there have to be major shifts in the distribution of factor inputs over time. Second, productivity levels and growth rates must differ considerably across manufacturing
branches. It is well known that in the East-Asian countries profound changes have taken place in the industrial structure since 1963, with an increasing share of heavy
industries and especially of the electrical machinery branch and a declining share of food manufacturing and textiles. In India and Indonesia, industrial growth did not
typically involve a change from light to heavy industries. Due to the emphasis on self-reliance after independence, India had already developed a sizeable heavy
industry in the 1960s, alongside a large textile sector the importance of which dwindled in the 1980s. In Indonesia the industrial sector only started to develop in
the 1970s. In the 1990s, the larger part of manufacturing output is still being produced in the light industries. Within light industry, the importance of food and
rubber manufacturing declined rapidly in favour of growth in branches such as textiles and wearing apparel.
3
Huge factor shifts combined with a large variance of productivity levels and growth rates across branches suggest a potentially important
role of structural change as a source of productivity growth.
4. Impact of structural change on labour productivity growth
To assess the contribution of structural change resource reallocation among sectors to the growth of aggregate productivity, a method known as the shift-share
method is often used see Syrquin 1984 for an overview and Paci and Pigliaru 1997 or Fagerberg 1999 for recent applications. In the shift-share analysis,
aggregate productivity growth is decomposed into effects due to productivity growth within sectors and effects due to structural change. The decomposition was
pioneered by Fabricant 1942 who was interested in measuring labour require- ments per unit of output, but later users of his method focused more on its
reciprocal: labour productivity.
Let LP denote the labour productivity level, subscript i denote manufacturing branches i = 1, . . . n, with n the number of branches, S
i
the share of branch i in total manufacturing employment and superscripts 0 and T the begining and end of
the period 0, T. Then aggregate labour productivity at time T can be written as: LP
T
= Y
T
L
T
=
n i = 1
Y
i T
L
i T
L
i T
L
T
=
n i = 1
LP
i T
S
i T
1 Using Eq. 1, the difference in aggregate labour productivity levels at time 0 and
T can be written as: LP
T
− LP
=
n i = 1
LP
i T
− LP
i
S
i
+
n i = 1
S
i T
− S
i
LP
i
+
n i = 1
S
i T
− S
i
LP
i T
− LP
i
2
3
See Timmer 2000 Chapter 7, for a detailed description of industrial development patterns in these countries.
Dividing both sides of Eq. 2 by LP , it follows that aggregate productivity
growth can be decomposed into intra-branch productivity growth the first term on the right-hand side and the effects of structural change which consist of a static
shift effect the second term and a dynamic shift effect the third term. Whereas the static shift effect measures productivity growth caused by a shift of labour
towards branches with a higher labour productivity level at the beginning of the period, the dynamic shift effect captures shifts towards more dynamic branches, i.e.
branches with higher labour productivity growth rates. The last term is an interaction effect that arises because of the use of a discrete fixed weight decompo-
sition. One could use mean weights to eliminate this term, as in Syrquin 1984, but we retain it because this term can be given an interesting economic interpretation.
As branches differ not only in terms of productivity levels, but also in terms of productivity growth rates, resource reallocation has both static and dynamic effects
and a distinction between the two is useful.
4
The results of the decomposition of aggregate manufacturing labour productivity growth are reported in Table 1. For each country and each of the four subperiods,
the annual growth rate of labour productivity in total manufacturing is given in the first column. The other columns show the percentages of this growth, which are
explained by growth in labour productivity within branches and by shifts in labour shares across branches. Note that a change in labour shares does not necessarily
involve a physical transfer of employees from one sector to another. This depends on the overall growth of the manufacturing labour force. In an expanding manufac-
turing sector, shares may change while employment is increasing in all branches.
The overriding conclusion derived from Table 1 is that the structural-bonus hypothesis for labour productivity should be rejected for the manufacturing sector.
Labour reallocation has been unimportant in explaining labour productivity growth in aggregate manufacturing. On the contrary, structural change often involved a
shift of labour to branches that had both lower productivity growth rates and levels. As such, in most periods it was a drag on aggregate labour productivity
growth, rather than a bonus.
5
This is most clear in the case of Indonesia. During 1975 – 1993, labour shares in branches with above-average labour productivity levels
rapidly declined, as comparative advantage was realised in more labour-intensive manufacturing activities. Structural change decreased labour productivity growth
by : 15 in the whole period. For India, the effects of structural change were positive but small. A major shift of labour to the chemical branch took place and
this contributed positively to aggregate labour productivity growth in the 1970s 15, but over the period 1973 – 1993 structural change contributed only 9.
In South Korea and Taiwan, profound changes in the structure of the manufac- turing sector have taken place and labour productivity grew at a rapid pace.
4
The size of the interaction effect will of course depend on the length of the period under consideration because it vanishes when the length approaches 0 see next footnote also.
5
This result is robust to the various ways in which the structural decomposition formula is applied. It can be applied to annual data or to data for the beginning- and end-year of a period. The results of
the latter are reported in Table 1. Using annual data, the positive effects of labour shifts on aggregate productivity growth become even smaller.
377 M
.P .
Timmer ,
A .
Szirmai Structural
Change and
Economic Dynamics
11 2000
371 –
392
Table 1 Decomposition of annual compound labour productivity growth in aggregate manufacturing based on 13 branches
a
Labour productivity growth annual Percentage of labour productivity growth explained by:
Total effect Intra-branch effect
Static shift effect Dynamic shift effect
India, registered sector 13
85 2
1.7 1973–1982
100 3
100 86
1982–1987 10
7.1 7.0
100 95
1 1987–1993
4 3
1973–1993
b
4.6 100
6 91
Indonesia, medium and large scale sector 57
100 100
1975–1982 −
56 0.7
3 100
1982–1987 3.8
103 −
6 −
9 100
1987–1993 6.5
120 −
12 3.5
− 2
100 −
13 1975–1993
b
115 South Korea, firms with fi6e employees or more
10.1 100
1963–1973 125
− 21
− 3
100 6.0
108 −
6 −
2 1973–1982
8 100
9.0 1982–1987
94 −
3 1987–1993
13.3 99
2 −
1 100
− 2
− 1
103 100
1963–1993
b
9.3 Taiwan, all firms
− 11
100 131
1963–1973 −
19 7.9
− 7
100 1973–1982
105 4.3
2 −
7 100
− 2
108 1982–1987
5.9 100
5.3 80
14 7
1987–1993 −
3 100
6.0 1963–1993
b
104 −
1
a
Decomposition of labour productivity growth into part due to labour productivity growth in branches intra-branch effect and shift of labour between branches shift effects using Eq. 2. Percentages may not add to 100 due to rounding.
b
The decomposition for the total period is given by a weighted sum of the sub-periods’ effects. Sources: India
:
Value added and number of persons employed from CSO, Annual Survey of Industries, annual issues. Deflators from CSO, National Accounts Statistics, various issues; Indonesia
:
Value added and number of persons employed from BPS, Printout on revised Statistik Industri data, September, 1997. Deflators from BPS, Indikator Ekonomi, various issues; South Korea
:
Value added and number of persons employed from EPB, Report on Mining and Manufacturing Survey, various issues; Deflators from Bank of Korea, National Accounts, various issues; Taiwan
:
Value added at constant prices from DGBAS, National Income in Taiwan Area of the Republic of China 1994. Number of persons employed from DGBAS, Monthly Bulletin of
Earnings and Productivity Statistics, various issues.
However, even for these countries shifts of labour were not important in explaining aggregate labour productivity growth. The contribution of labour shifts in the period
1963 – 1993 was even slightly negative in both countries, though the impact varied during the different phases of development. Especially in the early period 1963 –
1973, labour shifts had large negative effects, both static and dynamic. Also in the period after 1973, structural change did not play an important role in determining
aggregate labour productivity growth, whether positively or negatively, with the exception of Taiwan in the most recent period 1987 – 1993.
6
The finding of negligible or even negative contributions of structural change to aggregate labour productivity growth in manufacturing is not typical only for the
developing countries studied here. Dollar and Wolff 1993 in Chapter 8 found similar results for the manufacturing sectors in Brazil, Hong Kong, Singapore and Thailand.
Within manufacturing, there is no structural bonus comparable to that involved in the shift from agriculture to industry. However, the analysis so far is incomplete. The
labour productivity measure of structural change as presented in this section is a partial measure, as it does not consider other inputs besides labour. In the next section
we consider shifts of labour and capital simultaneously.
5. Impact of structural change on total factor productivity growth
