ArticlePDF Available

HOW BIG IS THE SPANISH MANUFACTURING PRODUCTIVITY GAP? ALTERNATIVE ESTIMATES

Authors:

Abstract and Figures

Comparing labour productivity levels across countries and industries is problematical, given the difficulties associated with the measurement of certain variables (working hours and labour quality), and specifically involves choosing an appropriate exchange rate. GDP-PPP is shown not to be suitable for this purpose. Two alternative conversion factors are reviewed, from the expenditure and production standpoints. Unlike GDP-PPP, these seek to reflect price differentials in manufacturing and not in the economy as a whole. The relative productivity levels estimated with these conversion factors considerably reduce the estimates obtained when GDP-PPP is used.
Content may be subject to copyright.
HOW BIG IS THE SPANISH MANUFACTURING
PRODUCTIVITY GAP? ALTERNATIVE ESTIMATES
Concha Artola
artola@bde.es
Tfno: 91 338.51.27
Banco de España
Madrid 28014
November 2003
I am grateful for comments from José María Bonilla, Juan José Camio, Ángel Estrada,
Esther Gordo, Pilar L'Hotellerie, Mario Izquierdo, Rafael Myro and Soledad Núñez
2
ABSTRACT
Comparing labour productivity levels across countries and industries is
problematical, given the difficulties associated with the measurement of certain variables
(working hours and labour quality), and specifically involves choosing an appropriate
exchange rate. GDP-PPP is shown not to be suitable for this purpose. Two alternative
conversion factors are reviewed, from the expenditure and production standpoints. Unlike
GDP-PPP, these seek to reflect price differentials in manufacturing and not in the economy
as a whole. The relative productivity levels estimated with these conversion factors
considerably reduce the estimates obtained when GDP-PPP is used.
JEL Classification: 047,057
Keywords: Purchasing Power Parity, Convergence, Balassa-Samuelson effect,
Productivity measurement.
RESUMEN
Resulta problemático comparar los niveles de la productividad del trabajo entre
países e industrias, dadas las dificultades inherentes a la medición de determinadas
variables (horas trabajadas y calidad del trabajo), y requiere, en particular, la elección de un
tipo de cambio adecuado. En este documento se argumenta que la PPA del PIB no es un
factor de conversión apropiado, analizando dos factores alternativos, construidos desde el
punto de vista del gasto y de la producción. A diferencia de la PPA del PIB, estos tratan de
reflejar los diferenciales de precios en la industria manufacturera y no en la economía en su
conjunto. Los niveles relativos de productividad estimados con estos factores de conversión
reducen considerablemente las estimaciones obtenidas con la PPA del PIB
Clasificación JEL : 047,057
Palabras Clave: Paridad de Poder Adquisitivo, Convergencia, Efecto Balassa-
Samuelson, Medición de la productividad
3
Numerous studies have addressed the question of the growth of labour productivity
in Spanish manufacturing from the standpoint of convergence with other countries, yet
studies giving satisfactory estimates of the relative levels of productivity in Spanish
manufacturing are few, which is quite understandable given the difficulty inherent in making
this type of estimate. However, an analysis of the relative growth of Spanish industry is
certainly incomplete if not accompanied by an estimate of the size of the productivity gap
between the various Spanish manufacturing industries and those of our competitors.
Section I of this study reviews the problems inherent in estimating the level of productivity in
manufacturing. Section II summarises the procedure used by the OECD to calculate the
PPP of GDP. The problems inherent in the use of the PPP of GDP as the factor of
conversion to a common currency are reviewed in Section III. Section IV presents two
alternative conversion factors, which are used in Section V to calculate the relative
productivity of Spanish manufacturing industries. Section VI estimates the growth and
relative level of the various manufacturing industries during the 1990s. Finally, Section VII
concludes, listing a number of extensions that could be made.
I. MEASUREMENT OF PRODUCTIVITY LEVELS IN MANUFACTURING
The calculation of productivity levels from a sectoral standpoint involves a number of
problems1. First, it is necessary to decide on the most appropriate definition of productivity.
Labour productivity is defined as the volume of output generated by a given volume of
labour. The first question to answer is what is the most appropriate concept of output: value
added or volume of production? Second, it is necessary to decide how to measure the
volume of labour: number of employees or number of hours worked. Finally, when the
object is to estimate relative levels of productivity across countries, it is necessary to define
the appropriate exchange rate for converting to a common currency.
As regards the first of these questions, there seems to be a certain consensus that,
when attempting to measure labour productivity at the sectoral level (e.g. for
manufacturing as a whole, or for the various manufacturing industries) the preferred
concept of output is the volume of production and not value added. For example, in the
United States, the Bureau of Labor Statistics (BLS), argues that although when attempting
to measure productivity for a broad aggregate (e.g. the market economy or the total
economy) the correct measurement of output should be based on final product, in order to
avoid the double-counting of intermediate goods, when the purpose is to calculate
productivity at the sectoral level (e.g. for manufacturing as whole or the various
manufacturing industries) the exclusion of intermediate goods is erroneous. In this context it
is more appropriate to include in the output of a sector the intermediate goods sold to other
sectors, although intra-sectoral sales of such goods must be excluded. This is the measure
of output proposed by Gollop (1979), and it is known as "sectoral output". It is used by the
4
BLS to calculate the labour productivity indices published quarterly for manufacturing as a
whole and annually for the 119 different manufacturing industries2.
When the purpose of the analysis is to make international comparisons, the labour-
productivity calculations normally use the value-added in each industry3. This is not only on
account of the greater ease with which these data are obtained as compared with the
complex estimation process needed to estimate "sectoral outputs", but also on account of
other kinds of consideration, including the different degree of vertical integration of the
industries in the various countries.
There is no doubt, from the viewpoint of productivity analysis, that the most
appropriate way of measuring labour4 is in terms of the total number of hours worked in the
sector. Taking the total number of workers employed in an industry ignores the existence of
differences in the average number of hours worked, which will depend, inter alia, on the
incidence of part-time work and the importance of overtime. The BLS, in line with this
argument, in its "International Comparisons of Manufacturing Productivity Programme",
constructs labour productivity indices in which labour is measured by the number of hours.
However, when the object is to obtain levels, and not only growth rates, of labour
productivity, the use of series of hours worked in different countries is not advisable since
they are based on different definitions. The OECD regularly publishes statistical series for
the average number of hours worked per annum per person employed and per employee
for a wide range of countries5, although it warns that these data should not be used to
compare the average level of hours worked in different countries, but only to analyse
trends6. In view of these considerations it has been decided to measure the labour factor in
this paper in terms of the number of persons employed, which displays a much higher
degree of homogeneity across countries than the data for hours worked7.
Having decided to estimate the level of labour productivity in terms of the ratio of
value added to the number of persons employed in each industry it must then be decided
what exchange rate to use to express these ratios in terms of a common currency. The
simplest choice, proposed by certain authors (e.g. Dollar and Wolff (1993)), is to use the
PPP of GDP estimated within the framework of the International Comparison Project (ICP)8
and published regularly by the OECD9 to convert the series of value added in the various
industries, denominated in national currency, into a common currency. This procedure has
been used in the Spanish case by Peñalosa (1994)10. The PPP of GDP is clearly the
appropriate exchange rate to estimate the level of productivity of the economy as a whole.
However, serious problems arise when it is used to convert the value added of
manufacturing as a whole or of the various manufacturing industries into a common
currency. Before describing the nature of these problems and their possible solution, the
following section briefly reviews the most relevant aspects of the OECD’s estimates of the
PPP of GDP.
5
II. ESTIMATION OF THE PPP OF GDP IN THE CONTEXT OF THE INTERNATIONAL COMPARISON
PROJECT (ICP)
The PPP of GDP is constructed on the basis of the collection of prices of final
products representative of the structure of spending in each country, which are
subsequently aggregated under different headings in accordance with a common
classification from the spending viewpoint. The joint OECD-Eurostat programme was
initiated in the early 1980s, with the regular publication of estimates of PPPs for GDP and
the various spending components. There are currently six rounds of estimates available
corresponding to 198011, 1985, 1990, 1993, 1996 and 1999, which cover a growing number
of countries (from the first round when PPPs were calculated for 18 countries, until the most
recent in which PPPs were calculated for 43 countries12). Until 1996 inclusive, the
disaggregation of the GDP on the expenditure side was carried out in accordance with the
1968 system of national accounts (SNA68), while in 1999 the results were based for the first
time on the SNA93 classification.
PPPs are computed in three phases. First the relative prices are calculated for
individual representative products selected from a common basket of goods and services13.
The number of products in this common basket hardly varies from one round to another; for
example, in 1996 this basket was made up of 2,900 consumer goods and services, 34
occupations in government, education and health services, 186 types of capital goods and
20 categories of construction projects.
Second, the PPPs are calculated at the product group or basic heading level14, each
of which is made up of a set of goods and services that are similar in nature. In practice,
they are defined as the maximum level of disaggregation for which information is available
on the level of spending. The PPPs for each basic heading are obtained as an unweighted
average of the PPPs of the products included in the group15. Finally, the PPPs for each
successive level of aggregation, up to GDP, are weighted averages of the PPPs of the
basic headings included in the aggregate in question. For each aggregate and each pair of
countries two weighted averages are calculated: namely a Laspeyres-type and a Paasche-
type. The weights used for the first type are the shares of the basic headings of the first
country in the total spending of the aggregate. The weights used for the second type are
those corresponding to the second country. The product of these two PPPs gives a single
(Fisher-type) PPP between the two countries. This procedure enables an intransitive matrix
of PPPs to be obtained for the successive levels of aggregation, which is transformed into a
transitive one by application of the EKS method16. The disaggregation of GDP into different
spending headings changes from one round to another. In 1996, for example, GDP was
broken down into 6 main categories, 19 spending groups and 65 spending categories that
were, in turn, broken down into 213 basic headings. The tables published by the OECD up
to the 1996 edition broke down the PPPs into around 50 spending categories, that year the
6
number of countries analysed was increased from 24 to 32. For some of these countries,
the quality of the PPPs estimated at the disaggregated level did not warrant such a level of
detail. As a result, in 1996 the OECD published a breakdown of the PPP of GDP into 17
spending categories.
III. THE DIFFICULTIES OF USING THE PPP OF GDP AS A CONVERSION FACTOR WHEN MEASURING
PRODUCTIVITY AT THE SECTORAL LEVEL
There are various reasons why the use of the PPP of GDP as a conversion factor in
international comparisons of the level of productivity at the sectoral level, and for
manufacturing industries in particular, is not advisable. First, the PPP of GDP is
inappropriate to measure levels of productivity in manufacturing since the sector represents
a relatively small percentage of GDP in the more developed countries, in which services
have a much greater weight. Moreover, owing to the fact that the prices of non-tradable
goods (which include a large part of services) are generally higher in the more advanced
countries (Balassa-Samuelson), the PPP of GDP in country i relative to the United States
(the leader country) can be expected to be lower than the conversion factor that reflects the
true price differences in manufacturing between country i and the United States. As a result,
estimates of labour productivity in manufacturing industries in country i based on the PPP of
GDP will be skewed upwards17. Second, when the PPP of GDP is the only conversion
factor used in the calculation of the productivity of the various manufacturing industries, the
differences in the relative prices in the various industries are being ignored.
The use of the PPP of GDP as a conversion factor at the sectoral level has
undesirable consequences when the aim is to determine whether there has been
convergence in productivity among a particular group of countries. Most studies estimating
convergence equations refer to the total economy (e.g. Baumol (1986), Barro and Sala-i-
Martín (1991, 1992) Mankiw, Romer and Weil (1992))18. Bernard and Jones (1996), using
data for a group of 14 industrialised countries over the period 1970-1987, try to determine
whether the pattern of convergence in productivity observed at the aggregate level is
maintained when the analysis is carried out at the sectoral level. Their conclusions
surprisingly suggest that manufacturing displays little or no convergence in productivity,
while other sectors, especially services, display a strong tendency to converge. Given the
loss of weight of the manufacturing sector, this means that there has been convergence at
the aggregate level. The variable used to measure labour productivity19 in the various
industries in the model of convergence estimated by Bernard and Jones is value added per
person employed converted into dollars using the PPP of GDP in 1980.
Sorensen (2001) argues that the failure of Bernard and Jones to find any
convergence in productivity in manufacturing was precisely because they used a conversion
factor that is inappropriate at the sectoral level. When the conversion factor is appropriate
the level of relative productivity of a country will be invariant to the choice of the base year
7
used to compare its prices, as will the convergence parameters and the measures of
productivity dispersion across countries. Accordingly, it may be concluded that a conversion
factor is not appropriate when it is observed that the level of relative productivity of a
country and the estimated convergence parameters depend on the choice of base year20.
Sorensen estimates a productivity convergence equation for the economy as a whole and
for the manufacturing sector of the type:
iii Pp
εβα
++= 0
)
ln
where i
pdenotes average productivity growth in country i relative to country 1 for
the period 1970-1993, 0i
P
)
is the level of productivity in country i relative to country 1 in the
initial year, and
β
is the convergence parameter;
β
convergence exists if
β
is significantly
below zero. The level of productivity in the initial year (1970), 0i
P, is valued in dollars using
as conversion factor the PPP of GDP in years 1970 to 1993. The upper panel of Table 1
presents the results of estimating the convergence equation when 0i
P is valued using the
PPP of GDP in various base years. When labour productivity in the initial year is valued at
the prices and PPPs of distant years (1970,1980) the estimate of the parameter
β
is not
significantly different from zero. However, when the base year chosen is close to the final
period, the estimated value of
β
becomes significantly different from zero. In this case,
therefore, it would be concluded that there is productivity convergence. In fact a systematic
pattern is observed whereby the evidence for the existence of convergence is higher the
more recent the base year used. Results of the same type are obtained when σ
convergence is analysed. σ convergence is said to exist when the standard deviation of
productivity (in logs) decreases over time21. Chart 1 shows that the existence of σ
convergence over the period 1970-1993 depends crucially on the PPP used to calculate
labour productivity. When the PPP of a base year close to the initial period is used an
increase is observed in the dispersion over the period as a whole, which would lead to the
conclusion (as in Bernard and Jones on the basis of results obtained with the PPP of 1980)
that there is no convergence in manufacturing as a whole. However, when the PPP chosen
corresponds to base years towards the end of the period, there is a decrease in the
dispersion, which is greater the more recent the base year chosen. In this case the
hypothesis of σ convergence in labour productivity between 1970 and 1993 would be
accepted.
Sorensen concludes that it is not sufficient to make the usual qualifications regarding
the use of the PPP of GDP when carrying out sectoral analyses, like the OECD does, but
rather that this conversion factor should not be used at all since it generates inconsistent
results. It is therefore necessary to find conversion factors that satisfy the necessary
8
condition that levels of relative productivity are invariant to the choice of base year. This is
the subject of the next section.
IV. ALTERNATIVE CONVERSION FACTORS AT THE SECTORAL LEVEL
As discussed more extensively in Section II, the PPPs of GDP obtained in the
framework of the International Comparison Program (ICP) are constructed on the basis of
the collection of a large number of prices of representative final products, which are
aggregated under various basic headings (213 in the 1996 round), defined from the
standpoint of spending. There is, in principle, an alternative way of aggregating the prices
corresponding to the basic headings, namely assigning them to the various industries so
that a measure of PPP can be approximated for each22. In fact, however, there is no
disaggregated data available for the basic headings23, but only the breakdown provided in
the tables published by the OECD, so it is not possible to calculate PPPs for the various
manufacturing industries, although it is feasible to approximate a measure of PPP for
manufacturing as a whole. Hooper and Larin (1989) (hereafter HL) construct a PPP for the
manufacturing sector using the PPPs disaggregated from the standpoint of spending
published by the OECD. This approximation to the construction of a PPP for manufacturing
as a whole assumes that the quality of the PPPs for the spending headings published by
the OECD is acceptable. The Castles report (1997)24 suggests that although the PPPs
estimated at the level of GDP seem reasonable, the results would not be so satisfactory at
less aggregated levels, where it cannot be assumed that the law of large numbers (which
would ensure that errors at the level of the PPP of GDP are offset) holds. Castles
specifically detects the existence of certain inconsistencies in the PPPs estimated for three
basic headings (telephone services, men’s footwear and furniture), presumably arising from
a lack of representativeness or comparability of the products in question25. Varjonen (2001)
examines the size of the inconsistencies at the level of spending categories published by
the OECD and concludes that the discrepancies are much larger in public consumption and
gross fixed capital formation than in private consumption; that they tend to be smaller at
higher levels of aggregation; and, in the particular case of GDP, they do not exceed 3% for
most countries26.
Besides the possible problems of quality of the PPPs for the various levels of
spending, HL’s procedure has a number of other problems:
1) The PPPs obtained from the spending viewpoint are based on comparisons of
retail prices (for most consumer goods) or wholesale prices (for investment goods). That is
to say, in both cases, distribution and transport margins are added to producer prices (the
truly relevant ones for calculating productivity in manufacturing). These margins may vary
greatly across countries, which would, in principle, introduce distortions into the estimates of
the countries’ relative productivity.
9
2) The PPPs also include indirect taxes net of subsidies, which means that
differences in VAT and in other indirect taxes across countries may introduce bias.
3) The comparison of productivity across countries should only reflect the observed
differences in the prices of goods produced in each country. However, in the ICP, the prices
of exported goods are not included, while those of imported goods are included.
4) The prices included in the ICP are all for final goods, the prices of intermediate
goods being ignored. Intermediate goods have a very significant weight in some industries,
accounting for more than one third of the value of output of manufacturing as a whole.
In order to reduce the problems described above, Jorgenson and Kuroda (1995) and
Hooper and Vrankovich (1995) base their estimates of productivity levels across industries
on ICP PPPs, “peeling off” distribution margins and indirect taxes and correcting for
international trade. However, the lack of information on the prices of intermediate goods in
the ICP is a problem that cannot be remedied.
An alternative approach to making international comparisons of productivity levels is
the so-called "Industry of origin approach", developed in the International Comparisons of
Output and Productivity Project (ICOP) initiated in 1983 in the University of Groningen
under the direction of Angus Maddison. Under this approach, the output of each industry is
converted into a common currency using an exchange rate that approximates the
differences between countries in the producer prices of the industry in question. These
conversion rates, PPPs for the various industries, are calculated between pairs of countries
for which a basket of products is constructed for a base year. Specifically, the different
PPPs are obtained as unit value ratios, dividing the total value of the sales of each product
by the quantity sold. One of the advantages of this procedure is that the base data used to
calculate the PPPs come from a single primary source, the country’s business survey,
which provides detailed information on the structure of production in the various industries
with a sufficient degree of disaggregation to ensure the intersectoral consistency of the
output and employment data.
It should be stressed that this method also has its problems. Notable among them is
the fact that many products are not fully comparable, owing to differences in quality. As a
result they are excluded from the set of products common to the two countries analysed,
which reduces the PPPs’ coverage of the output of the industry in both countries. This
problem is less important in industries producing basic goods (e.g. textiles, food, paper,
wood, basic metals) where the products are less heterogeneous. By contrast, in certain
industries producing durable consumer goods and investment goods, the high degree of
heterogeneity means that the coverage of the PPPs is frequently low27.
10
Before presenting the estimates of productivity levels for manufacturing as a whole,
using the two conversion factors discussed above, it would be desirable to be able to check
whether they pass the Sorensen test of invariance to the choice of base year in the
estimation of a productivity convergence equation. Schjerning (2001) uses estimates of the
PPP of manufacturing obtained with the HL procedure for 1985, 1990 and 199328. The
lower panel of Table 1 shows the estimates of the β convergence parameter obtained for
these base years. In each case it is significantly different from zero, although the statistical
significance is greatest when the base year is 1993. As seen in the lower panel of Chart 1,
even though the quantitative results regarding the existence of σ convergence differ from
one base year to another (the lines do not coincide), σ convergence is still found to exist in
every case; moreover, a systematic pattern showing greater convergence the more recent
the base year is no longer observed, as it was when the PPP of GDP was used as the
conversion factor. However, these tests of invariance of the results when the PPP proposed
by HL is used are not very robust as they are based on just three base years. As regards
the second method of calculating sectoral PPPs, the ICOP project, prices are compared for
a single base year for each pair of countries, so that it is impossible to apply the Sorensen
test.
Given the pros and cons of each of the methods proposed for obtaining appropriate
conversion factors for manufacturing, Pilat (1996) proposes a mixed approach, whereby
prices are compared from the viewpoint of output for those industries that produce relatively
homogeneous goods and for those for which detailed sectoral studies are available (for
example, the McKinsey Global Institute case studies). By contrast, in industries
characterised by the production of heterogeneous goods, prices are compared from the
spending viewpoint.
Table 2 shows the values of the various conversion factors analysed in order to give
an idea of the size of the differences, which will necessarily be reflected in the estimates of
the relative productivity of manufacturing in the various countries. The estimates included in
the first three columns are obtained within the framework of the CPI from the spending
viewpoint. The first column gives the values of the PPP of GDP, the second column gives
the estimates obtained using the HL procedure and the third shows the values obtained
using this procedure when margins and international trade are also corrected for. The fourth
column gives the estimates of PPP of the manufacturing sector from the viewpoint of the
industry of origin, while the fifth one shows the results of the mixed approach proposed by
Pilat. The differences are substantial, particularly in Spain, Japan and the United Kingdom.
V. THE RELATIVE PRODUCTIVITY OF SPANISH MANUFACTURING ACCORDING TO THE ICP AND
ICOP CONVERSION FACTORS
In this section a PPP is constructed for the manufacturing sector as a whole
following the methodology proposed by HL. As mentioned in Section II, until the 1996
11
edition, the OECD published a breakdown of the PPP of GDP into 50 categories, for which
spending levels were also provided. It is possible to construct a measure of PPP for
manufacturing as a whole based on this breakdown, by allocating some of these categories
to the manufacturing sector. Specifically, the PPP of the manufacturing sector of country i is
defined as
==
8
1j
ij
w
iji PPPPPP
where ij
PPP is the PPP for country i and the spending category j (expressed in
US$)29.
The weights ij
w are defined as the geometric mean of the share in country i’s GDP
of spending category j
(
)
ij
E and the share of spending category j in the GDP of the United
States
(
)
USj
E. That is to say
=
jjUSij
jUSij
ij EE
EE
w
·
·
The upper panel of Table 3 shows the ij
PPP published by the OECD in its 1990
edition30 as well as the aggregate PPPs obtained by using the weights ij
w which are shown
in the lower panel of the table. Chart 2 compares the PPPs of manufacturing with the PPPs
of GDP for the European Union countries, Japan, Canada and the United States. As was to
be expected (see Section III) the PPP of manufacturing is found to be higher than the PPP
for GDP in all the countries, with the difference being greater in countries with lower income
per capita in 1990. In particular, in Greece, Portugal, Ireland and Spain the PPP of
manufacturing exceeds the PPP of GDP by more than 25% (the PPP of manufacturing is, in
Spain, 30% higher than the PPP of GDP). Chart 3 shows the relationship between the level
of development of a country (as measured by its income per capita) and the gap between
the two measures of PPP31.
Chart 4 shows estimates of labour productivity in manufacturing using two
conversion factors: the PPP of GDP and the PPP of manufacturing32. The higher levels of
relative productivity obtained when using the PPP of GDP as the conversion factor are a
precise reflection of the understatement of the latter relative to the PPP estimated for
manufacturing by the HL procedure. Note that the relative levels of productivity in
manufacturing seem more plausible when the PPP of manufacturing is used as the
conversion factor than when the PPP of GDP is used. For example, Greece (despite its
12
much lower degree of development) has a higher labour productivity in manufacturing than
Sweden when the PPP of GDP is used to convert to a common currency, a result that is
corrected when the PPP of manufacturing is used. The same is the case with the
productivity of Spanish manufacturing which, according to the first measure, would exceed
that of countries that, at least at the beginning of the 1990s, were more developed (UK and
Sweden). Likewise, when the PPP of GDP is used as the conversion factor, Italian
manufacturing is slightly more productive than German, a result that is changed when the
PPP of HL is used. In sum, labour productivity in Spanish manufacturing in 1990, according
to the theoretically more plausible measurement, was 52% of the level in North America that
year and was only higher than in Portugal, Greece, Norway and Denmark.
It is worth asking to what extent these results translate to the competitive position of
Spanish manufacturing. Table 4 shows estimates of unit labour costs in manufacturing. This
is one of the variables traditionally used in analyses of competitiveness33, although normally
comparisons are made in terms of index numbers, i.e. they only reflect competitiveness
gains or losses, but not the level of the variable. The estimates of unit labour costs
contained in Table 4 are obtained in the following way
()
()
M
iitit
ititit
it PPPLY
eLW
ULC
1990
//
//
=
where the conversion factor applied to express the compensation of employees34
()
it
W in
dollars is the exchange rate of each year
()
it
e; while the conversion factor applied to labour
productivity
()
itit LY / is the PPP estimated by the HL procedure for the base year 1990
(
)
M
i
PPP1990 . As seen in Table 4, the dispersion of ULCs is less than that of wages or
productivity, reflecting the correlation existing between productivity and wages. Specifically,
in the case of Spain, the gap in compensation per employee relative to the US is somewhat
higher than the gap estimated for productivity, and in consequence the ULCs of Spanish
manufacturing are slightly lower than those observed in the US (92% of US ULCs over the
decade 1985-1995).
As mentioned in the previous section the other approach that enables to obtain a
PPP for manufacturing is the so-called “Industry of origin procedure". The last column of
Table 5 gives estimates of relative productivity obtained within the framework of the ICOP
project for a set of OECD countries in 1990. These estimates summarise the results of a set
of binary analyses which, in most cases, take the United States as the country of reference.
Generally, the comparative analysis is for the year 1987, although there are cases in which
the base year is 1975 or 1984. The results for 1990 are extrapolations obtained by applying
the growth rates of real gross value added and employment, contained in the national
13
accounts of the various countries, to the levels estimated for the base year. The estimates
given for Spain in this table are based on a comparative analysis of Spain and the United
Kingdom made in 1984 (van Ark 1995).
The ICOP estimates are compared with those discussed above where the
conversion factor is the PPP of GDP, or the PPP of manufacturing estimated by the HL
procedure. In relation to the estimates using the PPP of GDP as the conversion factor, the
same type of result is generally obtained as previously: i.e. the productivity of manufacturing
estimated with the PPP of GDP as conversion factor tends to overstate the levels of
productivity estimated under the ICOP procedure35. In some countries, such as Germany,
this overstatement seems to be modest and hardly changes the assessment of the size of
the productivity gap relative to the US economy. In other countries, such as France, the use
of the PPP of GDP as conversion factor leads to an appreciable reduction in the productivity
gap, which is 15 percentage points with respect to US manufacturing when the PPP of GDP
is used, widening to 22 percentage points when the ICOP procedure is used. Finally, in
some countries, including Spain, the widening of the productivity gap is much more
significant; the productivity gap in Spanish manufacturing according to this measure was 32
percentage points at the beginning of the 1990s, while the gap estimated by the ICOP
method stood at 56 p.p.
In short, it can be concluded from the results obtained in this section that it is not
advisable to use the PPP of GDP to calculate productivity levels for sectoral aggregates.
These measures are generally upwardly biased, as shown by the estimates made with
alternative conversion factors. This bias appears to be very significant in the case of certain
countries, including Spain, although it should be recalled that despite the fact that both the
HL and the ICOP methods are theoretically superior, in that they focus on the estimation of
conversion factors for manufacturing instead of the GDP as a whole, their implementation in
practice is not without problems, as mentioned in the previous section, and as a result the
estimated levels should be taken with caution.
VI. PRODUCTIVITY GROWTH IN MANUFACTURING
This section analyses the growth of productivity in the manufacturing sector as a
whole36. Chart 5 gives estimates of levels of labour productivity over the last thirty years for
the main European countries, the United States and Japan. The ICOP productivity series
come from the "Icop Industry Database, Summary Tables" and are constructed on the basis
of an estimate of productivity in the base year (generally 1987; 1984 in the case of Spain),
which is extrapolated over the whole period using the growth rates of real gross value
added and employment contained in the national accounts of the various countries. The HL
productivity series are based on 1990 estimates, which are extrapolated using data for
value added (at 1990 prices) and employment taken from the STAN database
14
According to these estimates, for the set of five countries considered, the gap
between the productivity of their manufacturing sectors and that of US manufacturing began
to grow from the early 1990s37, although the pattern of convergence observed up to that
time differs considerably from one country to another38. In France and Germany,
productivity convergence came to a halt at the beginning of the 1980s, when their relative
productivity reached a peak. Although in the late 1980s productivity growth exceeded that in
the US, during the 1990s the relative productivity of French manufacturing stabilised at
around 80% of the level in the US according to ICOP estimates, the level according to
estimates based on the HL method being somewhat lower, at around 72%. In Germany,
relative productivity deteriorated during the 1990s from the 74% level estimated by ICOP in
1991 to 67% in 2000 (according to estimates based on the HL method the levels were
somewhat lower, declining from 69% in 1991 to 62% in 1997).39
In Spain productivity convergence continued until the early or mid-1980s (according
to ICOP and HL, respectively). In any event, relative productivity declined gradually over the
1980s to stand at 46% (ICOP) and 54% (HL) of the level of productivity estimated for US
manufacturing. The decline in relative productivity continued during the first half of the
1990s to 41% (ICOP) and 51% (HL) in 1996. The ICOP data suggest that in the second half
of the 1990s this trend intensified considerably40.
The relative productivity of British manufacturing grew during the 1980s, to reach a
peak of 53% (HL) and 60% (ICOP) in 1992. Since then both sources point to a decline in
relative productivity which, according to ICOP estimates, was down to 48% of the US level
in 200041. Note that, according to the HL estimates, productivity in British industry was lower
than in Spanish manufacturing during most of the period analysed. This result is not
observed in the ICOP estimates. Finally, in Japan, the pronounced convergence which
characterised productivity growth in manufacturing in the 1970s came to a halt in the first
half of the 1980s, resuming again with great force in the second half of that decade. In 1991
the productivity of Japanese manufacturing stood at 73% (HL) and 86% (ICOP) of the US
level, exceeding the levels of German and French industry. This result contrasts with those
obtained in other studies; for example, van Ark and Pilat (1993), using value added per hour
worked as the measure of productivity, find that, despite the strong convergence that
characterised Japanese industry until the early 1990s, productivity levels in German
industry exceed those in Japan42.
Chart 6 shows productivity, wages and unit labour costs up to 1997. It can be seen
that the growth of compensation per employee in all the countries exceeded that recorded
in US manufacturing, except in the periods of sharp dollar appreciation, i.e. between 1981
and 1985, and more recently in the period 1996-1997. As a result, in some countries
(France, Germany and Japan), wages surpassed the US level in 1995 for the first time. In
Spain, compensation per employee held steady at around 54% of the US level over the
15
period 1993-1996. The higher growth of wages relative to the United States in the group of
countries analysed has been reflected in persistent growth of their relative unit labour costs,
which reached a high in 1995.
VII. CONCLUSIONS
The estimation of levels of labour productivity at the sectoral level for the purpose of
making international comparisons is problematic, given the difficulties associated with the
measurement of certain variables, such as the number of hours worked and the differences
in quality of the labour factor, and especially owing to the problems involved in obtaining an
appropriate exchange rate to express nominal magnitudes in a common currency. It has
been argued in this study that the PPP of GDP is not an appropriate conversion factor at the
sectoral level, generating upwardly skewed estimates of the relative productivity levels for
manufacturing as a whole. Two alternative conversion factors have been proposed, one
from the standpoint of spending (HL) and one from that of output (ICOP), which, unlike the
PPP of GDP, seek to reflect price differentials in manufacturing and not in the economy as a
whole. These specific PPPs for manufacturing are, at least in theory, preferable to the PPP
of GDP. However, in practice, they suffer from problems; namely, the PPPs from the
expenditure approach require an adjustment for distribution margins and taxes net of
subsidies and a correction for international trade, which could be done as proposed by
Jorgenson and Kuroda (1995) and Hooper and Vrankovich (1995). However, the fact that
the expenditure approach relies on prices of final goods means that it ignores altogether the
prices of intermediate goods, which account for more than one-third of the output of
manufacturing as a whole. The shortcomings of the production approach are due to the fact
that many products are not directly comparable because of quality differences, resulting in a
reduction in the coverage of industry output in bilateral comparisons, as well as the use of
unit value ratios instead of prices.
These methods generally coincide in producing lower estimates of the level of
productivity in manufacturing than do measures using the PPP of GDP. Specifically, with
regard to Spain, productivity in manufacturing is estimated to be 44% (ICOP) and 52% (HL)
of the level of US manufacturing in 1990, as against the 68% level obtained when using the
PPP of GDP.
Estimates of labour productivity in manufacturing in the main European countries
and Japan point to a widening gap relative to US manufacturing from the early 1990s. In
France, on ICOP estimates, the relative productivity of manufacturing is around 80% of the
US level, being somewhat lower according to the HL method (around 72%). In Germany,
relative productivity is lower, at slightly less than 70% of the level in US industry. The
productivity of Japanese manufacturing is estimated to be 73% (HL) and 86% (ICOP) of the
16
US level, exceeding the levels of Germany and France. This could be a result of the way in
which the labour factor is measured, i.e. ignoring the differences in hours worked per
person across countries. As discussed in Section I these differences are difficult to
measure, but the available data would suggest that hours worked per person in Germany
are far fewer than in Japan and the US. Finally, the growing divergence in the relative
productivity of British manufacturing from the early 1990s should be noted; its level is
estimated to have stood in the late 1990s at less than 50% of the US level.
1 This paper only addresses the measurement of levels of labour productivity. The estimation of levels of total factor
productivity in the context of international comparisons involves additional difficulties primarily associated with measurement
of the stocks of physical and human capital.
2 Dean E. and Harper, M. (1998).
3 This is the case of the “BLS International Comparisons of Manufacturing Productivity Program”.
4 Differences in labour quality are being ignored here.
5 OECD. Employment Outlook.
6 For some countries the data come from national labour force surveys of the (e.g. Spain, the UK), while in others the basic
data are obtained from business surveys (e.g. Germany, Italy), there being a third group of countries for which the data
come from a combination of both these types of statistic (e.g. Japan, Finland, the US). The problem of comparability across
countries of the series of hours worked arises from the heterogeneity of the sources. For example, the original series of
hours obtained from business surveys refer to “hours per job” and, moreover, normally reflect hours paid and not hours
actually worked, while the national labour force surveys provide information on hours worked per person irrespective of the
number of jobs held. This disparity will be particularly important when the incidence of moonlighting varies greatly from one
country to another. Maddison (1991) states that the series of hours worked per person are some of the poorest quality
series in labour market statistics. More recently, Nordhaus (2002) points to the lack of quality of the statistics for hours
worked as being one of the problems inherent in estimates of productivity.
7 A possible alternative would be to use the number of full-time equivalent jobs as a measure of the labour factor, instead of
the number of persons employed, which would enable the effect of the different incidence of part-time employment across
countries to be corrected for. However, as in the case of the estimation of hours worked, the methodology used to calculate
this statistic differs considerably from country to country. OECD (2001).
8 This programme was initiated in 1968 jointly by the United Nations and the University of Pennsylvania. Beginning with a first
round in which 10 countries were analysed in 1970, there were subsequently further rounds in 1975, 1980, 1985, 1990 and
1993, with the number of countries analysed increasing on each occasion until 118 were analysed in 1993. The OECD, in
collaboration with Eurostat has continued the process of collecting prices to estimate PPPs for its member states and, since
1993, performs this analysis every three years, while the World Bank has assumed the overall co-ordination of the ICP
programme in countries not belonging to the OECD.
9 OECD, Purchasing Power Parities and Real Expenditures.
10 Another study that analyses the relative productivity of Spanish manufacturing is Maté (1995) in which the exchange rate of
a base year (1985) is used to convert output into a common currency. The levels of productivity obtained in Maté’s study are
considerably lower than those estimated in Peñalosa (1994), although the growth rates are similar, as is to be expected.
11 The first round in 1980 was a pilot project that suffered from a lack of appropriate data. See Ward (1985)
12 The 30 OECD member countries plus another thirteen that include the EU candidate countries.
13 Each product is specified in detail to ensure that the various countries provide prices for comparable products and to avoid
any bias arising as a consequence of differences in quality. Ideally the specification of a product should include its brand
and model; in practice this is not feasible in many cases so that the product specifications are generic, describing their most
relevant characteristics without mentioning particular brands or models.
14 It is attempted when designing the common basket to combine two frequently incompatible objectives: representativeness
and comparability of the products. The method used by the OECD to calculate the PPPs at the basic heading level does not
require the same number of representative products for each country. The way to obtain “equal representativeness” is
requesting each country, for each basic heading, to nominate the price of at least one representative product. A
representative product is considered to be one sold in sufficient quantities to be able to assume that its price is indicative of
this type of product at the national level (note that the definition is sufficiently imprecise for the decision on the
representativeness of a product to be frequently subjective). For a nominated product to be included in the final basket of
goods and services there must be at least one other country that provides its price, preferably one in which the product in
question is sold in sufficient quantity. This procedure means that, at the basic heading level, each country is not directly
compared with all the rest, but only with those countries with which it has a certain affinity (gradualism), it being possible for
this pattern of affinity to change from one heading to another.
15 The method used by the OECD to calculate the PPPs at the basic heading level is called Elteto-Koves-Szulc (EKS), a
method which, by construction, guarantees the transitivity of PPPs between countries.
18
16 The way of achieving transitivity is replacing the Fisher PPP(i,j) obtained for each pair of countries i, j, using the aforesaid
procedure, by the geometric mean of PPP (i,j)2 and all the PPP(i,k)*PPP(k,j) for all k other than i,j.
17 The OECD warns explicitly of the inappropriateness of using the PPPs of GDP to made comparisons of productivity levels at
the individual industry level. See Schreyer P. and Koechlin F. (2002).
18 Baumol’s estimates point to the existence of absolute convergence for a group of 16 countries, while the rest of the studies
mentioned find conditional convergence for more than 100 countries. De Long (1988) criticises Baumol’s finding of absolute
convergence on the grounds that his sample suffers from serious “selection bias".
19 The results obtained by Bernard and Jones do not change when different measures of productivity are used: labour
productivity calculated both in terms of persons employed and hours worked, and two alternative definitions of total factor
productivity.
20 The condition is necessary but not sufficient. That is to say, it cannot be concluded that a conversion factor is appropriate
merely because it has been established that the levels of relative productivity are invariant to the choice of base year.
21 The existence of β convergence is a necessary, but not sufficient, condition for the existence of σ convergence.
22 Pilat (1996), using data from the OECD 1990 round, establishes a correspondence between the basic headings on the
spending side and 25 industrial branches on the side of production.
23 The OECD does not publish these data.
24 This report was commissioned by the OECD to assess that part of the ICP programme for which the OECD and Eurostat
are jointly responsible.
25 Castles’ main test of the quality of the PPPs estimated for the basic headings is based on a comparison between the PPPs
estimated for two base years (1990 and 1993) and the behaviour of the component corresponding to the heading in the
Consumer Price Index. In theory, the change in the PPP for a basic heading between two base years should reflect the
relative change (with respect to the reference country) in that heading in the CPI. In practice, substantial differences are
observed in some cases. In the absence of sudden changes in the structure of relative prices, this would suggest that the
representativeness and/or comparability of the products included in the calculations of the PPPs is not adequate.
26 This would justify the OECD recommendation (2002, p. 13) that "at the level of GDP, a broad and arbitrary rule of thumb is
that differences in indices of real final expenditure need to be at least five percentage points to be considered as statistically
significant"
27 Moreover, the fact that unit value ratios are used instead of observed prices may be problematic, in particular in the case of
goods for which there is a wide variety of qualities.
28 Although disaggregated PPPs are also available on the spending side for 1980, they are not included as they do not cover
all the 14 countries analysed; moreover the quality of the results for that year is much lower than for subsequent studies
(see Ward (1985))
29 In all the calculations the United States is taken as the reference country i=1
30 OECD (1993). As already mentioned, the most recent year for which the OECD publishes PPPs for a sufficiently
disaggregated level (50 spending categories) is 1993. In 1996 the disaggregation was reduced to 17 spending headings
which is insufficient to make the correspondence with manufacturing. 1990 has been chosen (instead of 1993) because the
STAN database provides value added at constant 1990 prices.
31 The estimated regression coefficient is –0.53 with a t statistic of 6.66.
32 The data for value added at 1990 prices in national currency and for employment used in the estimates presented in Chart 4
are taken from the OECD’s STAN (Structural Analysis Industrial Database) database. OECD (1999)
33 In fact, for the industrialised countries, relative unit labour costs in manufacturing are probably the best individual indicator of
the competitive position of a country (see Turner and Van't dack (1993))
34 The data for compensation of employees used in Table 5, from the STAN database, are compatible with the countries’
national accounts, so that there are no problems of lack of comparability across countries.
35 Only in Finland, Sweden and Japan is the productivity in manufacturing estimated by the ICOP procedure higher than that
obtained using the PPP of GDP as the conversion factor.
36 In a previous version of this paper estimations of the relative productivity levels in the main manufacturing industries were
computed using the ICOP method. Given that no comparable data are available from the expenditure viewpoint, and, as a
consequence there is no way to asses the quality of the ICOP estimates for the individual industries, this estimations have
been dropped from this version.
19
37 The dispersion of labour productivity falls slightly until 1992; thereafter the dispersion increases very rapidly.
38 Note that the results presented in this section differ, both as regards the time period considered (up to 1997 with HL
methodology and 2000 with ICOP) and the set of countries analysed with the results corresponding to the convergence
equations presented in Section III.
39 Recall that labour productivity is being measured here in terms of the number of persons employed and not the number of
hours worked as would be conceptually more appropriate (see Section I). This may be the reason why the level of
productivity is higher in French manufacturing than in German.
40 The OECD has not updated the STAN database beyond 1997. Currently a new database is being compiled (NEW STAN),
which is not fully comparable with “old” STAN, the database used in this study.
41 UK manufacturing productivity has been the subject of numerous studies, notable among them being those of the National
Institute of Economic and Social Research (NIESR) and the McKinsey Global Institute (1998). In one of the most recent
NIESR publications, O'Mahony and de Boer (2002) estimate relative productivity (in terms of value added per hour worked)
for a group of countries (US, UK, France, Germany and Japan). They obtain results that are not very different from the
ICOP estimates, except in the case of Japan (O'Mahony has advised me that the ICOP estimates for Japan are more
accurate than those obtained with the NIESR database). The Mckinsey results for British manufacturing as a whole are
based on the PPP of manufacturing estimated using the ICOP procedure.
42 These authors estimate that in 1990 the hourly productivity of Japanese manufacturing was 80% of that of US
manufacturing, while that of German manufacturing was 86%.
References
Ark, B Van (1992). "Comparative Productivity in British and American Manufacturing"
National Institute Economic Review, November, pp. 63-74.
Ark, B. Van (1995). "Producción y productividad en el sector manufacturero español. Un
análisis comparativo 1950-1992" Información Comercial Española, 746, pp. 67-77.
Ark, B. Van and Pilat, D. (1993). "Productivity levels in Germany, Japan, and the United
States: Differences and Causes”. Brooking Papers on Economic Activity: Microeconomics,
2, pp.1-69.
Ark, B. Van and Timmer, M. (2001). "PPPs and International Productivity Comparisons:
Bottlenecks and New Directions". Joint World Bank-OECD Seminar on Purchasing Power
Parities.
Ark, B. Van and Monnikhof (2000). "Productivity and Unit Labour Costs Comparisons: A
Data Base" International Labour Office.
Barro, R. and Sala-i-Martín (1991). "Convergence Across States and Regions" Brookings
Papers on Economic Activity, 1, pp. 107-182.
Barro, R. and Sala-i-Martín (1992). "Convergence" Journal of Political Economy, 100, pp.
223-251.
Baumol, W (1986). "Productivity Growth, Convergence and Welfare” American Economic
Review, 76, pp. 1072-1085.
Bernard, A. and Jones C. (1996) “Comparing Apples to Oranges: Productivity Convergence
and Measurement Across Industries and Countries” American Economic Review,
December, pp.1216-1238.
Castles I. (1997). "Review of the OECD-Eurostat PPP Program" At
http://www.oecd.org/pdf/M00019000/M00019094.pdf
Dean, E and Harper, M. (1998). “The BLS Productivity Measurement Program” Conference
on Research in Income and Wealth: New Directions in Productivity Analysis.
De Long, B. (1988). "Productivity Growth, Convergence and Welfare: Comment" American
Economic Review, 78, pp. 1138-1154.
Dollar, D. and Wolff, E. (1993). Competitiveness, Convergence, and International
Specialization, Cambridge. MIT Press.
Gollop, F. M. (1979). "Accounting for Intermediate Input: The Link Between Sectoral and
Aggregate Measures of Productivity" in Measurement and Interpretation of Productivity,
Washington, D.C: National Academy of Sciences.
Hooper, P. and Larin, K (1988) "International Comparisons of Labor Costs in
Manufacturing". International Finance Discussion Papers, No 330, Board of Governors of
the Federal Reserve System.
Hooper, P and Vrankovich, E. (1995) "International Comparisons of the Level of Unit Labor
Costs in Manufacturing" International Finance Discussion Papers, No 527, Board of
Governors of the Federal Reserve System.
Jorgenson, D.W. and Kuroda (1995). "Productivity and International Competitiveness in
Japan and the United States, 1960-1985" Chapter 9 of Productivity. International
Comparisons of Economic Growth, ed. Jorgenson, D.W. Cambridge. MIT Press.
Maddison, A. (1991). Dynamic Forces in Capitalist Development. A long-Run Comparative
View. Oxford University Press, Oxford.
Mankiw, G., Romer, D. and Weil, D. (1992) "A Contribution to the Empirics of Economic
Growth" Quarterly Journal of Economics, vol 107, No 2, pp 407-437.
Maté, J.M. (1995). "La productividad del trabajo en España y en la UE. Análisis comparado
del conjunto de la economía y de la industria manufacturera", Papeles de Economía
Española no 63, pp 112-125.
McKinsey Global Institute (1993) Manufacturing Productivity, McKinsey Global Institute,
Washington D.C.
McKinsey Global Institute (1998) UK Economic Growth Productivity and Employment,
London.
McKinsey Global Institute (2002) Reaching higher productivity growth in France and
Germany, London
Nordhaus, W. (2002). "An Economist's View of the Statistical State of the Nation" Testimony
Before the Joint Economic Committee
OECD. Employment Outlook.
OECD (1993). Purchasing Power Parities and Real Expenditures. EKS Results. 1990
benchmark year.
OECD (1998). International Sectoral Data Base 1970-1993
OECD (1999). The OECD STAN Database for Industrial Analysis. 1978-1997.
OECD (2001). Measuring Productivity. OECD Manual
OECD (2002). Purchasing Power Parities and Real Expenditures. 1999 benchmark year.
O'Mahony, M and de Boer, W. (2002) "Britain's Relative Productivity Performance: Updates
to 1999" Final Report to DTI/Treasury/ONS. National Institute of Economic and Social
Research.
Peñalosa, J.M. (1994). "The Spanish Catching-up Process: General Determinants and
Contribution of the Manufacturing Industry". Banco de España. Working paper 9428.
Pilat, D. (1996). "Labour Productivity Levels in OECD Countries: Estimates for
Manufacturing and Selected Service Sectors". OECD/GD(96)138
Sorensen, A. (2001) “Comparing Apples to Oranges: Productivity Convergence and
Measurement Across Industries and Countries: Comment” American Economic Review, vol
91, No 4, 1160-1167.
Schjerning, B. (2001) “Measuring Productivity Convergence in the Manufacturing Sector”.
Institute of Economics, University of Copenhagen. Mimeo.
Schreyer P. and Koechlin F. (2002). At
http://www.oecdwash.org/DATA/STATS/ppp_1999bench.pdf
Turner, P. and Van't dack, J. (1993). "Measuring International Price and Cost
Competitiveness" Bank for International Settlements Economic Papers, No 39.
Varjonen, S. (2001). " Consistency Between GDP Based on PPPs and National Accounts
Time Series" OECD STD/NA(2001)34
Ward, M (1985). Purchasing Power Parities and Real Expenditures in the OECD. OCDE,
Paris.
Table 1. Convergence equations
A. β convergence with PPP of GDP
0i
P calculated with the PPP of GDP for different base years
Base year β
t
R2
1970 -0.0031 -0.29 0.01
1980 -0.0135 -1.30 0.12
1985 -0.0187 -1.98 0.25
1990 -0.0231 -2.63 0.37
1993 -0.0253 -3.15 0.45
B. β convergence with the PPP of manufacturing (Hooper Larin)
0i
P calculated with the PPP (HL) for different base years
Base year β
t
R2
1985 -0.0173 -2.50 0.34
1990 -0.0187 -2.45 0.33
1993 -0.0215 -3.28 0.47
Source: OECD (1998). International sectoral Data Base. 1970-1993.
The fourteen countries included are: Australia, Belgium, Canada, Denmark, Finland, France, Italy,
Japan, the Netherlands, Norway, Sweden, the United Kingdom, the United States and Germany.
Note: The β parameter is estimated for the equation
iii Pp
ε
β
++= 0
ln
)
Table 2. Manufacturing. Alternative conversion factors. 1990
(unidades de moneda nacional por $US)
PPP of GDP PPP of manufacturing (ICP) PPP of manufacturing (ICP) PPP of manufacturing PPP mixed approach exchange rate
(ICP) (HL method) adjusted for (ICOP)
margins and
imports and exports
United States 1,00 1,00 1,00 1,00 1,00 1,00
Japan 195,30 203,03 217,90 154,30 156,67 144,80
Germany 2,09 2,28 2,36 2,11 2,21 1,62
France 6,61 7,76 8,07 7,04 7,36 5,45
Italy 1.421,00 2.004,80 1.198,10
United Kingdom 0,60 0,71 0,79 0,73 0,71 0,56
Canada 1,30 1,45 1,43 1,32 1,38 1,17
Netherlands 2,17 2,45 2,28 2,44 1,82
Sweden 9,34 9,60 8,44 8,94 5,92
Spain 109,50 142,00 139,00 101,90
Source: Pilat (1996) and Van Ark and Monnikhof (2000). The estimates of PPPs (HL method) are those discussed in Section V hereof.
Bel
g
ium Denmar
k
France German
y
Greece Ireland Ital
y
N
etherlands Portu
g
al S
p
ain U
K
A
ustria Finland Sweden Norwa
y
Ja
p
an Canada US
Purchasin
g
Power Parit
y
Food, bevera
g
es and tobacco 43,04 10,97 6,94 2,05 160,60 0,83 1494,35 2,21 134,25 121,33 0,65 15,18 8,62 12,30 13,83 269,76 1,54 1,00
Clothin
g
and footwear 61,65 13,26 11,08 2,88 217,04 0,90 2207,72 2,88 194,14 181,09 0,70 19,17 8,70 9,29 12,49 214,38 1,57 1,00
Fuel and
p
ower 55,60 13,52 9,63 2,83 204,22 1,07 2262,19 2,78 207,77 173,39 0,88 20,29 5,55 9,51 7,69 287,22 1,04 1,00
Household e
q
ui
p
ment 42,07 9,36 7,01 1,95 144,83 0,75 1549,43 2,14 115,86 123,45 0,61 14,37 6,41 8,59 9,20 189,66 1,36 1,00
Medical and
p
harmaceutical
p
roducts 25,08 7,94 3,02 1,74 62,10 0,58 767,74 2,12 64,27 55,32 0,43 11,29 4,07 4,95 5,63 99,19 1,15 1,00
Personal trans
p
ort e
q
ui
p
ment 38,63 14,82 7,15 1,95 342,86 1,09 1554,29 2,59 271,89 164,00 0,76 15,20 8,72 8,17 13,68 146,29 1,36 1,00
Recreational e
q
ui
p
ment 53,43 9,50 8,64 2,35 275,41 0,88 1910,17 1,99 152,84 166,55 0,62 19,62 8,13 10,80 12,60 159,57 1,51 1,00
Machiner
y
and e
q
ui
p
ment 10,92 9,28 2,64 260,87 0,94 2123,85 2,82 223,98 169,43 0,87 16,73 6,90 9,13 12,71 196,31 1,45 1,00
Total manufacturin
g
46,73 11,09 7,76 2,28 188,75 0,87 1727,67 2,45 162,46 142,00 0,71 16,29 7,50 9,61 12,04 203,03 1,45 1,00
Share in ex
p
enditur
e
Food, bevera
g
es and tobacco 0,25 0,27 0,27 0,25 0,37 0,29 0,27 0,26 0,33 0,29 0,25 0,26 0,25 0,23 0,25 0,23 0,23 0,23
Clothin
g
and footwear 0,13 0,12 0,12 0,13 0,15 0,14 0,15 0,13 0,14 0,15 0,14 0,15 0,11 0,14 0,13 0,14 0,13 0,15
Fuel and
p
ower 0,02 0,03 0,02 0,02 0,02 0,02 0,02 0,02 0,02 0,02 0,02 0,02 0,03 0,02 0,04 0,02 0,03 0,03
Household e
q
ui
p
ment 0,14 0,12 0,13 0,13 0,14 0,12 0,13 0,13 0,12 0,12 0,13 0,12 0,12 0,12 0,12 0,11 0,13 0,11
Medical and
p
harmaceutical
p
roducts 0,03 0,02 0,05 0,04 0,04 0,03 0,04 0,02 0,03 0,04 0,02 0,02 0,03 0,02 0,02 0,02 0,02 0,03
Personal trans
p
ort e
q
ui
p
ment 0,10 0,07 0,09 0,10 0,06 0,07 0,09 0,08 0,07 0,09 0,11 0,09 0,09 0,10 0,07 0,07 0,11 0,11
Recreational e
q
ui
p
ment 0,07 0,09 0,07 0,08 0,05 0,06 0,07 0,09 0,05 0,05 0,09 0,06 0,07 0,08 0,08 0,08 0,08 0,09
Machiner
y
and e
q
ui
p
ment 0,25 0,28 0,24 0,26 0,19 0,27 0,23 0,27 0,22 0,24 0,25 0,28 0,31 0,29 0,28 0,33 0,25 0,26
Total manufacturin
g
1,00 1,00 1,00 1,00 1,00 1,00 1,00 1,00 1,00 1,00 1,00 1,00 1,00 1,00 1,00 1,00 1,00 1,00
Source: OECD (1993). Purchasing Power Parities and Real Expenditures, EKS Results 1990.
This is the conversion factor for aggregate manufacturing proposed by Hooper and Larin (1989):
Table 3. PPPs and the industries' shares in expenditure in 1990.
1
1
Table 4. Productivity, Wages and Unit Labour Costs. Average 1985-1995 (US=100)
(in order of increasing productivity)
Productivity Compensation ULCs
per employee
Portugal 24 19 78
Denmark 42 71 167
Greece 43 36 83
Norway 48 87 182
United Kingdom 51 63 123
Spain 54 49 92
Finland 56 77 138
Sweden 57 82 144
Austria 58 80 136
Italy 60 60 100
Germany 66 88 135
Japan 67 81 122
Belgium 70 82 118
France 71 83 117
Netherlands 77 84 110
United States 100 100 100
Coefficient of variation 0,29 0,30 0,23
Note: the productivity has been converted to a common currency using the PPP of HL for 1990.
Compensation per employee has been converted to a common currency using the exchange rate for each year.
Source: GVA, Employment and Compensation per employee are taken from the STAN database.
Table 5. Productivity in manufacturing 1990
(in order of GDP per capita)
with with with
PPP of GDP ICP (HL) ICOP
Portugal 39,5 25,2 25,1
Spain 67,7 52,2 44,4
Netherlands 86,9 76,9 84,1
United Kingdom 62,2 52,7 57,0
Begium 85,27 72,0 83,7
Finland 64,2 54,7 72,1
Sweden 55,4 53,8 70,0
France 85,4 72,8 77,6
Japan 74,4 71,5 84,8
Germany 73,4 67,2 72,7
Canada 79,0 71,2 77,3
United States 100,0 100,0 100,0
corr. with GDP per capita 0,73 0,85 0,91
Source: The GVA and employment data are from the STAN database. PPP of GDP and its components on
the side of spending from OECD, 1992. Purchasing Power Parities and Real Expenditures, EKS Results 1990.
ICOP estimates from van Ark and Timmer, M (2001)
Chart 1
Source: OECD (1998) International Sectoral Data Base 1970-1993. The 14 countries included
in the sample are those reported in Table 2
Sigma convergence (with PPP of GDP)
(normalised std 1970=1)
0,60
0,70
0,80
0,90
1,00
1,10
1,20
1,30
1,40
1970 1972 1974 1976 1978 1980 1982 1984 1986 1988 1990 1992
1970 1980 1985 1990 1993
Sigma convergence (with PPP HL)
(normalised std 1970=1)
0,70
0,75
0,80
0,85
0,90
0,95
1,00
1,05
1,10
1970 1971 1972 1973 1974 1975 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993
1985 1990 1993
Chart 2
Source: OECD (1993). Purchasing Power Parities and Real Expenditures, EKS Results 1990.
1990. PPP of manufacturing/PPP of GDP
(in order of 1990 GDP per capita)
20
40
60
80
100
120
140
160
GRC
PRT
IRL
ESP
NLD
ITA
UK
NOR
BEL
FIN
AUT
DNK
SWE
FRA
JPN
DEU
CAN
US
Chart 3
Source:OECD (1993) Purchasing Power Parities and Real
expenditures. EKS Results 1990.
90
110
130
150
40 50 60 70 80 90 100 110 120 130
GDP per capita
PPP of manufacturing/PPP of GDP
%
DEU
ITA FRA
DNK
PRT
JPN
GRC
NLD
A
UT
FIN
%
SWE
BEL
IRL
ESP
CAN
US
Ratio between the PPP of GDP and the PPP of manufacturing against income per capita
Chart 4
Source: STAN database for GVA and employment. PPP from OECD (1993)
Purchasing Power Parities and Real Expenditures, EKS Results 1990.
Labour productivity in manufacturing in 1990
(in increasing order according to estimates with PPP of manufacturing)
0
20
40
60
80
100
120
PRT
DNK
GRC
NOR
ESP
UK
SWE
FIN
AUT
ITA
DEU
CAN
JPN
BEL
FRA
NLD
US
with PPP of GDP with PPP of Manuf
Chart 5
Source: Upper panel: STAN database for GVA and employment. PPP from OECD (1993) Purchasing Power
Parities and Real Expenditures, EKS Results 1990. Lower panel:ICOP Industry Database
Productivity in manufacturing with PPP HL 1990 (US=100)
30
40
50
60
70
80
90
1970 1972 1974 1976 1978 1980 1982 1984 1986 1988 1990 1992 1994 1996 1998 2000
FRA DEU JPN ESP UK
Productivity in manufacturing ICOP (US=100)
30
40
50
60
70
80
90
1970 1972 1974 1976 1978 1980 1982 1984 1986 1988 1990 1992 1994 1996 1998 2000
FRA DEU JPN ESP UK
Chart 6
Source: STAN database for GVA, employment and compensation per employee.
PPP from OECD(1993), Purchasing Power Parities and Real Expenditures, EKS Results 1990.
Productivity estimated with PPP HL in 1990 (US=100)
40
45
50
55
60
65
70
75
80
85
1970
1971
1972
1973
1974
1975
1976
1977
1978
1979
1980
1981
1982
1983
1984
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
FRA DEU JPN UK ESP
Compensation per employee estimated with current exchange rate
(US=100)
0
20
40
60
80
100
120
1970
1971
1972
1973
1974
1975
1976
1977
1978
1979
1980
1981
1982
1983
1984
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
FRA DEU JPN UK ESP
Unit labour costs (US=100)
0
20
40
60
80
100
120
140
160
180
200
1970
1971
1972
1973
1974
1975
1976
1977
1978
1979
1980
1981
1982
1983
1984
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
FRA DEU JPN UK ESP
ResearchGate has not been able to resolve any citations for this publication.
Article
Conventional wisdom and current practice in India appear to privilege total factor productivity (TFP), as a measure of productivity, rather than labour productivity. However, labour productivity is a measure of potential consumption and, as such, a leading claimant for the indicator of standard of living, which makes it important in any programme of poverty reduction. Estimates of labour productivity also give us useful additional information in evaluating the reforms undertaken in India in the 1990s.
Article
This article presents the results of a benchmark comparison of the relative level of manufacturing productivity in the United Kingdom and the United States for 1987. It shows value added per hour worked in the UK at 58 per cent of the corresponding level in the United States. The UK appears to show relatively high productivity levels in basic goods industries, including textiles, chemicals and basic metals. The benchmark results are extrapolated backwards to 1968 and forwards to 1990, showing a narrowing of some ten percentage points of the productivity gap between the two countries. It is also shown that differences in capital intensity account for only a small part of the productivity gap.
Article
________________________________________________________________ I am delighted to have the opportunity to discuss some of the important questions regarding the state of federal economic statistics in our rapidly evolving economy. Improving our nation's statistical infrastructure is an important goal. The emergence of new technologies and the growing importance of technologically advanced service industries have created new data-collection and measurement challenges for our statistical agencies. I will review some of the major issues and make a handful of recommendations for improvements. I might begin with a word of background on my interest in economic statistics. For most of the last decade, I served on the National Academy of Science's Committee on National Statistics. This body is charged "to select and study statistical topics to improve the effectiveness of the federal statistical system." Its reports have reviewed such issues as the definition of poverty, the American Community Survey, better measures of the cost of living, and augmented accounting.