Lhewhole

t toward
rrnational
comprise
:ensifying

ON THE RATIONALE OF' GROUPDECISION-MAKING

DTINCAN BLACK

re wants
nl collecuction;at
ark of reorizon.It
r between
of colleclemof the
i the task
:rful ana-

ical explanaI to shift the

: the second
n'hoietheory
along other
no emphasis
so,the elabodevelopedin
y about the
un,andB. M.
toward the
le wants, alby which the
nlly interact
)n the literaxely related,
;ory of "ecoi'n
Econont'ics
i. ed, Edwin
*, r93rl, pp.

dicate the course of the argument'r

I.

GENERAL

ASSUMPTIONS

theory will be set out at greater length in

ring book on The Pure Sci,enceo! Pol,itics.

To d,evelopour theorY, we must make

somefurthei assumptions'Our major assumptionwillbe that eachmemberof the
.o**itt.e ranks the motions in a definite
order of preference,whattYgt that order
*uy be. bo take a simple illustration, if
thete are four motions denotedby'o', a"'
&t, &4,say,beforea committee,the mema""t i may prefer a'2to any of the others'
-ut Ueittdifferentbetween03and ao,and
may prefer either of them to a,'
it io, /.'s valuation of the motions
of
could be representedby the schedutre
Fig'
preferenceson the left-hand side of
i, io which o" stands highest; ls a2d a,
next highest,eachat the samelevel; and
o, lo*Jut. And similar scalescould be
drawt for other membersof the'committee with (h . .. 04 appearingin some
defi.niteorder on each scale,though the
ordering of the motions might be different on the scaleof eachmember'
We are here using the theory of relative valuation of orthodox Economic
S.i*n.., whether the theory of reiative
utility or the theory of indifference
which have sig.nr.r.t. The only points
-d'irected
straight line
,rifi.un.. on the
representing a member's scheduie of
pr.f.r.rr.es are those at which motions
are marLed.,and his scale really consists
of a number of points placedin a certain
order in relation to each other' No significance attaches to the distance between the points on the scale, and any
two scaleswould be equivalent on which
the motions occurredin the sameorder'
Slhen a member values the motions
before a committee in a definite order,
it is reasonableto assume that, when
a
thesemotions are put againsteachother,
23

DUNCAN BLACK

oA

he votes in accordancewith his valuation, i.e., in accordancewith his schedule

of preferences. Thus the member .4
would be assumedto vote for o, when it
was put in a vote againsta,; or lI a3were
put against oo-since he is indifferent between the two and it would be irrational
for him to support either against the
other-he would be assumedto abstain
from voting.
A member's ievel of preferencebetween the different motions may also be
shown by denoting the motions put
forward by particular points on a horizontal axis, while we mark level of
oPAp 0F
PREFFEITCE"

While a member's preference curve

may be of any shapewhatever, there is
reasonto expect that, in someimportant
practical problems, the valuations actually carried out will tend to take the form
of isolatedpointson single-peakedcurves.
This would be particularly likely to
happen were the committee considering
different possible sizes of a numerical
quantity and choosingone sizein preference to the others. It might be reaching
a decision,say, with regard to the price
of a product to be marketedby a firm, or
the output for a future period, or the
wagerate of labor, or the height of a par-

of any cur
for the m
mum.
Anothe
occurrenc
where iht

oRaEP oF
PFFEPE*CE

oz
o3,o4
O1

preference along the vertical axis. For

instance, the same set of valuations of
the individual .4 is shown in the right
and left parts of Fig. r. The only points
in the diagram having significancewould
be those for the values e'r, a2, a3, e,4,orr
the horizontal axis, correspondingto the
motions actually put forward. We have
joined these points standing at various
levels of preferenceby straight-line segments, but this is done merely to assist
the eye, since the curve would be
imaginary except at the four points. In
this dia,gram,as in the caseof the preference schedule, it is only the reiative
heights of different points which have
meaning, not their absolute heights.'
" C{. F. H. Knight, Ri,sh,Uncertaintyand Prof.t,
pp. 68-7o.

PotNf ,SEr
AEPRESEIttrrlO

t ovopa

ticular tax, or the legal school-Ieaving

age,and so on.
In such casesthe committee member,
in arriving at an opinion on the matter,
would often try initially to judge which
sizeis for him the optimum. Oncehe had
arrived at his view of the optimum size,
the farther any proposal departed from
it on the one side or the other, the less
he would favor it. The valuations carried
out by the member would then take the
form of points on a single-peaked or
O-shapedcurve.
In working out our theory we shall devote considerableattention to this class
of curves which slope continuously upward to a peak and slope continuously
downward from that peak. We shall refer
to the motion correspondingto the peak

whom t

;ilil

ON TIIE RATIONALE OF GROUP DECISION-MAKING

e curve
there is
rportant
ns actubheform
I curves.
kely to
sidering
rmerical
r preferreaching
he price
, fi.rm, or
, or the
of a par-

c Eonops

[-leaving
member,
: matter,
3ewhich
:ehe had
.um size,
ted from
t}le less
s carried
take the
:aked or
shall de-.,,
:his class,,:

usly up'
inuousll
hall refer
the peak

25

put forward and valued by

of any curve-the most-preferredmotion motions is
for the member concerned-as his opti- the members.
We assumethat the committee with
mum.
which we are concernedmakes use of a
Another caselikely to be of frequent
simple majority in its voting' In practice,
occurrencein practice-especially, again,
voting would be so conducted that, after
where the committee is selecting a pardiscussion,one motion would be made
ticular size of a numer.{palQuantity-is
and, after further discussion,another
that in which the valudtions carried out
("tt "amendment," that is)
-otioo
by a member take the form of points on
might tte moved. If so, the original moa single-peakedcurve with a truncated
tion and amendment would be Placed
top. Such a case would arise when the
against each other in a vote. One of the
inclividual feels uncertain as to which
two motions having been disposedof,
of two or more numerical quantities proleaving a single motion'in the field, a
posed represents his optimum choice'
further amendment to it might be
He cannot discriminate in choicebetween
movedl then a further vote would be
(say) two of these numerical quantities;
taken between the survivor of the first
but the farther the proposal made falls
vote and the new motion; and so on' If
below the lower of these vaiues, or the
z motions were Put forward, r vote
higher it rises above the larger of them,
would be taken; if 3 motions, z votesl
the lesshe esteemsthe motion concerned'
and, in general, if. m motions were put
We shall work out the theory first for forward.,there would be (m - r) votes'3
,
the case in which the members' preferNow it will be found to simPlifY the
't ,,
encecurves ale single-peaked,and, after development of the theory if, in the flrst
''
that, we shall show how the answer to instance, we suppose that the voting
'
any problem can be obtained no matter
is difierent from this. We wish
I r,' , *nu, the shape of the members' curves procedure
io make the assumPtion that when m
..,,,,1 may be. When any matter is being conmotions a,, a", ' . . atu (saY) have been
sideredin a committee,only a finile num- put forward, the committee places each
'
r r------^-l
^*.J
..,:;,
.r.:lberof motions will be put forward and of th"t" motions against every other in a
vote and picks out that motion, if any,
which is able to get a simple majority
against nery otlter motion. The motion
o, it to be envisagedas beingput against
all the other motions 02 . . . ami a, will
already have been put against o,, and we
assumethat it will then be put against
1s,.curvcs
and-since there are an infi- a3 . . . a*; and so on' a--, finallY being
te number of points on a.ny continupitted against a^. On this assumption
ls curve-we imply that the person for
the number of votes taken will be the
hom the curveis drawn has carriedout number of ways of choosingz things out
ion of each of an infinite numof motions in regard" to each of the

:rs. This is unrealistic,it is true, but,

'n the theory is worked out for this
re can easily get the answer for
;e in which only a finite number of
ii,
iii

be supplied.

DUNCAN BLACK

z6

ii,
:t

i
iir

ii

r ii.l
i,:.ti:
t: i::,

.li;

I i:,
'i'1,

:,
i

'ii

of m,i.e.,m(m-t) /z votes,insteadof the

(* - t) votes which would be taken in
practice.
This assumptionenablesthe theory to
proceedmore smoothly and quickly than
the assumption that only (m - r) votes
are held. When we'have worked out the
theory on this basis, we can go on to
prove that-in the classof casesin which .
we are mainly interested-the same answerwould be given whether m(m- r) 1 z
votes were held, as we assume,or oniy
the (rn - r) votes of reality. The assumption is a kind of theoreticai scaffolding which can be discarded once it
has servedits turn.
These,tb.en,are our assumptions:that
in a committee m motions are put forward, that each member carriesout an
evaluation of each motion in regard to
every other, that in the voting each motion is put against every other, and that
the committee adopts as its decision
("resolution") that motion, if any, which
is able to get a simple majority over
every other.
It can be shown that, at most, onlY
one motion wiil be able to get a sirnple
majority over every other. To prove this,
let us assrrmethat or is such a motion,
i.e., that an carrget a simple majority
over every other. And let us assumethat
this is alsotrue of someother motion, ar,.
By our first assumption,howevet,a1 czrr
get a simple majority over every other
motion, including ar. Therefore or csrlnot get a simple majority o.vera6.I{ence,
at most, only onemotion can get a simple
majority over every other.

, ,iL

.ii

ii
li

"i 1..

II.

TIiTF'MBERS' PREFERNNCE

CURVES ALL

SINGLE-PEAKED

The method of reasoningwhich we employ can be seenmost easily from a particular example. Figure z shows the
preferencecurves of the 5 members of a

value in
simplem
seen (en,
only a.
adopted
motion t

committee. Only part of each curve has

been drawn, and the curves are supposed
to extend over a common range of"the
horizontaLaxis.
Then if aa is put against oa (where
at I an (O,), the preferencecurve of
each member-irrespective of what its
preciseshapemay be-is upsloping from
ah to o,k; and azo,standing at a higher
level of preferenceon the curve of each
member,will get a 5:5 (5 out of 5) majority against aa. If o7,is put againstoe,
(where an I at ( O,), at least 4 members-viz., those w-ith optimums at or
above O,-rvill have preference curves
which
ar will
ad.If
a 3:5

are upsloping from a5 to

get at least a 4: 5 majority
an { or ( Ou at will get
majority against on. And

ORDEIQ
PREFERE

a1,; and
against
at least
similar

relationshold for motions corresponding

to values above Or. If two values above

O, are placecl aga,inst each other in a

l
l

vote, the nearer of the two values to O"

wiil get a majority of at least 3:5
againstthe other.
ff a value a6 (where an I Ot) is put ,
against a value a1e(where a6) O),
before we could find which of the values
would win in a vote, we would have to
draw the courplete preference curve for
eaclr member, fi.nd whether {Ln ot an
stood higher on the preferencecurve of
eachmemberand count up the votes cast ;
tor an and an. But even though a value :
below the median optimum O, should de- "
feat all values to the left of itself, and ,
should defeat some of the values above
Or, this wouid be without significance..,
What we are looking for is that motion ''
',
which can defeat every other by at least.
a simple majority. And we notice that ;
the preferencecurvesof at least 3 mem- :
bers are downsloping from O, leftward, .
and the preference curves of at least 3 r;
membersare downslopingfrom O, rightward. Therefore O, can defeat any other' ,

aRoe
PREFE,
t

ON THE RATIONALE OF GROUP DECISION-MAKING

has
rsed
the
lere
:of
its
:om
Jher
ach
ma-

value in the entire range by at least a

simplemajority. And, aswe have already
seen (end of Sec. I), this can be true of
only a single value. The resolution
adopted by the committee must be the
motion correspondingto theuvalue-Or.

or7

the horizontal axis correspondingto the

members' optimums are named O,, O",
Or, . . ., in the order pf their occurrence. The middle or median ontimum
will be the (n * r)/r", and, in Flgure 3,
only this median optimum, the one im-

ORnEF oF
PACFERETICE.

Akt

em.or
'ves
and
inst
:ast
ilar
ting
ove
n a
,Os
3:5

PotNT sE7
Re U RESEn rrilq ktu//Ottl6

Fro. z
ORaEQ oF
PREFEQEilCE.

put
)'),
Lues
:to
for

to/Atr sE7

A*

:of
last
rlue
de-.
end
ove
1Ce.

:ion
:ast
hat
:rnrj
rrd, l
IL

tht

To give the generai proof, two cases

tnust be worked out-that in which the
of members in the committee is
tdd and that in which it is even. We will
'nsidereach in turn.
Let there be z members in the com-tjee,where z is odd. We supposethat
rrderingof the points on the horizonaxis representing motions exists,
{6ring the preference curves of all
single-peaked.The points on

mediately above it and the one immediately below it are shown.

Then Op4,y7zwill be the motion
adopted by the committee. Suppose
Opa,11.were placed against any lower
value, sa,.!,ap.Since(a * r)/z members
have optimums at or above Op141z1&s
'we move from left to right from on to
Osq'y1",at least @ * r) /z curves are upsloprng, viz., those of members with
At
optimurns at or to the right of O14-.17,.

^Y

'l
:f'
.' : .
:lr

: L .,'
ir

DTINCAN BLACK

Ieast (n * r)/z members prefer Op4'11, ence curve must be downsloping from
to at"and,in avote againstan,Og+r11"will On1"to a7r.
Thus, when the chairman's oPtimum
get a majority of at Ieast (n I r)/z:n,
and this is sufficient to.give it at least a is situated at or below On/",O,/" will be
simple majority. {lherefore Os+;y, ca;r- able to get at least a simple majority
get at ieast a simple majority against any against any other value which may be
lower value which is put against it. Simi-. proposed.
Similarly when rz, the number of
larly it can get at least a simple majority
against any higher value. Thus it can members in the committee, is even, and
get a simple majority against any other the chairman's optimum is at or above
value which can be proPosed.And bY Op1"1y',it can be shown that01"7a"',wili
previous argument, it is the only value be able to get at least a simple majority ,
against every other value.
which can do so.
One cannot leave the theorem of the ,
When the number of members, n, in
pointing I
the committee, is even, there maY be a preceding paragraphs without
principle
central
the
tie in the voting; and we will suppose out its analogy with
how price is '
that an additional personacting as chair- of economics-that showing
The
suPPlY.
and
man, in the event of a tie has the right fixed by demand
the I
that
proved
shows
have
theorem we
to cast a decidingvote.
be- '
committee
the
by
decision adopted
Let us suppose,first, that this memposias
the
as
soon
comes determinate
ber who acts as chairman has his optition of one optimum-which we can I
mum at Ony2ot at, one of the lower
refer to conveniently enough as the ,,
ootimums. It can be shown that the
median optimum-is given. No matter in
motion correspondingto the value Oo1"
what manner the preference curves or l
wiil be able to defeat any lower value
optimums of the other membersalter or
(Fig. +). Let an be such a value, that is,
move about, if it is given that one opti-',
an l On/2. Then (n/z * t) members
mum remains the median oPtimum, the
have optimums at or above On1,;and at
decision of the committee must remain
least (n/z { r) preferencecurves will be
fixed. The analogy with economicscience
upsloping as we move from left to right
is that, in the determination of price in a.r
fiom anto Onp.At least (n/, * r) memmarket, price remainsunchangedso long
bers will vote for O^1" against an, and
as the point of intersectionof the demand'
this is suftcient to give On/" & simple
and supply curves is fixed and given, ir-u
majority.
respectiveof how thesecurves may alter
If Onf is put against anY value o6
their shapesabove and below that point'
(where an ) O^n)-since there are nf z
Or, in the version of the theorY due
optimums at or below O,p-the prefer- Bcihm-Bawerk, which brings out the
ence curves of at least nf z memberswill point very clearly, price remains un:
be downsloping from Onp to &4 3,\d altered,o io.tg u, th. n*utginal pairs" of
O,p vmTlget at Ieast nf z votes against buyers and sellers and their price attlr
i.e., wili at least tie with or. In the tudes remain unchanged.
o,10,
event of a tie O*1,will defeat a* with the
But the analogy exists only betwee'
vote
bedecicling
chairman's
the two theories; ihere is a marked dil
aid of the
is
his
optimum
ference in the materials to which the
cause, by hypothesis,
situated at or below Onp a.ndhis prefer- relate. In the caseof market Price,

the price of z
mined, & ser
part of the
existencea s
commodity,
purchase, w
cance at the
This is one
monies runn
economic lif,
one who und
pressed-an<
of,the iast ge
impressed.I.
oPAA
PREFE]

decisions, (
in general)
grand harmr
the persiste
cord is as s'
the certaint
In reachir
we assumed
mittee voter
forward in
on his schec
shown that,
r,, would.defea
,, voted in thi
',,,,, ,ff1gfn!g1,

9y

,.r,,ing in conc
,i.,r,that some r
preferredb;
': ,fesolutiono
:,

: ]f oniy (z
".;1:; no longer holds
i:,t

29

ON THE RATIONALE OF GROUPDECISION.MAKING

however, to vote in suclr a way

the price of a commodity is being deter- to them,
motion wiil be able tb get a
mined, a seriesof adjustments on.the that no
over all the others'
oart of the consumers will bring into majority
it ali me-bers voted as we have suPi"irt".t.. a state of affairs in which this
posed,the motion adoptedby the com,nitt." would be that corresponding to
the median optimum, 0*66, S&]' Let us
supposenow that one or more members
with optimums above O*"a-by voting
otherw^isethan directly in accordance
with their schedule of preferences-atto give some other value' sa'! an,
pressed-and by which the economists tempt
over all the others, where
of the last generationwere perhapsover- a majority
O^.a'
impressed.ln the material of committee &n)

OPDR OT
PREFERTICE.

g
D

.e

is
IE
IC

ritn
in
or
or
tihe
lin,
ICe,i
1&r
'ngl

"E

o+

2',
,
,.
i.,,

Pohlr 6Er
RePPESENtl'yq

,ronoils

d e c i s i o n s , ( o r o f p o l i t i c a l p h e n o m e n a B u t w h e n t h e m e m b e r s v opreference
tedirectly
accordance with their
in general) on the othe, itana, no such in
who have anhigher on their
grand harmony .*i.it- ift" possiUility-of scales'those
and dis- scalesthan O-"a would already be supthe persistence"r-Jirrr-i-i"y
against,of"u
cord is as striking in the one caseas is porting-oa
Td'*ii:::
iiit
o-"a' Before

r"'tr'" lrn.i.

"t tffi"-

;#:ili'U

bv
iould"be defeated

the
conclusions, could.defeat O*a, or, would require
1., . In reaching th" f;;;;ing
iie
optimums
r.r.,;
,, we assumedthat a member of the com- support of memberswhose
who-by

members
i',,, llrlJi "ii"u ""'irt" variousmotionsput u"io* o*"o.Theonly
than in accordance
i,.r,.r forward in accordancewith their order

voting otherwise

or prererences-courd
scares
i-.,;;;lJ:.illu"rl?il;;:;;;"fr;;1,1;; which witrr"their
makeanthe,resolution^1f,^*:.:."f-nlti'
shownthat, whenil;i;;;;ists
i!;i1i1,,,
o'"4' i'e''
other if the members arethosewith optimumsbelow
iljfi ffi;;;;t
lr l^ +^ J^ .^
to do so'
is
it
interest
*"i, it is not opento any thoseagainstwfiose
. . ;otj ilril
for a
course.,
of
It would be possible,
Jt^""t i;*ber of memb"r,u"i;i: ;.i'i";
Vr

. .IrwIlIULI,

AUJ

"^

*^-"_"

ftg, i" concert,to atrertheir.vo.tingso nu1ler of 1e*":ti:,::::

::,"T::"1:
over every
tfi;";
"ifrli -"tion which is more motion would get a majority
*
j;.IJft;;i;;;;te
asthe other'If, for eTa,mpl:t1ffi:i'1^":-:
adopted

$fuil;b""ir
ffi'ffiffi;ii;l'
". '-l:-:*,'-

;o;;

__

o*aa
above
berot voterswithoptimums
were to vote against O'ua when it was

."-'..
1i 1r-:^^-^,..^:^thisconclusion
are,held'
r) votes
pt"..a against ,o*.
,iJj-:lt{"rk.
. . :
tiri..,i
i:i :

It,,

uulrr. which stood

DUNCAN BLACK

one of the motions put forward and it

must enter into. the series of votes at
somepoint. When it does,it will defeat
the fiist motion which it meets' It will
likewise defeat the second and every
other motion which is put against it'
rrlust enter the voting
it
can
That is, O1r+'172
are single-peaked,as we suppose,
stage, and, when it
some
differat
process
be shown that voting betweenthe
the other motions put
defeat
propit
wiil
.lo.*,
ent motions obeys the transitive
the decisionof the
become
o,,
and
it
against
ertys and that if-of any three values
\Mereached
conclusion
The
a,
and
.J**itt.e.
a2, a3-or c&rrdefeat o' in a vote
imaginary
the
for
only
not
good
or
can
holds
can defeat ar, then, of necessity,
everY
placing
of
procedure
defeatar.
Pot19+
for the
This can be proved by consideration against every other but also
real life'
of the orderings of the points ar, oz, as, aitual committee procedure of
we
conclusions
It
of
the
The same is true
in relation to the median optimum'
number
the
the
in
which
of
case
reachedfor the
can be shown that each ordering
of membersin the committee was even'
In the committee procedure of real life
Onp ot Os1,1a'wiilbe the motion actually
adopted.
- t) / z
The assumPtion, that m(m
feats or.
votes were held, enabled us to give a
The transitive property can easily be mathematical Proof which was both
extended to show that, if o, can defeat definite and"short. But our conclusions
azt a2 can defeat at, . .. and aF can are true independentiy of this assumpclefeat41,then at can defeat o1'
tion.
:
It follows from the transitive property
As an examPle of the use of this
a,
that a, can defeat or. BY hYPothesis,
technique, we may suPPgry that the,i
can defeat ao. Hence ar can defeat oo' three directorsof a monopolisticfirm are';
applicationswe
Proceedingby successive
fixing the price of their product for a';
can seethat a, can defeat o7'
forthloming period. Let us further as--';
In arriving at the above-mentionedre- sume that neither future salesnor future.rl
sults, we assumedthat every motion was costs can be calculated with certainti':
placed against every other and that in all and that there is no possibility of a.
m(m - i /z votes were held. We can choice of price being made purely by.
now remove this assumption and show means of cost accounting. Subjective''
that the samemotion will be adoptedby
ofj
- r) votes factors enter, and' varying estimates
a committee when onIY (m
tbg
by
formed
are
position
the future
are taken as in the committee practice of
different directors. If, on their difierenti
real life.
is views of the situation, the directors'1
For the casewhen n is odd, O1,ar17z
scalesof preferenceare as shown (Fig:,S)i
S'
L'
in
s The transitive property is defined
the price fixed will be that corresponding

Iower on their scalesof preferences'O-"4

i : '

Stebbing,A Modern Introd'uctionto Logic, pp' rr2

and 168.

to the rnotion ar'

NI. WIIE}
ENCE C'

When the
are not of 1
solution to
arrived at :
number of r
ORoE
PREFE,

ON THE RATIONALE

OF GROUP DECISION-MAKING

3r

rE. wrrEN THE MEMBERS' pRETERENCE CITR\rES ARE SUsJECT TO

NO RESTRICTION

is voting in a committee in which,the

four motions ar. . . a* have 6.sn, put
forward. Along the top row and down,the
When the members' preferencecurves left-hand column are shown the rnotions
are not of the single-peakedvariety, a a\ . . . an.Tn each cell of the matrix, we
-'-recordthe individual's
solution to any problem can always,ibe
vote for one moarrived at arithmetically, provided"the tion when it is placed against another.
number of motions put forward is finite. Looking to the topmost row of figures,

::"1f::,f",.

Poutrset
4,2

44

Frc. 5

0naenor
PREFERENCE.

(against)

d3

&z

otro*.

4r

(o, t)

(t, o)
(t, o)
O,a (t, o)

(o, o)

8z

(for)

4t

43

ai

(o, r)
(o, t)

(., o)

?4
(o, t)
(o, o)
(t, o)

(o, r)

A
Frc. 6

ii,-,.,

ii"'i.r,',To begin with, we return to the asthat every motion is placed.

1i;1,;Surnption
vote
against every other. The reilim.e
of the seriesof votes can be shown
iii.-q1{,ts

when o, is placed against a", A votes for

a., and we enter in the cell (a., a,) the
figures (o, r). When o, is placed against
ar, he votes for a, and, in the cell (ar,
l#ry. readily by the constructionof a or), we enter the figures (o, r) standing
matrix.6
for o votes for and r against. The other
Constructionof a matrix is illus- cells are filled in the sameway. SinceL is
in Figure 6, which gives the indifferent in choicebetweena" andao,he
Fi,:c0rrqsponding
to the scheduleof will abstainfrom voting when a, is placed
g19n..sof the single member L who against ao, and the cell (ar, ao) will show
indebted to Dr. R. A. Newing for sug- (o, o).The figures in the cell(oo,o,) will
: use ot a matrix notation.
also be (o, o).
-O---

r:,,.1,,r.:

!.:'ir;;'

,i-

rri{-^

^ --

m,

32

DUNCAN BLACK

Along the main diagonal of the matrix, and join them by a straight line. In
instead of having cellsof the usual type, the cell (a,, a") we enter the figure (t, S)
we have simply placed a series of zeros becauseon the scalesof z members or
and joined them by a straight line. This stands higher than a,, and on the scales
is to indicate that the cells (4,, a'), (a,, of the remaining 3 members a, stands
a,), .. . , which would,denotethat a, was higher. The other ceils are filled in the'
placed against art ar&iagainst a"t . . . t sameway and, as before, the half of the
have no meaning. In constructing the matrix on one sideof the diagonalcan be
matrix in practice, it is usually easiestto obtained by reversalof the frequenciesin
enter thesezerosalong the main diagonal the correspondingceils on the other side.
From the group matrix we can read off
first and join them by a straight line.
Each row to the right of the main immediately that-when the motions
j
diagonal is a reflection in the diagonal (h . . . a6 a,tE placed each against every
of the column immediately beneath,with other, as we suppose-o, wiil be able to
tJrefigures in the cells reversed.Thus the get a simple majority over each of the '
cell (.a,", er) immediately to the rlght other motions put forward. For this com- ,
of the diagonal shows (o, t), the reflec- mittee a3 would be the resolution ::
tion in the diagonal of the figure (r, o) adopted.
If a. motion exists which would be
immediately below the diagonal. The
cell (a,, oo), two placesto the right of the able to get a simple majority over all
diagonal, is the reflection of the ceil (oo, the others when the members voted dia,) two places below the diagonal. The rectly in accordancewith their schedules
reason for this is that the figures in any of pieferences,it would not be open to
cell (ar,, ax) must be those of the cell any member or group of members. bY ,
i
(or,,an)on the other side of the diagonal, voting in some other fashion-to bring
placed in the reverseorder. This feature into existence as the resolution of the
roughly halves the work of constructing committee a motion which stood higher
a matrix: we can filI in the figures on one on the scalesof all of them. Proof of this
side of the diagonal and then complete proposition is almost identical with that ::
the matrix by reflection of these figures bt o,rt earlier analysis (seeabove,p. z6). ,'
If, when m(m - r)/z votes are held,
in the diagonal.
The construction of an individual ma- a motion exists which is able to get at
trix would be gratuitous labor since it least a simple majority over eachof the I
merely gives, in a clumsier form, infor- other motions put forward, it can be i
mation which is shown clearly enough proved, as before, tJrat, when the mem- I
in the member's scheduleof preference. t.tt vote d.irectly in accordance with "i1
When, however, we have a group of indi- their schedulesof preferences'this would
,i
viduals voting on a particular topic and be bound to be the motion adopted even
'i:
had
been
(*
,)
votes
only
the preferencescheduleof eachis known, though
"i:
the matrix for the group presentsin very held.
But when the members' Preferencei
convenient form the information that we
'
need. For instance, for the group of curves are not single-peaked,no motion
schedulesshown in Figure 7, the accom- need exist which is able to get at least a,
panying matrix has been constructed simple majority over every other. ThiSr+
precisely as describedabove. Along the can be seen very'quickly from the ac*
main diagonal, as before, we enter zeros companying group bf scheduies(Fig. 8).'
LLt.

in which the
(Lrt Azt A3, On

metrical. Wl
feated by a,
and C; whe:
feated by o,
ORaee ar
FREFERENCI

Q,

a2
ag
la+
ag
a6

ON THE RATIONALE OF GROUP DECISION-MAKING

ne. In
t (2, 3)

berso..
: scales
stands
in the
iof the
canbe
nciesin
er side.
rcadoff
notions
I every
able to
of the
riscom;olution
ruld be
lver all
rted dihedules
opento
ers-by
o bring
of the
I higher
f of this
ith that
, P' 26)'
re held,
> get at
h of the
can be
rememce with :
iswould,
:edeven ,
ld been,.
eference
motion
t least a,i

er. This,;i
the ac=
(Fig. B)i

in which the arrangementof the motions

a,, a"t ar, on the members'scalesis syrnmetrical. When a, is put forward, it is defeated by or, which gets the votes of B
and C; when a, is put forward, it is defeated by o,; when a, is put forward, it

number of motions put forward in a

committee of any given size,the greater
will be the percentageof the total number of possiblecasesin which there exists
no motion which is able to get a simple
majority over each of the others.

's?

qRoee or
PREFREI{CE.

a2

a
,

as

as

ag

as

a4

o4

4c

?t

a3

a a

JJ

a
5

o4

o6

a4

42

o6

"s

ag

44

a-'o

*6

&z

o6

a3
44

4g

8.5

os

G'

oz

4/

@t

6rl

a6

43

L4

?5

a6

b, t) (2,s) Q,s) Q, s) Q,s)

(2, g) k, r) (2,s) k, ")
k, z)
')
r)
k, r) k, r) G, z)
k, k,
(2,
(r,
g)
r)
Q,s) (5, o)
k,
s)
(g, ,) G , ' ) Q,s) k, z)
(q, t)
b, ") b, s) (2,s) (o,s) ( t , q )

Frc. 7

OnaaB oe
PREFERENCE.

oe

qt

a2

ag

oJ

al

a
3
&z
&3
(2, r) (r, r)
(e, r)
o.z (r, z) \e--(r,z)\
a3 ( r , t )
2,r

2,r
-e-t.-

oz
I

c
Frc. B

In this state of affairs, when no one

of the
simple motion can obtain a simple majority
over eachof the others, the procedureof
a committee which holds only (m - r)
votes will arcive at the adoption of a
particular motion, whereas-if the requirement were that a motion should be
able to get a simple majority over every
other-no motion would be adopted.The
particuiar motion which is adopted by
the committee using the proceclureof

DUNCAN BLACK

34

practice will depend on chance-the

chance of particular motions coming
earlier or later into the voting process.
For Figure 8, if only (m - , : z) votes
were taken, that motion, a\ ot a2 or a3,
would be adopted which was introduced
last into the voting process.If, for example, or were first Put against a", a,
would be eliminated; and, with the field
thus cleared,o, would defeat o..
If, then, only (m - r) votes are held
and if no motion exists which is able to
get a simple majority over every other,
we cannot read off directlY from the
matrix the decisionadopted by the committee. But when, in addition to the
matrix, we know the order in which the
motions are put against one another in a
vote, again we can deduce what the
decisionof the committeemust be.
Referenceto Figure 8 will show that,
when the shapesof the preferencecurves
are subject to no restriction, the transitive property does not necessarilyhold
good.
tt.

rl.
i:

IV, CONCLUSION

The technique of this paper applies irrespectiveof the topic to which the motions may relate. TheY maY refer to
price, quantity, or other economicphenomena; they may relate to motions put
forward in regard to colonial government, to the structure of a college cur-

riculum, and so on.'.The tleory applies

to a decisiontaken on any topic by means
of voting-so far, of course, as the assumptions which are made correspondto
reality. And it is possible to widen the
assumptions, for example, to include
cases of complementary valuation; to
make allowance for the time element;
and to cover the cases of committees
making use of special majorities of any
stipulated size. With theseextensionsin
the assumptionsthere would be a widening of the fie1d of phenomenato which
the theory applies.
The theory, indeed, would aPPearto
'
present the basis for the development of
a pure scienceof politics' This would emptoy the same theory of relative valuation as economic science' It would employ a different definition of equilibrium.
Equilibrium would now be defined in
terms of voting, in place of the type of
definition employed in economicscience.
We could move from the one scienceto
the other with the alteration of a single
defi.nition.This, in the view of the writer,
would be the main function of the theory'
It fairly obviously, too, enables solne
oarts of economics-those which relate
to clecisions taken by groups-to be
carried a stage beyond their present
development.
Gr-asoowU*rvrnsrtv

Tffi;
I ttyunder the
Businessar
University
praisal of tJ
dustriai opr
able for ar
economist'
perfect con
full critiqu
businessbe
dustry in
level of emr
lotment of
ficiency of r
distributior