Moreover, the database includes data on Investment, Energy Prices, Compensation of
Employees, Export and Import - all at the sectoral level. The sector-specific energy prices are
constructed by dividing sector-specific expenditures on energy over total sectoral energy
consumption. The sector-specific expenditures are calculated as the product of the sectoral
consumption of the four main energy carriers (Coal, Natural Gas, Electricity, Oil) - available
from the Energy Balances - and the (annual) price of each energy carrier at the aggregate
industrial sector - available from the IEA Energy Prices and Taxes series. In addition, some
missing aggregate energy price data series have been constructed.6
All currency-denominated variables are in 1990 US$ and have been converted by the
OECD using 1990 purchasing power parities. In principle the theoretically most appropriate
conversion factors for productivity comparisons at the sectoral level are to be based on a
comparison of output prices by industry of origin, rather than on expenditure prices (see, for
example, van Ark and Pilat 1993). Expenditure PPPs exclude the part of output that is
exported, while they include imported goods produced elsewhere; they take account of
differences in trade and transport margins and indirect taxes between countries, and they do
not cover intermediate products. The main problem in using the production or industry-of-
origin approach, however, is the limited availability of producer-price based PPPs, in
particular for non-Manufacturing sectors (van Ark 1993).7 Moreover, we have no a priori
reason to presume that the drawbacks of expenditure PPPs differ substantially across
countries. Hence, we follow most studies including cross-country productivity comparisons
using expenditure PPPs, which enables us to do a systematic cross-country convergence
analysis of energy- and labour-productivity performance at a high level of sectoral detail.
Obviously, the results presented in this paper should be interpreted with caution, bearing in
mind the before mentioned issues.8
6 Further details on the dataset, including descriptive statistics, are available upon request.
7 This limited availability is due to some problems inherent to the industry-of-origin approach: producer prices
(i.e., production values divided by output quantities) may not properly account for cross-country quality
differences and imply aggregation problems for they are available only for a sample of goods (partly because of
confidentiality problems), and because the production structure among countries tends to be less comparable
than the consumption structure due to specialization tendencies in production according to comparative
advantage (Pilat 1996).
8 For a discussion of this issue in empirical analyses of convergence at the sectoral level see S0rensen (2001)
and Bernard and Jones (2001).