mative at all. This problem is avoided in a specification between employment (l) and
output growth,
(2.4) lt =α0 +α1yt, α0 =-β0,α1 =1-β1,
that has been already favoured by Kaldor (1975). A high correlation between productiv-
ity and output growth does not imply the same for employment and productivity
growth. In the case of perfect correlation of the former variables, for example, the latter
variables are not correlated at all. Thus, spurious regression arising from common
trends driving productivity and output growth is avoided with Equation (2.4). Moreover,
the best linear predictor of employment growth is given by a regression with the right-
hand-side of Equation (2.4) as its systematic part.
A Verdoorn coefficient of 0.5 in (2.1) implies a marginal employment intensity α1 of
the same size in (2.4). This means that a 1 percent growth in output would stimulate
employment by half a percent on the average. The underproportional reaction is due to
efficiency gains, which can be realized more easily in periods of higher output growth.
They may be traced to inter alia manpower reserves, increases in working hours and
higher labour intensities.
The threshold of employment (yE) indicates output growth for which employment is
constant (lt=0). In terms of the model parameters the threshold level yE reads:
(2.5) yE =-α0 .
α1
According to the parameter -α0, it is positively related to the rate of technological pro-
gress (τ) and negatively related to the production elasticity of labour (η). In addition, a
higher marginal employment intensity (α1) will reduce the threshold. Provided that out-
put growth is above this bound, employment will be stimulated. If output growth falls
beyond the threshold, losses in employment are expected on the average. In this case,
output growth is not sufficient to compensate for the rise in productivity due to techno-
logical progress and employment shrinks. According to (2.4) and (2.5), the evolution of
employment,
(2.6) lt =α1(yt-yE),