Printer Friendly

Using event history analysis to model delay in grievance arbitration.

A persistent criticism of grievance arbitration is that far too much time elapses between the filing of the grievance and the rendering of a decision by an arbitrator. Such criticism strikes at the heart of the modern North American grievance arbitration system, which evolved initially to provide informal and speedy resolution of employment disputes. In 1973, William Usery stated:

It has been said that time is money. But in the highly human area of resolving disputes that have a direct impact on the lives of people, time is a far more cherished commodity. For a worker to wait nine months and longer to find whether a discharge will stick is cruel to the worker, places a strain on his family and friends, and is costly to management both in dollars and in employee. morale. (Usery 1973)

When Usery made his comments, FMCS records showed that the average elapsed time from the filing of a grievance to an arbitrator's award was 257 days (Mills 1986:292); twenty years later the elapsed time had increased to 317 days (Records of the Federal Mediation and Conciliation Services 1992).

In this paper, using an event history analysis approach, we systematically examine the factors that contribute to delay in grievance arbitration. While there have been numerous attempts to reduce delay, most of the research to date has been concentrated on examining the extent of delay and the problems it creates. Analysis of the causes of delay in grievance arbitration has been limited, either because the research has been directed to a practitioner audience or because the focus is on the consequence rather than the cause of delay. Thus, studies of the factors that contribute to delay have been limited in scope and methodology. In this paper, we review the results of previous studies of time delay and the methods used in those studies, describe event history analysis, and apply it to a sample of arbitration cases from the province of Alberta.

Previous Research

Delay in arbitration is not a new issue. In a seminal article, Ross (1958) presented data on the "well-aged arbitration case." Comparing cases published by Labor Arbitration Reports in two time periods, 1945/46 and 1955/56, Ross found that elapsed time from grievance filing to arbitrator decision had increased by more than 50% in a decade. Among various types of cases, those involving discharge and discipline took almost two months less time than others, a result attributed to the pressure placed on the grievant and the increased exposure of the employer to retroactive pay with the passage of time. Ross cited a number of factors that might be contributing to delay, including the increased use of attorneys, a shortage of acceptable arbitrators, and hearing postponements. The article concluded with a cautionary note about the implications for arbitration if the trends reported in the article continued.

Some thirty years later, Stieber, Block, and Nichol (1990) attempted to determine the extent to which these trends had continued. In the most comprehensive examination of the delay in the time between grievance filing and arbitration decision, the authors presented data from four different data sources. Although these sources produced varying results, with average delay ranging from 253 days to 364 days depending on the data set,(1) the authors concluded that "the trend noted by Ross for a single decade is shown to have continued in the ensuing three decades" (Stieber et al. 1990:131). Unlike Ross, however, they found, based on an analysis of one of their data sets, that the presence of attorneys did not increase delay.

Studies in Canada, which has a system of grievance arbitration similar to that of the United States (Thornicroft and Eden 1995), have produced comparable results. Research in Ontario showed that the elapsed time from grievance filing to arbitration award increased from 250 days in 1973 (Goldblatt 1974) to 301 days in 1980 (Winter 1983) and 345 days (expedited awards excluded) by 1983 (Rose 1986). Investigations in Alberta indicated that elapsed time for grievance resolution grew from approximately seven months in the early 1970s (Fricke 1976) to 345 days by the mid-1980s (Olson 1990). A review of the experience in Newfoundland from 1980 to 1992 found that elapsed time increased an average of eight days per year over the twelve-year period (Thornicroft 1995). Ponak and Olson (1992) examined two time periods in the most widely used Canada-wide reporting service, Labour Arbitration Cases. Cases reported in 1975 took an average of 283 days from filing of the grievance to issuance of the arbitration award; by 1987/88 the average elapsed time had increased to 428 days.

With respect to sources of delay, the Canadian studies showed that delay was increased by the use of three-person boards of arbitration (Goldblatt 1974; Winter 1983; Olson 1990; Thornicroft 1995); attorneys (Goldblatt 1974; Olson 1990; Thornicroft 1995); and the filing of preliminary objections over arbitrability (Olson 1990; Thornicroft 1995).(2) The research also found that discharge arbitrations were handled more expeditiously than other types of cases (Goldblatt 1974; Olson 1990; Thornicroft 1995), public sector cases took longer than private sector cases (Olson 1990; Thornicroft 1995), and the busiest arbitrators also were the slowest (Olson 1990; Thornicroft 1995).

Three studies examined delay at different stages of the process (Winter 1983; Stieber et al. 1990; Olson 1990). They showed that completing the pre-arbitration grievance steps and the scheduling of the arbitration hearing were the most time-consuming parts of the process and that the preparation of the award and the selection of the arbitrator were the most expeditious. Both Winter (1983) and Stieber et al. (1990) suggested that delay might be better understood through an analysis of the different stages of the process. Olson (1990) found some support for the proposition. that different factors could account for delays at different steps of the process.

Researchers' interest in the extent of delay and its sources is motivated by a wide consensus that delay is harmful to employees, the arbitration process, and the union-management relationship. Negative consequences of delay that have been cited include harm to formal contract negotiations; inequities created for the grievant in terms of financial remedies available; a possible inverse relationship between elapsed time and likelihood of reinstatement (Rose 1986); financial loss to the employer, especially in discharge cases, if the grievance is sustained (Sloane and Whitney 1985); a strong correlation between the length of a reinstated employee's suspension in discharge cases and elapsed time (Bemmels 1988); harm to the arbitration hearing itself as memories of the material events dim with the passage of time (Prasow and Peters 1983); employee restlessness and uncertainty among supervisors (Ross 1958); and a relationship between the incidence of wildcat strikes and a slow and backlogged grievance procedure (Brett and Goldberg 1979, 1983; Rose 1986; Kochan and Katz 1988).

Given these findings, it is unsurprising that a survey of over 350 union and management arbitration advocates reported that 44% of the respondents ranked delays at various stages of the process as the most serious fault in the arbitration system (Berkeley 1989). The study showed that delays were of far more concern to union advocates (51% ranked delays first) than to management advocates (38%). The widespread interest in time-saving (and cost-saving) devices like expedited arbitration and grievance mediation further attests to the concerns of the parties over the length of the entire process.

In summary, research to date carried out in the United States and Canada confirms that the grievance arbitration process takes close to one year from initiation to completion, that the elapsed time is increasing, and that delay has negative implications for the participants. The research further indicates a number of factors that may be associated with delay (such as type of issue) and suggests that disaggregating the grievance arbitration process into distinct stages may help clarify the causes of delay.

Research Model

The modern grievance arbitration system has been discussed in detail by numerous students of industrial relations, and we will not review it here (see, for example, Slichter, Healy, and Livernash 1960; Lewin and Peterson 1988). There is little question that a certain amount of delay is inevitable no matter how strong the parties' desire to expedite the process. Time must be spent assessing and discussing the issues involved, arranging meetings, writing letters, collecting relevant information, researching cases, and so forth. Some parties will be more efficient at reducing these delays and some collective bargaining relationships will contain features that will increase delay. For example, using a three-member arbitration board (a common practice in some parts of Canada) increases scheduling time, and having longer time limits or more steps in the grievance procedure also tends to increase delay.

Stages of Delay

We propose that the factors affecting delay vary in different stages of the grievance arbitration process. This reflects the fact that the kinds of activities undertaken by the parties vary depending on the stage of the process. Following Winter (1983) and Stieber et al. (1990), we have broken the grievance arbitration process into four discrete stages: pre-arbitration grievance steps; arbitrator selection; hearing scheduling; and preparation of the arbitration award.

Pre-arbitration grievance steps. This is the initial stage of the process. Typically, an individual employee initiates discussion of the matter that is to be grieved, usually after consulting with a union representative. The matter is then discussed, in a multi-step process, by increasingly senior union and management representatives, who attempt to resolve the issue in dispute. The collective agreement normally specifies in some detail which levels of the union and management hierarchy are involved in the discussion at each step and also stipulates time limits for each step. Though grievances may be lost for failure to adhere to the specified time limits, it is more common for the sides to extend time limits by mutual agreement. If discussions fail to resolve the issue in dispute, the matter is referred to arbitration. In this paper, we define this stage as the number of days from the date the grievance was filed to the date it was referred to arbitration.

Arbitrator selection. In the event the grievance is referred to arbitration, the parties must choose a mutually acceptable arbitrator. An arbitrator may be chosen on an ad hoc case-by-case basis or may be pre-selected through, for example, a rotation list stipulated in the collective agreement. In situations where the arbitration selection occurs by case, considerable positioning may occur before agreement is reached on the particular arbitrator to be used.(3) Selection of the arbitrator is considerably simplified in situations where the collective agreement contains a list of acceptable arbitrators who are chosen in a set rotation or by lot. As well, the parties may agree beforehand to allow a neutral agency, such as the American Arbitration Association, to appoint the arbitrator. Where a three-person arbitration board is used, an additional step is added as each side appoints one member to the arbitration board at its discretion. The third person, who serves as the arbitration board chair, is then selected in the same manner as described above. The arbitrator selection stage is defined as the number of days from the referral of the grievance to arbitration to the arbitrator's notification that he or she has been selected.

Hearing scheduling. Once the arbitrator (or arbitration board) has been appointed, the next step is scheduling the hearing. In the case of a sole arbitrator, this means finding a mutually acceptable date for three people - the arbitrator, the union representative, and the management representative - as well as ensuring that necessary witnesses also are available. Where a three-person arbitration board is used, the schedules of five people must be coordinated - the three members of the board plus the union and management representatives. This stage is defined as the number of days from arbitrator notification to the date of the first hearing.

Preparation of the arbitration award. The last stage of the process occurs following the hearing, as the arbitrator prepares the decision. Time is spent by the arbitrator reviewing the evidence and arguments, reading any relevant precedent cases submitted by the parties, awaiting any post-hearing briefs, and drafting the award.(4) Where a three-person board is involved, the board will usually convene following the hearing to discuss the case. The chair then drafts a decision, which is circulated to the other two board members. Further discussion and even meetings may take place among the three board members until the decision is finalized. Most frequently, the appointee of the losing party dissents from the decision; unanimity is rarely achieved. This stage is defined as the number of days from the last hearing day to the announcement of the award.(5)

Based on this description of the different stages of the process, in which the underlying dynamics of the processes at each stage of arbitration are different, we propose two general hypotheses about the nature of delay: (H1) the factors that affect delay will differ at each stage of the process; and (H2) some factors that appear to be related to overall elapsed time (that is, total delay) will not be related to elapsed time at any particular stage. With regard to the latter hypothesis, we reason that this is a case in which the whole could easily be different from the sum of its parts. For a specific case, total delay is the aggregation of its four components. Total delay is thus affected to some extent by the factors that affect the components. But because each component varies in length, the aggregation is staggered, and this staggering can differ for each case. Therefore, across cases total delay is not a simple aggregation of its components but a complex staggered aggregated function. Factors may affect this aggregated function yet not affect its components. Conversely, factors that affect a particular component may not affect the aggregate delay.

Causes of Delay

We propose that delay at each stage will be related, although possibly in a different manner at each stage, to the nature of the parties, the type and complexity of the issue being arbitrated, the structure of the arbitration, and the workload of the arbitrator. With respect to the nature of the parties, we propose that a public/private sector distinction is most important. Previous research suggests that public employers are more bureaucratic and have more complex decision-making processes than their private sector counterparts (Ponak and Thompson 1995; Derber 1979). Public sector unions often mirror these tendencies (Rose 1995). Accordingly, we hypothesize that (H3) arbitrations in the public sector will take more time than those in the private sector. Further, (H4) we expect sector to have the most impact in the pre-arbitration grievance steps and little or no impact at the award preparation stage (which we expect to be largely immune from the parties' internal decision-making constraints).

The issue at stake in the arbitration and the complexity of the case also should affect delay. Because of the special human and financial costs involved in discharge, and specific contract language that, commonly, expedites termination cases, we expect that (H5) discharge cases will be handled more expeditiously than other cases at all stages of the process.(6) Job competition cases also have special personnel implications, and thus we expect that (H6) job competition grievances will receive the second-fastest handling (after discharge cases) throughout all stages of the process.(7) We have no specific hypotheses regarding other categories of grievances.

In addition, for obvious reasons, the more complex the case (no matter what type of case it is), the longer it can be expected to take. Complexity can be manifest in several different ways. We know whether either the union or employer had registered a preliminary objection, whether more than one day of hearing was required, and the total page length of the award. Each of these variables is an indicator of case complexity. We hypothesize that (H7) complexity will increase delay in the pre-arbitration grievance steps, the arbitrator selection stage (the more complex the case, the more difficult it is to choose an arbitrator), and the award preparation stage. Complexity should have less impact on scheduling, unless more than one day of hearing is required.

A third set of factors that is expected to be related to delay is structural aspects of the arbitration process itself. In particular, previous researchers have noted that using a three-person arbitration board, rather than a sole arbitrator, greatly increases delay. In particular, we hypothesize that (H8) use of a three-person board will lengthen the time needed to schedule the arbitration hearing, and (H9) use of an arbitration board also lengthens the preparation stage because of the need to discuss the decision among the board members. The role of the legal profession is another structural variable that has been associated with delay, as it is believed that lawyers take a more legalistic and time-consuming approach than do lay people. We hypothesize that (H10) the use of legal counsel by one or both sides will contribute to delay in all stages of the process.

Finally, we expect that the workload of the arbitrator will have an impact on delay at the scheduling and preparation stages. While there may be many active arbitrators in any given jurisdiction, the usual pattern is for a small proportion of popular arbitrators to handle the majority of the cases. In our sample, for example, seven of the 47 arbitrators issued 60% of the awards. We hypothesize that (H11) choosing one of the busier arbitrators will add to both scheduling and preparation time.

Statistical Methods

To test our hypotheses, we analyze a sample of arbitration cases using event history analysis, a procedure that is particularly appropriate for analyses of dependent variables that measure time to a certain event. This procedure deals well with issues such as the "censoring" of data, involving, for example, questions concerning how to consider incomplete information on the timing of events. Event history analysis also can address explanatory variables that change in value over the observation period, repeated versus non-repeated events, and multiple kinds of events, and it allows the base likelihood of an event's occurrence to vary over the observation period (Allison 1984). Organizational phenomena that have been investigated using event history analysis include turnover (Morita, Lee, and Mowday 1989), waiting time to first product introduction (Schoonhoven Eisenhardt, and Lyman 1990), attendance behavior (Fichman 1989; Harrison and Hulin 1989), illegal corporate behavior (Baucus and Near 1991), auditor-client attachment (Levinthal and Fichman 1988), and the time to settlement when strikes are involved (Kiefer 1988).

Event history analysis centers around the estimation of two related rates, the survival rate and the hazard rate. These terms come from biostatistics, in which the typical event is death. The survival rate at time t in biostatistics estimates the probability that an individual will survive beyond time t. The hazard rate in biostatistics is the instantaneous rate of death at a given time t, given that the individual survives up until time t.

Since event history analysis acknowledges that these rates are most likely to vary over time, they are described as functions of time. The survival function describes a set of survival rates for a given time interval. In the case of turnover research, the event is voluntary turnover, and for smoking cessation research it is relapsing. For grievance arbitration research, the survivor function represents the probability that an arbitration will be delayed beyond a given time. Therefore we will refer to this function as the delay function.

The hazard function specifies the hazard rate at time t. Rather than use the term hazard, which has a negative connotation, we will use the term transition. Thus the transition rate at time t is the instantaneous rate of completion of an arbitration stage at time t, given that completion has been delayed at least until time t.

Note that the delay and transition rates are not merely inverses. The transition rate captures the likelihood of a transition from non-completion to completion at a given point in time, whereas the delay rate captures the prevalence of delay past a point in time.

While a great deal can be learned from the inspection of these functions, an important aspect of event history analysis is its evaluation of the effect of possible explanatory variables on these functions. Statistical regression models have been developed for event history analysis, similar to multiple linear regression models, that can be used to model the relationship between predictor variables and a normally distributed dependent variable (Cox and Oakes 1984).

We use regression models to assess the simultaneous effects of the predictors of delay at each stage. Two major classes of regression analyses are possible: parametric regression models and semiparametric regression models. Foremost among semiparametric models is the proportional hazards model (Cox 1972). The major differences between the two classes are the functional form of the dependent variable and the statistical assumptions underlying each model. Parametric regression models use the logarithm of the survival time as the dependent variable, whereas the proportional hazards model uses the log of the hazard function as the dependent variable. Parametric regression models require that a parametric distribution be specified for the error terms (for example, log-normal, gamma, log-logistic, or Weibull distributions). The proportional hazards model does not require that such a distribution be specified, but this model assumes that the ratio of the hazard rates of any two individuals at any point in time is constant.


Arbitration data for our analysis cover the four-year period 1985-88 in the province of Alberta. Under Alberta labor law, arbitrators are required to file all grievance arbitration awards with the Department of Labour. Approximately 600 awards were filed in the period under review. Union density in Alberta during the study was approximately 30%.

An analysis of the arbitration awards provided data on total elapsed time from the grievance filing to the arbitrator's decision, elapsed time from the hearing to the decision (that is, preparation of decision stage), and some of the independent variables (for example, use of legal counsel, three-person board, sector, issue). Additional data were obtained from collective agreements, interviews, and the records of unions, employers, and legal counsel. Information on the elapsed time at each stage of the arbitration process was not always available; this accounts for the varying number of cases analyzed at each stage of the process.



Descriptive statistics for the variables used in this study are shown in Table 1. The average total delay was 333.6 days, the median was 307 days, and the maximum was 936 days, indicating positive skewness. A similar pattern was evident for each delay measure. This should not be surprising given the nature of delay: it has zero as its lower bound and the value of an infrequent but extreme case as its upper bound.

Zero order correlations among variables are shown in Table 2. These data show that there is some intercorrelation among the predictor variables, but low correlation between the delay at different stages of the process. For example, there is a correlation of only -.03 between delay at the pre-arbitration grievance stage and delay at the selection stage. These low correlations suggest that delay at any one stage is not related to delay at other stages. This finding provides preliminary support for our approach of disaggregating the grievance arbitration process into distinct stages.

Analyses of the time delays at each stage of the grievance arbitration process were performed using the LIFEREG procedures in SAS (SAS institute 1990). The total delay rate created with the LIFETEST procedure is shown by time interval in Table 3. As defined earlier, the delay rate is the number of unfinished arbitrations at a given point in time as a proportion of all arbitrations. Table 3 shows that 552 out of 564 grievances had not had awards issued by 100 days following the occurrence of the grievance; the delay rate at that point was .98. At 200 days the survival rate was .80, meaning that overall there is an 80% probability that an arbitration will be delayed beyond 200 days. Figure 1 plots the delay rate over time to form the delay function.


Hazard Rate of Arbitration Delay

The transition function is shown in Figure 2. The shape of the function shows that the transition rate increased steadily from the time of the grievance to about 350 days.(8)

The transition function of each of the four stages is presented in Figures 3-6. Simple observation reveals that the four individual transition functions have much flatter profiles than the transition function for overall time delay shown in Figure 2, reinforcing our decision to disaggregate time delay in arbitration into discrete stages.(9)

The fairly flat slope of the transition function in the first three stages indicates that the transition rate changes little with the passage of time; thus, contrary to what one might expect, any pressure caused by the increasing delay does not appear to translate into action to move the grievance to the next stage of the process. In fact, the transition functions in the first two stages have a slight downward slope. For the pre-arbitration [TABULAR DATA FOR TABLE 3 OMITTED] grievance steps, shown in Figure 3, this suggests that the likelihood of a grievance being referred quickly (that is, the next day) to arbitration actually declines with the passage of time. A similar pattern is evident for the arbitrator selection stage, shown in Figure 4. As time passes, the likelihood of agreement on an arbitrator also declines. After approximately 275 days, however, the hazard function begins to slope upward, suggesting that at this point pressure builds to choose an arbitrator.

The transition function for scheduling delay is reported in Figure 5. The graph indicates that the likelihood of a first hearing being scheduled increases slightly from the date of notification of the arbitrator for about 125 days and then declines very slowly but steadily. This profile is not surprising, as the actual process of scheduling is a onetime event rather than a recurring one. Once a hearing is scheduled for some date in the future, the date chosen becomes fixed and rarely changes.

Of the four stages, the transition function for delay in preparation of the decision, shown in Figure 6, shows the most change as time passes. The likelihood of an award being issued increases moderately for the first 100 days after the hearing, then declines moderately for the next hundred days. After 200 days the likelihood of an award being issued increases fairly sharply. The shape of the function suggests that the initial pressure on the arbitrator to quickly issue the award dissipates after three months, but then re-emerges once six months have passed.(10)

While the transition function graphs for the four stages are far from identical, the degree of similarity among the curves in the first three stages was not expected. Our earlier description of the different stages of the grievance arbitration process led us to believe that the transition functions would be dissimilar, a reflection of the different activities undertaken at each stage. The rationale for disaggregating the process rests on the assumption of dissimilar dynamics at each stage of the process. The similarity of the transition functions (at least in the first three stages) raises questions about these assumptions. In the next section, we model the factors that are expected to be systematically related to delay at each stage. This will provide a more rigorous test of whether the underlying causes of delay vary at different stages of the grievance arbitration process.

Regression Models

Our initial assessment of the fit of the proportional hazards model to the data indicated a violation of the proportional hazards assumption. Conversely, the parametric regression model provided a good fit to the data. We used a log normal distribution to model total delay, arbitrator selection delay, and decision preparation delay, and a Weibull distribution to model the grievance procedure delay and scheduling delay.(11)

The coefficients for each of the covariates as predictors of total delay and delay at each stage are shown in Table 4. The coefficients reported in Table 4 can be interpreted much like unstandardized regression coefficients, with the log of time as the dependent variable. For example, in the equation predicting total delay, the coefficient of -.20 for the size of the board means that if the case involves a three-person board, the log of time is decreased by .20 compared to that for a sole arbitrator, controlling for other variables. Perhaps more intuitively apprehensible are the results obtained by exponentiating (taking the antilog of) the coefficients. The exponentiation of .20 is 1.22, which means that the predicted time for cases with a three-person board exceeds that for cases with a sole arbitrator by about 22%.

Interpretation of the statistical significance of the coefficients also follows traditional regression procedures. For moderate to large samples, the ratios of the estimates to their standard errors can be treated like t-statistics in ordinary multiple regression (Allison 1984). Similarly, the relative size of these ratios can be used to gauge the relative importance of the variables.

The data in Table 4 provide support for our first two general hypotheses. As is discussed below in more detail, it can be seen that different sets of factors are responsible for delay at different stages of the arbitration process (H1). For example, delay in the pre-arbitration grievance steps is related to the type of issue, hearing duration, and award length, while delay in arbitrator selection is a function of board size and use of legal counsel. There is also support for our second hypothesis (H2). Variables related to total delay, such as sector, are not necessarily related to delay at particular stages. Conversely, and consistent with this hypothesis, some variables related to delay at several stages, such as legal counsel, are not related to total delay.

Total delay. Table 4 shows that four of the eight variables in the model have a statistically significant association with total delay. As predicted (H5), discharge cases, which provide high potential costs for all participants, are processed more quickly than other grievances. This is shown by a statistically significant negative coefficient for discharge cases and statistically significant positive coefficients for the other issue categories. Also as predicted (H6), cases involving job competition are next quickest to be completed, as shown by the relative size of the coefficient. However, the differences between non-discharge coefficients are not statistically significant. Hearing length is associated with increased delay, also as expected (H7). To the extent that hearing length is a good proxy for case complexity, these data support the conclusion that case complexity contributes to total delay. Neither preliminary objections nor award length is associated with total delay, however.

The size of the board, which we would expect to contribute to the number of interpersonal transactions and considerations involved in scheduling and decision preparation, is positively associated with total delay, and this association is statistically significant. The analysis also shows that, as predicted (H3), the process takes longer in the public sector than in the private sector.

None of the other variables exhibits a statistically significant association with total elapsed time.

Pre-arbitration grievance steps delay. Three [TABULAR DATA FOR TABLE 4 OMITTED] of the eight measures in the model are statistically significant. As expected, discharge cases are handled more expeditiously than non-discharge cases, although - contrary to expectations - job competition cases are not handled more quickly than the others (H6). The results with respect to case complexity are ambiguous, however. On the one hand, cases that produced arbitration hearings of two and three days in duration (a proxy for complexity) take more time to move through the pre-arbitration grievance steps (H7). However, cases in which hearings are longer than three days (a very small portion of the sample) do not predict elapsed time in a statistically significant manner, although the signs of the estimated coefficients are in the expected direction. Contrary to expectations (H7), length of award, another proxy for complexity, has a small but statistically significant negative association with delay. These results suggest that caution is required both with respect to our hypothesis that complexity contributes to delay and with respect to our use of length of hearing and award length as measures of complexity. The measure of preliminary objections is not associated with grievance procedure delay.

None of the other variables is statistically significant, a particular surprise with respect to public sector cases. We expected (H4) that the greater complexity of organizational decision making in the public sector than in the private sector would lead to slower processing of grievances in the public sector. While the sign of the estimated coefficient is in the expected direction, this association is not statistically significant at standard test levels.

Arbitrator selection delay. Selection of the arbitrator is affected most strongly by the use of a three-person board, a relationship we did not expect. Reliance on a three-person board of arbitration (rather than a sole arbitrator) increases the time for selecting the neutral arbitrator (that is, the board chair) by approximately 50%. The increased delay is partly explained by the fact that some time must be spent by each side appointing its partisan nominee prior to selecting the neutral board chair. Contrary to our prediction (H10), when both sides use legal counsel, the selection process is shorter than when neither or only one does.

Scheduling delay. Two covariates have a statistically significant association with scheduling delay, each in the predicted direction. Discharge cases are scheduled more quickly than other kinds of cases (H5), possibly a reflection of the special human and financial costs that arise in a termination. The use of legal counsel substantially increases the delay at this stage of the process (H10). This finding is consistent with the belief that lawyers are more difficult to schedule than non-lawyers, thus contributing to delay.

There is no statistically significant relationship, however, between scheduling delay and board size (H8) or arbitrator workload (H11). Busier arbitrators are no more likely to contribute to delay than their less busy colleagues. More surprising is the absence of a statistically significant relationship between board size and scheduling delay. Previous studies have indicated a positive relationship between use of a three-member board and scheduling delay. Though these studies do not disaggregate their data or employ multivariate procedures, anecdotal evidence suggests that the dilatory effect of a board is greatest at the scheduling stage. Furthermore, common sense suggests that scheduling five busy people will take longer than scheduling only three. Not only do our results fail to confirm this hypothesis, but the sign on the relevant estimated coefficient is actually the opposite of what we expect.

Decision preparation delay. As predicted, arbitrators issue their decisions more quickly in discharge cases than in other types of cases (H5), with job competition cases the next fastest (H6). Also as hypothesized, a sole arbitrator is able to issue a final decision more quickly than a three-person board (H9), possibly a reflection of not having to circulate and discuss the award with board members. More popular (that is, busier) arbitrators, however, are slower at issuing their decisions than their less popular colleagues (H11). Thus, the desire of the parties to rely on the handful of highly acceptable arbitrators does have some consequence in terms of a slightly longer delay in receiving the final award. Award length also is associated to a statistically significant degree with delay (H7). Lengthier and more complicated cases apparently take longer to prepare and involve a longer gestation period for the arbitrator.

The effect of outside counsel, which is to reduce delay, is again contrary to our expectations (H10). We predicted that outside legal counsel would increase legal maneuvering as well as transaction time, and hence the length of each stage of the process. At the scheduling stage this expectation is borne out, but in the pre-arbitration grievance steps there is no statistically significant relationship between use of outside legal counsel and elapsed time. However, at the arbitration selection stage, use of legal counsel by both sides actually reduces delay. The same finding, even more strongly confirmed, is evident in the decision preparation stage. One explanation for this result is that legal counsel, rather than making the issue in dispute more complicated, may actually serve to sharpen the matters to be decided, thereby reducing delay. This post hoc explanation is admittedly speculative; the whole issue of the impact of legal counsel on arbitral delay needs to be examined more carefully.


We have investigated the factors that contribute to the length of the grievance-arbitration process. Our analysis of data on approximately 600 grievance arbitration awards in Alberta shows that a different combination of factors explains time delay at each stage of the process. In fact, only two of the variables predict elapsed time in more than two stages of the process. Furthermore, a comparison of the results for total delay with the results for each stage of the arbitration process reveals that the variables that predict total delay differ from a simple summation of the variables that predict delay at each stage. Our results suggest that the causes of delay in grievance arbitration cannot be understood by examining only total delay. While the time taken in the overall process is certainly important, this paper demonstrates that a more fruitful approach to understanding the causes of delay is an examination of each stage of the process.

With respect to individual variables, the most consistent predictor of delay is the nature of the case. Discharge cases are handled more expeditiously than non-discharge cases at every stage of the process except arbitrator selection. These findings suggest that the parties are capable of moving more quickly when it is in their mutual interests to do so.

The presence of legal counsel is another variable with a statistically significant relationship to delay at three stages of the process. The results, however, are not always in the predicted direction. During arbitrator selection and decision preparation, the presence of legal counsel reduces delay, a finding contrary to our hypotheses, most previous research, and almost all anecdotal evidence. Only during the scheduling stage does legal counsel contribute to delay.

These findings suggest that perceptions of how attorneys affect delay may be skewed by the adverse effects they have on scheduling, a relatively transparent process in which it is easy to identify the individual whose schedule is the busiest. Why decision preparation and arbitrator selection apparently occur more rapidly when attorneys are present than when they are absent is puzzling. Perhaps attorneys bring certain efficiencies to bear in those processes (for example, sharpening the issue in dispute). More research on the impact of attorneys on the process of arbitration (as opposed to their impact on arbitration outcomes) would be useful, focusing in particular on the roles played by legal counsel at the different stages of the process.

A second variable toward which future research attention should be directed is hearing length. This variable had a substantial impact on total delay and some impact on delay during the pre-arbitration grievance steps, but little effect on delay in the other three stages of the process. Perhaps these findings stem, in part, from the way we defined hearing time.(12) We simply calculated hearing time as the total number of hearing days. However, there is a big difference between a three-day hearing that occurs over three consecutive days and one in which the three days are spread over a period of some months. The latter situation, which would greatly increase total delay, typically arises where the parties have initially underestimated the length of time necessary to complete the case and have scheduled insufficient hearing days at the outset. They are then forced to scramble for additional hearing days on the completion of the originally scheduled hearing. Sometimes there is no mutually available date for several months. To properly address this possibility (which we were unable to do using our data set), future research should consider elapsed time between the first and last day of the hearing as a specific stage and carefully examine situations in which the parties have failed to initially schedule sufficient hearing time. The frequency of this occurrence, the factors associated with it, and its relationship to total delay all could prove to be important variables.

In addition to the variables that are good predictors of time delay, there are several variables that, contrary to expectations, have little or no impact. Among them are sector, the filing of preliminary objections regarding arbitrability, and arbitrator workload. The apparent unimportance of arbitrator workload is of particular interest because it suggests that the parties may not necessarily expedite the arbitration by choosing an arbitrator for whom there is below-average demand.


Many suggested approaches to reducing time delays in arbitration, such as the introduction of grievance mediation (Brett and Goldberg 1983) or statutory expedited arbitration (Rose 1986), have been aimed at a fundamental alteration of the system. This study suggests that attempts to speed up grievance arbitration can be directed at changing specific stages of the process rather than the entire process. However, such approaches must be carefully considered. Solutions that might reduce, for example, time spent in the selection of the arbitrator will not necessarily reduce the amount of time needed to schedule the hearing (and indeed may have the opposite impact). The need to search for solutions at each step of the grievance arbitration process will not surprise those experienced in that process. Yet much of the criticism of the length of arbitration is directed at how long the process takes in total. Such criticism may deflect attention from where it should be directed - the length of time taken at each stage.

Further technical details on the methodology and the computer programs used to generate the results are available from Allen Ponak.

1 The authors provided little discussion of why the results varied by data source. Some of the variation probably results from the fact that two of the data sources were subsamples of larger data sets, one by region and one by arbitration issue. The authors commented as well about some procedural differences in arbitration appointments by the American Arbitration Association (which provided one of the data sets) and the Federal Mediation and Conciliation Service (which provided another data set), which also could explain differences in elapsed time.

2 Such objections, usually raised at the outset of the hearing (advance notice of intention to raise an objection is usually provided), might take the form of allegations, for example, that the grievance procedure time limits have been violated, the matter at issue is being pursued in another forum, or the grievance lacks particulars. Arguments and evidence over such objections may necessitate additional hearing time and may add time to the arbitration award preparation.

3 Many articles have examined arbitrator predispositions. See, for example, Deitsch and Dilts (1989) and Thornton and Zirkel (1990).

4 It is extremely rare in Canada to have transcripts made of the hearing. Thus, this is a variable that will not be considered in this study. Readers should note, however, that the use of a formal transcript may well be more common in the United States.

5 We did not include the hearing itself as a separate stage because the vast majority of cases in our sample had hearings of one day's duration.

6 The financial issue takes on special importance for an employer in discharge cases because of the possibility that the grievant will be reinstated with back pay. The faster the case is heard, the lower the employer's potential financial loss.

7 In promotion, transfer, and layoff cases, one employee is typically pitted against another. For example, a grievant may claim that he or she should have won a promotion that was instead given to other worker. The longer the case takes to resolve, the longer the employee who received the promotion remains in the job, and the greater the human toll if the original decision is reversed by an arbitrator.

8 It is important to note that because cases are being resolved with the passage of time, the transition function is based on fewer and fewer cases as time passes. Therefore, the confidence level of the function decreases over time because the standard error of the transition rate in later intervals is based on fewer cases.

9 The curve for total delay is not merely the concatenation of the four component curves. First, because of differences in data availability, the component curves represent subsets of the total number of cases. More important, for each case, each stage has a different length. Accordingly, while the progress of a single case through all four stages is equivalent to their concatenation, the same is not true for all cases taken together.

10 The observed pattern with respect to preparation of the decision struck a responsive chord with one of the authors, an experienced arbitrator. He commented that if he does not get the award out fairly quickly, it tends to sit for a while. After several months elapse, the parties begin lobbying him for a decision, and their persistence, combined with his mounting guilt, ensures that the award is issued soon thereafter.

11 Our choice of model was based on a residual analysis. Residuals describe the difference between the observed values and the fitted values. Examination of residuals plots can turn up patterns that indicate whether the chosen model provides a good fit to the data. We chose the parametric model because it provided the best fit to the data based on these residual plots.

12 We thank an anonymous referee for alerting us to this issue.


Baucus, Melissa, and Janet Near. 1991. "Can Illegal Corporate Behavior Be Predicted? An Event History Analysis." Academy of Management Journal, Vol. 34, No. 1, pp. 9-36.

Bemmels, Brian. 1988. "Gender Effects in Discharge Arbitration." Industrial and Labor Relations Review, Vol. 42, No. 1, pp. 63-76.

Berkeley, Arthur. 1989. "The Most Serious Faults in Labor-Management Arbitration Today and What Can Be Done to Remedy Them." Labor Law Journal, Vol. 40, pp. 728-33.

Brett, Jeanne, and Stephen Goldberg. 1979. "Wildcat Strikes in Bituminous Coal Mining." Industrial and Labor Relations Review, Vol. 32, No. 4, pp. 465-83.

-----. 1983. Grievance Mediation in the Coal Industry: A Field Experiment." Industrial and Labor Relations Review, Vol. 37, No. 1, pp. 49-69.

Chamberlain, Neil, and James Kuhn. 1986. Collective Bargaining. 3rd ed. New York: McGraw Hill.

Cox, D. R. 1972. "Regression Models and Life-Tables." Journal of the Royal Statistical Society, Series B, Vol. 34, pp. 187-220.

Cox, D. R., and D. Oakes. 1984. Analysis of Survival Data. London: Chapman & Hall.

Deitsch, Clarence, and David Dilts. 1989. "An Analysis of Arbitrator Characteristics and Their Effects on Decision Making in Discharge Cases." Labor Law Journal, Vol. 40, No. 2, pp. 112-16.

Derber, Milton. 1979. "Management Organization for Collective Bargaining in the Public Sector." In B. Aaron, J. Grodin, and J. Stern, eds., Public Sector Bargaining. Madison, Wis.: Industrial Relations Research Association.

Federal Mediation and Conciliation Services. 1992. Forty-Fourth and Forty-Fifth Annual Reports. Washington, D.C.

Fichman, Mark. 1989. "Attendance Makes the Heart Grow Fonder: A Hazard Rate Approach to Modelling Attendance." Journal of Applied Psychology, Vol. 74, No. 2, pp. 325-35.

Fricke, John. 1976. An Empirical Study of the Grievance Arbitration Process in Alberta. Edmonton: Alberta Labour.

Goldblatt, Howard. 1974. Justice Delayed ... The Arbitration Process. Toronto: Labour Council of Metropolitan Toronto.

Harrison, David, and Charles Hulin. 1989. "Investigations of Absenteeism: Using Event History Models to Study the Absence-Taking Process." Journal of Applied Psychology, Vol. 74, No. 2, pp. 300-316.

Kiefer, Nicholas. 1988. "Economic Duration Data and Hazard Functions." Journal of Economic Literature, Vol. 26, No. 2, pp. 646-79.

Kochan, Thomas, and Harry Katz. 1988. Collective Bargaining and Industrial Relations. 2nd ed. Homewood, Ill.: Richard D. Irwin.

Lewin, David, and Richard Peterson. 1988. The Modern Grievance Procedure in the United States. New York: Quorum.

Levinthal, Daniel, and Mark Fichman. 1988. "Dynamics of Interorganizational Attachments: Auditor-Client Relationships." Administrative Science Quarterly, Vol. 33, No. 2, pp. 345-69.

Mills, Daniel. 1986. Labor-Management Relations. 3rd ed. New York: McGraw Hill.

Morita, J. G., T. W. Lee, and R. T. Mowday. 1989. "Introducing Survival Analysis to Organizational Researchers: A Selected Application to Turnover Research." Journal of Applied Psychology, Vol. 74, No. 2, pp. 280-92.

Olson, Corliss. 1990. "Time Delays in Grievance Arbitration in Alberta." MBA thesis, University of Calgary.

Ponak, Allen, and Corliss Olson. 1992. "Time Delays in Grievance Arbitration." Relations Industrielles, Vol. 47, No. 4, pp. 690-708.

Ponak, Allen, and Mark Thompson. 1995. "Public Sector Collective Bargaining." In Morley Gunderson and Allen Ponak, eds., Union-Management Relations in Canada. 3rd ed. Don Mills, Ont.: Addison-Wesley.

Prasow, Paul, and Edward Peters. 1983. Arbitration and Collective Bargaining. New York: McGraw Hill.

Rose, Joseph. 1986. "Statutory Expedited Grievance Arbitration: The Case of Ontario." Arbitration Journal, Vol. 41, No. 4, pp. 30-45.

-----. 1995. "The Evolution of Public Sector Unionism." In Gene Swimmer and Mark Thompson, eds., Public Sector Collective Bargaining in Canada. Kingston, Ont.: Queen's University Industrial Relations Centre.

Ross, Arthur M. 1958. "The Well-Aged Arbitration Case." Industrial and Labor Relations Review, Vol. 11, No. 2, pp. 262-71.

Schoonhoven. C. B., K. M. Eisenhardt, and K. Lyman. 1990. "Speeding Products to Market: Waiting Time to First Product Introduction in New Firms." Administrative Science Quarterly, Vol. 35, No. 2, pp. 177-207.

Slichter, Sumner, James Healey, and E. Robert Livernash. 1960. The Impact of Collective Bargaining on Management. Washington, D.C.: Brookings Institution.

Sloane, Arthur, and Fred Whitney. 1985. Labor Relations. 5th ed. Englewood Cliffs, N.J.: Prentice Hall.

Stieber, Jack, Richard Block, and Victor Nichol. 1990. "New Perspectives on Old Problems: Elapsed Time in Grievance Arbitration." Proceedings of the 43rd Annual Meeting of the National Academy of Arbitrators (San Diego, May 24-June 2, 1990), pp. 128-41.

Thornicroft, Kenneth. 1995. "Sources of Delay in Grievance Arbitration." Employee Responsibilities and Rights Journal, Vol. 8, No. 1, pp. 57-66.

Thornicroft, Kenneth, and Genievieve Eden. 1995. "Grievances and Their Resolution." In Morley Gunderson and Allen Ponak, eds., Union-Management Relations in Canada. 3rd ed. Don Mills, Ont.: Addison-Wesley.

Thornton, Robert, and Perry Zirkel. 1990. "The Consistency and Predictability of Grievance Arbitration Awards." Industrial and Labor Relations Review, Vol. 43, No. 2, pp. 294-307.

Usery, William. 1973. Daily Labor Report No. 233 (December 4). Washington, D.C.: Bureau of National Affairs, p. A13. (Cited in Daniel Mills, Labor-Management Relations, p. 293.)

Winter, Catherine. 1983. Grievance Arbitration Costs and Time, 1980. Toronto: Research Branch, Ontario Ministry of Labour.

Allen Ponak is Professor of Industrial Relations and Wilfred Zerbe is Associate Professor and Chair of Management of Organizations and Human Resources, both at the Faculty of Management, University of Calgary. Sarah Rose is Professor of Biostatistics, Department of Community Health Sciences, Faculty of Medicine, University of Calgary. Corliss Olson is a doctoral candidate, Industrial Relations Research Institute, University of Wisconsin-Madison. The authors thank the Industrial Relations Research Group (University of Calgary) and the Alberta Labour Relations Board for financial assistance, and Hoyt Wheeler for helpful comments on an earlier draft.
COPYRIGHT 1996 Sage Publications, Inc.
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 1996 Gale, Cengage Learning. All rights reserved.

Article Details
Printer friendly Cite/link Email Feedback
Author:Ponak, Allen; Zerbe, Wilfred; Rose, Sarah; Olson, Corliss
Publication:ILR Review
Date:Oct 1, 1996
Previous Article:Do injured workers pay for reasonable accommodation?
Next Article:The effect of government-mandated benefits on youth employment.

Terms of use | Privacy policy | Copyright © 2022 Farlex, Inc. | Feedback | For webmasters |