
L 8 4.35 1
M 8 5.00 1
N 7 mo 1
0 7 5.155 I
G 6 4.80
4.00 01

it :
5 3.00
2.50 1
4 4.00 0
4 4.00 1
----w ---

Table fi
Fxpcrhmd decision behavior in ..asy games.
---_-__l_ ----.
c = amount
to be Demand of
dimibutcd pla: :‘r : Ikision of
Game fDM) @Mj player 2
.--.-_a
A 10 7.00 1
B 10 7.50 1
C 9 4.50 1
D 9 6.00 1
‘E 8 5.00 1
f7 8 3.00 I
G 7 4.00 1
H 7 5.00 1

: 4 3.00 0
K 5 tz! 0
I, 5 3:o0 1
h4 6 5.00 0
N 6 3.80 1
0 IO 6.00 1
p 9f 4.50
4.00
6SO 01

s 6 3.00 1
T 5 4.00 0
u 4 3.00 1
-_- -
 W. 4?W e:?d., Ati tsqwhentd atudysfs of nltimafm hrgainin,p 377

rqgw&xi as the uosts of player 2 for choosi:lg conflict. The decision of player
2 may also depend on the share (cc-a& elf player 2 according to player l’s
‘propsaL Qne would expect that player 3 is more likely to refuse a given
&man&of player 1 if his pai (c--al) as wefl as his share (c - LFJC in case
of wptance bre comparatively low. &idly: one exceptional case (piayer R2
in table 5) *where the rather moderate de’nand aI =DM 4 was refused at
costs of DM 3 for player 2, it can be seen j#ith the help of figs. I and 2 that
the exnwr&nntai results are in line with our intuitive cxpc:tatioz.s.

33. Consistency of demands in easy games

After testing twice the behavior in easy games we became interested to
learn how the demand behavior of a subject, i.e., his decisions as player 1, is
related to his acceptance behavior, i.e., his de&ions as player 2 [similar
questions for other game situltions are analyzed by Stone (1958)J Would a
certain subject acccpt as player 2 an ofler to distribute c which he would
suggest as player I? In order to investigate this question, we performed a
third experiment of the easy game with c:= 7 DM in the following way: Ah of
the 37 subjects participated in the experiment as pla!rer 1 as well as player 2.
First every subject had to decide as player I which amount a, he demands

Hz/C
0.6
I
t” I 1 1 1 1 11
i
378 W. Giith et al., An experimental analysis of ultimatum bargaining

11

10
0.4

0.3
101

0.2
o
o

0.1

o 2 3 4

Fig. 2. Experienced acceptance behavior in easy games.

for himself. Then every subject got another form which asked him to state
his minimal acceptance payoff a2 as player 2. If c - a 1 ~ a2, player 2 accepts
player l's demand which yields the payoff a 1 for player 1 and c-a 1 for
player 2. Conflict results if c-al <a2' The subjects were told in advance that
it will be determined by chance which of the other 36 player 2's decisions a 2
will be opposed to the own decision a 1 as player 1. Since every subject had
to hand in his sign as player 1 and as player 2, we were able to identify
uniquely his decisions as player 1 and as player 2.
It should be mentioned that the decision of player 2 in this experiment is
more complicated compared to the former experiments of easy games. Here a
player 2 has to consider all possible decisions of player 1, while in the former
experiments player 2 was only asked to react to a specific choice of player 1.
Although a subject had to expect a different subject as his opponent, we
were mainly interested how a subject's decision a 1 as player 1 is related to
his decision a2 as player 2. In table 7 the decisions of one subject are listed
in one line. The second column gives the demand al as player 1 and the
third column the acceptance level a2 as player 2, whereas the sum a 1 +a2 of
demands appears in the fourth column. If this sum is greater than/equal
to/smaller than c = 7 DM, this is indicated by '+ '/,0'/' -' in the fllth column
of table 7.
 y4 qw4 at id., A?cexjw~~iul anulysis of ultimatumburgainw 379

...‘I
.A. I.
..,” Tablo 7
.,; ”
: 1 ,) __

I :_; CoPr(r*toncy.?fpPyoadCmsnda
inay games*
” 1 ‘,l‘*x (2f : ‘8,4e&md s,hiem&. ur +az =sum Chsistcncy
1 ,,’ :’ :~~‘. ::sal@layc?
1 ~#w2 bpdcmlands of dem&s
_--
,- ;,‘-‘, / ‘6.00 ( ’ ia 7.00 0
3.50 230 6.0 -
3 3.50 , 3.50 7.00 0
‘8 3.50 7.00 0
:;g ?.OO
4 3.00 0
6 3:50 7.06 0
7 4.00 ;z 7.00 0
’ 8 3.50 8.50
9 ’ i-ii 3.50 7.00 0’
‘, ’ 10 i50 3.50 7.00 0
11 4 7.00 0
I2 :z tz
13 ’ 5:oo Ii0 ;; z
.j~ 14’ 3.50 1.00 i50 _-
15 3s 5.00
8 16 4.w 2.50 ;:g :’
17 - 4.00 3.00 7.00 0
’ 18, 4.00 ’ mcl 7.00 0
19 5.00 1.00 6.w __
20 0.01 7.00 0
21 tz 5.50
&IO z 6.50 --
ii 4.00 3:50 _ 7.50 +
24 3.50 3.00 6.50
5.013 2.00 7.00 0
,ii 4&O 1.00 s.uI) -’
. 27 II 3.CG 2.00 5.50 --
25, +#.OO 1.00 5.0
29) 3.50 3.00 6.50
30‘ 3s 2.50 6.00
31. 4.50 8.00 -1.
qlo Eii 0
:5 4.00 610 ;z -.
34 3.50 7:00 I,
:z
ii 7:oo Ei G-: 7
37 4.00 2:sa &I -
--

‘. ,I“
‘j,, I
5 d&on are in conflict (+), 15 consistent ([Cl)and 17 in
vtxtom
micmf%t (-) Thus 32.1of ‘thi; 37 sub+ revealed a nwimt demand
hehwi~r in tilwesense that; ,&he,payoff c-u1 was not smalkr than their
stance ley?! .a2 i)sspl&~ + *Nearlyhal$ of the 37 veckws &, a,) were
~wc;‘& a&k&&k F ‘j$&ct~ were a to accept demkis of player
1 Iji$qK~.~gher than tI#r .Kti aspiratim kweis a,.
’ ,Jg: r;;ric+z+flkt sub~;Wvw lesv to My& 2 than t&y themselves are
w&!g ~+FM?U!s_ ph+G /!hy must consider themseks as exceptionally
“111’
/
( \“/
coiumn of table 8. The actual choice I or II of player 2 is listed in the last
column irbtable ‘8;It can be seen from tabic 6 that players 2 always chose; the
bundle which yielded a higher r.:ayoff Hz. In the pilot study of complicated
games only one player 1: 11=ziely subject Il, proposed the equilihri-J;n
solution.
The same subjects who participated in the main experiments of easy games
were afterwards confronted with the complicated game. The results of the
main experiments with the complicated game are Zisted in tables 9 and 1C;.In
a first test the payoffs were the *sameas in the pilot study. The -esults of this
first test are listed in table 9; we refer to thrum as decision behavior in
con;gGeated games with low payoffs. After 3ne week the experiment was
repated with the rather high payoffs as determined by the description of the
game. These results - we refer to them as decision behavior in complicated
games with high payoffs - are listed in table 10.
Compared to an easy game situation the equilibrium payoff vector (1X0
DW, 1.00 DM) in table 9 or (18 DM; 10 DM) in table 10 is less extreme in
complicated games since it yields comparatively high payoffs for both
players. There are two possibilities I =(ntl, ne,) for player I to suggest the
rational solution, namely (ml, mz)==/5,0) and (m,, m2)= (0,9). In 6 of the 17
games in tabk 9 players 1 suggested the rational solution, whereas in table
10 this was done in 9 of 15 games. Thus compared to our results for easy
games players .I in complicated games rely more often on the rational
decision behavior although it is more difficult to derive. This indicates t‘nat

Table 9
ZtisiQn behavior in complic&d games with low payofps.
_-- -..LI__.-_---
r&xisi0a
I=(w?d (H,(I); H,(I)) (WIO; fi,W)) Decision of
Gwne of player 1 VW PM) plqer 2
- .. __-.-c_--
A (5>0)* (l.%o;1.00) (1.00; 0.90) 1
Ii.2 y; (1.00; 0.90) 9:
E IZ. (1.40; 0.70) I
$1;; (1% 1:10) (l.f%;0.80) 1
-! . yg y$ I;.$ ;g; II
.L’F (42) . I
’ (J *. 6-B ’ (1:40: 1:20) (1:40; 0.70) I
H. ,(W y$ :.t; ykw;;; I
1, .~ HP31 /’
621 . 11:4&1:20) (1:&: 0.7;) :
.& (4+ ‘! / (L20;1.20) (l.ao; 0.79) I
I *.,. -L
i : L. * ‘@s,Q) , (l.&W.Qti) Poo;W I
>W $4 3) g4Qj 1.10) w;g{ I
N ’ ‘(4,z) 1 ’ % (l&q 1.00)
(S3) ~ (l;sosQ9q’ . [y&$ :
, ;‘k. I (5&f . I
’ : , I&#@ MO)
5 I
,: Q v: :+. WB $R s*ou1> _ <1:00;&)
.
-‘A,’ --1___1
 lif player I wa.;~ts as, zqt& spht, he (;a~ propose this either by k=(4,4) 0x
BHre~~rres~~~d~~~ budk 11 or by F= 4, f j and the
ff player 2 accepts it, the payoff vector is
the first case and 90) in the !Xcond one.
fcr player 2 to ac split; if he dek~afes player 2 vdould s&fer a
bile player 1 wuuld gain by such a devirrtisn. I\ IS, of course, better to
, yp2J==(I, 5) or the corresponding bundle ;,; J) since this implies
offs for both piaq2rs.
9 ot1ly c!ne playrs 1 suggests m equa’ split, nameiy the one with
s in table 10 three players , suggest the equal split with
is indicates that in the r~p~~i~d experi 2ent there is a stronger
;ii2 equal split 3rd that not alI players 1 in the repeated
experiment were fully aware of the payoff structure. At leash for these players
9 it is doubtFA whether they have analyscd the game situation earefuully
enough.
In a compfic-ated game player 1 chaos a maximin-strategy if he designs a
bundle H= im ipm*> with Vzll+ n22~ 7. ue to the special structure of
aximin-strategy by player 1 determines
and DM 14 in table 10).
in table 101this occurs
nt plajers 1 in the first

eriment of complicateIS.
games more players 1 tended towards the normative solution while more
~~ay~r~ 2 were ~~~~i~~ to block unbalanced pay& vectors. This behavior sf
players 2 has 2s counterpart in a stronger tendency of some experiericed
egign bundles which allow more balanced payofl vectors.

ai~~g games are special bargaining games since

i~t~ra~t~~~ of players occurs only in the form of ant:cipatisn. In order to
make the ultimatum aspect obvious. we concentrated cm the easiest non-
385
 387

about pIdyer l’s decision (m,,m,J. Knowing player

can choose between the bundle (ml,m,) of ml
the residual bundle (5-mm,+mz) with S-m1
s. Player t receives the bundle which has not

‘I% pay& of oath player is determined by the v,‘riu~of all the chips which
:x&& If, for instanee, player 2 chooses the bundle (m,,m,),hispayoff is
m, - D&t ii+ m;, IX++41. Player 2% payoff is DIM 2 times the number of chips
which he receivc;d.
(Rustration lof bargaining rules various numerical examples). The
experiment will proceed as follows:
Ther: will be k=... bargaining games. First it will be decided by chance who
of you will be players 1 and who of you will be players 2 in the k bargaining
games. All players 1 will be seated at the (isolated) desks on one side,
whereas players 2 .wiIl be seated at the (isolated) desks on the other side of
the room.
Each player 1 will receive a decision form. Every player 1 has to determine a
bundle I==(ml,m,) of m, black and mz white chips. By this he u?ers player 2
to choose between the bundle I =(ml,mt) and the residual bundle II =
(5 -ml, 9- m2)of 5 -m, black and 9 - m, white chips. When de’ermining his
decision I=(mr, mz),player 1 does not know who of the k = . . . l Iayers 2 will
be his opponent.
After ail players 1 have made their decision, their decisio:l forms are
distributed by chance among the k = . . . players 2. Knowing the two bundie>s
I =(ml,m2) and II = (5 -ml, 9-mz) each player 2, has to decide whether he
wants the bundle I =(ml,m,) or the bundle II =(5 -ml, 9-m,).
Each player has 15 minutes for his decision. When all decisia ns have been
made, the decision forms will be collected. As described abov: your payoff
will be determined by the bundle of black and white chips which you
received. To get your money you have t 7 keep the tickei which is attached to
your decision form.

If you have any questions, we will be happy to answer them now. During the
experiment it is forbidden to ask questions or to make remarks.

R
Four&q L,E, and S. Siegel, 19/j& Bargaining behavior (New York).
oilth, l#., 1976, Toward a mow general study of v. Stackelberg-&u&ions, irRitschriA fiir die
&mute Sta&4!i$mscM 13a,592-60&
oistb, W., 1978, Zut ‘Rxmie koifektiver Lohnverhandbgen (Badcn-Baden).
388 W. Giith et al.. An experimental analysis of ultimatum bargaining

Giith, W., 1979, Kriterien fUr die Konstruktion fairer Aufteilungsspiele, in: W. Albers, G.
Bamberg and R. Seiten, eds., Entscheidungen in kleinen Gruppen, Mathematical Systems in
Economics 45, 57-89.
Giith, W. and B. Schwarze, 1983, Auctioning strategic roles to observe aspiration levels for
conflict situations, in: R. Tietz, ed., Aspiration levels in bargaining and economic decision
making, Lecture Notes in Economics and Mathematical Systems (Berlin-Heidelberg-New
York).
Harsanyi, J.e., 1968, Games with incomplete information played by Bayesian players,
Management Science 14, 159-182,320-334,486-502.
Harsanyi, J.C., 1980, Noncooperative bargaining models, Working papers in Management
Science, CP-421 (Center for Research in Management Science, University of California,
Berkeley, CA).
Krelle, W., 1976, Preistheorie, Part II, Ch. 9.4 (Tiibingen).
Kuhn, H., 1967, On games of fair division, in: M. Shubik, ed., Essays in mathematical
economics in honor of Oskar Morgenstern, 29-35.
Pazner, E.A. and D. Schmeidler, 1974, A difficulty in the concept of fairness, Review of
Economic Studies XLI, 441-443.
Selten, R., 1975, Reexamination of the perfectness concept for equilibrium points in extensive
games, in: International Journal of Game Theory Bd. 4, 25-55.
Selten, R., 1978, The chain store paradox, Theory and Decision 9, 127-159.
Selten, R., 1982, Einfiihrung in die Theorie der Spiele mit unvollstiindiger II'lformation, in: E.
Streissler, ed., Schriften des Vereins fUr SociaJpolitik, N.F. 126, 81-148.
StAhl, J., 1972, Bargaining theory (Stockholm).
Steinhaus, H., 1948, The problem offair division, Econometrica 17, 101-104.
Stone, J.J., 1958, An experiment in bargaining games, Econometrica 26, 286-296.