coefficient β, for each sector. A negative estimated coefficient β indicates the existence of so-
called β-convergence. We start in section 5.1 by testing for unconditional convergence,
assuming that energy- and labour productivity converge towards a unique steady state for all
countries included in the data set. In section 5.2 we relax this assumption and test for
conditional convergence, assuming productivity levels to converge towards multiple steady
states that are conditional on country-specific characteristics. Finally, as part of this analysis
we try to identify the country-specific characteristics that determine (differences in) energy-
and labour-productivity growth across countries.
5.1 Unconditional β-convergence
We test for unconditional β-convergence by regressing for each sector the growth rate (g) of,
respectively, energy- and labour productivity (y), on its initial level (and a constant a),
generating an estimate of β, according to:
git = log(y)i,t - log(y)i,t-1 = a + β ln(y)i,t-1 + eu (1)
with i and t denoting, respectively, the cross-country and the time-series dimension,
while εit is the standard error. Following Islam (1995) we use five-year time intervals in order
to reduce the influence of business-cycle fluctuations and serial correlation of the error term.
Hence, the growth rate (g) in equation (1) is an average over a five-year period (if t = 1975,
for example, t-1 = 1970). The results are presented in Table 2.
(Table 2, Page 28)
From the table it can be seen that we obtain a statistically significant negative estimate of β
for energy-productivity growth in most sectors, except for Total (i.e., the macroeconomic
level), Chemicals, Iron and Steel, Non-Ferrous Metals, Paper and Wood. In terms of labour-
productivity growth we found β to be statistically significant in all sectors, except for
aggregate Manufacturing and Non-Ferrous Metals.14
14 We also estimated equation (1) including a period-specific fixed effect η t according to git = α + β ln(y ) i,t-1 + η t
+ εit . The regression results with these period dummies included do not substantially improve the estimates in
most sectors, except for Non-Ferrous Metals and in terms of labour productivity also for Chemicals, Iron and
Steel and Machinery. These findings suggest, that in spite of a few exceptions, in general there is not much
evidence for substantial differences in growth rates between the time periods included. Details are available
upon request.
11