"Chilling" Under Arbitration and Mixed Strike-Arbitration Regimes F R E D E R I C C. C H A M P L I N University o f O k l a h o m a , N o r m a n , O K 73019 M A R I O F. B O G N A N N O * University o f Minnesota, Minneapolis, M N 55455
This paper investigates the "'chilling effect'" o f arbitration hypothesis using data from public sector collective bargaining activity in Minnesota. Settlement rates under a regime requiring the compulsory arbitration o f impasses are compared to rates under a regimepermitting a limited right to strike. The authors f i n d a significantly lower settlement rate under thepure arbitration regime. Other factors that could result in this rate differential are discussed. While not ruling out the possibility o f unmeasured differences, the authors conclude that the differential is due to differences in regime and not to the confounding factors discussed. I.
Introduction
Finding a suitable mechanism for public sector labor dispute resolution is a perennial problem for policy makers. The tradition of voluntarism places a high value on mechanisms that do not impose third-party solutions, while the social costs of the public sector strike lead many to conclude that it should not be tolerated. The result has often been a compromise, procedures with progressively more intrusive thirdparty interventions culminating in some form of compulsory interest arbitration. There has long been a suspicion that the twin goals of encouraging voluntary settlement and eliminating strikes are not entirely compatible. Arbitration mechanisms that take the place of a strike may tend to discourage voluntary settlement. A number of studies have been conducted comparing settlement behavior under various final resolution conditions, and their results have tended to support the view that compulsory interest arbitration may discourage bargaining.' As noted by Anderson (1981), however, these results are not entirely persuasive. In many cases the studies lack both representative samples and an adequate specification of control variables.
*The authors wish to thank Paul Schumann and an anonymous referee for comments resulting in substantial improvements in this paper. ' A representative selection o f these studies includes Anderson (1981), Feuille (1975), Grodin (1974), Lipsky and Barocci (1978), Stern, et al. (1978), and Wheeler (1978). JOURNAL OF LABOR RESEARCH Volume VI, No. 4 Fall, 1985
376
JOURNAL OF LABOR RESEARCH
In the first part of this paper, we outline a simple theory o f bargaining and show that this theory has provided the basis for most o f the recent theoretical work on the effects of dispute resolution regimes on bargaining incentives. In the second part, we compare the relative frequency o f settlement under two dispute resolution regimes: one mandates compulsory conventional arbitration in the event of impasse and is termed a " p u r e " arbitration regime, and the other permits the employers o f certain groups to choose between the strike and arbitration as the terminal step in the dispute resolution procedure and is termed a " m i x e d strike-arbitration" regime. This statutory arrangement permits an evaluation o f important hypotheses suggested by the bargaining theory presented earlier. II.
The Theory o f Bargaining Under Strike and Compulsory lnterest Arbitration Regimes
Much of the theory of bargaining in the presence o f compulsory interest arbitration schemes is best interpreted in light o f the theory o f bargaining games first advanced by Nash. 2 Two important ingredients o f this bargaining model are presented in Figure 1. The first is the collection of available negotiated outcomes evaluated in utility terms, represented in the region OCD. The second is the beliefs that each side holds about the utility value o f the outcome resulting from failure to agree. Failure to agree implies some form o f conflict, and these utility values are net o f any costs or losses, such as strike costs, associated with this conflict. This is termed the threat point and it varies with this outcome. For our purposes, two possible outcomes will be considered: the strike and compulsory interest arbitration. In Figure 1, the utility value of the strike is arbitrarily set at the origin yielding a value of zero to each side, and the utility value of arbitration is A yielding ,4, to the union and A,, to management. Thus, the threat point associated with the strike is the origin, and the threat point associated with arbitration is A. A third element of Nash's model, the contract zone, may be derived from these ingredients. The contract zone is the collection o f available outcomes that (1) both sides prefer to the threat point and (2) is Pareto efficient. An outcome is Pareto efficient if there is no other available outcome that both sides prefer to it. The contract zone resulting from the strike threat point is the curved line CD in the figure and the contract zone resulting from the arbitration threat point is the line C'D'. Given the set o f available outcomes, the threat point determines the contract zone. A number o f theorists have used the contract zone as the point o f departure in their analyses of bargaining under arbitration schemes. It has usually been assumed that the likelihood of agreement depends on the " s i z e " o f the contract zone. In general, the lower or less attractive the threat point, the greater the size o f the contract zone. For example, compare the two zones CD and C'D'. In any
2SeeNash (1950)and (1953). Also see Harsanyi (1977).
FREDERIC C. CHAMPLIN and MARIO F. BOGNANNO
377
Union Utility C C'
\
\ \ \ \ \
',E'_
A~
D'
\ \ \ \ \ \ \ D
A~
Management Utility
Figure 1
case, the lower or less attractive the threat point, the more attractive, relatively speaking, is a potential negotiated agreement in the contract zone. Thus, both the opportunity and the incentive to reach negotiated agreement increase as the threat point becomes lower and less attractive to the two sides. An examination of recent literature on the theory of bargaining under compulsory arbitration schemes shows that a major concern of this work is with variables that affect the attractiveness or utility value of the arbitration threat point. Stevens (1966) argued that arbitration is "compatible" with bargaining depending on (1) the decision criteria each side uses to select its final prearbitration position and (2) the parties' pessimism or optimism about arbitration outcomes - - that is, the utility value of the threat point. Farber and Katz (1979) emphasized the extent of the parties' risk aversion and the riskiness o f arbitration. Risk is seen as reducing the value of the arbitration outcome and, therefore, the threat point for risk averse parties. Both Bloom (1981) and Hirsch and Donn (1982) extended the Farber and
378
JOURNAL OF LABOR RESEARCH
Katz model by considering the effects o f exogenous costs o f arbitration, such as fees and case preparation expenses. Such costs clearly reduce the relative value of the arbitration threat point. Bloom argued the matter further, however, by introducing the cost of negotiation. In the model presented above, such costs increase the value of the arbitration threat point relative to negotiated outcomes and decrease the size of the contract zone. This is illustrated in the figure by the dashed line, which portrays the reduced utility value of negotiated outcomes due to the imposition of negotiation costs. The new contract zone with respect to the arbitration threat point is C"D". In summary, the factors emphasized in this literature that affect the threat point associated with arbitration outcomes are: (1) the parties' pessimism (or optimism) concerning likely arbitration outcomes, (2) the perceived riskiness of the arbitration procedure and the extent of the parties' risk aversion, and (3) the exogenous costs of both arbitration and negotiation. If we assume that the strike is associated with a less attractive threat point than arbitration for all parties, then, following the theories described above, we would expect to observe relatively more settlement with the strike rather than with arbitration as the final step. It is the empirical manifestation of this difference in settlement rates that is termed the "chilling effect" of compulsory interest arbitration. The existence of this chilling effect is an implication of the bargaining model considered here, and it is this model that informs much recent theoretical work that analyzes the effects of arbitration on bargaining. Failure to find a chilling effect would be evidence against the threat model of bargaining and damaging to the theories that rely upon it. Therefore, a search for the chilling effect is a search for evidence bearing on these theories. It is reasonable to ask why an employer would ever choose the strike option with its less attractive threat point. The more costly threat will be chosen if the employer believes this will induce the union to settle on more favorable terms without its actual use (e.g., bluffing). In short, if the employer expects to reach agreement he will choose the threat that maximizes the value to him o f this expected agreement and not the threat that minimizes his conflict costs. If the threat model has any validity we should observe more settlement under a " p u r e " strike as compared to a pure arbitration regime. The problem is that there are little data available to directly test this hypothesis. Data gathered from case activity under the Minnesota Public Employment Relations Act (MPELRA) do permit an indirect attack on this problem, however. This statute (described in the next section) allows a comparison of the relative frequency of settlement under a pure arbitration regime to that under a mixed strike-arbitration regime. One could, however, question whether the above conclusion would follow if the pure strike alternative were replaced by a mixed strike-arbitration regime. Later, it will be argued that the theory presented in this section does indeed imply a chilling effect even with the substitution of the mixed regime. Thus, the comparison of results under pure arbitration and mixed strike-arbitration regimes also yields evidence bearing on these theories.
F R E D E R I C C. C H A M P L I N and M A R I O E B O G N A N N O
III.
379
The M P E L R A Procedure
The Minnesota Public Employment Labor Relations Act (MPELRA) was a "choice of procedures" statute in the sense that during the period under investigation, management had a limited option to choose either arbitration or the strike as the impasse or terminal step? This option was available to employers of employee units that were not classified as "essential." Prior to the 1979 amendments to the act, essential employees were, defined as those "whose employment duties involve work or service essential to the health or safety of the public and the withholding of such services would create a clear and present danger to the health or safety of the public." For purposes of the strike prohibition, certain other supervisory employees were also included in the "essential" category. In 1979, this definition was altered to include certain specifically named employee groups.' This change affected the status of only two bargaining units during the period of this study. As it existed in the period 1973 to 1980, the MPELRA impasse procedure was as follows: Upon opening contract negotiations, parties were supposed to file a "negotiation notice" with the Bureau of Mediation Services (BMS). Parties jointly desiring mediation services then filed a "mediation petition" with the BMS director. If mediation efforts failed, either party could petition the director to certify the dispute to impasse. If the employee unit was essential or if it was nonessential and the employer petitioned, the director would certify the impasse to binding conventional arbitration. If, however, the employee unit was nonessential and the union was the petitioner, the employer had fifteen days to accept or reject arbitration. If the employer rejected arbitration, the employee organization was free to strike. Thus, the strike could occur but only if the employee group was nonessential and the employer rejected arbitration. IV.
Empirical M e t h o d
Data. In what follows the presence of the chilling effect is investigated using data from all cases handled by the BMS during the period 1973 to 1980. We will compare the relative and absolute frequencies of cases reaching each point in the
3The Minnesota statute was most recently changed effective August 1, 1984. The law now makes the strike available in the nonessential sector unless both parties agree to artibration. Arbitration is mandatory in the essential sector. In both sectors arbitration is conventional unless both parties agree to either the issue or package final offer system. These changes followed an extended period of experimentation with various arbitration structures. 'Minnesota Statutes, 1979, Section 3-855 (Section 1). The change in definition is found at Minnesota Statutes, 1979, Section 179.63, Subdivision 11. During the period of this study the numerically most important essential groups included licensed peace officers as well as law enforcement officers employed by the state, firefighters, supervisory employees, school principals and assistant principals, administrative heads and their assistants in municipalities and municipal utilities, employees of hospitals other than state hospitals, and health care professionals employed by the state.
380
JOURNAL OF LABOR RESEARCH
M P E L R A impasse procedure under the essential and nonessential regimes. These data represent a census o f cases processed by the BMS during the period fiscal year 1973 through fiscal year 1980 and were gathered as part o f a larger study o f the M P E L R A conducted by the authors (Bognanno and Champlin, 1981). The ideal base for the relative frequency computation is the total number o f negotiations conducted by public sector bargainers under the two regimes during this period. The negotiation notice requirement has not been enforced, however, so there is no precise and strictly comparable data on the number o f negotiations actually conducted. As a substitute for the "exposure r a t e " that data on number of negotiations would provide, we use the number o f mediation petitions filed.
Other Factors Requiring Control. A major problem with comparisons o f this kind is the possible existence of systematic differences other than the final step o f the dispute resolution procedure between pairs bargaining under the two regimes. One difference could arise if the parties were free to alter the regime under which they bargain - - that is, if the regime was itself endogenous. The essential and nonessential categories, however, are defined by statute, and neither party has any ability to alter its status. Another difference could arise if the two groups were subject to different impasse procedures prior to the final step. Under this statute, however, the procedures are identical for all cases down to the final step. Anderson (1981) specifically mentioned six other features o f the bargaining environment requiring control. Especially significant would be differences between the essential and nonessential sectors in their economic, legal, or political environments. But these cases all arose in the same state and during the same time period. While there are undoubted differences between particular bargaining pairs on these factors, the temporal and geographic distribution o f cases in the data set argues against any systematic differences between pairs bargaining under the two regimes. Furthermore, public employee collective bargaining in Minnesota has existed in both sectors at least since 1971, and parties have had ample opportunity to gain skill and experience in it. Again, there are individual differences between bargaining pairs in skill and experience, but there is little reason to suppose that these vary systematically between the two regimes. Other factors considered by Anderson include employer ability to pay, the structure o f management decision making, and the presence o f internal management conflict. These factors are relevant to particular employer jurisdictions and to control for them we disaggregate to the level of employer types. Since, at this disaggregated level, the body of employers dealing with essential units is largely the same as those dealing with nonessential units, it is hard to convincingly argue that they systematically differ by final step regime.
'The relationshipbetweenthe parties is a final factor mentioned by Anderson. It might be argued that such attitudinal factors are independentvariablesaffecting the negotiation outcomes. It seemsapparent, however,that these factors are themselvesendogenous.
FREDERIC C. CHAMPLIN and MARIO F. BOGNANNO
381
Another important problem in comparing bargainers under these two regimes is one of selectivity. Essential groups - - the ones for whom the strike is forbidden by law - - may be more militant, aggressive, or bellicose than nonessential groups. If so, they may tend to arbitrate more frequently; that is, their base rate for final step use may be greater. The test would thus be biased in favor of the chilling effect. In our analysis o f the disaggregated data, comparisons within and between employer categories will be made, showing that this is not a serious problem.
Effects of the Mixed Regime. The main question remaining in this analysis relates to the effects of the nonessential sector employer's option o f choosing either the strike or interest arbitration as the impasse procedure. We address the question o f whether arbitration as compared to the strike has a chilling effect. To answer this question we should compare the marginal relative frequency of impasse under a pure arbitration regime to that under a pure strike regime. Since nonessential sector employers have a choice between strike and arbitration, however, the data presented here are not from a pure strike regime but rather from a mixed strikearbitration regime. Therefore, the observed relative frequency o f impasse in the nonessential sector is a joint relative frequency, that is, the joint result of impasse procedure selection and the actual occurrence o f impasse. The question is whether this joint relative frequency in the nonessential sector offers any information about the marginal relative frequency o f impasse under a pure strike regime. In our view, this relative frequency represents an upper bound on the relative frequency o f impasse under a pure strike regime. This follows from the observation that the threat presented to nonessential sector bargainers is some combination of the strike and arbitration. Prior to the employer's decision at the final step, this is a risky combination of the two potential outcomes. After this decision, o f course, the outcome is certain. Since it has been assumed that the strike is more costly for all parties than arbitration, it follows that the threat value of the strike-arbitration combination will always fall somewhere between that o f the pure strike and pure arbitration regimes (including the two polar cases as extreme points). Consequently, it is expected that the impasse rate for parties under the strike-arbitration regime would be less than that under a pure arbitration regime but greater than that under a pure strike regime. As a corollary, it would also be expected that there is a strong correlation between strike and arbitration use. That is, we would expect to find that those cases that would result in a strike if that opportunity were present would nearly always result in arbitration if this were substituted for the strike as the impasse outcome. If the impasse rate under the mixed strike-arbitration regime is an upper bound on the impasse rate under a pure strike regime, then the difference in impasse rates between a pure arbitration and a mixed strike-arbitration regime must be a lower bound on the difference in rates between the two pure cases. Since in this paper the difference in impasse rates between a pure arbitration and a mixed regime is measured, the results presented here represent such a lower bound. If there is a difference in rates between these two regimes, it may be concluded that
382
JOURNAL OF LABOR RESEARCH
the difference between the pure arbitration and the pure strike regimes and, hence, the chilling effect is even greater. Thus, we conclude that this joint relative frequency does provide important information about the relative frequency of impasse under a pure strike regime. It might be objected, however, that the nonessential sector employer could discriminate between units, offering some the strike and others arbitration. This arguably could be done in such a way as to reduce the incidence o f impasse for this sector in total below that which would occur under a pure strike regime. In essence, this argument postulates a class of nonessential units that would settle more readily under a pure arbitration regime than under a pure strike regime. (If such a class did not exist, the total impasse incidence for the sector under the mixed regime could not be less than that under the pure strike regime.) To maintain this postulate one must suppose that the strike has a chilling effect relative to arbitration.
V.
Results
Use Rate Data. Table 1 contains a summary of the disposition o f all cases processed by the BMS during the period fiscal year 1973 through fiscal year 1980. The table shows that o f those cases for which mediation petitions are filed, about 17 percent result in impasse certification and about 13 percent result in strike or arbitration. Of those in the essential sector who file mediation petitions, about 30 percent result in an arbitration award. For the nonessential sector, the proportion ending in either strike or arbitration is about 9 percent. The null hypothesis that these last two proportions are generated by the same process is tested using the Pearson Chi-square test of association (with Yate's correction). ~ The null hypothesis is easily rejected at a = .01 (X2 = 211.3, d.f. = 1). Since essential employees are absolutely prohibited from striking while nonessential employees have a limited potential for the strike, it is hard to resist the conclusion that the chilling effect is operating here. A closer inspection reveals that an important source of the difference between these sectors is the number of settlements that occur prior to impasse certification. Of the 254 essential sector cases reaching impasse certification, 76.4 percent result in an arbitration award. The corresponding proportion (including strikes) of the 348 nonessential sector cases is 72.4 percent. The null hypothesis that these two proportions are equal cannot be rejected at o~ = .25 (X2 -- 1.0, d.f. = 1). By contrast, 8.2 percent of those in the essential sector filing mediation petitions settled prior to the first mediation session and another 52.8 percent settled during mediation. For those in the nonessential sector, the comparable proportions are 23.4 percent and 64.5 percent. Not surprisingly, the null hypothesis that the propor-
6With minor exceptions, this is the statistic used to test all hypothesesabout relative frequency data.
FREDERIC C. CHAMPLIN and MARIO F. BOGNANNO
383
Table 1
Public Sector Collective Bargaining Activity in Minnesota Case Disposition by Procedure Stage and Bargaining Unit Category 1973-1980 All Sectors
Essentiala % of all Sectors
No
%
No
%
3,525
100.00
650
100.00
725
20.56
53
8.15
Settled during med iation but before impasse certification 2,198
Mediation Petitions Settled before first mediation
Nonessentiala % ofall Sectors
No
°70
18.44
2,875
100.00
81.56
7.31
672
23.37
92.69
62.35
343
52.77
15.61
1,855
64.52
84.39
Certified to Impasse
602
17.08
254
39.08
42.19
348
12.10
57.81
Settle prior to hearing/strike
156
4.43
60
9.23
38.46
96
3.34
61.54
Ar bitration awards
381
10.80
194
29.85
50.92
187
6.50
49.08
65
1.84
0
0
0
65
2.26
100.00
446
12.65
194
43.50
252
8.76*
56.50
Strikes A ward or strike
29.85*
aFor sector definitions see text. *The null hypothesis that these two proportions are generated by the same process is rejected at c~ = .01. X2 = 213.18; X2o,., = 6.635.
tions of cases in the two sectors reaching impasse certification are equal is rejected at o~ = .01 (x ~ = 270.5, d.f. = 1). The nonessential groups seem to have a stronger proclivity for settling early in the process. This result provides further support for the chilling effect hypothesis. It also suggests that these results cannot be accounted for by the hypothesis that these differences arise from a nonessential employer strategy o f offering arbitration only to those units that are more likely to settle under arbitration than under strike regimes. If this hypothesis were true one would expect to find substantial settlement late in the process when the employer's intentions were either relatively clear or already revealed. Observation o f the reverse pattern o f settlement, therefore, argues against this hypothesis. As has already been noted, many important factors that might distinguish between essential and nonessential sectors are relevant only to the employer. To control for these, the data in Table 1 are disaggregated to the level of employer category. Turning to Table 2, we find that, except for the state, the difference be-
384
JOURNAL
OF LABOR RESEARCH
tween final step use rates persists and is significant at ~ = .01 for each category. Controlling for employer specific factors, then, does not alter our conclusions.7 The data in Table 2 also allow us to pursue the selectivity caveat raised earlier. Note that each occupational group (e.g., sheriff's deputies, firefighters, or supervisors in county government) has individual attributes that result in a base rate for final step use which is characteristic for that group. There is little reason to suppose that these various occupational groups have the same base rate (e.g., compare sheriffs' deputies to school administrators). Note further that the employer categories in Table 2 differ in the occupational composition o f their essential sector bargaining units (e.g., school districts do not employ sheriffs' deputies). All of this implies that if higher base rates cause the observed differences in final step use between essential and nonessential sectors, then variation should also be observed within the essential sector. In particular, we should observe variation in final step use rates across essential employer categories. On the other hand, if the chilling effect is the dominant factor affecting this observed difference, it would be expected that these rates would be relatively homogenous
Table 2
Proportion of Cases Resolved Through Arbitration or Strike by Employer Category and Sector Employer Category
Essential %
Nonessential %
:~***
State
21.4
34.8
Counties
29.7
10.9
23.4*
Municipalities
30.6
11.5
41.8*
778
School districts
28.7
--
--
2060
1.0"*
(Nonessential = Nonteachers)
--
5.0
58.9*
(Nonessential = Teachers)
--
8.1
44.0*
N 80 449
*Significant at ot = .001, x-2.o~,, = 10.8 **Not significant at a = .25, x 2,~., = 1.3 The test statistic reported is developed from a likelihood ratio test for problems of association in categorical data. See Winkler and Hays (1975), p. 831. ***Except for the state data, all tests are Pearson Chi-square tests of association for categorical data (with Yate's correction). See Winkler and Hays 0975), pp. 825-831.
'State data represent an a n o m a l y in Table 2 since nonessential use rates exceed the essential. The difference, however, is not statistically significant. The explanatio n is that duri ng the period under study, the state followed an " u n w r i t t e n " but generally k n o w n policy o f not exercising its strike opt i on. Thus, from the bargaining viewpoint, nonessential units behaved as if their final step options were regulated by the non-strike regime.
F R E D E R I C C. C H A M P L I N a n d M A R I O F. B O G N A N N O
385
across employer categories. Thus, if there is a selectivity bias resulting from higher base rates for essential sector employees, these base rates should reveal themselves as variation in final step use across employer categories. To test these hypotheses, we examine the variation in essential sector final step use rates across employer categories in Table 2. The null hypothesis that there is no difference between employer categories in the rate o f final step use is tested against the alternative that at least one differs. The null hypothesis cannot be rejected at o~ = .75 (X~ = .45, d.f. = 2), and we conclude that the four employer categories are homogenous in their essential sector final step use rates? This suggests that the effect o f the dispute resolution regime dominates differential final step use base rates. In particular, consider the case of school districts. Virtually the only essential units in this employer category are made up o f school administrators, yet their final step use rate is 28.7 percent. This may be compared to teachers whose rate is 8.1 percent and to nonteachers whose rate is 5.0 percent. Both these differences are statistically significant at ct = .001. To explain these differences in terms o f the characteristics o f essential sector employees, one must suppose that school administrators are more militant, aggressive, or bellicose than either teachers or other school district employees. Further recall that school administrators do not differ significantly from other essential groups in their final step use rates. VI.
Discussion and Conclusions
This analysis covers a period of Minnesota's public employee collective bargaining history when essential employee units were barred from striking while nonessential units could strike at the employers' option. Nonessential sector negotiators always faced a distinct possibility of strike until the final m o m e n t o f negotiations; essential sector negotiators never faced such a threat. It is found that 30 percent of the essential cases filing mediation petitions ended in arbitration, a proportion that stands in sharp contrast to the 9 percent impasse rate figure reported for nonessential cases. The evidence of final step use rates presented here supports the chilling effect hypothesis. These results also lend support to the bargaining theories discussed at the beginning o f this paper. A major problem with comparisons o f this kind is control. Factors other than the impasse procedure but correlated with it may affect bargaining outcomes and bias the results. To deal with this problem, it has been shown that these results persist under disaggregation to employer type, which controls for employer specific factors. It has also been shown that the rate o f final step use is the same across essential employer categories with widely differing mixes o f essential bargaining units. If differential base rates are significant in determining final step use
*The test statistic is the usual Chi-square "goodness of fit" statistic. See.Winkler and Hays (1975), pp. 817-822.
386
JOURNAL OF LABOR RESEARCH
rates for essential units, then it should show up in differential rates across employer categories. Finally, it has been shown that the results persist in comparisons where differential base rates for final step use are not a plausible explanation. The nonessential employers' option of selecting either the strike or arbitration as the final step of the procedure presents another problem for this analysis. To test the chilling effect hypothesis, we would ideally compare results from a pure strike regime to those from a pure arbitration regime. It has been shown, however, that under conventional assumptions the difference between impasse rates under a pure arbitration regime and a mixed strike-arbitration regime is a lower bound on the difference in these rates between the two pure cases. The only condition that could invalidate this conclusion is the existence of a substantial body of cases that would settle more readily under a pure arbitration regime than under a pure strike regime. In the absence of such a group of cases, the difference between the impasse rates under a pure arbitration regime and a mixed strike-arbitration regime implies an even larger chilling effect. A number of factors, therefore, have been controlled. It must be recognized, however, that unmeasured and possibly unknown factors other than the dispute resolution regime may account for the differences in impasse rates. For this reason, these and similar results must be counted as preliminary. Improvement awaits a better theory of bargaining behavior that would permit a more precisely specified model of impasse and settlement. Even though the bargaining model discussed earlier has been supported by the data presented here, it may still be criticized. It is a much better theory of agreement than disagreement. Rational parties could always be expected to reach agreement on the Pareto frontier rather than invoking the conflict outcome. In this way, disagreement and conflict are relegated by default to the realm of irrationality and misperception. This is unquestionably an undesirable state of affairs and represents an important challenge to the theoretical ingenuity of those interested in designing regimes for managing conflict. Recent views of the impasse resolution process have emphasized that many procedures, including those terminating in compulsory conventional arbitration, may encourage compromise, concession, and settlement. But not all are equally effective in this, which implies that the chilling effect is a relative matter. This view is certainly supported by the essential-nonessential impasse evidence presented here. Substantial settlement occurs under both regimes, even though the possibility of strike seems to present greater agreement incentives. While the chilling effect appears to exist and ought not to be ignored, it may not be as serious a problem as once thought. We believe that relatively more attention should be devoted to studying other components of dispute resolution policy. Feuille (1979) has already listed marly of these, including the impact of compulsory arbitration and arbitration awards on the "public interest," the impact of awards on employer/employee preferences, and the success of arbitration as a
F R E D E R I C C. C H A M P L I N a n d M A R I O F. B O G N A N N O
387
means of regulating inter-group conflict without imposing excessive stress on the principles of representative government.
REFERENCES Anderson, John C. "The Impact of Compulsory Arbitration: A Methodological Assessment" Industrial Relations 20 (Spring 198 l): 129-148. Anderson, John C. and Thomas A. Kochan "Impasse Procedures in the Canadian Federal Service: Effects on the Bargaining Process" Industrial and Labor Relations Review 30 (April 1977). Bloom, David E. "Is Arbitration Really Compatible with Bargaining?" Industrial Relations 20 (Fall 1981): 233-244. Bognanno, Mario F. and Frederic C. Champlin "A Quantitative Description and Evaluation of Public Sector Collective Bargaining in Minnesota: 1973-1980" Unpublished consulting report submitted to the Legislative Commission on Employee Relations, Minnesota State Legislature, August 1981. Farber, Henry S. and Harry C. Katz "Interest Arbitration, Outcomes, and the Incentive to Bargain" Industrial and Labor Relations Review 33 (October 1979): 55-63. Feuille, Peter "Final Offer Arbitration and the Chilling Effect" Industrial Relations 14 (October 1975): 302-310. Grodin, Joseph R. "Arbitration of Public Sector Labor Disputes: The Nevada Experiment" Industrial and Labor Relations Review 28 (October 1974): 89-102. Harsanyi, John C. Rational Behavior and Bargaining Equilibrium in Games and Social Situations Cambridge: Cambridge University Press, 1977. Hirsch, Barry T. and Clifford B. Donn "Arbitration and Incentive to Bargain: The Role of Expectations and Costs" Journal o f Labor Research 3 (Winter 1982): 55-68. Lipsky, David B. and Thomas A. Barrocci "Public Employees in Massachusetts and Final Offer Arbitration" Monthly Labor Review 101 (April 1978): 34-37. Nash, John P. "The Bargaining Problem" Econometrica 18 (1950): 155-162. "Two Person Cooperative Games" Econometrica 21 (1953): 120-140. Stern, James L., Charles Rehmus, J. Joseph Lowenberg, Hirschel Kaspar, and Barbara D. Dennis Final Offer Arbitration Lexington, Massachusetts: D.C. Heath, 1975. Stevens, Carl M. "Is Compulsory Arbitration Compatible with Bargaining?" Industrial Relations 5 (February 1966): 38-52. Wheeler, Hoyt N. "How Compulsory Arbitration Affects Compromise Activity" Industrial Relations 17 (February 1978): 80-84. Winkler, Robert L. and William L. Hays Statistics: Probability, Inference, and Decision New York: Holt, Rinehart, and Winston, 1975.