Stage II of the policy lasted only from April to October
4 customary timing of settlements of different size groups.
Therefore, whilst one might expect $ not to be dominated
by purely economic factors (and this is the conclusion of Johnston
and Timbrell, 1973, amongst others) the very presence of incomes
policy will have an important influence. Evidence on contract
length from the Aberdeen database finds some support for the
idea that contract length is reduced in times of rapid inflation
but no conclusive evidence to draw inferences regarding the role
of incomes policy on contract duration. ^ For example, a wage
freeze might be expected to substantially reduce the level of wage
settlements. To deflate the wage variable b y the proportion of
workers settling in a given period would clearly be an inappropriate
way to measure the influence of incomes policy since the main aim
of policy is to reduce the rate of growth of the wage bill regardless
of whether this comes from a reduction in the size of average
6
settlement or the number of settlements. In Chapter S the
influence of policy rules on the timing of settlements was
discussed and this influence would be increased by any
anticipation or catch-up effects. In principle, incomes policies
whi c h involve a twelve-month rule for settlements should also
affect the timing of settlements but if such a rule follows a
period of freeze then the rule may not have such an effect since
the freeze would have already extended the length of contract.
7
Tarling and Wilkinson (1977) using data on national negotiations
■ I
6.5
behaviour with a consistently higher proportion of bargaining
groups settling in each year and with a near
100
per centsettlement rate since 1974. This might be explained by the
institutionalisation of the
12
-month rule by the fact that priceand wage inflation was significantly higher after 1970 and under
these circumstances one might expect groups to settle more 8
rapidly than otherwise.
The preferred approach to this question is to isolate
the normal (if there are any) institutional patterns in wage
settlements from the total as deferred settlements, anticipation
and catch-up effects can be dealt with explicitly within the equation.
The issue regarding the differencing of the wage variable
is also related to the timing of settlements. Rowley and Wilton
(1973) set out a model where wages are set annually for all
workers and then fixed until the next annual settlement and the
relative change in the wage rate over its value four quarters
earlier is given by a moving average of the importance of the
groups settling within each quarterly period. On the assumption
that these weights are equal the model then induces a fourth-order
moving average serial correlation process which results in
biased estimates of the standard errors.
Returning to (6.1) we have Alog Alog W S C
if Alog WSt - a+
8
X t + et(6.2)
then Ashenfelter and Pencavel (1975) show that:
Alog - a $ t B$tX t ♦ v fc ( S 3 )
where v £ — q>ce t . The disturbance term therefore depends systematically
6.6
They go on to show that for:
logWt - log W
t_4
- « ♦0
Z * t-i V ii - 0
. X^ . + E
i
-0
t-i(6.3)
and ♦ - i for all periods; i.e. workers settle evenly throughout
the year then:
3 3
log W t - logW
t_4
- a +6
Z iXfc_ i ♦ E . v ^ (6.4)i
-0
i-0
- « + i SEX . + Z
Ashenfelter and Pencavel then argue that there is no reason
why (6.3) should not be fitted directly. However, the same
objections as outlined earlier remain. The major question is whether
the lags on the price and real wage variables should be
1
quarteror 4 quarters. If we were dealing with a group of workers settling
at period t, having previously settled in t-4, then the four-quarter
lag might seem the more appropriate. However, the logic of the real
wage catch-up implies that it is the current size of the real wage
gap that is relevant not the gap at the time of the last settlement.
Therefore the real wage lagged one quarter is the correct variab le.
In terms of the real wage model there is no additional role for
price inflation since this is already incorporated in the real wage
catch up. Therefore its role must relate more to expected inflation.
In earlier versions of the model the specification used was
largely that of equation (3. 12) ; i.e.
w t - bQ (
1
-k) + b 1 (1
-k) u t + b2
(l-k)pt6.7
but including also the term w c_^> The rationale for this term was
that it reflected the influence of settlements made by other workers
in the previous period. A term in the proportion of workers
settling might be included on the right hand side of the equation.
However, when the settlements term is decomposed into non-seasonal
and seasonal elements (using seasonal dumnies to represent the
seasonal elements) it is the latter which provide all of the 9
statistical explanation Therefore it is clear that the
proportion of settlements is not a useful addition to the equation,
rather it is the seasonal factors reflecting regular changes in
the proportion of workers settling in each period which helps to
explain variations in the rate of wage inflation.
6.2 Anticipation and catch-up effects
Previous chapters have emphasised the importance of allowing
as far as possible for timing effects relating to policy. In