length. Though all of the aforementioned studies test for a relationship between nominal uncertainty and contract length, none of the studies address the point
made by Danziger concerning uncertainty regarding aggregate real shocks to the economy.
3. Theoretic underpinnings
The theoretical work of Gray and later of Danziger on contract duration identify two key factors as the chief determinants of labor contract length: the cost of
negotiating a labor contract and uncertainty. Negotiation costs
2
are hypothesized to be directly related to contract duration. The effect of uncertainty in the
economic environment on contract length, on the other hand, depends on the type of uncertainty involved, with nominal uncertainty predicted to be associated with
contracts of shorter duration and real shocks associated with longer contracts.
A third factor not addressed by Gray or Danziger that likely plays a role in determining contract length is bargaining power.
3
The older, industrial relations literature on the subject does, however, emphasize the role of bargaining power in
Ž .
this context. Stieber p. 140 , for example, notes that a long term agreement ‘‘is a very substantial union concession and ‘must not be sold cheaply’’’. This follows
because the union’s objective is to provide services to its constituency. One of the chief ways that a union demonstrates its usefulness is by negotiating and winning
increases in compensation for the rank and file. If a long term contract is signed, then this very important channel through which the union attends to the needs of
its members is lost for a period of years. This is not to say that the union will prefer a series of short contracts in order to remain continually in the bargaining
process, for the union has some degree of preference for length also.
4
But this
2
Ž .
Ž .
Gray 1978 and Wallace and Blanco 1991 refer to these costs as ‘‘fixed costs of negotiation’’. They are fixed only in the sense that the firm would like to amortize these costs over as long a period
as possible so as not to have to incur them very often. They can vary at the bargaining pair level if, for example, a strike is incurred and higher bargaining costs are borne than would have been the case had
no strike taken place. Such costs also vary across bargaining unit pairs to the extent that the degree of Ž
complexity in negotiations varies across bargaining unit pairs e.g., a contract negotiated for 50,000 .
workers will be more complex than one negotiated for 500 workers . Other authors use different Ž
. terminology to identify the same concept. Christofides and Wilton 1983 refer to these costs as
Ž .
Ž .
‘‘negotiating costs’’; Vroman 1989 refers to these costs as ‘‘contracting costs’’; and Murphy 1992 refers to the concept as ‘‘transaction costs of bargaining’’. For present purposes, the term ‘‘negotiating
costs’’ seems most descriptive of the underlying concept and will be used throughout the remainder of the paper.
3
Ž .
Ž .
Papers that discuss this issue at length are Hendricks and Kahn 1983 and Murphy 1992 .
4
For example, it would be difficult to negotiate complex fringe benefit packages on an annual basis. In addition, strategic reasons may exist inducing a union to prefer a longer contract over a shorter
contract.
does imply, ceteris paribus, that the union will prefer agreements of shorter duration than will the firm.
The effect of bargaining power on duration may show up both directly and indirectly. According to the argument advanced in the preceding paragraph, union
bargaining power and contract length should be inversely related. The indirect channel through which bargaining power may manifest itself is via the union’s
Ž .
willingness or unwillingness to make wage concessions in return for a shorter contract. Conversely, the union may have to be compensated with larger annual
wage increases than would otherwise be necessary to agree to a longer term contract. In essence, the union should need to be paid a compensating wage
premium in order for it to accept a contract of greater duration.
Because of the indirect link noted in the preceding paragraph, it follows that the length of the contract and the rate of annual wage change specified in the contract
are determined simultaneously in the negotiation process and, therefore, this simultaneity should be taken into account in formulating an empirical model of
contract duration. In specifying the wage change equation below, I draw on previous work analyzing the wage determination process in the unionized sector.
In particular, I rely on two recent papers that speak to this issue in some detail: the
Ž .
previously cited paper by Christofides and a paper by Prescott and Wilton 1992 . According to these authors, negotiated wage settlements depend on demand
conditions in the labor market and on expected shifts in labor supply or labor demand during the period in which the contract will be in effect. Furthermore, if
workers and firms negotiate over real wages, then the annual wage change in the
Ž contract should also be determined by the amount of inflation both expected and
. unexpected prevalent in the economy.
A final issue that must be addressed concerns how parties negotiating a labor agreement deal with inflation. Inflation is of concern to both labor and manage-
ment since both care about real rather than nominal compensation. As noted Ž
. above, one way to deal with inflation both anticipated and unanticipated is to
negotiate contracts of short duration on a frequent basis. Frequent negotiations entail clear efficiency costs, however. Alternatively, if inflation is expected over
the life of the contract, then a nominal wage change that takes such inflation into account may be negotiated directly into the contract. It will be impossible,
Ž .
however, to allow for unanticipated inflation it is unanticipated after all in this fashion. Though the costs imposed by inflation may be addressed through either or
both of the aforementioned channels, both channels have clear drawbacks. It is not surprising, therefore, that unions and firms have developed a formal device — the
Ž .
cost of living allowance COLA — to mitigate the problems caused by inflation. As in the case of the rate of wage change, I draw on previous work in specifying
the determination process of whether a contract contains a COLA clause. A Ž
. seminal paper on this issue is Hendricks and Kahn 1983 , but other papers on the
Ž .
Ž .
topic include Cousineau et al. 1983 , Ehrenberg et al. 1983 , Kaufman and Ž
. Ž
. Woglom 1986 , and Prescott and Wilton 1992 . This literature views workers as
more risk averse than firms. As a consequence, workers have a preference for protection from inflation and this is something they will have to ‘‘pay for’’ either
by accepting a lower rate of explicit wage change or a longer contract or some combination of the two. Hence, union bargaining power should play a role in
determining whether or not a contract contains an escalator clause. Additionally, the degree of uncertainty about future inflation should be positively related to the
probability that a contract contains an escalator clause. If it does not appear that the price level will remain stable during the contract period, then workers will
have a stronger preference for protection from inflation.
Because contract duration, the rate of wage change, and whether or not the contract contains a COLA clause are determined simultaneously in the process of
negotiating a labor contract, a model that takes such endogeneity into account is appropriate. Accordingly, I report below on estimation of the following model:
DUR s g DWAGE q g COLA
U
q X b q u
1
Ž .
i 12
i 13
i i1
1 i1
DWAGE s g DUR q g COLA
U
q X b q u
2
Ž .
i 21
i 23
i i2
2 i2
COLA
U
s g DUR q g DWAGE q X b q u
3
Ž .
i 31
i 32
i i3
3 i3
where DUR is the length of contract i; DWAGE is the annual rate of change in
i i
pay in contract i; COLA
U
is the propensity for the i-th contract to be indexed by a
i
cost of living escalator clause; COLA s 1 if the i-th contract is indexed by a cost
i
of living escalator clause and COLA s 0 otherwise; g is the coefficient of the
i jk
Ž .
k-th endogenous variable in the j-th equation j, k s 1,2,3 ; X
is the 1 = K
i j j
vector of exogenous regressors pertinent to the determination of the j-th endoge- nous variable; b is the K = 1 vector of regression coefficients corresponding to
j j
the exogenous regressors in the j-th equation; and u is a random disturbance
i j
term for the i-th contract in the j-th structural equation.
The vector X is of key interest since it will contain the various measures of
i1
nominal and real uncertainty as well as other proxies reflecting negotiation costs and bargaining power.
4. Econometric specification
