Determinants of grievance outcomes: a case study.
Most empirical studies of the grievance process have concentrated on identifying the determinants of grievance filing activity and the factors associated with the early resolution of grievances. To the best of our knowledge, no study has attempted to identify the factors that may influence the outcome of a grievance. That omission is surprising, since the parties to a grievance are usually interested not only in how quickly the grievance is resolved but also in whether the grievance is fully granted, partially granted, or denied.
The purpose of this study is to empirically examine, using data from 15 Canadian federal government work sites, the variability of grievance outcomes across grievance steps, the salary grades and occupations of the grievants, and grievance issues. The study contributes to the grievance literature in several ways. First, since little is known about the factors influencing grievance outcomes, the empirical findings of the paper should be of interest in themselves. Second, the findings may help explain why grievances are usually settled at lower levels of the grievance procedure, rather than at higher levels. Specifically, if it is found that the probability of having a grievance granted at higher levels of the grievance procedure is very low, it may be reasonable to speculate that unions are discouraged from pushing grievances beyond a certain level of review. Third, unlike most studies, in which firms are the unit of observation, the present study focuses on grievance cases as the unit of observation.
The Grievance Literature
Despite the centrality of the grievance process in industrial relations, and the fact that it is viewed by some as the most important North American innovation in collective bargaining (Thomson and Murray 1976), the empirical evidence on its operation is widely perceived as disjointed and the relevant theory is criticized as lacking a unifying framework (Peterson and Lewin 1981; Gordon and Miller 1984; Knight 1985).
Most studies of the grievance process have concentrated on the determinants of grievance filing activity. Among the factors that have been studied are the demographic and personal characteristics of the grievants, group and organizational characteristics, and features of the union-management relationship.
The studies of demographic characteristics of grievants and non-grievants suggest that age, gender, and work experience are among the factors that have an impact on the grievance filing behavior of individuals (Ash 1970; Price et al. 1976). Studies concentrating on group, organizational, and work environment determinants have found that the type of group and the nature of the work (Ronan 1963), technological change (Peach and Livernash 1974), the type of technology used (Nelson 1979), decision-making centralization (Weiss 1957), and general work environments (Wynne 1973; Muchinsky and Maassarani 1980) are major factors affecting the number of grievances filed. Regarding the impact of the union-management relationship on grievance filing activity, Gandz and Whitehead (1981) and Gandz (1979) found that good union-management relationships are associated with lower grievance activity. The effects of both managements' and unions' leadership styles on grievance rates have also been studied (Fleishman and Harris 1962; Glassman and Belasco 1975).
Another line of research focuses on the settlement of grievances. Since early resolution of grievances is usually believed to be to the advantage of all concerned, the emphasis of this research has been on identifying the determinants of early settlements. Among the factors that have been identified are the existence of a cooperative bargaining relationship (Turner and Robinson 1972), the attitudes and experience of first-line supervisors (Knight 1986a), and the parties' learning from previous grievance outcomes (Knight 1986b). Deitsch and Dilts (1986) also found that when attorneys are involved prior to the arbitration stage, a grievance is less likely to be resolved before arbitration.
The above studies have undoubtedly improved our understanding of the factors affecting the resolution of rights disputes. Apparently still missing, however, is an examination of the nature of the settlements. How many grievances are granted, how many are denied, and under what circumstances? One unexamined question is whether grievance outcomes vary at different levels of the grievance procedure. For example, are the grievances that are resolved at the lower levels more or less likely to be settled in favor of the grievant than those resolved at higher levels? The purpose of the present study is to fill this apparent gap in the grievance literature.
Our general theory is that grievance outcomes are affected by the levels at which the grievances are settled, the salary grades and occupations of the grievants, and the nature of the grievances.
Grievance level. Regarding the impact of grievance steps on outcomes, we hypothesize that a grievance is less likely to be granted at higher levels of the grievance procedure than at lower levels. Our reasons for this hypothesis require clarification.
In all 15 work sites examined in this study, the grievance procedure consists of two stages--the informal stage and the formal stage.(1) During the informal stage, the complaint is made orally to the first-line supervisor. A grievant who is dissatisfied with the action (or inaction) of the supervisor can take the complaint to the next stage (the formal stage) by submitting it in writing to the supervisor. Only then does the complaint become officially a grievance.
Within the formal stage, there are five separate steps, and all grievances must go through the first steps before they can advance to the higher ones. In step 1, the grievance is filed by the shop steward, and the department head denies or accepts it. In step 2, the power to deny or accept the grievance is passed along to the director of the institution, and the union participant is the president of the local union. Management and the union are represented by the Regional Director and the regional union representative, respectively, in step 3, and by the National Director and the national union representative in step 4. The final step in the grievance procedure is arbitration. Information collected during our interviews suggests that the parties do, in practice, follow the procedures prescribed in the bargaining contract.
(1) In the Canadian federal government, a grievance procedure is prescribed by statute for each department. Since all the work sites analyzed in the present study belong to the same government department, they have the same grievance procedure.
Although a management representative officially decides for or against the grievant in each of the first four steps of the formal stage, often a denial is obligatory. Federal law specifies the issues over which managers of various ranks have jurisdiction, and a manager below the rank required for deciding a given case must deny the grievance and refer it to the next level. A department head, for example, is authorized to make decisions regarding such matters as working hours; but if faced with a grievance involving discharge, which can be decided only by managers at or above the level of Director of the institution, the department head has no choice but to deny that grievance and refer it to the next level.
Each successive step in the grievance procedure involves a manager (and a union participant) of a higher rank. Several considerations suggest that management representatives at successively higher steps will be less and less likely to decide cases in favor of grievants.
First, a higher-level manager who grants a grievance is, in effect, overturning the decisions made by lower-level management, which may cause resentment within management or affect the confidence of lower-level managers in handling grievances. Second, and more important, reversals of lower-level denials may encourage unions to pursue rejected grievances to higher levels, increasing the cost in time and money of resolving grievances. Third, if the senior managers frequently overturn lower-level managers' decisions by granting grievances, the lower-level managers may react by becoming more lenient, and may eventually begin granting grievances that they should not. And fourth, as Knight (1986a:294) pointed out, the intervention of higher-level management in the grievance process may reduce the rate of informal grievance resolution. Because of these potential problems, we expect that higher-level managers will, as a rule, reverse the decisions of their subordinates only when those decisions ignore the contractual merit of a grievance or are inconsistent with previous decisions.(2)
There are still other reasons to suppose that grievances are more likely to be denied at higher levels. One factor is the personal relationships (if any) among the parties. At the lower steps of the grievance procedure, the participants usually know each other and may interact with each other on a daily basis. We expect this type of relationship to be conducive to tradeoffs and concessions when a grievance is aired, and to result in a fairly high proportion of decisions favorable to the grievants. In contrast, when grievances reach higher levels, the participants to the process are often strangers, and in disputes among strangers there is a tendency for a hardening of position (McPherson 1983). Grievances are therefore more likely to be denied at these levels.
Finally, Knight (1986a:286) noted that grievances that are settled at lower levels tend to involve job-oriented issues, such as work conditions and overtime, whereas grievances that reach higher levels tend to involve policy issues, such as subcontracting. In other words, there is usually more at stake at higher grievance levels. Management may therefore be less inclined to grant a grievance at higher levels than at lower levels.
Salary. We expect that grievances filed by higher-paid employees are more likely to be granted than those filed by lower-paid employees. The former are usually more satisfied than the latter (Borjas 1979:28) and tend to view their organizations more positively (Payne and Mansfield 1973:525). Hence, they are less predisposed to file grievances than are lower-paid employees, and the grievances they do file are more likely to be well founded.
(2) These problems are likely to occur in smaller organizations, where supervisors are usually not trained in handling grievances. In the work sites we are examining, however, lower-level managers commonly attend seminars on the handling of grievances, and access to legal help is always provided to all industrial relations specialists.
Grievance outcomes may also be affected by the types of jobs held by the grievants. Among employees in the same salary grade, however, it is not clear a priori whether a grievance filed by, say, a professional employee is more likely to be granted than a grievance filed by a non-professional employee.
Subject of grievance. The nature of the grievance issue may also affect the settlement. In general, we would expect job-related grievances to have a higher likelihood of being granted than policy-oriented grievances. A job-related grievance usually affects a single employee, whereas a policy-oriented grievance is usually a challenge to some aspect of management rights, and may affect the operation of the entire organization. It is therefore reasonable to expect management to be especially reluctant to "give in" to policy-oriented grievances.
Organizational type. Finally, grievance outcomes may vary across organizations, depending on the functions of the organizations. For example, if most of the tasks in a given organization are necessarily unpleasant or stressful, employees in that organization are likely to have low job satisfaction. The grievances filed by such employees are likely to be less well founded, and therefore more easily rejected, than the grievances filed by more satisfied workers (see the discussion above). On the other hand, in organizations with jobs that pose safety risks for employees, grievances tend to be safety-related (Muchinsky and Maassarani 1980:411); and to the extent that improving the safety of the employees (or avoiding accidents and injuries) also benefits the employer, the success rate for grievants in such organizations may be relatively high.
The sample used in this study consists of 1,174 grievance cases, for the period 1984-85, from 15 federal government work sites in Canada. These work sites are all part of one federal department, and all provide similar human and social services to their geographical constituencies. As a condition of data collection, we agreed not to divulge the identity of the government department or the work sites under study.(3) Three of the work sites are in Manitoba, six in Saskatchewan, and the remaining six in Alberta. Together, they employed a total of 2,170 people in 1984, and approximately 2,050 at the end of 1985. The smallest work site had, on average, 10 employees over the 1984-85 period, and the largest had over 300 employees. More than 90 percent of the employees were unionized. Two national unions represented them: the Public Service Alliance of Canada and the Professional Institute of Public Servants. Since these unions negotiate contracts at the national level, the contract negotiated by each union covers all its members across the 15 work sites.
The grievance data were collected from the records kept at the regional headquarters of the work sites. In addition, a series of interviews was conducted with the management and the industrial relations specialists in the six largest work sites, as well as with the industrial relations staff at the headquarters of those work sites. The interviews were aimed at obtaining additional information about the nature of grievance procedures and processes, and at gaining an understanding of the context within which the grievance activities took place.
For each grievance case,(4) the data that are available are the salary grade and job category of the grievant, the work site, the grievance subject, the step at which the grievance was settled, the outcome of the resolution, and the year of the grievance resolution. There are eleven salary grades, with grade 1 being the lowest-paid category. The job categories are professional employees (such as accountants), administrative employees (such as administrative assistants), administrative support employees (such as typists), and operational employees. There are two kinds of operational employees: maintenance employees, and employees who perform the primary services offered by the department to the public.(5) There are six categories of grievance subjects: discipline, pay, leave, working conditions, assignment of duties, and others. The grievance outcomes are classified as granted, partially granted, denied, or withdrawn.
(3) The data are, however, available on request from the authors.
(4) Only written grievances are considered in this study.
(5) For example, if the department in question were a police department, then the operational employees would include the police officers.
Over the two-year period, the grievance rate for the sample was 541 per 1,000 employees. Sixty-three percent of the grievances filed were from two work sites that together employed about 26% of the total sample of workers, thus confirming that grievance filing varies considerably across work sites. Ninety-three percent of the grievances filed were from individuals below salary grade 5, who constituted only 73% of the workers sampled. Thus, we find evidence that lower-paid workers were more likely to file a grievance than higher-paid workers. Both relatively low pay and low job satisfaction may help account for this result.
With respect to the resolution of the grievances, 98 percent were settled prior to the arbitration stage. This figure is consistent with the results of Gandz's (1982:300) study of grievances in a mining company. A more detailed breakdown, however, reveals that 26 percent of the grievances were resolved in step 1, 25 percent in step 2, 16 percent in step 3, and 31 percent in step 4. These figures indicate that, contrary to conventional wisdom, more grievances were settled in step 4 than at any other level. One possible explanation for this pattern is that grievances with department-wide implications (for example, grievances over retirement age and contracting-out) could be resolved only by the National Director of the department, and were thus automatically forwarded to that level by lower-ranking managers.
Regarding the outcomes of the grievances, 24 percent of the grievances were fully or partially granted, 63 percent were denied, and the remaining 13 percent were withdrawn. Table 1 summarizes the grievance outcomes for each step. The first column reveals that the percentage of grievances that are at least partially granted does tend to decline at higher grievance levels.(6) To examine whether this conclusion holds when the salary grades of grievants and the nature of the grievance issues are controlled for, we performed an estimating equation using the logit model. (Variable definitions are given in Table 2.)
In our empirical model, which is based on 1,174 grievance cases, the dependent variable takes the value of 1 if the grievance was at least partially granted and 0 otherwise.(7) Ordinary least squares regression is not appropriate in this specification. Not only would it suffer from the problem of heteroscedasticity (given the dichotomous dependent variable), but its coefficients might yield a probability estimate lying outside the 0-1 range (Kennedy 1979:144). Because of these problems, the logit model is chosen to estimate the coefficients.
(6) The same pattern is obtained when the grievances that are at least partially granted are expressed as a percentage of total grievances settled. In this case, the figures are 7, 9, 4, and 4 for steps 1-4, respectively.
(7) Table 2 presents definitions of the dependent variable and other variables discussed below.
The independent variables in the estimating equation can be divided into four categories. In the first group, there are four dummy variables, each representing one step in the (internal) grievance procedure.(8) As noted before, we expect a lower probability that a grievance will be at least partially granted the higher the grievance level. In other words, the estimated coefficient for the variable representing higher grievance levels should be smaller than the one representing lower grievance levels.
The second set of variables represents the salary grades and occupations of the grievants. Three salary grades are considered here, and it is expected that lower-paid grievants are less likely to have their grievances granted (or partially granted) than higher-paid grievants. The occupational categories examined are administrative, administrative support, and operational, with professional being the control group.
The third set of independent variables includes five variables, each representing a particular type of grievance: pay-related grievances, discipline-related grievances, grievances over working conditions, grievances over assignment of duties, and grievances over leaves. We included these variables to test whether grievance outcomes vary across grievance issues.
Finally, four variables are used to differentiate the work sites, which, though part of the same industry, have different goals and operational mandates. Depending
(8) Step 5, the arbitration stage, is excluded to avoid perfect collinearity. That is, if all the steps of the grievance procedure had been included in the regression analysis, we would not have been able to estimate the equation. We chose to exclude the final step because of the small number of cases reaching arbitration and also because that step differs fundamentally from the others in allowing the employer no voice in the decision and the employee no discretion to appeal.
Table 1. Grievance Outcomes, by Steps of the Grievance Procedure.
% Fully or % Partially With- Step(a) n Granted % Denied drawn Step 1 314 26 61 13 Step 2 297 35 43 22 Step 3 191 28 54 18 Step 4 358 12 86 2 Arbitration 14 15 85 --
(a) Step 1 involves the shop steward and the department head, step 2 involves the president of the local union and the work site director, step 3 involves the regional representatives of the union and of the department, and step 4 involves the national representatives.
Table 2. Variable Definitions.
Variable Definition GRANTED 1 if a grievance is granted or partially granted; 0 otherwise STEP1 1 if a grievance is settled in step 1 of the grievance procedure; 0 o therwise STEP2 1 if a grievance is settled in step 2 of the grievance procedure; 0 o therwise STEP3 1 if a grievance is settled in step 3 of the grievance procedure; 0 o therwise STEP4 1 if a grievance is settled in step 4 of the grievance procedure; 0 o therwise GRADE1 1 if the grievant holds a salary grade 1 position; 0 otherwise GRADE2 1 if the grievant holds a salary grade 2 position; 0 otherwise GRADE3 1 if the grievant holds a salary grade 3 position; 0 otherwise ADMIN 1 if the grievant holds an administrative job; 0 otherwise SUPPORT 1 if the grievant is a support employee; 0 otherwise OPER 1 if the grievant is an operational employee; 0 otherwise PAY 1 if the grievance concerns pay; 0 otherwise DISP 1 if the grievance concerns discipline issues; 0 otherwise COND 1 if the grievance concerns working conditions; 0 otherwise DUTIES 1 if the grievance concerns assignment of duties; 0 otherwise LEAVES 1 if the grievance concerns leaves of absence; 0 otherwise INSTA 1 if the grievance is filed in a type A work site; 0 otherwise INSTB 1 if the grievance is filed in a type B work site; 0 otherwise INSTC 1 if the grievance is filed in a type C work site; 0 otherwise
on their operational mandate, the work sites provide different levels of services, with widely varying skill requirements, responsibilities, and safety risks.(9) In category A are those work sites at the "high" end of the spectrum in terms of skills, responsibility, and risk to employee safety, and in categories B and C are those in the "moderate" and "low" ranges on the spectrum, respectively. The two work sites that are primarily administrative are categorized as "Others." Thus, to capture the differences in the level of services provided, three dummy variables are added to the estimating equation.
The estimation results are reported in a step-wise manner in Table 3. Equation 1 contains only the variables of interest, equation 2 contains all the variables described in the previous section except for the dummy variables representing types of work sites, and equation 3 includes the entire set of variables. A comparison of equations 2 and 3 suggests that the estimated coefficients are quite stable. The results from both equations also indicate that decisions favorable to grievants are less likely at step 3 than at step 2, and least likely of all at step 4.
Since the estimating equation is of the form (1) P = 1/(1 + e(-xb)), where P is the probability that a grievance is at least partially granted, x is the vector of explanatory variables, and b is the vector of logit coefficients, it follows that the estimated logit coefficients only provide a measure of the logarithm of the odds that a grievance will be at least partially granted.(10) To determine the impact of a change in a particular independent variable on the actual probability that a grievance will be at least partially granted requires a number of steps to convert the estimated logit coefficient into a change in actual probability.
For example, consider the impact of a unit change in STEP1.(11) The adjusted probability as a result of the change in STEP1 can be expressed as (2) P(1) = 1/(1 + e(-xb-b(1)), where b(1) is the logistic coefficient of STEP1. The change in P for a unit change in STEP1 is therefore: (3) dP = 1/(1 + e(-xb-b(1)) - P The above equation implies that the change in P is a function of the probability itself, since xb is a function of P.(12) Thus, to determine dP, we need to assume that the value of P is given. We therefore propose to evaluate the impact of a change in an explanatory variable on the change in the probability that a grievance will be at least partially granted at three probability levels--0.1, 0.24, and 0.5.(13)
(9) A comparable case is the health care sector, in which different hospitals, although in the same sector, provide different levels of services, ranging from acute care to general and auxiliary services.
(10) This conclusion follows from the fact that the estimating equation can be rewritten as Ln(P/(1 - P)) = xb. Expressed in this form, the equation also implies that xb varies with given levels of P. For example, if P = 0.05, then xb = -2.944, and if P = 0.5, xb = 0.
(11) Since STEP1 is dichotomous in nature, a unit change in STEP1 implies going from 0 to 1.
(12) See footnote 11.
(13) By convention, the evaluation is usually done at three probability levels. The choice of the particular probability levels is, however, ad hoc, and there is no reason why other probability levels could not be chosen.
Table 4 presents the actual change in the probability that a grievance will be at least partially granted at each step of the grievance procedure.(14) Assuming an underlying probability of 0.24, our calculations indicate that if a grievance is settled in step 1, the probability that the grievance will be at least partially granted will increase by 0.10. If a grievance is settled in step 2, the probability increases to 0.43. For steps 3 and 4, the probabilities are 0.38 and 0.23, respectively. The results therefore indicate that a grievance is most likely to be at least partially granted in step 2 and least likely in step 4. Thus, grievances that have been denied at grievance levels 2 and 3 are not likely to be granted if they are pursued to level 4.
Our calculations also show, however, that it may be worthwhile to pursue those grievances that have been denied in step 1. The probability of having a grievance partially granted at this level is only 0.34 (assuming an underlying probability of 0.24), whereas in step 2 this probability is 0.43. As mentioned earlier, department heads, who are responsible for handling grievances in step 1, are more limited in the types of grievances they can grant than is the director of the work site. To the extent that the department head's authority is more limited than the director's, grievances are more likely to be denied at step 1 than at step 2.
By the same token, of course, managers at step 2 have less discretion to grant grievances than those at step 3, and managers at step 3 have less discretion than those at step 4. The higher denial rate at step 1 than at step 2 evidently reflects other factors as well.
Information collected from our interviews suggests some additional explanations as to why grievances are more likely to be granted at step 2 than at step 1. In the work sites under study, most of the decisions that one would expect to be taken by department heads (for example, decisions over purchasing new equipment for the unit) are in fact made by the director. This centralization of decision-making may extend to the granting of grievances. For example, grievances over discipline and discharge, paid sick leave exceeding 60 days, and leave for attending union business can be decided only by the directors or their superiors. Another explanation is that department heads are often reluctant to deal with identical (or similar) grievances filed by a number of individuals separately. In such cases, the employees usually attach particular significance to the subject of the complaint, and by filing grievances together they may intend to send a message to the director.
With regard to the other variables, equation 3 suggests that grievances over pay and working conditions are more likely to be granted at one step or another than grievances over discipline and assignment of duties. A lower-paid grievant is less likely to be granted a grievance than a higher-paid grievant. The dummies for type of work site are statistically insignificant.
(14) The estimated coefficients used to calculate Table 4 are from equation 3 of Table 3.
Table 5 presents the actual probability that a grievance will be at least partially granted, by grievance issue and by step. To calculate these probabilities, the estimates of equation 3 (Table 2) are used, and it is assumed that the grievant is in administration and holds a salary grade 2 position. Furthermore, it is assumed that the grievant works in a category B work site.(15)
The results show that for all five grievance issues examined, the probability that a grievance will be at least partially granted is smaller at step 3 than at step 2, and smallest at step 4. Our calculations also indicate that for any given step, grievances over working conditions (for example, over faulty ventilation and the lack of privacy in washroom facilities) are most likely to be granted, whereas grievances over assignment of duties (for example, over the interpretation of job descriptions) are least likely to be granted. This finding probably reflects the greater cost of granting a grievance over assignment of duties than of granting a grievance over working conditions. Working conditions tend to be specific to the work site; issues concerning the assignment of duties have more general application, and changes in work duties may affect all the work sites in the department.
Our results suggest that grievances are more likely to be at least partially granted by management in the first stages of a grievance procedure than in the later stages. This finding obviously does not mean, however, that unions should never pursue grievances denied in the initial steps. After all, many grievances in this study were won in the later stages of the grievance procedure, and even grievances that are lost might serve important purposes, such as setting the stage for a contract demand in future negotiations.
We have also found that higher-paid employees were more likely than lower-paid employees to have their grievances at least partially granted. Furthermore, grievance outcomes apparently varied with grievance issues; grievances over working conditions were most likely to be granted and grievances over assignment of duties least likely. If this finding is supported by further studies, union representatives should recognize that their chances of winning a grievance short of arbitration will depend in part on the subject of the grievance; most employers, that is, are probably more likely to resist granting grievances over some subjects than over others.
The results presented above are subject to a number of qualifications. First, of course, the study is limited to the public sector and to a single employer's relationship with two unions. Since the grievance process here will differ from that in many other cases, one must be careful in generalizing the findings of the present study. Second, our analysis does not take into account the quality of the union-management relationship, which previous studies have shown influences the rate of grievance filing and the outcomes of grievance reviews. Finally, we did not examine the role of industrial relations specialists at various levels of the grievance process. Clearly, depending on their involvement and influence, such specialists may affect the process and outcomes of grievance filing.
(15) These assumptions are necessary because the results presented in Table 5 require knowledge of the salary grade and occupation of the grievant, the work site, and so on. It should also be noted that the probabilities presented in Table 5 would change depending on the assumptions made. For more details on the calculation of these probabilities, see Pindyck and Rubinfeld (1981), Chapter 10.
Table : Table 3. Determinants of Grievance
Table : Outcomes (Logit Results).
Table : (N = 1,174)
Table : Table 4. Estimated Coefficients and Changes in the Probabilities of Favorable Grievance
Table : Outcomes Associated with Four Grievance Steps.
Table : Table 5. Probability of a Favorable Outcome,
Table : by Grievance Steps and by Issues.(a)
|Printer friendly Cite/link Email Feedback|
|Author:||Ng, Ignace; Dastmalchian, Ali|
|Date:||Apr 1, 1989|
|Previous Article:||The impact of bargainer experience on teacher strikes.|
|Next Article:||The unemployment rates of men and women: a transition probability analysis.|