Professional Documents
Culture Documents
( ) i i
p x
( )
1
i i
p x
=
where
represents partial prestige independently earned
esource i
in disciplines 1, , i N = , with x
i
being the correspond-
ing financial r nput. We assume d d 0
i i
p x >
and that an institution allocates its limited reso
1
N
i
urces
i
X x =
(1)
with the required optimality conditions
( ) ( ) ( ) ( ) ( ) ( )
d
d d
k
i i j i i k
j j k k
p
p x x p x x
p x p x
=
c c
(2)
1
,
N
i
i
d
j
p P P c c
X x
=
=
(3)
1 , j k N s s . Then, it is an elementary matter tha
solution is g
N N
persurface
t for
the raphically presented where the possible
prestige hypersurface, represented by a vector
( ) ( ) ( )
, , p x p x , and the iso-prestige hy
1 1
( ) i
P p =
of financi
P , are tange
i
nt with each other, for a given level
resources
1
N
i
al X x =
.
=
g the optim ation problem with
th
In contrast to solvin iz
e additively separable prestige function, an obvious di-
fficulty with the problem (1) is brought out by
j
P p c c
in the first-order condition (2) which may in gen -
volve partial prestige
i
p in fields i j = . In order to
cope with the technical ficulty, we no hat the possi-
ble prestige hypersurface ( ) ( )
eral in
dif te t
( ) 1 1
, ,
N N
p x p x is drawn
solely with allocation dep ffected by
the functional form of
endency and is not a
( ) i
P p . That is, the effect of
changing allocation arrangement { } 1
,
N
x x is captured
by the possible prestige hypersurf as the effect
of the change in the functional form of the prestige func-
tion
ace, where
( ) i
P p is captured by the iso-prestige hypersurface.
Thus ffects of these two changes on the optimiza-
tion may be analyzed and discussed separately.
, the e
3. Analysis
out loss of generality, a heuristically We consider, with
simple case with 2 N = . The loci of attainable prestige
combinations, i.e., ble prestige curve, for a given
amount of institutional funding
1 2
possi
X x x = + , may be
traced by
( )
( ) reas the i
ressed by
2 1 2
;
p
p p p X = , whe so-prestige
curve is exp
( )
( ) 2
;
s
p P for a fixed level
of prestige
2 1
p p =
P .
1
Then, entities are satis-
fied by cons ction:
( ) p X =
the following id
tru
( )
( ) ( ) 2 1 2
;
p
x p p x X (4)
(
( )
( )) 1 1 2
, ; .
s
P P p p p P = (5
Using the identity (4), we derive
( )
)
( )
( )
( ) ( )
( )
( )
( )
2
2
1
2 2
1
1 1
1
1
d
d d
d
d d
d
d d
p
p
d
p
p
X x
p x
p x
x
p
p p
p
x x
x x
= = (6)
from which we obtain
( )
1 2
d d 0
p
p p s since d d 0
i i
p x >
for 1, 2 i = . Similarly, g both s
iden with respect to
1
p gives
differentiatin ides of the
tity (5)
( )
( )
( )
( )
( )
( )
( )
( )
( )
( )
( )
( )
( )
( )
( )
( )
2
1 1 1 1 1 2 2
1 2 1
1 1 2
1 2
1
1
1 1 2
2
d
0 , ,
d
,
d
d
,
s
s s
s
s
s
p
p p p p p p p
p p p
P
p p p
p p
p
P
p
p p p
p
= +
c c
c
c
=
c
c
(7)
which also yields
( )
P P c c
1 2
d d 0
s
p p s . As the optimization is
o curves are
( )
sought where the tw tangent,
( )
( )
( ) ( )
( )
2
d P p
( )
( )
( )
( )
1 1 2
1 2 2 2
1
1 1
1 1 2
1 2
,
d d d
d
d d
,
d
s
p s
s
p p p X x
p p p x
p P
p p
x p p p
x p
c
= =
c
c
(8)
certainly yielding the same optimality condition as given
( )
( ) 2 1
;
p
c
in Equation (2) above.
1
For notational simplification, we denote p p X and
( )
( ) 2 1
;
s
p p P
( )
( ) 2 1
p
p
( )
( ) 2 1
s
and p , respectively. p p as
Copyright 2012 SciRes. TEL
Y. ABE, S. P. WATANABE 339
Suppose now that the functional form of P is altered
sl
0
P
ightly from the original form ( ) 1 2 0
, P p p P = which
has an additively separable form of
1 2
p p prestige = + ,
to a more generalized form ( ) ( ) 1 2
, P p p
first-order ap
1 2 0
, P p p P = + .
We then examine the effect -
proximation. The situation is depicted in Figure 1 below.
For ( ) 1 2 0
, P p p P = , the optimality condition at
of P to its
( ) 1 2
, p p
requires
( )
( ) 1 1
d 1 p = as
2
d
p
p p 1
i
P p c c = . For
( ) 1 2
, P p p = + maximize ge is ( , P p p , the d presti ) 1 2
P
attained at
0
( 1 1 2
, p p p + + ) 2
p , where the condition
requires
( )
( )
( )
( )
( )
( )
1 1
1 2
1 1
1
1 1
2
1
2
,
d
d
,
1
p
P
p
p p
p p
P
p
p
p
P
p
p
P
p
p
c
c
+ =
c
c
c
+ +
c
=
c
+ +
c
2 2
2 2
1 1 2 2
1 1 2 2
,
1 ,
p p p
p p p
p p p
p p p
+ +
+ +
+
+
(9)
The new maximum is attained on the identical possible
pr
Figure 1. Shift in the optimizing point as a result of
change in the functional form of prestige function.
estige curve, and the relation between
1
p and
2
p
is written by
( )
( ) 2 1 1 1
p p = = (10)
Expanding the Equation (9) to the first-order approxi
m
2
1
d
d
p
p
p p
p
-
ation with respect to
1
p ,
2
p , and P , and using
the result (10) gives
( )
2 1
2
2
2
1
d
d
p
P P
p p
p
p
c c
c c
= (11)
The slope of the iso-prestige c rve at
1
p
u ( ) 1 2
, p p is
( )
1 2
2
d
1
d
s
1 2 1
p P P p
P
p p p
p
c c c
= +
c
c c
c
(12)
so
1
p may be rewritten as
( )
P c
( )
2
1
1
2
2
2
1
d
1
d
d
d
s
p
p
p
p
p
p
+
= (13)
( )
Since we have
2 2
1 2
d d 0
p
p p < at the maximum
a
Equation (10)) if
( )
1 2
d d 1
s
p p < , and
1
p 0 < (and
0 p > ) for
( p is determ
slope o
) 1 2
, p p , the sign of ined by whether the
the iso-prestig urve at
1
e c f ( ) 1 2
, p p is greater or
smaller than 1 after the change in tional form P. the func
We know with certainty that nd
2
0 p
1
0 p > (a < by
2
( )
1
d d 1 p p > .
4. Impact on Functional Differentiation
ollege
omics
2
s
For a heuristic analysis with N = 2, consider a c
with only two academic departments, e.g., econ
and physics, maximizing the objective function
( ) 1 2
, P p p which explicitly involves the correlation term
( ) ( ) 1 2 1 2 1 2
, , P p p p p p p = + + (14)
( The last term ) 1 2
, p p represents the
between the two disciplines and is assumed to satisfy 1
correlation
)
( ) ( ) 2 1
0, , 0 p p 0 = = ; 2)
respect to p for 1, i
( ) ( ) 1 2 2 1
, , p p p p = ; 3)
differentiable for an appropriate number of times with
i
2 = ; and 4)
1
0 p c c > ,
2
0 p c c > .
Using th esult in Equation (7), e r
( )
( )
) ( ) (
( )
( )
( )
( )
1 1 2
1 2
1
1
1 1 2
2
,
d
1 ,
s
s
p
p
p
p p p
p
=
c
+
c
(15)
yielding
( )
( )
1
d
s
p p
p p
c
+
c
1 1 2
d d 0
s
p p p s . Since is symmetric in
i
p , i.e., ( ) ( ) 1 2 2
, , p p p p = , we have
1 1
p c c =
2
p c c for
1 2
p p = . Therefore, we obtain
( )
1
d d
s
p p
1
2
en
( )
2
s
p p = wh =
1
.
We c a situ rsity enco
ages ciplinary activities and further assume that
onsider ation in which a unive ur-
cross-dis
( ) 1 2
, p p is monotonically decreasing in
1 2
p p ,
which means that the correlation term takes larger values
els of prestige for the two disciplines e
closer. In order to represent this property more precisely,
we use the following variables instead of
1
p and
2
p :
as the lev becom
Copyright 2012 SciRes. TEL
Y. ABE, S. P. WATANABE
Copyright 2012 SciRes. TEL
340
( )
( ) 1 2
1
2
p p p
+
= +
( )
In ter s of p
and p
+
, the above-mentioned condition
1 2
1
,
2
p p p
=
16
m
n by
is give for 0 p
c c > 0 p
s and 0 p
c c s
for 0 p
> . Since
( )
( ) 1 2
1 2 p p p
c c = c c c c ,
it follows that
1 2
p p > c c c for
1 2
c p p s , and
1 2
c c fo t p p c s c r
1 2
p p > . This indicates in turn tha
( )
1 2
d d 1
s
p p s d
( )
for
1 2
p p s , an
1
s
> r
1 2
p p > .
Therefore, c ing the t with the fin -
tained in the third section, the optimizing solution with
co
s that
2
d d 1 p p fo
ombin resul ding ob
rrelation is necessarily located closer to the diagonal
array
2 1
p p = than the optimization attainable by the
prestige function without the correlation term. The result
indicate , when the conditions 0 p
c c > for
0 p
s and 0 p
c c s for 0 p