rates apply together with the level of the premium are hxed by legislation. By
contrast, changes in ʌ in (11) are employment driven, dependent on the pro-
portion of workers choosing to undertake overtime. Together, these variables
recognise an essential overtime breakdown underlined in the work of Trejo
(1993) dealing with union effects on overtime working.
Why is it important to differentiate between W, μ and ʌ in (11)? Stan-
dard and overtime components in the wage may well differ with respect to the
cyclical indicator with which they most strongly со-vary as well as the degree
to which they are in- phase with their dominant cyclical influence. Suppose
at the trough of an output or employment cycle firms hold underutilized labor
stock. The initial phase of the upturn may be dominated by labor dishoard-
ing which involves effective hours being bought into line with paid-for hours.
Planned increases in the stock of employment will depend on the anticipated
time required to restore normal work intensity together with the expected de-
gree and length of the ensuing growth period. In the later phases of the upturn,
employment adjustment lags associated with search, hiring, and training may
require firms to resort to temporary hours increases as an employment buffer
(Nadiri and Rosen, 1973).
Overtime hours would be expected to feature prominently in this latter
adjustment process. The degree to which the firm will extend overtime working
will depend on the relative costs of alternative buffers, such as a greater than
planned run-down of inventories ( Topel, 1982). To the extent that overtime
is used to offset shortfalls in planned employment growth - especially at times
of exceptional demand peaks - overtime cycles are likely to be relatively short
and, given potential substitution, may well correlate strongly with changes
in business inventories. Moreover, since premium rates apply automatically
to changes in overtime hours - with no pay negotiation involved - we would
expect that, to a large degree, overtime pay would be in-phase with the cycle
(see the discussion in Hall and Lilien, 1979).
Overtime adjustments occur not only through hours adjustments, μ, but
also through employment adjustments, ʌ. Whether or not these two vari-
ables are close substitutes is an empirical question. Speedier responses may be
achieved through changing the hours of workers already committed to over-
time rather than through persuading marginal workers to move in and out of
overtime completely. In this case, μ rather than ʌ is more likely to represent
12