This article was downloaded by: [Ohio State University Libraries]

On: 18 June 2012, At: 07:47

Publisher: Routledge
Informa Ltd Registered in England and Wales Registered Number:
1072954 Registered office: Mortimer House, 37-41 Mortimer Street,
London W1T 3JH, UK

Journal of Development
Studies
Publication details, including instructions for
authors and subscription information:
[Link]

Approaches to identifying
key sectors empirically
by means of input‐output
analysis
a
Siegfried Schultz
a
Department ‘Western Industrialised and
Developing Countries’, Deutsches Institut fur
Wirtschaftsforschung, Berlin

Available online: 23 Nov 2007

To cite this article: Siegfried Schultz (1977): Approaches to identifying key

sectors empirically by means of input‐output analysis, Journal of Development
Studies, 14:1, 77-96

To link to this article: [Link]

PLEASE SCROLL DOWN FOR ARTICLE

Full terms and conditions of use: [Link]

terms-and-conditions

This article may be used for research, teaching, and private study
purposes. Any substantial or systematic reproduction, redistribution,
reselling, loan, sub-licensing, systematic supply, or distribution in any
form to anyone is expressly forbidden.

The publisher does not give any warranty express or implied or make any
representation that the contents will be complete or accurate or up to
date. The accuracy of any instructions, formulae, and drug doses should
 be independently verified with primary sources. The publisher shall not
be liable for any loss, actions, claims, proceedings, demand, or costs or
damages whatsoever or howsoever caused arising directly or indirectly in
connection with or arising out of the use of this material.
Downloaded by [Ohio State University Libraries] at 07:47 18 June 2012
 Approaches to Identifying Key Sectors Empirically by
Means of Input-Output Analysis
by Siegfried Schultz*

On the basis of input-output tables from developing countries sectors

Downloaded by [Ohio State University Libraries] at 07:47 18 June 2012

are distinguished according to the degree of their interdependencies.

Using 20 standard sectors for all 22 tables analysed, the paper takes
the intensity of interindustrial linkages as an indicator of a sector's
ability to spread growth impulses to its economic environment.
Backward and forward linkages are calculated; in addition, spread
effects are computed via the inverse matrix. Then the sectors are
classified according to their total (direct and indirect) primary input
requirements per unit of final demand. The analysis is supplemented
by the determination of the sectoral employment impact, i.e.
applying figures for the persons engaged sectorwise.
Particularly under the linkage aspect, the obtained rankings are
checked for similarity. Although some rankings are highly
correlated, none of the criteria under consideration proves superior
to all others.

Taking up an earlier attempt,1 this paper specifies the methodology and

presents some results of a trial application of input-output techniques to
identify key sectors for some developing countries. Aiming at an empirical
determination of the relative importance of individual sectors, the mutual
dependence of sectors is measured first by the extent of their intermediate
demand and supply. Besides these direct spread effects in terms of
backward and forward linkages, use is made of the inverse matrix to catch
the indirect effects as well. Moreover, the sectoral effects on central
economic aggregates like import dependency, GDP and the balance of
foreign trade are considered. Finally, by use of statistical information from
outside the input-output system, the analysis is supplemented by including
sectoral employment effects as well.
The results of the computation, i.e. the size of the several linkage ratios,
the primary input requirements and the employment impact on a per unit
level, are translated into an ordinal sequence. This ranking is carried out on
the basis of a standardised 20 x 20 sectoral scheme for a number of
input-output tables from developing—predominantly Asian—countries.

1. VALUATION AND CLASSIFICATION, COUNTRY COVERAGE

With reference to their valuation, the input-output tables used in the

*Deutsches Institut fur Wirtschaftsforschung (Berlin); Department 'Western Industrialised

and Developing Countries'. A short version of this article was presented at the VIth
International Conference on Input-Output Techniques, Vienna 22-26 April 1974.
 78 JOURNAL OF DEVELOPMENT STUDIES

analysis generally are valued in producer's prices. The level of aggregation2

is identical for all tables throughout the study: following ISIC definitions, a
scheme of 21 standard sectors (see Table 1) has been applied, out of which
two sectors (1,2) stand for primary and three sectorr (18-20) for tertiary
production. Industrial activity is subdivided into 13 manufacturing
branches (3-15) as well as energy (16) and construction (17). In another
sector (21) all those transactions have been collected that could not be
allocated elsewhere. This residual has been ignored in the subsequent
ranking of sectors.
Downloaded by [Ohio State University Libraries] at 07:47 18 June 2012

TABLE 1
STANDARDISED SECTORS (UN—CLASSIFICATION)

ISIC1
No. Sector 19582 19683

1. Agriculture, forestry and fishing 01-04 11-13

2. Mining and quarrying 11-14, 19 21-23, 29
3. Processed foods 20-22 31
4. Apparel, incl. textiles and footwear 23, 24 32 ./. 323
5. Leather and—products 29 323
6. Wood and paper prod., printing 25-28 33, 34
7. Rubber products 30 355
8. Chemicals 31 351, 352, 356
9. Petroleum products 32 353, 354
10. Non-metallic mineral products 33 36
11. Basic metals, metal products 34,35 37, 381
12. Non-electrical machinery 36 382
13. Electrical machinery 37 383
14. Transport equipment 38 384
15. Industry n:e.c. 39 385, 39
16. Electricity, gas, water 51, 52 41,42
17. Construction 40 50
18. Trade 61 61,62
19. Transp., storage and communic. 71-73 71, 72
20. Services 62-64, 81-85 63, 81-83, 91-94, 952,
953, 959
21. Undistributed 90 00

1. International Standard Industrial Classification of All Economic Activities.

2. UN, Statistical Papers, Series M, No. 4, Rev. 1.
3. UN, Statistical Papers, Series M, No. 4, Rev. 2, Add. 1.

Of the tables analysed, the majority is from Asian developing countries.

For comparison, additional tables have been included from developing
countries outside this region and from economically more advanced
countries. A complete analysis of input-output tables used is given in Table
2.
2. METHODOLOGY
2.1 Linkage Indicators
Typically, countries with high per capita incomes have a high degree of
division of labour, expressed by a dense network of intermediate supplies
or deliveries. Those sectors of the economy which, owing to their close
 IDENTIFYING KEY SECTORS EMPIRICALLY 79
TABLE 2
ANALYSED INPUT-OUTPUT TABLES

Number of
sectors in Imports as
Country Year Abbrev. orig. table row col.

India 1963 IND63 33 X

India 1964/65 IND65 77 X
Indonesia 1969 IDO69 43 X
Iran 1965 IRN65 29 X
Korea (S) 1963 KOR63 43 X
Downloaded by [Ohio State University Libraries] at 07:47 18 June 2012

Korea (S) 1966 KOR66 43 X,

Malaysia (W) 1965 MLY65 31 X
Pakistan 1960/61 PAK61 30 X
Pakistan 1963/64 PAK64 54 X
Philippines 1961 PHI 61 29 X
Philippines 1965 PHI 65 97 X
Sri Lanka 1963 SLA 63 38 X
Sri Lanka 1965 SLA 65 41 X
Taiwan 1964 TAI64 55 X
Taiwan 1966 TAI66 76 X
Taiwan 1969 TAI 69 76 X
Turkey 1963 TUR63 37 X

Israel 1965/66 ISR 66 80 X

Israel 1968/69 ISR69 30 X
Yugoslavia 1966 YUG66 29 X

Germany (W) 1966 GER66 56 X

Japan 1965 JAP 65 56 X

technology-related ties, are in a position to promote or generate growth in

other sectors through that of their own can be called strategic for achieving
higher income levels. Therefore, the extent of interdependence can serve as
a suitable criterion for a ranking of sectors.

2.1.1 Standard notion of backward and forward linkages. In the simplest

approach, a sector's degree of interdependence with its economic
environment can be expressed by the relation of intermediate to total
transactions. Depending on whether the input or the deliveries are used in
the numerator, the obtained ratios express the well-known sectorial
linkages:3 Backward linkages (u) indicate to what extent the economic
branches have been specialising as yet: A tight network of intermediate
flows is transmitting production stimuli to other sectors.
Thus, coefficients are high if the sectors observed are drawing heavily on
the system of industries or, vice versa, if the value added by use of primary
input is relatively small. Forward linkages (w)— generally weaker since the
output produced is not necessarily met by adequate demand—give an
indication of the direction of supply: High coefficients will typically be
found with those sectors producing relatively little directly for final demand
but rather for intermediate demand of other sectors.
 80 JOURNAL OF DEVELOPMENT STUDIES

2.1.2 Modifications. The above notion of linkages (ua, wa) can easily be
computed and has been used before.4 Besides the technological
relationships expressed in these ratios it may be desirable, however, also to
consider the size of a sector's transactions. For this reason, the analysis was
supplemented by attaching weights to the linkage ratios:
(x) the sum of intermediate input or output per sector (ub, wb) and
(xx) the share of the sectoral gross output in the output of the entire
economy (uc, wc).
In a further variation, intrasectoral transactions were deducted before
computing the linkages. The coefficients thus obtained (u<i, wd) refer only to
Downloaded by [Ohio State University Libraries] at 07:47 18 June 2012

deliveries among different sectors. In analogy to the basic linkage version,

the weighting procedure was repeated here (ue, we; uf, wf) (cf. definitions in
appendix A).
Allowance was made for the differing treatment of imports in the I.
quadrant of input-output tables affecting the size of the calculated
linkages. If alternative versions were on hand or the matrix cells had several
entries, preference was given to the one containing domestic transactions
only. In those—fortunately few— cases where competing imports were not
compiled in a separate row, i.e. the matrix elements comprise both domestic
and imported inputs, the u's. are relatively too high while the w's are
properly computed after the gross output has been adjusted by the
inclusion of imports (w*) (cf. definitions in appendix A).

2.1.3 Power of dispersion (p). By using the inverted Leontief matrix, total
(direct and indirect) production effects can be taken into account:5 the
vertical sum over the matrix elements of the inverse indicates the extent of
the increase in output in the entire economy caused by the rise of final
demand in the sector under consideration. If this variation in final demand
takes place alternately, the inverse's vertical sums show each sector's
impact on overall production. If the sum over all column totals is divided
by the number of sectors, \ C.} denotes the average rise in output—in a
sector chosen at random— which is necessary as a consequence of the
increase in the observed sector j.
By standardising the column totals for the purpose of intersectoral
comparison in terms of the average of all elements of the inverse, the
indicator thus obtained, p, is expressing the 'power of dispersion' of the
observed sector in the entire economy (see appendix A). So p is a measure of
the effects of increased output in one sector relative to those of all sectors:6
If p < 1, the sector in question produces only weak output stimuli for the
economy; a value of p > 1, however, would signal that this sector is
transmitting above-average impulses to others via its intermediate input
requirements.

2.1.4 Hypothetical extraction (s). Again taking advantage of the features

of the inverse matrix, it is attempted in yet another approach to determine
each sector's significance by alternating fictitious extraction from the
interdependent system.7 This does not only mean a hypothetical
production shut-down in the observed sector but also intermediate supplies
from and deliveries to other sectors would be affected. Depending upon
 IDENTIFYING KEY SECTORS EMPIRICALLY 81

whether the effect on production of other sectors or the effect on the sector
itself is greater, the isolated sector can be described as stimulating or
dependent.
The indicator can be derived as follows. A sector's exclusion begins by
deleting its row and column from the input-output table. The sought
sector-specific impact can be measured by comparing numerically the
output levels of the economy before and after the hypothetical extraction,8
adjusted by the gross production of the isolated sector (for details see
appendix A).
The ratio of these net effects on other sectors and the intermediate
Downloaded by [Ohio State University Libraries] at 07:47 18 June 2012

deliveries of the isolated sector to others is called indicator s. If s < 1, the

sector is dependent on impulses of other sectors; if s > 1, the sector can be
called stimulating. Under this approach, keysectors obviously are those for
which the indicator is largest.

2.2 Effects on Some Economic Aggregates

The linkage coefficients help us to identify sectors that—due to their
important position in the interindustrial network—are significant for
initiating or distributing growth impulses. However, they do not indicate
what variations of central economic aggregates like national income, trade
balance or employment are affected. For this reason, the total requirements
of primary input (domestic and foreign) also have been calculated by
means of the inverse matrix. Its multiplication by the row vectors of the
direct coefficients gives us—under inclusion of the primary input content
of intermediate supplies—the total of primary inputs necessary per unit of
sectoral gross output.

2.2.1 Gross domestic product, foreign trade. The sum of each sector's net
contribution to the GDP indicates the total variation of national income
equalling the initial increase of gross production in sector j minus imports
caused directly and indirectly. Therefore, if all initial increases are of the
same size, the sectoral ranking under the GDP and the import aspect must
be identical. (In case intermediate inputs were mixed irrespective of their
(domestic or foreign) origin, both effects could not be calculated.) Due to
modest disaggregation in quadrant III of the input-output tables from
developing Asian countries in our sample—only about every other had an
entry for wages and salaries— none of the GDP components were
compared per sector and country.9
As to the foreign trade aspect, by balancing exports and imports per
sector the value of sectoral net exports (imports) has been used to determine
whether the sector under consideration is drawing on or contributing to
foreign exchange reserves.10
2.2.2 Employment. Finally, as a further criterion for identifying key
sectors the employment aspect has been introduced for those countries
which do have the necessary extra data. The rough approximation via the
sum of wages and salaries is unfit for intersectoral comparison due to
substantial differences in wages. Also, the real employment impact may—
in developing more than in advanced countries—be underestimated
 82 JOURNAL OF DEVELOPMENT STUDIES

because relatively more own-account workers (farmers, artisans) are to be

found at lower levels of development.
Basically the data have been taken from international sources. 11 These
figures generally refer to the total civilian labour force employed.12 Since
their sectoral breakdown does not go beyond the 1-digit ISIC level—while
manufacturing industry in this study is subdivided into several branches—
the structure of the more detailed data in the UN industrial statistics13 has
been transplanted to the ILO-figures for manufacturing after being fitted to
the scheme of standard sectors applied here. 14
Downloaded by [Ohio State University Libraries] at 07:47 18 June 2012

3. RESULTS
3.1 Backward and Forward Linkages
The sectors can be grouped according to the size of their u's and w's in such
a manner that all combinations of high and low 15 values are easy to survey.
(It ought to be mentioned, however, that the classification is country-
specific; a sector with the same values for u and w may in another country
belong to another group. 16 ) Sectors can be arranged according to their
linkage ratios as follows:

Group Sector Classification Dominant Direction

I manufacturing final demand
II manufacturing intermediate demand
HI primary, tertiary final demand
IV 17 primary, tertiary intermediate demand

Analysis of the statistical coherence between the ratios for the backward
and forward linkages in their basic type a and the weighted variants b and c
revealed only a loose connection: For the Asian developing countries under
consideration the average correlation between ua and Ub (wa and wb) was
r = 0-63 (0-64), between ua and uc (wa and wc) r was 0-62 (0-51). Despite
their different weights, however, the ratios of variant b and c correlated
highly; their mean values for r were 0-95 and 0-94 respectively.18
Because in. some of the input-output tables transactions were
concentrated in the cells on the main diagonal, these items were deducted
before computing the linkages (version d). Yet the ranking of sectors varied
only slightly: For the analysed sample of developing Asian countries, the
correlation between linkages in version ua and ud (wa and wd) was r = 0-97
(0-98). Since intrasectoral transactions frequently are substantial, their
share obviously is similar from sector to sector. 19
Finally, attention must be called to an interesting possibility regarding
the interpretation of backward and forward linkages in the entire economy
which is raised by version c. Whereas version c represents a—weighted—
quantification of the degree of mutual dependence of various economic
sectors, the national averages of all sectors (u*, w*) (see definitions in
appendix A) can be taken as an indicator of the division'of labour and thus
the level of development within a country. From a ranking according to
economic integration (see Table 3) countries can be grouped in a way which
by and large conforms to ranking according to conventional indicators of
the stage of development (per capita GNP or GDP). There were, however,
 D
tn
Downloaded by [Ohio State University Libraries] at 07:47 18 June 2012

i
o
m
•<

TABLE 3 tn
COUNTRY RANKING ACCORDING TO (COMBINED) BACKWARD AND FORWARD LINKAGES
9
o
0-30 035 0-40 0-45 73
+ w*
rn
India 1963 India 1965 Indonesia 1969 Germany 1966 Japan 1965 2
Philippines 1961 Iran 1965 Pakistan 1961 Yugoslavia 1966
Sri Lanka 1963 Israel 1966 Pakistan 1964 2
o
Sri Lanka 1965 Israel 1969 Philippines 1965 >
Korea 1963 Taiwan 1964 r
Korea 1966 Taiwan 1966
Malaysia 1965 Taiwan 1969
Turkey 1963
 84 JOURNAL OF DEVELOPMENT STUDIES

TABLE 4
SECTORS WITH VALUES OF INDICATORS P AND S ABOVE AVERAGE

Country IND IND IDO IRN KOR KOR MLY PAK

Sector 63 65 69 6. 6t) 65 61

(1) Agriculture, forestry and fishing s.

(2) Mining and quarrying P= s.
(3) Processed foods p = ; s:p s: s: p = ; s:p = ; s. p = , s:p = ; s: P = s:
(4) Apparel, incl. textiles and footwear s:p = s. s p = ; s.p P s: P = s:
(5) Leather and products s.p = s. s p = ; s.p = ; s. P = . s. P= s:
Downloaded by [Ohio State University Libraries] at 07:47 18 June 2012

(6) Wood and paper prod., printing P s. s p= p = ; s. P = s.

(7) Rubber products P= p= s: p = s. s.p = ; s: P =
(8) Chemicals p = ; s.p = s s p = ; s. P = s:
(9) Petroleum products p= s p = ; s: s P=
(10) Nonmetallic mineral prod. P= s p= P= p= P=
(11) Basic metals, metal prod. P p s .P P p = ; s: P s.
(12) Nonelectrical machinery p = ; s:p = s: s P P = S. P s:
(13) Electrical machinery s:p = s: s p = ; s P = S. s:
(14) Transport equipment p = ; s:p = s. s p= p = ; s P = S. P= s:
(15) Industry n.e.c. s: s s:p = ; s P = s.p = ; s P = s:
(16) Electricity, gas, water p s p=
(17) Construction p= P= s:p = , s:p = ; s:p = s:p = ; s s:
(18) Trade s s: s
(19) Transp., storage and communic. p=; s s s S\
(20) Services s s: s s:
Notes = even dispersion of impulses;
share of final demand > 50 ^75 per cent;
: share of final demand > 75 per cent.

certain striking departures; Israel's position in this classification is

relatively low, while Indonesia 20 would appear to be ranked too high. The
surprisingly high ranking of Yugoslavia and the position of Japan relative
to that of Germany may be explained by an intensive, growth-oriented
sectorial integration which has yet to be reflected to a corresponding degree
in per capita income. .

3.2 Linkage criteria allowing for indirect effects

3.2.1. Power of dispersion (p). Findings based on the empirical results of
the power of dispersion (indicator p in Table 4) are as follows. Relatively
strong intersectoral impulses are emitted by the various industrial sectors
of Asian developing economies. In the primary and tertiary sectors, on the
other hand, values of p > 1 are seldom (India 1963, Philippines, Turkey).
In principle, the same is true of the more highly developed countries. The
most stimuli are forthcoming from the secondary sectors, although in the
case of Israel, the intensity with .which agriculture is conducted apparently
creates considerable stimulus for other sectors.
Of the secondary sectors, processed foods predominate, followed closely
by apparel and construction. In frequency of appearance, the sectors
leather and leather products, wood and paper products, printing, and
chemicals are on nearly equal footing. 21 When the subsidiary criterion of
the greatest possible evenness of dispersion of impulses (indicator v) is
applied, only minor changes result with regard to the ranking of sectors
 IDENTIFYING KEY SECTORS EMPIRICALLY 85
TABLE 4
SECTORS WITH VALUES OF INDICATORS P AND S ABOVE AVERAGE

PAK PHI PHI SLA SLA TAI TAI TAI TUR ISR ISR YUG GER JAP
64 61 6. 6. 1 6. 64 66 69 6. ! 61 69 66 66 65

P s:p p P= s P = ; s
p = s. s. p=; s
P= s: s:p = . s:p = ; s:p = , s:p = ; s:p = ; s: P = ; s:p = ; s: P = < s. P = ; s p = ; s: P = s. p = ; s:
P= s.p s:p s.p = . s P P P p p . s. P p P P s.p ; s.
P= s:p = s.p = p = ; s P = . s:p = ; s:p ; s. s:p = ; s. P P p=; s. P s.p = ;s
P s.p = s.p = p = P . p = ; s p = ; s p = ; s.p = p=; s P
Downloaded by [Ohio State University Libraries] at 07:47 18 June 2012

P= s.p = s:p = s P= s.p = P= P= p=; s P= p = ;s

P= s.p = s.p = p = P= p P= p P= p P P
s*
P= p = s.p = p= p=; s p= P=
s.p . s: P p P p P P p P P
P= s:p = s: P = p = ; s p = ; s. P = : s.p = P= P= s. p = ; s.
p = s:p s.p = ; s. s: P= p = ; s. P = s. p = ; s
P= s:p = s: s: P = . s:p = ; s.p = ; s: s.p = ; s p = ; s: P = s:p = ; s.
P= s:p = s. P = P = p = ; s:p = ; s. s: s: p = ; s s: P = s.p = ; s.
P = s.p . s: P = P= P
P- s: s : p - , s : p - , s p - s : p - ; s : p - ; s: p - ; s : p - P - s: p = ; s p - ; S.' s: p = ; s:
s s.
s: s s: p = ; s: s.
s s: s s: s P = : s. P s. s:

according to their frequency of appearance. Of the leaders, only apparel

receives a lower rating; this apparently because only few other sectors are
drawn upon to cover intermediate input requirements (e.g. in Korea and
Taiwan). Sectors in which there is a high degree of vertical integration will
not register high p and v values. This is clearly demonstrated in the example
of basic metals and metal products. Stimuli forthcoming from transport
equipment, non-electrical machinery, and other industries, by comparison,
are powerful and relatively evenly distributed, not only in the industrialised
countries but in some of the developing countries as well.

3.2.2 Hypothetical extraction (s). A calculation of the impact of a

hypothetical production shut-down in each sector to determine that
sector's economic importance in intersectoral flows (indicator s in Table 4)
produces results similar to those for p. 22 The secondary sectors
predominate here as well, while primary production is only modestly
represented. There is a distinct departure, however, in the tertiary sectors,
which show up more strongly.
Of the secondary sectors, processed foods has an above-average
indicator value in all countries in the study. Also high on this list is leather
and leather products. Transport equipment is found comparatively often
among the stimulator sectors, as is construction.23

3.3 Correlation of Sector Rankings with the Various Linkage Criteria

A comparison of results for the indicators p and s leads to the general
 86 JOURNAL OF DEVELOPMENT STUDIES

observation that on average, i.e. including industrialised countries, a

number of sectors are ascribed a lesser importance by s than by p. This
applies in particular to basic metals and metal products, non-metallic
mineral products, and chemicals and to a lesser extent to apparel, a sector
which is one of the stimulator sectors, even in the two industrialised
countries—presumably due to the high share of synthetic fibre processing.
The p and s sector rankings were tested for correlation using the
Spearman coefficient.24 The coefficient of correlation between the p and s
ordinal values of each sector averaged r = 0-49 for the Asian developing
countries. The national coefficients ranged from —0-21 (India) to 0-79
Downloaded by [Ohio State University Libraries] at 07:47 18 June 2012

(Pakistan). 25
A much stronger correlation was found to exist between the ua and p
sectoral rankings, the coefficients averaging r = 0-95. Among the Asian
developing countries, individual figures ranged from 057 (Indonesia) to
0-99 (Malaysia, Philippines). Coefficients for the other two groups were
similarly high: 0-99 (Israel, Yugoslavia) and 0-96 (Germany, Japan). Thus,
a good approximation of p sector rankings can be obtained by way of the
more simply calculated backward linkages in the basic version.
As expected, a corresponding test of the correlation of ua and s rankings
did not turn out as good. The correlation between ua and p is generally so
close—apart from Indonesia—that indicators can be interchanged without
significantly altering the result. The average coefficient of correlation is 0-49
for the Asian developing countries and even lower for the other two
country groups.
The similarity of wa and p rankings is also quite limited (—0-18). Between
the wa and s rankings the correlation is also negative, but higher. In view of
the mean correlation of —0-69, however, an approximation of s results
with—more simply derived—reciprocal wa values is not advisable.

3.4 Gross Domestic Product, Foreign Trade

Under the income aspect, the analysis of the average sectoral position in the
international comparison is necessarily restricted to gross domestic product.
The original tables vary in the extent to which they break down national
income components—in some cases there exists only one primary input
entry. A country-to-country cross section comparison is thus pointless.
As explained earlier, the sectors are ranked on the basis of their
combined coefficients of direct and indirect effects. These coefficients are
standardised on the basis of gross sectoral output. Unless imported
intermediate inputs are shown in quadrant III, all GDP coefficients will
equal one—and thus offer no means for comparison. For this reason,
analysis under this aspect concentrates on those tables which include an
import row (cf. Table 2).
In the average of the Asian developing countries, 26 the highest rankings
are in the tertiary and primary sectors. In manufacturing, processed foods
is found most often among the leaders. By and large, a similar constellation
is seen in the more advanced countries. In these countries, however,
construction also occupies a comparatively high position.
As a general observation on this criterion, it is to be noted that here the
 IDENTIFYING KEY SECTORS EMPIRICALLY 87

weight of the direct effects of a sector, as expressed in its input coefficients,

is the dominant factor on the overall impact and the ranking.
The cross section comparison of balance of trade effects27 gives, at first
glance, about the same picture. Rankings of the primary sectors in the
Asian developing countries average highest. More than anything else, this
is attributable to exports of mining products such as ores and crude oil and
agricultural products. Of the manufacturing sectors, only processed foods
has a high ranking; the others are found farther down on the scale of net
foreign exchange earners. In view of the country-to-country variations in
resources, in industrialisation policy, and in competitiveness on export
Downloaded by [Ohio State University Libraries] at 07:47 18 June 2012

markets, a determination of average sectoral rankings would not be very

useful. This would mean attaching an interpretation to a balance which
derives from factors which are variously powerful from country to country.

3.5 Employment
Evaluation of the results of the employment impact reveals the following
characteristics: As expected, a marginal change in final demand, indirect
effects included, produces a comparatively strong employment effect in
agriculture 28 and trade in all Asian developing countries. As regards the
dispersion of effects (see Table 5), impulses from wood, paper and printing
and from construction are generally powerful and relatively evenly
distributed to other sectors.
Strong effects on output and employment as well are apparently had by
sectors for which, because of production-related ties or their specific
employment requirement, indicator values for p and i_p > 1 (specially noted
in Table 5). In the Asian developing countries, processed foods is most
frequently one of these sectors.
Due to the central importance of the employment aspect, the detailed
sectoral rankings, based on total effects rather than just rough groupings of
above and below average positions are also shown (Table 6). Owing to the
strong direct effects, the greatest total employment effect is that of
agriculture; in second place is processed foods. Averaged for the Asian
developing countries, the employment effect of apparel nearly equals that
of trade—a reflection of the large extent of indirect employment by the
apparel industry.
Whereas the ratio of indirect to direct employment varies widely from
sector to sector, these ratios were shown in addition to the total
employment effect rankings. These quotients give an indication of the
extent to which a sector's employment effect is underestimated whenever
only direct effects are considered. While the direct effects may be regarded
as sufficiently characterising the tertiary sectors—as well as the primary
sectors in the more advanced countries—a true grasp of the potential
employment effect of most manufacturing sectors is gained only when
adequate allowance is made for the induced effects.
For lack of employment data, the wages and salaries ranking might be
assumed to approximate that of employment. In order to obtain some
statistical evidence as to whether this would be a justified supposition, the
ranking of sectors was checked for similarity in countries having data for
wages and salaries as well as for the number of persons employed. The test
 oo
oo
Downloaded by [Ohio State University Libraries] at 07:47 18 June 2012

TABLE 5
SECTORS WITH VALUES OF jP ABOVE AVERAGE

YVG 66
KOR63

TVR63

GER66
KOR66

JAP 65
69IV1

99HS1
o

TAI66
^o
Country *», OS
Sector a. a. S3
(1) Agriculture, forestry and fishing X X Xv XT X X Xv Xv xv X X
(2) Mining and quarrying X X X X
(3) Processed foods xv xv X v
xsv xs v
xX* x«v X9
(4) Apparel, incl. textiles and footwear Xv X"
x xv x xv Xv
xXv9
(5) Leather and -products X" xs xvT
(6)
(7)
Wood—and paper prod., printing
Rubber products
X'
X-
X- X- xs
X
X X" X
X-
xXv9
u
(8) Chemicals o
(9) Petroleum products c
(10) Non-metallic mineral prod. X" z
50
(11) Basic metals, metal prod. xvv >
(12) Non-electrical machinery
xv xs xv oTl
(13)
(14)
Electrical machinery
Transport equipment
xv X"
X"
x D
X xs
(15) Industry n.e.c. X* X" Y_ X xv XJ? m
(16)
(17)
Electricity, gas, water
Construction X- X X- X*
xv X
X-
m
O
(18) Trade X X X X
xs
X X X
xs
X X X •o
(19) Transp., storage and communic. X X tn
(20) Services X X X X X X X" Z

STUDIES
v
Note: X i; " = even dispersion; = also p > ,
 TABLE 6
SECTORAL EMPLOYMENT EFFECTS COMPARED INTERNATIONALLY a
m
Downloaded by [Ohio State University Libraries] at 07:47 18 June 2012

H
Total employment effect Ratio of indirect/direct
ranking employment effect 3
o
Country groups1 m
Sectors C A C
m
(1) Agriculture, forestry and fishing 1 6/7 1 0-28 0-62 0-25 O
(2) Mining and quarrying 11 11 9 1-12 0-59 0-51
(3) Processed foods 2 10 6 5-75 1 92 5-64 <*)
m
(4) Apparel, incl. textiles and footwear 4 2 2 1 39 0-90 1-64 2
(5) Leather and -products 14 19 10/11 4-34 1-90 200
(6) Wood—and paper prod., printing 6 12 4 1-80 0-68 1-78 jo
(7) Rubber products 9 16 7 110 118 1-51
(8) Chemicals 15/16 17 17 1-81 0-99 2-86
(9) Petroleum products 19 20 19/20 3-35 0-58 1103
(10) Non-metallic mineral prod. 10 18 12 0-76 0-74 0-91
(11) Basic metals, metal products 17 15 15/16 110 106 209
(12) Non-electrical machinery 15/16 8 10/11 1-24 0-42 1-30
(13) Electrical machinery 12 13 5 103 0-35 1-27
(14) Transport equipment 13 6/7 13/14 100 0-35 201
(15) Industry n.e.c. 7 4 15/16 0-99 010 4-86
(16) Electricity, gas, water 20 5 19/20 113 0-56 2-40
(17) Construction 5 9 8 0-91 0-81 118
(18) Trade 3 3 3 012 0-21 0-26
(19) Transp., storage and communication 18 14 13/14 0-46 0-45 0-49
(20) Services 8 1 18 0-22 012 0-57
l
Note: A = Korea 1963 and 1966, Philippines 1961 and 1965, Taiwan 1966 and 1969, Turkey 1963;
B = Israel 1966 and 1969 plus Yugoslavia 1966;
C = Germany 1966 and Japan 1965.
 90 JOURNAL OF DEVELOPMENT STUDIES

for rank correlation showed widely dispersed values with a maximum of

around r = 0-60. Supposedly, this is because the statistical coverage of both
approaches is too different. But even if only wage earners and salaried
employees were compiled in the employment category (as with
Yugoslavia), the observed correlation is very moderate (r = 0-23). thus, it is
not advisable to substitute the one ranking for the other.

OUTLOOK
The above analysis might also be conducted on the developing countries of
Africa and Latin America. The necessary data are available for quite a
number of countries. However, it is recommended that future studies
Downloaded by [Ohio State University Libraries] at 07:47 18 June 2012

should not be based only on marginal changes in final demand. As unused

capacity cannot be assumed generally, capital requirements per unit of
production in each sector would have to be determined—either by
transformation of limitedly available empirical data or estimation of
sectoral capital productivity.
Since capital is not the only factor whose availability is restricted— limits
are also set in qualified labour and natural resources—such limitations
should be explicitly taken into account and incorporated into a
programming approach. Such programming, either linear or non-linear,
would comprise restrictions and planning goals (e.g. maximisation of
growth or consumption, minimisation of unemployment or foreign
exchange loss), thus producing consistent results.

APPENDIX A
DEFINITIONS
Within the open static Leontief model,
Xj intermediate input if sector j
Xj total input of sector j
Xi. = intermediate demand for products of sector
total demand for products of sector i
n n
X Z< ZJ ij

n n
X = X Xj-X

n
X Mi = M import of products of sector i
i=l

M+ X = Z
 IDENTIFYING KEY SECTORS EMPIRICALLY 91
Indicator u and w, different versions:

Xj

U b = Ua • X.j wa •

Xj Xi
U c = Ua • wl = w* • —
Downloaded by [Ohio State University Libraries] at 07:47 18 June 2012

X.i - - Xn
U<j = - Wd = -
Xj Xi +|M||

U e = Ud • wd • Xi. ! = w5 • Xi.

Uf = Ud Wf = Wd • F = w5- —

X
u* = w* =
X
resp.

X + M

Indicator p and v:
n
Cg = C.j column total of the inverse matrix
i= 1

i=l
Pi = (j = 1, 2 . . . , n)
1 n n
- 2 2 Cij
n2 i = 1 j = 1

1 n
L Cii - - Ci
n - 1 i<
U = 1, 2 . . . n)
 92 JOURNAL OF DEVELOPMENT STUDIES

Indicator s:

m e — 1 m
Z Xi - Z X* (e)i - Z X* (e)i - Xe
i = I i = 1 i = e + 1
$c =
X e - Ye

m
Z (Xi - X* (e)i )
i = 1
Downloaded by [Ohio State University Libraries] at 07:47 18 June 2012

i ? e
= (e = 1, 2 . . . m),
Xe - Y e

whereby X*(e)i is an element of a vector defined as

X:<e) = C*(e) " Y*(«)

and

Xi = gross production of sector i

C*(e) = inverse of the reduced matrix after elimination of column and row of the extracted
sector

Y*(C) = final demand vector as shortened by omission of the extracted sector

Xe = gross production
, of the extracted sector.
Ye = final demand

APPENDIX B
SOURCES

Input-output tables

Germany (W)
R. Staglin and H. Wessels, 'Input-Output-Tabelle fur die Bundesrepublik Deutschland
1966', in Vierteljahrshefte zur Wirtschaftsforschung des DIW, no. 3/1971.
India
Gokhale Institute of Politics and Economics, Poona; P. N. Mathur e.a. Input-Output Flow
Table, 1963, (reprint from: Artha Vijnana, vol. 11, no. 2; 1969).
'Perspective Planning Division, Planning Commission Structure of Indian Economy: Inter-
Industry Flows and Pattern of Final Demand 1964-65' in Sankhya. The Indian Journal of
Statistics, series B, vol. 30, part 1 and 2, 1968.

Indonesia
Leknas/Kyodai, Input-Output Table of Indonesia, 1969, Djakarta 1972.
 IDENTIFYING KEY SECTORS EMPIRICALLY 93

Iran
A. S. Shaheen, 1965, Input-Output Table for Iranian Economy (no place, no date).
Israel
Bank of Israel, Input-Output Tables for 1965-66, (computer print-outs).
Central Bureau of Statistics; D. Chen, Input-Output Tables 1968/69 (Special Series, no. 380),
Jerusalem 1972.
Japan
Inter-Industry Relations Table for 1965, Tokyo 1969.
Korea (S)
The Bank of Korea, Research Department, Input-Output, Interindustry Relations Tables for
1963, Seoul 1965.
Downloaded by [Ohio State University Libraries] at 07:47 18 June 2012

The Bank of Korea, Research Department, Input-Output, Interindustry Relations Tables for
1966, Seoul 1968.

Malaysia
Department of Statistics, Malaysia Barat Kiraw Perusahaan' Perantaraan 1965, Kuala
Lumpur (no date).

Pakistan
'Inter-Industry Relationships, Pakistan, 1960/61', (Harvard Advisory Group; G. Rasul) in
The Pakistan Development Review, vol. V, no. 3 (autumn 1965).
For sector names S. H. Katano, in UN-ECAFE, Economic Bulletin, vol. XVIII, no. 2 (Sept.
1967).
'Pakistan Inter-Industry Flow Table 1963/64', in W. Tims, Analytical Techniques for
Development Planning, Annex VIII, Karachi 1968.
Philippines
Department of Commerce and Industry, Bureau of the Census and Statistics 'Input-Output
Analysis of the Philippine Economy, 1961'in Journal ofPhilippine Statistics, vol. 19,no. 1
(Jan-March 1968).
Department of Commerce and Industry, Bureau of the Census and Statistics; Tito A. Mijares,
The 1965 Inter-Industry Relations Study of the Philippine Economy.
(U.P. School of Economics—National Economic Council Workshop Series, Paper no. 71-2).
Manila 1971.

Sri Lanka
D. Perera, 'A Preliminary Input-Output Table for Ceylon 1963' in Central Bank of Ceylon
Bulletin (July 1967).
Ministry of Planning and Economic Affairs, Perspective Planning Division. Input-Output
Table for Ceylon—1965 (UNDP Special'Fund Planning Project). Colombo 1969.
Taiwan
Council for International Economic Cooperation and Development; J. S. Chiu. The Taiwan
Economy: An Input-Output Study (reprint from: Industry of Free China, vol. XXX, no. 5).
November 1968.
Council for International Economic Cooperation and Development, Taiwan's Interindustry
Transactions Table for 1966. Taipei 1969.
Taiwan's Interindustry Transactions Table for 1969. Taipei 1972.
Turkey
Prime Ministry, State Planning Organisation; A. K. Chakraverti, C. Cinar, G. Canalp,
Structural Interdependence of the Turkish Economy: 1963. Istanbul 1970.
S. Senel, Aufstellung einer Input-Output Tabelle und Input-Output-Analyse fur die Turkei,
Bonn 1971.
Yugoslavia
Savezni Zaved za Statistiku, Medusobni odnosi privrednih delatnosti Jugoslavije u 1966,
Beograd 1969.
 94 JOURNAL OF DEVELOPMENT STUDIES

Employment data
UN, Dept. of Econ. and Soc. Affairs, Statistical Office, The Growth of World Industry, vol. I
(General industrial statistics), part 1 (National tables); 1968-1970 editions New York
1970-1972.
International Labour Office, Yearbook of Labour Statistics, 31. and 32. edition Geneva 1971,
1972.
UN, Economic Commission for Asia and the Far East Stat, Yearbook for Asia and the Far
East, 1970 and 1971 Bangkok 1971, 1972.

NOTES
Downloaded by [Ohio State University Libraries] at 07:47 18 June 2012

1. Schultz [1973: 27-54].

2. As to the formal procedure see Benz [1971; Appendix C, Matrix Aggregation via Matrix
Multiplication: 55-60]. Both tables for Israel were slightly changed: Non-positive intrasectoral
deliveries in the aggregated version (1966: sectors 9 and 15; 1969: sectors 2 and 7) were deleted
and the gross outputs accordingly adjusted. These changes did not go beyond 2.6 per cent. In
the case of Japan, two variations of this nature hardly affected the gross output figures (0.01
per cent).
3. Hirschman [1961 (1958): 100, 105].
4. See Chenery and Watanabe [1958]; Rasul [1964]; Chakraverti [1965]. For more recent
applications see Yotopoulos and Nugent [1973] and Thoburn [1973].
5. This approach was originated by Rasmussen [1956]. For an application to Indian data
see Hazari [7970]. Cf. also Yotopoulos and Nugent [op. cit., 161 et seq.].
6. Another indicator (v), derived analogously, for the average of the row totals ('sensitivity
of dispersion'/Rasmussen) has been considered here as an auxiliary criterion only, since the
prime concern was the power of dispersion irrespective of its distribution.
7. See Strassert [1968a: 100 et seq. or 1968b: 211-14]. Cf. also Lehbert [1970: 63 et seq.].
8. Respectively, this is done by subtracting the reduced matrix from the unity matrix,
inverting the new Leontief matrix and multiplying it by the also-reduced final demand vector.
9. The effect on gross output has not been computed separately since the criterion for the
ranking of sectors—column totals of the inverse matrix—has been used before (cf. indicator
p), and the results would have been the same.
10. The grading is based on the difference between the (direct) export ratio in sector j and
the total (direct and indirect) import effects caused by the extension of production in j.
11. ILO, Yearbook of Labour Statistics; UN-ECAFE, Statistical Yearbook for Asia and
the Far East.
To a limited extent, use was made directly of some national employment statistics which
were already adjusted to the sectoral classification of the respective input-output table
(Korea, Germany, Japan).
12. Main exceptions: a—includes military personnel; b—employees only.
13. UN [1968-70]. Figures from ILO and from the UN Statistical Office generally do not
coincide because of differing coverage (employed persons above a certain age; size of the
establishment in terms of number of persons engaged). However, major deviations occur only
if the entry age for the labour force is very low on the one hand and predominantly large
establishments are included on the other.
14. Indicator p can also be used here. Analogously interpreted for the 'employment
inverse',
1
- L C.j
n
stands for the average employment increase touched off in an arbitrarily chosen sector by the
rise in final demand in sector j. For intersectoral comparison LP indicates the relative
importance of each sector with reference to its employment effect; e.g. ,LP > 1 means that this
sector is transmitting impulses that are above normal.
15. 'High' resp. 'low' stands, synonymously, for values above resp. below (or equal to)
average.
16. The non-homogenous residuals in sector 21 are—here and in the following—
disregarded. The same applies to sectors with no entries.
 IDENTIFYING KEY SECTORS EMPIRICALLY 95

17. By definition, backward linkages in the primary sectors can only be limited as
intermediate inputs are few. Low forward linkages indicate a lack of processing and would
thus attach a low priority to those sectors which produce, e.g., predominantly for immediate
export—justifiably so with regard to the desired stimulation of domestic processing.
18. For comparison the correlation coefficients for Germany, Israel and Japan were always
in the neighbourhood of the top values of the developing countries in Asia; Yugoslavia was to
be found in the middle range.
19. Correlation between linkages of type d and their weighted versions e and f was found to
be of the same intensity as between the basic type a and its weighted versions; analogously, the
results according to e and f correlated strongly.
20. This first table for Indonesia is to be regarded with some reservation. The original
version could not be inverted due to singularity; for this analysis, a marginal alteration was
made in the intrasectoral flow of sector 21.
Downloaded by [Ohio State University Libraries] at 07:47 18 June 2012

21. Leather and leather products' indicator is considerably higher in Asian developing
countries than in the more developed countries. This finding is confirmed by Yotopoulos and
Nugent [1973] in their study of six industrialised countries and five developing countries. Their
finding[p. 163] that values for the standard sectors 1, 2, 16 and 20 in the developing countries
range considerably below those of the industrialised countries is not corroborated by the
results of the present study.
22. Without further refinements, this method tends to result in extremely high values for
those sectors produce heavily for final demand.
23. For India and Turkey the difference in frequency between construction and food
processing is due only to a peculiarity in the present computer programme if the final demand
share increases to the extreme of 100 per cent.
24. The subsidiary criterion v was not allowed for in the test (cf. note 6).
25. All sectors of the Indonesian economy showed above-average s values and below-
average p values. Presumably, this is to be explained by the lack of linear independence in the
rows and columns of the original table. By making marginal corrections in the aggregated
version it was possible to obtain an inverse matrix but not plausible results here.
26. In place of the official Turkish table a different one—based on official figures—in which
imports are presented as a row was used here in order that Turkey could be included in this
section of the study.
27. Ranking is based on direct export less total (direct and indirect) import effects. The
countries examined are the same as those for the income aspect, with the exception of Pakistan
1961 (where exports are not shown as a subgroup of final demand).
28. Due to a difference in statistical concepts (i.e. labour force as opposed to employed
persons), the employment effect is presumably overestimated in the case of Turkey.

REFERENCES
Benz, C. W., 1971, Passion (Programme for Algebraic Sequences Specifically of Input-Output
Nature), San Francisco.
Chakraverti, A. K., 1965, Input-Output Techniques in National Planning (with special
reference to India)', Opsearch, vol. 2.
Chenery, H. B., and T. Watanabe, 1958, 'International Comparisons of the Structure of
Production', Econometrica, vol. 26, no. 4.
Hazari, B. R., 1970, 'Empirical Identification of Key Sectors in the Indian Economy', The
Review of Economics and Statistics, vol. LII, no. 3.
Hirschman, A. O., 1961 (1958), The Strategy of Economic Development, New Haven: Yale
University Press.
Lehbert, B., 1970, 'Bedeutung und Auswertung regionaler Input-Output-Tabellen', Kieler
Studien (Tubingen), no. 105.
Rasmussen, P. N., 1956, Studies in Inter-Sectoral Relations, Copenhagen/Amsterdam.
Rasul, G., 1964, Input-Output Relationships in Pakistan, 1954, Rotterdam: University Press.
Schultz, S., 1973, 'Quantitative Criteria for the Determination of Sectoral Priorities', Asian
Economies (Seoul), no. 5.
 96 JOURNAL OF DEVELOPMENT STUDIES

Strassert. G., 1968a, 'Möglichkeiten und Grenzen der Erstellung und Auswertung regionaler
Input-Output-Tabellen", Schriften zu Regional—und Verkehrsproblemen in Industrie—und
Entwicklungsländern (Berlin), vol. 2.
Strassert, G., 1968b, 'Zur Bestimmung strategischer Sektoren mit Hilfe von
Input-Output-Modellen', Jahrbiicher für Nationalökonomie und Stalistik, vol. 182, no. 3.
Thoburn, J. T., 1973, 'Exports and the Malaysian Engineering Industry: A Case Study of
Backward Linkage', Oxford Bulletin of Economics and Statistics, vol. 35, no. 2.
UN, 1968-70, The Growth of World Industry, vol. I, part I of each edition.
Yotopoulos, P. A., and J. B. Nugent, 1973, 'A Balanced-Growth Version of the Linkage
Hypothesis: A Test', Quarterly Journal of Economics, vol. LXXXVII, no. 2.
Downloaded by [Ohio State University Libraries] at 07:47 18 June 2012