Models of Arbitrator Behavior Theory and Evidence ORLEY ASHENFELTER - - PowerPoint PPT Presentation

Models of Arbitrator Behavior： Theory and Evidence

ORLEY ASHENFELTER & DAVID E. BLOOM THE AMERICA N ECONOMIC REVIEW 1984

報告人：高培儒

20090429

1

Introduction

• Arbitration‘s use in the settlement of disputes is

arising, for example：

• 1. the settlement of grievances in union‐management

contract

• 2. disputes between buyers and sellers in commercial

contract

• 3. disputes between baseball players and club owner
• 4. settlement of public sector labor dispute

2

Introduction

• Vincent Crawford (1979)：
• if the arbitrator‘s notion of a fair settlement were

known to the two parties, then

• both conventional and final‐offer arbitration would

almost certainly lead to the same outcome which is the arbitrator's preference .

• If the parties bear the arbitration costs, the actual

arbitration wouldn’t happen.

3

Introduction

• The key to Crawford's conclusion is, of course, the

assumption that both parties know with certainty the arbitrator's preferred outcome.

• Henry Farber (1979, 1980)、Farber and Harry Katz

(1979)：explored the case where the parties are uncertain to the arbitrator's preferences.

• In this setting, under conventional and final‐offer

arbitration, the outcomes will generally differ.

4

Introduction

• Our purpose in this paper is to open up the

empirical analysis about：

distinguishing whether the parties are certain about arbitrator preferences is equivalent to testing whether the arbitrator's decisions under a final‐offer are predictable once information on the final offers and the environment of the bargaining unit is known.

5

Introduction

• Our purpose in this paper is to open up the

empirical analysis about：

set out simple models of arbitrator behavior under both final‐offer and conventional arbitration provide a method for testing the empirical implications of these models for observed arbitration awards and for the "fairness" of arbitrator behavior In addition, they also want to investigate how the differences in arbitration systems may affect outcome.

6

New Jersey Fire and Police Arbitration Act

• Approved in May 1977
• Followed from the passage in 1968 of

legislation

that granted New Jersey's public sector employees the right to organize and bargain collectively but not the right to strike. in the absence of the right to strike, negotiations

• ften went on long after annual municipal budget

cycles had been closed.

7

New Jersey Fire and Police Arbitration Act

• The purpose of this act is to ensure that contract

negotiations were final in time： – The collective bargaining must begin at least 120 days before employer’s budget submission date. – If parties do not reach an agreement by 60 days prior that, they may select conventional arbitration. – If parties cannot or prefer not to agree to conventional arbitration, they are compelled to final‐offer arbitration.

8

New Jersey Fire and Police Arbitration Act

• New Jersey Public Employment Relations

Commission (PERC)：

It is responsible for administration of the act. Arbitrators are assigned to cases by it. A list of seven members from PERC’s panel of arbitrators is first circulated to parties for comment. One of the seven is then appointed to the case. Arbitration fees borne jointly by the parties.

9

New Jersey Fire and Police Arbitration Act

• Arbitrators are directed to give due weight to

variety of factors：

1. Interests of the public 2. Relevant comparisons of wage and working condition 3. Overall level of compensation already received 4. Financial impact of the settlement on the municipality 5. Cost of living 6. Continuity of employment

10

Arbitrator behavior under final‐

• ffer
• Notation：

Wu=union’s final offer We=employer’s final offer Wa=arbitrator’s preferred settlement which is unknown to outsider and never reveal under final‐

• ffer
• The arbitrator would select the employer’s
• ffer if：

11

Arbitrator behavior under final‐

• ffer
• Since We < Wu：
• The probability of an employer victory, P, is：
• F(。) is the value of the cumulative

distribution function that describes arbitrator’s preference.

12

• Let F(。) to be normal with
• where N(。) indicates the value of the

cumulative standard normal distribution function

constant in this probit function： coefficient of (We+Wu)/2：

Arbitrator behavior under final‐

• ffer

13

• Testable model for evenhanded arbitrator：
• Arbitrators demonstrate：

– Pro‐union： – Pro‐employer： – No bias：

Arbitrator behavior under final‐

• ffer

14

• Whether arbitrator notions of fairness

explicitly take account of the parties‘ final

• ffer？
• The employer's offer is accepted if
• or

Arbitrator behavior under final‐

• ffer

15

• Condition (8) is, of course, identical to (1).
• In this sense, the decision rule (1) may be a

very robust empirical device

• Combine these together：

Arbitrator behavior under final‐

• ffer

16

• The sum of the coefficients of We and Wu is

an estimate of

• implying that we can identify：
• the equality of the coefficient provide a test of

the hypothesis：

• Whether some specified variables affect

Arbitrator behavior under final‐

• ffer

17

• For the i th observation, we may write
• Our estimates of the coefficients in the vector

will indicate how the arbitrators think what is fair

Arbitrator behavior under final‐

• ffer

18

• three main candidate variables to enter the

vector：

• 1. a measure of the wage rate in some alternative
• ccupation may influence an arbitrator's

decision

• 2. arbitrators may act to narrow(or widen) the

wage differentials of police workers across municipal

• 3. impact of the municipality's financial well‐being
• n the arbitrator's decision

Arbitrator behavior under final‐

• ffer

19

• Under conventional arbitration, the arbitrator

is free to fashion a settlement of his own choosing

• In these cases, Wa is observable, but We and

Wu are not.

Arbitrator behavior under conventional arbitration

20

• Whether the arbitrator fashions a compromise

based on the arguments presented to him by the parties：

Arbitrator behavior under conventional arbitration

21

• omitted variable bias
• We and Wu are unobservable
• In New Jersey, the same panel of arbitrators is used

in both conventional and final‐offer arbitration cases

• For this group of arbitrators, we may assume
• we may fit equations (6') and (11) separately and use

a test for the equality of parameters as evidence to favor one or the other of the formulations of (11) or (12)

Arbitrator behavior under conventional arbitration

22

Empirical Methods

• The likelihood function for the final‐offer

arbitration sample：

• The likelihood function for the conventional

arbitration sample：

• When

23

Table 1

24

Table 1

• 1. the mean of the union and employer final offers

is quite low when compared to the mean of the conventionally arbitrated settlements.

• 2. in all three years, the union offers were

accepted in roughly two‐thirds to three‐quarters

• f the cases.
• 3. the proportion of cases reaching arbitration

dropped considerably between 1978 and 1980.

25

Table 2

26

Table 2 column 1

• In column (1) simplest results for each year

probit function (4) and the regression equation (10)

• The estimated mean of arbitrator preferences

from the final‐offer arbitration cases

– higher than the mean of the final offers of the employer groups – only slightly lower than the mean of the final

• ffers of the union groups indicated in Table 1

27

Table 2 column 2

• columns (2) are obtained by maximizing the likelihood

function (15)

• Suppose that we take the mean and standard deviation of the

conventional arbitration awards as measures of the mean and standard deviation of arbitrators' preferences.

• These assumptions imply that we may predict the actual

fraction of employer victories from knowledge of the final

• ffers.
• The result suggest these predictions are essentially identical

to the maximum‐likelihood predictions based on the final‐

• ffer data alone.

28

Table 2 column 3

• test the hypothesis：
• In 1980, the estimate is well above unity
• In 1979, the estimate is slightly below unity
• but in no case can we reject the hypothesis
• The exception is for 1978, where the

hypothesis is rejected at any conventional test level.

29

30

Table 3

• ESTIMATES OF EQUATIONS( 6') AND (11), variables in the

vector xi

• Lagged percent change in county income per capita：
• Salary level in previous year (in dollars)：
• Municipality‘s state equalized tax rate in previous year (in

dollars per 100 dollars of：

• Net debt per capita in municipality in previous year (in

dollars)：

31

Table 3

• In sum, we have found statistically significant,

but temporally unstable.

• It may simply be the case that the main

predictable variability in the central tendency

• f arbitrator awards is in their movements

through time.

• Or, there are stable relationships between

arbitrator preferences and some set of variables, but have been omitted from xi.

32

Table 5

• attempt to account for the individual influence of

specific arbitrator

• In the probit model, these fixed effects are

econometrically identified only if an individual arbitrator selected at least one employer offer and

• ne union offer in a given year

33

Table 5

• The introduction of the arbitrator dummies

considerably reduces the estimate of the unpredictable variability in arbitrator preferences in 1980 and 1978

• This strongly suggests that successful models
• f the strategic behavior of the parties will

almost certainly have to account explicitly for the uncertainty the parties face in arbitrator decision making

34

Conclusion

• 1. under the New Jersey statute, union offers

have been selected most of the time in final‐

• ffer arbitration cases

There is evidence of similar behavior in Michigan and Massachusetts, as noted in our 1983 paper. In general, we should expect a fair arbitrator to be one that considered the objective considerations in a particular case and then settled on what, in the arbitrator's own mind, seemed a preferred settlement

35

Conclusion

two different reasons why union offers may be disproportionately accepted by the arbitrator

Arbitrators may not follow the decision process set

• ut above , and the integrity of the arbitration system

is being seriously undermined Parties may not typically position themselves equally distant from, and on opposite sides of, the arbitrator's preferred award, either because unions have a more conservative view of what arbitrators will allow, or because unions may be more fearful of taking the risk

• f loss than are employers.

36

Conclusion

• 2. conservative union behavior resulted in

lower average wage increases under the final‐offer arbitration provisions than under the conventional arbitration provisions of the New Jersey statute

• 3. Considerably less variability in the actual

awards under the final‐offer arbitration provisions than under the conventional arbitration provisions of the statute

37