Content uploaded by L. Rachel Ngai
Author content
All content in this area was uploaded by L. Rachel Ngai on Jul 21, 2015
Content may be subject to copyright.
Content uploaded by L. Rachel Ngai
Author content
All content in this area was uploaded by L. Rachel Ngai on Jun 01, 2015
Content may be subject to copyright.
Structural Change in a Multisector Model of Growth
By L. RACHEL NGAI AND CHRISTOPHER A. PISSARIDES*
Economic growth takes place at uneven
rates across different sectors of the economy.
This paper has two objectives related to this
fact: (a) to derive the implications of different
sectoral total factor productivity (TFP)
growth rates for structural change, the name
given to the shifts in industrial employment
shares that take place over long periods of
time; and (b) to show that even with ongoing
structural change, the economy’s aggregate
ratios can be constant. We refer to the latter as
aggregate balanced growth. The restrictions
needed to yield structural change consistent
with the facts and aggregate balanced growth
are weak restrictions on functional forms that
are frequently imposed by macroeconomists
in related contexts.
We obtain our results in a baseline model
of many consumption goods and a single cap
ital good, supplied by a sector that we label
manufacturing. Our baseline results are con
sistent with the existence of intermediate
goods and many capital goods under some
reasonable restrictions. Production functions
in our model are identical in all sectors except
for their rates of TFP growth, and each sector
produces a differentiated good that enters a
constant elasticity of substitution (CES) util
ity function. We show that a low (below one)
elasticity of substitution across ﬁnal goods
leads to shifts of employment shares to sec
tors with low TFP growth. In the limit, the
employment share used to produce consump
tion goods vanishes from all sectors except
for the one with the smallest TFP growth rate,
but the employment shares used to produce
capital goods and intermediate goods con
verge to nontrivial stationary values. If the
utility function in addition has unit intertemporal
elasticity of substitution, during structural change
the aggregate capitaloutput ratio is constant and
the aggregate economy is on a balanced growth
path.
Our results contrast with the results of Cris
tina Echevarria (1997), John Laitner (2000),
Francesco Caselli and Wilbur Coleman II
(2001), and Douglas Gollin, Stephen Parente,
and Richard Rogerson (2002), who derived
structural change in a two or threesector
economy with nonhomothetic preferences.
Our results also contrast with the results of
Piyabha Kongsamut, Sergio Rebelo, and Dan
yang Xie (2001) and Reto Foellmi and Josef
Zweimuller (2005), who derived simulta
neous constant aggregate growth and struc
tural change. Kongsamut, Rebelo, and Xie
(2001) obtain their results by imposing a re
striction that maps some of the parameters of
their StoneGeary utility function onto the
parameters of the production functions, aban
doning one of the most useful conventions of
modern macroeconomics, the complete inde
pendence of preferences and technologies.
Foellmi and Zweimuller (2005) obtain their
results by assuming endogenous growth
driven by the introduction of new goods into
a hierarchic utility function. Our restrictions
are quantitative restrictions on a conventional
CES utility function that maintains the inde
pendence of the parameters of preferences
and technologies.
Our results conﬁrm William J. Baumol’s
(1967) claims about structural change. Baumol
divided the economy into two sectors, a “pro
gressive” one that uses new technology and a
“stagnant” one that uses labor as the only input.
He then claimed that the production costs and
prices of the stagnant sector should rise indeﬁ
nitely, a process known as “Baumol’s cost dis
* Ngai: Centre for Economic Performance, London School
of Economics and CEPR (email: l.ngai@lse.ac.uk); Pissar
ides: Centre for Economic Performance, London School of
Economics, CEPR, and IZA (email: c.pissarides@lse.ac.uk).
We have beneﬁted from comments received at several presen
tations (the CEPR ESSIM 2004 meetings, the SED 2004
annual conference, the NBER 2004 Summer Institute, the
2004 Canadian Macroeconomic Study Group, and at several
universities), and from Fernando Alvarez, Francesco Caselli,
Antonio Ciccone, Nobu Kiyotaki, Robert Lucas, Nick Oulton,
Danny Quah, Sergio Rebelo, Robert Shimer, Nancy Stokey,
Richard Rogerson, Jaume Ventura, and two anonymous refer
ees. Funding from the CEP, a designated ESRC Research
Centre, is acknowledged.
429
ease,” and labor should move in the direction of
the stagnant sector.
1
In the more recent empirical literature, two
competing explanations (which can coexist)
have been put forward for structural change:
our explanation, which is sometimes termed
“technological” because it attributes structural
change to different rates of sectoral TFP
growth; and a utilitybased explanation, which
requires different income elasticities for differ
ent goods and can yield structural change even
with equal TFP growth in all sectors. Baumol,
Sue Anne Batey Blackman, and Edward N.
Wolff (1985) provide empirical evidence at the
twodigit industry level, consistent with our
model. Irving B. Kravis, Alan W. Heston, and
Robert Summers (1983) also present evidence
that favors the technological explanation, at
least when the comparison is between manufac
turing and services. Two features of their data
satisﬁed by the technological explanation pro
posed in this paper are: (a) relative prices reﬂect
differences in TFP growth rates; and (b) real
consumption shares vary a lot less over time
than nominal consumption shares.
2
Our model
is also consistent with the observed positive
correlation between employment growth and
relative price inﬂation across twodigit sectors
3
and with historical OECD evidence presented
by Simon Kuznets (1966) and Angus Maddison
(1980) for onedigit sectors.
4
Section I describes our model of growth with
many sectors and Sections II and III, respec
tively, derive the conditions for structural
change and balanced aggregate growth. In Sec
tions IV and V, we study two extensions of our
baseline model, one where consumption goods
can also be used as intermediate inputs and one
where there are many capital goods. The Ap
pendix discusses the implications of one more
extension and differences in capital intensities
across sectors, and contains proofs of the main
results.
I. An Economy with Many Sectors
The baseline economy consists of an arbi
trary number of m sectors. Sectors i 1, ... ,
m 1 produce only consumption goods. The
last sector, which is denoted by m and labeled
manufacturing, produces both a ﬁnal con
sumption good and the economy’s capital
stock. We derive the equilibrium as the solu
tion to a social planning problem. The objec
tive function is
(1) U
冕
0
e
t
vc
1
, ... , c
m
dt,
where
0, c
i
0 are per capita consump

tion levels, and the instantaneous utility func
tion v is concave and satisﬁes the Inada
conditions. The constraints of the problem are
as follows.
The labor force is exogenous and growing
at rate
, and the aggregate capital stock is
endogenous and deﬁnes the state of the econ
omy. Sectoral allocations are controls that
satisfy
(2)
冘
i 1
m
n
i
1;
冘
i 1
m
n
i
k
i
k,
where n
i
0 is the employment share and k
i
0 is the capitallabor ratio in sector i, and k 0
is the aggregate capitallabor ratio. There is free
mobility for both factors.
All production in sectors i 1, ... , m 1is
consumed, but in sector m production may be
either consumed or invested. Therefore:
(3) c
i
F
i
n
i
k
i
, n
i
i m;
(4) k
˙
F
m
n
m
k
m
, n
m
c
m
k,
1
Baumol controversially also claimed that as more
weight is shifted to the stagnant sector, the economy’s
growth rate will be on a declining trend and eventually
converge to zero. This claim contrasts with our ﬁnding that
the economy is on a balancedgrowth path. We get our
result because we include capital in our analysis, ironically
left out of the analysis by Baumol (1967, 417) “primarily for
ease of exposition ... that is [in]essential to the argument.”
2
See Rodney E. Falvey and Norman Gemmell (1996) for an
update of some of their results. Falvey and Gemmell ﬁnd a unit
income elasticity and a small (negative) price elasticity for ser
vices in a cross section of countries, consistent with our results.
3
These correlations are shown in the working paper
version of this paper, Ngai and Pissarides (2004).
4
Kuznets (1966) documented structural change for 13
OECD countries and the USSR between 1800 and 1960, and
Maddison (1980) documented the same pattern for 16 OECD
countries from 1870 to 1987. They both found a pattern with
the same general features as the predictions that we obtain
when the ranking of the average historical TFP growth rates is
agriculture followed by manufacturing followed by services.
430 THE AMERICAN ECONOMIC REVIEW MARCH 2007
where
0 is the depreciation rate. Production
function F
i
( 䡠 , 䡠 ) has constant return to scale,
positive and diminishing returns to inputs, and
satisﬁes the Inada conditions.
The social planner chooses the allocation of
factors n
i
and k
i
across m sectors through a set
of static efﬁciency conditions:
(5) v
i
/v
m
F
K
m
/F
K
i
F
N
m
/F
N
i
i.
The allocation of output to consumption and
capital is chosen through a dynamic efﬁciency
condition:
(6) v˙
m
/v
m
F
K
m
,
where F
N
i
and F
K
i
are the marginal products of
labor and capital in sector i.
5
By (5), the rates
of return to capital and labor are equal across
sectors.
In order to focus on the implications of dif
ferent rates of TFP growth across sectors, we
assume production functions are identical in all
sectors except for their rates of TFP growth:
(7) F
i
A
i
Fn
i
k
i
, n
i
; A
˙
i
/A
i
i
; i.
With these production functions, we show in the
Appendix that static efﬁciency and the resource
constraints (2) imply
(8) k
i
k; p
i
/p
m
v
i
/v
m
A
m
/A
i
; i,
where p
i
is the price of good i in the decentral

ized economy.
The utility function has constant elasticities
both across goods and over time:
(9) vc
1
, ... , c
m
1
1
1
;
冉
冘
i 1
m
i
c
i
1/
冊
/ 1
,
where
, ,
i
0 and ¥
i
1. Of course, if
1, v ln
, and if 1, ln
¥
i1
m
i
ln c
i
. In the decentralized economy, de
mand functions have constant price elasticity
and unit income elasticity. With this utility
function, (8) yields
(10)
p
i
c
i
p
m
c
m
冉
i
m
冊
冉
A
m
A
i
冊
1
⬅ x
i
i.
The new variable x
i
is the ratio of consumption
expenditure on good i to consumption expendi
ture on the manufacturing good, and will prove
useful in the subsequent analysis. The intuition
behind this formula is in terms of price elastic
ities, given that all goods have unit income
elasticity. The ratio of consumption expenditures
is a weighted average of the ratio of the weights of
each good in the utility function and of their relative
prices. A higher price ratio p
i
/p
m
raises the ratio of
expenditure on good i to good m by one minus their
common price elasticity.
We also deﬁne aggregate consumption ex
penditure and output per capita in terms of
manufacturing:
(11) c ⬅
冘
i 1
m
p
i
p
m
c
i
; y ⬅
冘
i 1
m
p
i
p
m
F
i
.
Using static efﬁciency, we derive
(12) c c
m
X; y A
m
Fk,1,
where X ¥
i1
m
x
i
.
II. Structural Change
We deﬁne structural change as the state in
which at least some of the labor shares are
changing over time, i.e., n˙
i
0 for at least some
i. We derive in the Appendix (Lemma A2) the
employment shares
(13) n
i
x
i
X
冉
c
y
冊
i m,
(14) n
m
x
m
X
冉
c
y
冊
冉
1
c
y
冊
.
The ﬁrst term on the right side of (14) parallels
the term in (13) and so represents the employ
ment needed to satisfy the consumption demand
for the manufacturing good. The second brack
eted term is equal to the savings rate and rep
5
The corresponding transversality condition is lim
t3
k
exp(
0
t
(F
k
m
) d
) 0.
431VOL. 97 NO. 1 NGAI AND PISSARIDES: STRUCTURAL CHANGE IN A MULTISECTOR MODEL OF GROWTH
resents the manufacturing employment needed
to satisfy investment demand.
Conditions (13) and (14) drive our structural
change results. To see the intuition behind them,
note that by aggregation over all i, we obtain
that in our economy the employment share used
to produce consumption goods is equal to c/y,
and the employment share used to produce cap
ital goods is 1 c/y. Conditions (13) and (14)
state that the same holds for each sector i. From
(10) and (12), the consumption expenditure
share of each sector is p
i
c
i
/p
m
c x
i
/X. So the
employment share of consumption good i is
the consumption share of good i multiplied by
the employment share of total consumption.
Equivalently, the employment share of con
sumption good i is the average propensity to
consume good i : n
i
p
i
c
i
/p
m
y.
Condition (13) has the important implica
tion that the growth rate of two sectors’ relative
employment depends only on the difference be
tween the sectors’ TFP growth rates and the
elasticity of substitution between goods:
(15)
n˙
i
n
i
n˙
j
n
j
1
j
i
i, j m.
But (8) implies that the growth rate of relative
prices is:
(16)
p˙
i
p
i
p˙
j
p
j
j
i
i
and so,
(17)
n˙
i
n
i
n˙
j
n
j
1
冉
p˙
i
p
i
p˙
j
p
j
冊
i, j m.
PROPOSITION 1: The rate of change of the
relative price of good i to good j is equal to the
difference between the TFP growth rates of
sector j and sector i. In sectors producing only
consumption goods, relative employment shares
grow in proportion to relative prices, with the
factor of proportionality given by one minus the
elasticity of substitution between goods.
6
The dynamics of the individual employment
shares satisfy
(18)
n˙
i
n
i
c
˙
/y
c/y
1
i
; i m;
(19)
n˙
m
n
m
冋
c
˙
/y
c/y
(1 )(
m
)
册
c/yx
m
/X
n
m
冉
c
˙
/y
1 c/y
冊冉
1 c/y
n
m
冊
,
where
¥
i1
m
( x
i
/X)
i
is a weighted aver

age of TFP growth rates, with the weight
given by each good’s consumption share.
Equation (18) gives the growth rate in the
employment share of each consumption good
as a linear function of its own TFP growth
rate. The intercept and slope of this function
are common across sectors, but although the
slope is a constant, the intercept is in general
a function of time because both c/y and
are
in general functions of time. Manufacturing,
however, does not conform to this rule, be
cause its employment share is a weighted
average of two components, one for the pro
duction of the consumption good, which con
forms to the rule, and one for the production
of capital goods, which behaves differently.
The properties of structural change follow
immediately from (18) and (19). Consider, ﬁrst,
the case of equality in sectoral TFP growth
rates, i.e., let
i
m
@i. In this case, our
economy is one of balanced TFP growth, with
relative prices remaining constant but with
many differentiated goods. Because of the con
stancy of relative prices, all consumption goods
can be aggregated into one, so we effectively
have a twosector economy, one sector produc
ing consumption goods and one producing cap
ital goods. Structural change can still take place
in this economy, but only between the aggregate
of the consumption sectors and the capital sec
tor, and only if c/y changes over time. If c/y is
increasing over time, the investment rate is fall
ing and labor is moving out of the manufactur
ing sector and into the consumption sectors.
Conversely, if c/y is falling over time, labor is
moving out of the consumption sectors and into
manufacturing. In both cases, however, the rel
ative employment shares in consumption sec
tors are constant.
If c/y is constant over time, structural
change requires 1 and different rates of
6
All derivations and proofs, unless trivial, are collected
in the Appendix.
432 THE AMERICAN ECONOMIC REVIEW MARCH 2007
sectoral TFP growth rates. It follows imme
diately from (16), (18), and (19) that if
(c
˙
/y) 0, 1 implies constant employment
shares but changing prices. With constant em
ployment shares, fastergrowing sectors pro
duce relatively more output over time. Price
changes in this case are such that consump
tion demands exactly match all the output
changes due to the different TFP growth rates.
But if 1, prices still change as before but
consumption demands are either too inelastic
(in the case 1) to match all the output
change, or are too elastic ( 1) to be sat
isﬁed merely by the change in output due to
TFP growth. So if 1, employment has to
move into the slowgrowing sectors, and if
1, it has to move into the fastgrowing
sectors.
PROPOSITION 2: If
i
m
@i m, a
necessary and sufﬁcient condition for structural
change is c˙/c y˙/y. The structural change in
this case is between the aggregate of consump
tion sectors and the manufacturing sector.
If c˙/c y˙/y, necessary and sufﬁcient condi
tions for structural change are 1 and ?i 僆
{1, ... , m 1} s.t.
i
m
. The structural
change in this case is between all sector pairs
with different TFP growth rates. If 1, em
ployment moves from the sector with the higher
TFP growth rate to the sector with the lower
TFP growth rate; the converse is true if 1.
Proposition 2 for 1 conﬁrms the struc
tural change facts identiﬁed by Baumol, Black
man, and Wolff (1985). When demand is price
inelastic, the sectors with the low productivity
growth rate attract a bigger share of labor, de
spite the rise in their price. From the static
efﬁciency results in (8) and (12), we ﬁnd that
the nominal output shares (deﬁned as p
i
F
i
/p
m
y)
are equal to the employment shares in all sec
tors, and by (10) the nominal consumption
shares are given by x
i
/X, so the results obtained
for employment shares also hold for nominal
consumption and output shares. But real con
sumption growth satisﬁes
(20) c˙
i
/c
i
c˙
j
/c
j
i
j
; i, j,
an expression also satisﬁed by real output shares
@i, j m.
A comparison of (15) and (20) reveals that
a small can reconcile the small changes in
the relative real consumption shares with the
large changes in relative nominal consump
tion shares found by Kravis, Heston, and
Summers (1983). The authors concluded that
their ﬁnding is evidence in favor of the tech
nological explanation of structural change.
More recently, Daniel E. Sichel (1997) found
the same pattern for relative output shares,
and Falvey and Gemmell (1996) found that
the real consumption share of services (a sec
tor with low TFP growth rate) falls very grad
ually with income, both of which are
consistent with our model when 1.
III. Aggregate Growth
We now study the aggregate growth path of
this economy, with the objective of ﬁnding a
sufﬁcient set of conditions that satisfy struc
tural change as derived in the preceding sec
tion, and in addition satisfy Kaldor’s stylized
facts of aggregate growth. Recall that, for the
analysis of structural change, we imposed a
Hicksneutral technology. It is wellknown
that with this type of technology, the econ
omy can be on a steady state only if the
production function is CobbDouglas. We
therefore let F(n
i
k
i
, n
i
) k
i
n
i
,
僆 (0, 1).
7
With TFP in each sector growing at some rate
i
, the aggregate economy will also grow at
some rate related to the
i
s. The following
proposition derives the evolution of the ag
gregate economy.
PROPOSITION 3: Given any initial k(0), the
equilibrium of the aggregate economy is a path
for the pair {c, k} that satisﬁes the following two
differential equations:
(21)
k
˙
k
A
m
k
1
c
k
,
7
Daron Acemoglu and Veronica Guerrieri (2005) exam

ined the implications of different capital intensities for
economic growth and structural change. They show that
capital deepening can cause both structural change and
unbalanced growth. We examine in the Appendix the im
plications of different capital shares and a ﬁxed factor for
our model.
433VOL. 97 NO. 1 NGAI AND PISSARIDES: STRUCTURAL CHANGE IN A MULTISECTOR MODEL OF GROWTH
(22)
c˙
c
1
m
A
m
k
1
.
We deﬁne an aggregate balanced growth path
such that aggregate output, consumption, and
capital grow at the same rate. It follows from
Proposition 3 that a necessary condition for the
existence of an aggregate balanced growth path
is that the expression (
1)(
m
)bea
constant. To show this, let
(23)
1
m
⬅
constant.
Deﬁne aggregate consumption and the capital
labor ratio in terms of efﬁciency units, c
e
cA
m
1/(1
)
and k
e
kA
m
1/(1
)
, and let g
m
m
/(1
), the rate of laboraugmenting tech

nological growth in the capitalproducing sector.
The dynamic equations (21) and (22) become
(24) c˙
e
/c
e
k
e
1
/
g
m
;
(25) k
˙
e
k
e
c
e
g
m
k
e
.
Equations (24) and (25) parallel the two dif
ferential equations in the control and state of
the onesector Ramsey economy, making the
aggregate equilibrium of our manysector
economy identical to the equilibrium of the
onesector Ramsey economy when
0, and
trivially different from it otherwise. Both
models have a saddlepath equilibrium and
stationary solutions (cˆ
e
, k
ˆ
e
) that imply bal

anced growth in the three aggregates. The
capitallabor ratio is growing at the rate of
growth of laboraugmenting technological
progress in the sector that produces capital
goods, g
m
. Aggregate consumption and out

put deﬂated by the price of manufacturing
goods are also growing at the same rate.
Proposition 2 and the requirement that
be constant yield the following important
proposition.
PROPOSITION 4: Necessary and sufﬁcient
conditions for the existence of an aggregate
balanced growth path with structural change are:
1, 1;
and i 僆 1, ... , m 1 s.t.
i
m
.
Recalling the deﬁnition of
following equation
(19), Proposition 3 implies that the contribution of
each consumption sector i to aggregate equilib
rium is through its weight x
i
in
. Because each x
i
depends on the sector’s relative TFP level, the
weights here are functions of time. So
cannot be
constant during structural change and the only
way that
can be constant is through
1,
which yields
0. In this case, our aggregate
economy in c and k becomes formally identical to
the onesector Ramsey economy with growth rate
m
. There are two other conditions that give a
constant
and so yield balanced aggregate
growth:
i
m
@i or 1. But as Proposition
2 demonstrates, neither condition permits struc
tural change on the balanced growth path, where
c/y is constant.
Proposition 4 requires the utility function to be
logarithmic in the consumption composite
,
which implies an intertemporal elasticity of sub
stitution equal to one, but also to be nonlogarith
mic across goods, which is needed to yield
nonunit price elasticities. A noteworthy implica
tion of Proposition 4 is that balanced aggregate
growth does not require constant rates of growth
of TFP in any sector other than manufacturing.
Because both capital and labor are perfectly mo
bile across sectors, changes in the TFP growth
rates of consumptionproducing sectors are re
ﬂected in immediate price changes and realloca
tions of capital and labor across sectors, without
effect on the aggregate growth path.
To give intuition for the logarithmic inter
temporal utility function, we recall that bal
anced aggregate growth requires that aggregate
consumption be a constant fraction of aggregate
wealth. With our homothetic utility function, this
can be satisﬁed either when the interest rate is
constant or when consumption is independent of
the interest rate. The relevant interest rate here is
the rate of return to capital in consumption units,
which is given by the net marginal product of
capital,
y/k
, minus the change in the relative
price of the consumption composite,
m
. The
latter is not constant during structural change. In
the case 1,
is falling over time (see Lemma
A3 in the Appendix for proof), and so the real
interest rate is also falling, and converging to
y/k
. With a nonconstant interest rate, the
434 THE AMERICAN ECONOMIC REVIEW MARCH 2007
consumptionwealth ratio is constant only if con
sumption is independent of the interest rate, which
requires a logarithmic utility function.
8
Under the conditions of Proposition 4,
there is a steady state characterized by aggre
gate balanced growth, in the sense that in this
steady state the aggregate ratios are constant.
In order to achieve this balance, the aggre
gates c and y are divided by manufacturing
price, to conform to the aggregate k. If some
other price index is used as deﬂator, the rate
of growth of the aggregates is constant only if
the rate of growth of the price index is con
stant, but of course the aggregate ratios are
still constant. The published aggregate series
studied by macroeconomists usually use an
average price as deﬂator which does not have
ﬁxed weights. If the price index used to de
ﬂate national statistics is some p˜ , the pub
lished real aggregate income is y/p˜ . If the
weights used to construct p˜ are the sector
shares, p˜ changes during structural change.
But because sector shares do not change rap
idly over time, visually there is virtually noth
ing to distinguish the “stylized fact” of
constant growth in reported per capita GDP
with another “stylized fact” of constant
growth in our per capita output measure.
9
Next, we summarize the dynamics of em
ployment shares along the aggregate balanced
growth path.
PROPOSITION 5: Let sector l denote the sec
tor with the smallest TFP growth rate when
1, or the sector with the biggest TFP growth
rate when 1. On the aggregate balanced
growth path, n
l
increases monotonically.
Employment in the other sectors is either
humpshaped or declines monotonically. As
ymptotically, the economy converges to an
economy with
n
*
m
ˆ
冉
g
m
g
m
冊
;
n
*
l
1
ˆ ,
where
ˆ is the savings rate along the aggregate
balanced growth path.
Proposition 5 follows immediately from
(18)–(19) and Lemma A3. Consider the case
1, the one for 1 following by a
corresponding argument. For 1, sector i
expands if and only if its TFP growth rate is
smaller than
, and contracts if and only if its
growth rate exceeds it. But if 1, the
weighted average
is decreasing over time (see
Lemma A3 in the Appendix). Therefore, the set
of expanding sectors is shrinking over time, as
more sectors’ TFP growth rates exceed
. This
feature of the model implies that sectors with
TFP growth rates below the initial
exhibit a
humpshaped employment share, an implication
that we believe is unique to our model. These
employment shares ﬁrst rise, but once
drops
down to their own
i
, they fall.
10
In contrast to each sector’s employment
share, once the economy is on the aggregate
balanced growth path, output and consump
tion in each consumption sector grow accord
ing to
(26)
F
˙
i
F
i
A
˙
i
A
i
k
˙
i
k
n˙
i
n
i
i
g
m
1
.
If 1, the rate of growth of consumption and
output in each sector is positive (provided
i
0), and so sectors never vanish, even though
their employment shares in the limit may van
ish. If 1, the rate of growth of output may
8
After reexamining the evidence, Robert Barro and
Xavier SalaiMartin (2004, 13) concluded, consistent with
our model, “it seems likely that Kaldor’s hypothesis of a
roughly stable real rate of return should be replaced by a
tendency for returns to fall over some range as an economy
develops.” In our model, it is converging from above to a
positive value.
9
Nicholas Kaldor (1961, 178) spoke of a “steady trend
rate” of growth in the “aggregate volume of production.” In
Ngai and Pissarides (2004, ﬁg. 4) we plot our series of per
capita real incomes and the published chainweighted series
for the United States since 1929, and show that they are
virtually indistinguishable from each other.
10
Maddison (1980, 48), in his study of historical OECD
data, found a “shallow bell shape” for manufacturing em
ployment for each of the 16 OECD countries, which can be
reproduced by our model if the manufacturing TFP growth
rate takes values between the TFP growth rates of agricul
ture and services.
435VOL. 97 NO. 1 NGAI AND PISSARIDES: STRUCTURAL CHANGE IN A MULTISECTOR MODEL OF GROWTH
be negative in some lowgrowth sectors, and
since by Lemma A3
is rising over time in this
case, their rate of growth remains indeﬁnitely
negative until they vanish.
Finally, we examine brieﬂy the implications of
1. When
1, balanced aggregate growth
cannot coexist with structural change, because the
term
(
1)(
m
) in the Euler condition
(24) is a function of time. But as shown in the
Appendix Lemma A3,
is monotonic. As t 3 ,
converges to the constant (
1)(
m
l
),
where
l
is the TFP growth rate in the limiting
sector (the slowest or fastest growing consump
tion sector depending on whether or 1).
Therefore, the economy with
1 converges to
an asymptotic steady state with the same growth
rate as the economy with
1.
What characterizes the dynamic path of the
aggregate economy when
1? By differen
tiation and using Lemma A3, we obtain
(27)
˙
11
冘
i 1
m
x
i
/X
i
2
,
which is of second order compared with the
growth in employment shares in (15), given that
the
s are usually small numbers centered
around 0.02. Therefore, the rate of growth of the
economy during the adjustment to the asymp
totic steady state with
1 is very close to the
constant growth rate of the economy with
1, despite ongoing structural change in both
economies.
IV. Intermediate Goods
Our baseline model has no intermediate
inputs and has only one sector producing cap
ital goods. We now generalize it by introduc
ing intermediate inputs and (in the next
section) by allowing an arbitrary number of
sectors to produce capital goods. The key
difference between intermediate goods and
capital goods is that capital goods are reus
able, while intermediate goods depreciate
fully after one usage. The motivation for the
introduction of intermediate inputs is that
many of the sectors that may be classiﬁed as
consumption sectors in fact produce for busi
nesses. Business services is one obvious ex
ample. Inputoutput tables show that a large
fraction of output in virtually all sectors of the
economy is sold to businesses.
11
As in the baseline model, sectors are of two
types. The ﬁrst type produces perishable goods
that are either consumed by households or used
as intermediate inputs by ﬁrms. We continue
referring to these sectors as consumption sec
tors. The second type of sector produces goods
that can be used as capital. For generality’s
sake, we assume that the output of the capital
producing sector can also be processed into both
consumption goods and intermediate inputs.
The output of consumption sector i is now
c
i
h
i
, where h
i
is the output that is used as
an intermediate good. Manufacturing output
can be consumed, c
m
, used as an intermediate
input, h
m
, or used as new capital, k
˙
.We
assume that all intermediate goods, h
i
, are
used as an input into an aggregate CES pro
duction function (h
1
, ... , h
m
) [¥
i1
m
i
h
i
(
1)/
]
/(
1)
which produces a single
intermediate good , with
0,
i
0, and
¥
i
1. The production functions are mod

iﬁed to F
i
A
i
n
i
k
i
q
i
, @i, where q
i
is the
ratio of the intermediate good to employment
in sector i and
is its input share, with
,
0 and
1. When
0, we return to
our baseline model. We show in the Appendix
that a necessary and sufﬁcient condition for
an aggregate balanced growth path with struc
tural change is
1, i.e., should be
CobbDouglas.
12
When is CobbDoug
las, our central results from the baseline
model carry through, with some modiﬁca
tions.
The aggregate equilibrium is similar to the
one in the baseline model:
(28)
c˙
c
Ak
1/1
,
11
According to inputoutput tables for the United States,
in 1990 the percentage distribution of the output of two
digit sectors across three types of usage, ﬁnal consumption
demand, intermediate goods, and capital goods, was 43, 48,
and 9, respectively. In virtually all sectors, however, a large
fraction of the intermediate goods produced are consumed
by the same sector.
12
Nicholas Oulton (2001) claims that if there are inter

mediate goods, and if the elasticity of substitution between
the intermediate goods and labor is bigger than one, Bau
mol’s “stagnationist” results could be overturned (in the
absence of capital). No such possibility arises with Cobb
Douglas production functions.
436 THE AMERICAN ECONOMIC REVIEW MARCH 2007
(29)
k
˙
k
1
Ak
1/1
c
k
,
where A [A
m
(
m
)
]
1/(1
)
and
m
is the
marginal product of the manufacturing good in
. The growth rate of A is constant and equal to
m
(
¥
i1
m
i
i
)/(1
), where
i
is the
input share of sector i in . Therefore, we can
deﬁne aggregate consumption and the aggregate
capitallabor ratio in terms of efﬁciency units
and obtain an aggregate balanced growth path
with growth rate (
m
¥
i1
m
i
i
)/(1
), which is the sum of laboraugmenting tech
nological growth in the capitalproducing sector
and a
fraction of laboraugmenting techno
logical growth in all sectors that produce inter
mediate goods. Recall the aggregate growth rate
in the baseline model depended only on the TFP
growth rate in manufacturing. In the extended
model with intermediate goods, the TFP growth
rates in all sectors contribute to aggregate growth.
The employment shares (13) and (14) are
now modiﬁed to
(30) n
i
x
i
X
冉
c
y
冊
i
; i m;
(31) n
m
冋
x
m
X
冉
c
y
冊
m
册
冉
1
c
y
冊
.
For the consumption sectors, the extra term in
(30) captures the employment required for pro
ducing intermediate goods.
i
is the share of
sector i’s output used for intermediate purposes
and
is the share of the aggregate intermediate
input in aggregate output. For the manufactur
ing sector, the terms in the ﬁrst bracket parallel
those of the consumption sectors. The second
term captures the employment share for invest
ment purposes.
Our results on structural change now hold for
the component of employment used to produce
consumption goods, ( x
i
/X)(c/y). The deﬁnition
of x
i
and X is the same as in the absence of
intermediate goods. The contribution of inter
mediate goods to sectoral employment dynam
ics is the addition of the constant employment
share
i
, with no impact on the other two
components of employment. Following on from
this, the asymptotic results in Proposition 5 are
also modiﬁed. Asymptotically, the employment
share used for the production of consumption
goods still vanishes in all sectors except for the
slowest growing one (when 1), but the
employment share used to produce intermediate
goods,
i
, survives in all sectors.
V. Many Capital Goods
In our second extension we allow an arbitrary
number of sectors to produce capital goods. We
study this extension with the baseline model
without intermediate inputs.
We suppose that there are
different capital
producing sectors, each supplying the inputs
into a production function G, which produces a
capital aggregate that can be either consumed or
used as an input in all production functions F
i
.
Thus, the model is the same as before, except
that now the capital input k
i
is not the output of
a single sector but that of the production func
tion G. The Appendix derives the equilibrium
for the case of a CES function with elasticity
,
i.e., when G [¥
j1
m
j
(F
m
j
)
(
1)/
]
/(
1)
,
where
0,
m
j
0 and F
m
j
is the output of
each capital goods sector m
j
. G now replaces the
output of the “manufacturing” sector in our
baseline model, F
m
.
It follows immediately that the structural
change results derived for the m 1 consump
tion sectors remain intact, as we have made no
changes to that part of the model. But there are
new results to derive concerning structural
change within the capitalproducing sectors.
The relative employment shares across the
capitalproducing sectors satisfy
(32) n
m
j
/n
m
i
m
j
/
m
i
A
m
i
/A
m
j
1
;
i, j 1, ... ,
.
(33)
n˙
m
j
n
m
j
n˙
m
i
n
m
i
1
m
i
m
j
;
i, j 1, ... ,
.
These equations parallel (13) and (15) of the
baseline model and the intuition behind them is
the same.
When there are many capital goods, the A
m
of
437VOL. 97 NO. 1 NGAI AND PISSARIDES: STRUCTURAL CHANGE IN A MULTISECTOR MODEL OF GROWTH
the baseline model is replaced by G
m
j
A
m
j
for
each sector m
j
, where G
m
j
denotes the sector’s
marginal product in the production of aggregate
capital, and A
m
j
is the sector’s TFP level. This
term measures the rate of return to capital in the
jth capitalproducing sector, which is equal
across all
sectors because of the free mobility
of capital. In the Appendix we derive the ag
gregate growth rate
(34)
m
冘
j 1
j
m
j
;
j
⬅
m
j
A
m
j
1
冒
冉
冘
i 1
m
i
A
m
i
1
冊
,
which is a weighted average of TFP growth
rates in all capitalproducing sectors. The dy
namic equations for c and k are the same as in
the baseline model, given the new deﬁnition
of
m
.
If TFP growth rates are equal across all
capitalproducing sectors, c and k grow at a
common rate in the steady state. But then all
capital producing sectors can be aggregated into
one, and the model reduces to one with a single
capitalproducing sector. If TFP growth rates
are different across the capitalproducing sec
tors and
1, there is structural change within
the capitalproducing sectors along the transi
tion to the asymptotic state. Asymptotically,
only one capitalproducing sector remains. In
the asymptotic state, c and k again grow at a
common rate, so there exists an asymptotic ag
gregate balanced growth path with only one
capitalproducing sector.
A necessary and sufﬁcient condition for the
coexistence of an aggregate balanced growth
path and multiple capitalproducing sectors
with different TFP growth rates is
1. The
reason for this result is that a balanced aggre
gate path requires a constant
m
, which is un

attainable if the relative TFP levels in the
capitalproducing sectors are allowed to inﬂu
ence it. From (34), the inﬂuence of the produc
tivity levels disappears only when
1. The
aggregate growth rate in this case is
m
/(1
),
where
m
¥
j1
m
j
m
j
. Using (32), the relative
employment shares across capitalproducing
sectors are equal to their relative input shares in
G. There is no structural change within the
capitalproducing sectors, their relative employ
ment shares remaining constant independently
of their TFP growth rates.
The model with 1 and
1 has clear
contrasting predictions about the relation be
tween the dynamics of sectoral employment
shares and TFP growth (or relative prices). Sec
tors that produce primarily consumption goods
should exhibit a welldeﬁned linear relation be
tween their employment share growth and their
TFP growth rate; sectors that produce many
intermediate goods should still have a positive
linear relation, but less welldeﬁned because of
the constant term due to the production of in
termediate goods. But sectors that produce pri
marily capital goods should exhibit no linear
relation at all between their employment share
growth and their relative TFP growth rate.
VI. Conclusion
We have shown that different TFP growth
rates across industrial sectors predict sectoral
employment changes that are consistent with
the facts if the substitutability between the ﬁnal
goods produced by each sector is low. Balanced
aggregate growth requires, in addition, a loga
rithmic intertemporal utility function. Underly
ing the balanced aggregate growth there is a
shift of employment away from sectors with a
high rate of technological progress toward sec
tors with low growth, and eventually, in the
limit, all employment converges to only two
sectors, the sector producing capital goods and
the sector with the lowest rate of productivity
growth. If the economy also produces interme
diate goods, the sectors that produce these
goods also retain some employment in the limit,
which is used to produce the intermediate
goods.
Our results are consistent with the observa
tion of simultaneous growth in the relative
prices and employment shares of stagnant sec
tors such as community services, with the near
constancy of real consumption shares when
compared with nominal shares. It is also con
sistent with the longrun evidence of Kuznets
(1966) and Maddison (1980) concerning the
decline of agriculture’s employment share, the
rise and then fall of the manufacturing share,
and the rise in the service share. The key re
quirement for these results is again a low sub
stitutability between ﬁnal goods. Of course, at a
438 THE AMERICAN ECONOMIC REVIEW MARCH 2007
ﬁner sector decomposition, the elasticity of sub
stitution between two goods may reasonably
exceed unity, as for example between the output
of the sector producing typewriters and the out
put of the sector producing electronic word pro
cessors. Our model in this case predicts that
labor would move from the sector with low TFP
growth to the one with high TFP growth. The
approach that we suggested for intermediate and
many capital goods, namely the existence of
subsectors that produce an aggregate that enters
the utility or production function, is an obvious
approach to the analysis of these cases. Within
the subsectors, there is structural change toward
the high TFP goods, but between the aggregates
the ﬂow is from high to low TFP sectors.
A
PPENDIX:PROOFS
LEMMA A1: Equations (2), (5), and (7) imply equation (8).
PROOF:
Deﬁning f(k) F(k, 1), and omitting subscript i, (7) implies F
K
Af(k) and F
N
A[ f(k)
kf(k)]. So F
N
/F
K
f(k)/f(k) k, which is strictly increasing in k. Hence, (5) implies k
i
k
m
@i
m, and together with (2), results follow.
LEMMA A2: @i m, n
i
satisfy (13) and (18), and n
m
satisﬁes (14) and (19).
PROOF:
n
i
follows from substituting F
i
into (10), and n
m
is derived from (2). Given x˙
i
/x
i
(1 )(
m
i
) and X
˙
/X (1 )(
m
), the result follows for n˙
i
, i m. Using (2),
n˙
m
冘
im
n˙
i
冉
.
c/y
冊
c/y
1 n
m
1
冉
c/y
X
冊
冘
im
x
i
i
,
so result follows for n˙
m
by substituting n
m
.
PROPOSITION 3
PROOF:
Use (2) and (8) to rewrite (4) as
k
˙
/k A
m
k
1
冉
1
冘
im
n
i
冊
c
m
/k
.
But p
i
/p
m
A
m
/A
i
and by the deﬁnition of c,
k
˙
/k A
m
k
1
c/k
.
Next,
is homogenous of degree one:
冘
i 1
m
i
c
i
冘
i 1
m
p
i
c
i
m
/p
m
m
c/p
m
.
But
m
m
(
/c
m
)
1/
and c c
m
X; thus
m
m
/(1)
X
1/( 1)
and v
m
m
(
m
/(1)
X
1/( 1)
)
1
c
, so (6) becomes (22).
439VOL. 97 NO. 1 NGAI AND PISSARIDES: STRUCTURAL CHANGE IN A MULTISECTOR MODEL OF GROWTH
LEMMA A3: d
/dt 0 N 1.
PROOF:
Totally differentiating
as deﬁned in equations (18) and (19),
d
/dt
冘
i 1
m
x
i
/X
i
冉
x˙
i
/x
i
冘
j 1
m
x˙
j
/X
冊
;
1
冘
i 1
m
x
i
/X
i
冋
m
i
冘
j 1
m
( x
j
/X)(
m
j
)
册
;
1
冉
2
冘
i 1
m
( x
i
/X)
i
2
冊
1
冘
i 1
m
x
i
/X
i
2
.
Since the summation term is always positive, the result follows.
Capital Shares and Fixed Factors.—We now discuss the structural change results (equations (8)
and (13)–(17)) when capital shares are different across sectors and there is a ﬁxed factor of
production in at least one sector. The production function is F
i
A
i
k
i
i
z
i
1
i
i
n
i
, where
i
is labor
share and z
i
Z
i
/N
i
is a ﬁxed factor per unit of labor. Suppose
i
i
1 for i 1, and
1
1
1, i.e., the ﬁxed factor is used in sector 1 only. Static efﬁciency implies
n
i
/n
j
i
/
j
i
/
m
p
i
/p
j
1
, i, j m,
so the result that relative employment shares grow in proportion to relative prices, equation (17), is
independent of different factor shares and the existence of a ﬁxed factor. Our other results are
modiﬁed as follows. The static efﬁciency condition (8) is replaced by
k
i
i
k
m
, p
i
/p
m
A
m
/A
i
m
/
i
k
m
m
i
z
i
i
i
1
;
i
⬅
m
i
i
m
,
i
⬅
i
i
j
1
i
.
Different capital shares add the term (
m
i
)k
˙
m
/k
m
in (16). In a growth equilibrium with k
m
growing, lower
i
is another reason for higher relative price in sector i. Combining the relative price
and relative employment equations, different capital shares add the term (1 )(
j
i
)k
˙
m
/k
m
in
(15). The existence of a ﬁxed factor modiﬁes (15) to
1 1
1
1
1
n˙
1
n
1
n˙
j
n
j
1
j
1
1
j
1
k
˙
m
k
m
, j m.
If n
1
is falling, then the presence of a ﬁxed factor implies that n
1
falls at a faster rate. Finally, (13)
and (14) are modiﬁed to
n
i
x
i
X
c
y
冘
j
i
j
n
j
, i m; n
m
x
m
X
c
y
冘
j
i
j
n
j
1
c
y
冘
j
i
j
n
j
,
where c Xc
m
, y A
m
k
m
m
¥
i
(
m
n
i
/
i
), and x
i
(
i
/
m
)
[
i
k
m
m
i
z
i
i
i
1
( A
m
/A
i
)]
1
. The new
system implies n
1
, ... , n
m
can be solved simultaneously.
440 THE AMERICAN ECONOMIC REVIEW MARCH 2007
Intermediate Goods.—@i, F
i
A
i
n
i
k
i
q
i
,
,
僆 (0, 1),
1. We have
(A1) F
m
c
m
h
m
k k
˙
, F
i
c
i
h
i
, i m.
The planner’s problem is similar to the baseline with (A1) replacing (3) and (4), {h
i
, c
i
, q
i
}
i1, ... ,
m
as additional controls, and ¥
i1
m
n
i
q
i
(h
1
, ... , h
m
) as an additional constraint, where is
homogenous of degree one,
i
0, and
ii
0. The static efﬁciency conditions are
(A2) v
i
/v
m
F
K
m
/F
N
i
F
N
m
/F
N
i
F
Q
m
/F
Q
i
i
/
m
; i,
which implies k
i
k, q
i
, and p
i
A
m
/A
i
@i,so
y A
m
k
,
冘
i 1
m
i
h
i
冘
i 1
m
m
p
i
h
i
m
h; h ⬅
冘
i 1
m
p
i
h
i
.
Optimal conditions for h
m
and q
m
imply
m
A
m
k
1
1, so h
y and (A1) is
k
˙
A
m
k
冉
1
冘
im
n
i
冊
h
m
c
m
k h1
/
c
k.
The dynamic efﬁciency condition is v˙
m
/v
m
A
m
k
1
(
), so
(A3) c˙ /c
h/
k
, k
˙
/k 1
h/
k c/k
.
Constant c˙/c requires constant h/k, and constant k
˙
/k requires constant c/k. Thus, h
˙
/h must be constant.
To derive constant h
˙
/h, consider a CES (¥
i1
m
i
h
i
(
1)/
)
/(
1)
; then (A2) implies z
i
p
i
h
i
/h
m
(
i
/
m
)
( A
m
/A
i
)
1
, @i.So
h Zh
m
,
m
m
/
1
Z
1/
1
,
A
m
k
m
/
1
Z
1/
1
1/1
,
where Z ¥
i1
m
z
i
. Hence,
h /
m
A
m
k
1/1
m
/
1
Z
1/
1
/1
,
and so
1
h
˙
/h
m
k
˙
/k
冉
冘
i 1
m
(z
i
/Z)
i
m
冊
,
constant if ¥
i1
m
z
i
i
is constant. Given
i
differs across all i, constancy requires
1, so
写
i 1
m
h
i
i
, Z 1/
m
, z
i
i
/
m
i.
(A2) implies h
m
i1
m
(z
i
A
i
/A
m
)
i
and so
m
m
/h
m
i1
m
(
i
A
i
/A
m
)
i
. But
[
A
m
k
m
]
1/(1
)
,soh /
m
(
A
m
k
)
1/(1
)
m
/(1
)
. (A3) becomes
c˙ /c
Ak
/1
1
; k
˙
c
k 1
Ak
/1
,
441VOL. 97 NO. 1 NGAI AND PISSARIDES: STRUCTURAL CHANGE IN A MULTISECTOR MODEL OF GROWTH
where A [A
m
(
m
)
]
1/(1
)
. Deﬁne
c
e
⬅ cA
1
/1
; k
e
⬅ kA
1
/1
;
⬅ A
˙
/A.
We have
[
m
¥
i1
m
i
(
i
m
)]/(1
)
m
(
¥
i1
m
i
i
)/(1
), and
c˙
e
/c
e
k
e
1/1
g; k
˙
e
1
k
e
/1
c
e
gk
e
,
which imply the existence and uniqueness of an ABGP with growth rate
g ⬅ 1
/1
冉
m
冘
i 1
m
i
i
冊
冒
1
.
We obtain n
i
using F
i
c
i
h
i
, @i m, i.e.,
A
i
n
i
k
p
i
p
i
c
i
h
i
x
i
c
m
z
i
h
m
cx
i
/X
i
h.
Substitute p
i
and h to obtain n
i
y cx
i
/X
i
y, so (30) and (31) follow.
Many CapitalProducing Sectors.—@j, F
m
j
A
m
j
n
m
j
k
m
j
, which together produce good m through
G
冋
冘
j 1
m
j
(F
m
j
)
1/
册
/
1
,
m
j
0,
0,
冘
j 1
m
j
1.
The planner’s problem is similar to the baseline model with (4) replaced by
k
˙
G c
m
k
and (k
m
j
, n
m
j
)
j1, ... ,
as additional controls. The static efﬁciency conditions are
F
K
i
/F
N
i
F
K
m
j
/F
N
m
j
, i m, j,
so k
i
k
m
j
k. Also
G
m
j
/G
m
i
F
K
m
i
/F
K
m
j
A
m
i
/A
m
j
, i, j,
which implies n
m
j
/n
m
i
(
m
j
/
m
i
)
( A
m
i
/A
m
j
)
1
and grows at rate (1
)(
m
i
m
j
). Let n
m
¥
j1
n
m
j
; we have n
m
n
m
1
¥
j1
(
m
j
/
m
1
)
( A
m
1
/A
m
j
)
1
. Next,
p
i
v
i
/v
m
A
m
/A
i
, i m,
where A
m
G
m
1
A
m
1
. Thus, n
i
/n
j
and p
i
/p
j
are the same as in the baseline.
To derive the aggregate equilibrium, note that G ¥
j1
F
m
j
G
m
j
A
m
k
n
m
,soc˙/c and k
˙
/k are the
same as the baseline, so the equilibrium is the same as the baseline if
m
A
˙
m
/A
m
is constant, which
we now derive. Given
G
m
1
m
1
G/F
m
1
1/
; G/F
m
1
冋
冘
j 1
m
j
(A
m
j
n
m
j
/(A
m
1
n
m
1
))
1/
册
/
1
,
442 THE AMERICAN ECONOMIC REVIEW MARCH 2007
using the result on n
m
j
/n
m
1
we have G/F
m
1
[¥
j1
m
j
(
m
1
A
m
1
)
1
A
m
j
(
1)
]
/(
1)
, thus A
m
G
m
1
A
m
1
[¥
j1
m
j
A
m
j
(
1)
]
1/(
1)
,so
m
冘
j 1
j
m
j
,
j
⬅
m
j
A
m
j
1
冒
冉
冘
j 1
m
j
A
m
j
1
冊
,
constant if (
1) ¥
j1
j
(
m
j
m
)
2
0, i.e., if (1)
m
i
m
j
, @i, j,or(2)
1. If (1) is true,
the model reduces to only one capitalproducing sector. Thus, coexistence of multiple capital
producing sectors and an ABGP requires (2), i.e., G
j1
(F
m
j
)
j
and
m
¥
j1
m
j
m
j
.
REFERENCES
Acemoglu, Daron, and Veronica Guerrieri. 2005.
“Capital Deepening and NonBalanced Eco
nomic Growth.” Unpublished.
Barro, Robert J., and Xavier SalaiMartin. 2004.
Economic Growth: Second Edition. Cam
bridge, MA and London: MIT Press.
Baumol, William J. 1967. “Macroeconomics of
Unbalanced Growth: The Anatomy of Urban
Crisis.” American Economic Review, 57(3):
415–26.
Baumol, William J., Sue Anne Batey Blackman,
and Edward N. Wolff.
1985. “Unbalanced
Growth Revisited: Asymptotic Stagnancy
and New Evidence.” American Economic Re
view, 75(4): 806 –17.
Caselli, Francesco, and Wilbur John Coleman II.
2001. “The U.S. Structural Transformation
and Regional Convergence: A Reinterpreta
tion.” Journal of Political Economy, 109(3):
584 – 616.
Echevarria, Cristina. 1997. “Changes in Sectoral
Composition Associated with Economic
Growth.” International Economic Review,
38(2): 431–52.
Falvey, Rodney E., and Norman Gemmell. 1996.
“Are Services IncomeElastic? Some New
Evidence.” Review of Income and Wealth,
42(3): 257– 69.
Foellmi, Reto, and Josef Zweimuller. 2005.
“Structural Change, Engel’s Consumption
Cycles and the Kaldor’s Facts of Economic
Growth.” Unpublished.
Gollin, Douglas, Stephen Parente, and Richard
Rogerson.
2002. “The Role of Agriculture in
Development.” American Economic Review
(Papers and Proceedings), 92(2): 160 – 64.
Kaldor, Nicholas. 1961. “Capital Accumulation
and Economic Growth.” In The Theory of
Capital, ed. F. A. Lutz and Douglas C.
Hague, 177–222. New York: St. Martin’s
Press.
Kongsamut, Piyabha, Sergio Rebelo, and Dan
yang Xie.
2001. “Beyond Balanced Growth.”
Review of Economic Studies, 68(4): 869 – 82.
Kravis, Irving B., Alan W. Heston, and Robert
Summers.
1983. “The Share of Services in
Economic Growth.” In Global Econometrics:
Essays in Honor of Lawrence R. Klein, ed. F.
Gerard Adams and Bert G. Hickman, 188 –
218. Cambridge, MA, and London: MIT
Press.
Kuznets, Simon. 1966. Modern Economic
Growth: Rate, Structure, and Spread. New
Haven: Yale University Press.
Laitner, John. 2000. “Structural Change and
Economic Growth.” Review of Economic
Studies, 67(3): 545– 61.
Maddison, Angus. 1980. “Economic Growth and
Structural Change in the Advanced Coun
tries.” In Western Economies in Transition:
Structural Change and Adjustment Policies
in Industrial Countries, ed. Irving Leveson
and Jimmy W. Wheeler, 41– 65. London:
Croom Helm.
Ngai, L. Rachel, and Christopher Pissarides.
2004. “Structural Change in a MultiSector
Model of Growth.” Center for Economic Pol
icy Research Discussion Paper 4763.
Oulton, Nicholas. 2001. “Must the Growth Rate
Decline? Baumol’s Unbalanced Growth Re
visited.” Oxford Economic Papers, 53(4):
605–27.
Sichel, Daniel E. 1997. “The Productivity Slow
down: Is a Growing Unmeasurable Sector the
Culprit?” Review of Economics and Statis
tics, 79(3): 367–70.
443VOL. 97 NO. 1 NGAI AND PISSARIDES: STRUCTURAL CHANGE IN A MULTISECTOR MODEL OF GROWTH