predicts increasing returns to scale if the Verdoorn coefficient β1 turns out to be greater
than 0, see Fingleton (2001). A positive, but declining slope parameter is reported in
most empirical studies, see Harris and Lau (1998) and Léon-Ledesma (2000) for the UK
and Spanish economy, respectively. Increasing returns may be explained by a variety of
endogeneous growth models, see Aghion and Howitt (1998) for a survey.
A serious issue with Verdoorn’s law lies in the ignorance of the role of capital, that can
be substituted for labour. Because of the omitted variable problem, estimation of the
parameters β0 and β1 from the relationship (2.1) seem to be biased. Suppose output is
produced by a Cobb-Douglas technology,
(2.2) yt =τ+ηlt +λkt ,
where l, k and τ are the growth rates of labour, capital and technology, respectively.
Since employment growth is the difference between output and productivity growth, the
relation
(2.3) Pt =τ/η+[(n- 1)/n]yt + (λ/η)kt
is implied. In general, the bias is proportional to the coefficient from regressing capital
growth on output growth, see e.g. Greene (2003, pp. 148). However, the connection
(2.1) between productivity and output growth can be defended, if capital growth equals
output growth. This is in line with the stylised fact of a more or less constant capital-
output ratio, see e.g. Jones (1998, pp. 12) and Mauβner und Klump (1996, pp. 7). In this
case, the parameters β0 and β1 are specified as β0 = τ / η and β1 = (η + λ - 1)/ η, for
which unbiased estimates can be obtained from Equation (2.1). The returns to scale pa-
rameter cannot be revealed form the Verdoorn coefficient β1 without knowledge of the
production elasticities. The constant β0 corresponds with the rate of technological pro-
gress, divided by the production elasticity of labour.
Verdoorn’s law was originally considered for the industrial sector. Since the service
sector have become of increasing importance, the hypothesis should be examined using
total output data. Due to the high correlation of output and productivity growth, how-
ever, spurious regressions can easily occur. If employment growth is constant, a perfect
correlation between productivity and output growth would appear, which is not infor-