Monday, February 6, 2023

Research: Grievance Arbitration Project

Back in November, I posted a bit about a research project I’m involved with that is looking at grievance-arbitration decisions in Alberta. As of this morning, we’ve coded about 441 decisions by arbitrators (spanning 2006 to 2011) and I’m in a position to talk a bit more about the project and the themes we’re seeing.

When a union and an employer are unable to resolve a disagreement about how the employer has applied the collective agreement (or law, policy, or past practice), they can remit the dispute to an arbitrator for a decision. Arbitration is a form of adjudication akin to the courts. One of the main differences between arbitration and the courts is that, in arbitration, the two parties normally jointly select the arbitrator who will hear the matter (although, sometimes other appointment processes are used).

During the selection process, it is common for a side to internally discuss the merits of proposing or accepting particular arbitrators to hear a matter. This discussion appears premised on the assumption that arbitrators can be (un)sympathetic to certain arguments, evidence, types of grievances, and kinds of grievers. This belief is consistent with a constructivist view of the world, wherein there is infinite stimulus and what one pays attention to and how one interprets that stimulus is driven, in part, by one’s thoughts, beliefs, and expectations.

If the “who you get affects what you get” hypothesis is true, it suggests that identifying patterns in arbitrators’ decision-making can be used to increase the odds of success. This hypothesis (and, if true, the efficacy of various strategies that lever it) is part of what we’ll be examining once we’ve finished coding the dataset (ideally by Christmas 2023, but who knows).

In coding the dataset, we’re assigning the outcome of decisions one of three codes: union win, employer win, or mixed decision. An example of a mixed decision might be a termination grievance. The employer might seek to have the termination upheld, the union might seek to have it overturned and the worker reinstated without penalty, and the arbitrator may eventually decide there was grounds to discipline the worker, but that termination was unreasonable in the circumstances and then substitute some lesser penalty (e.g., a short suspension).

This coding allows us to visually (and statistically, I suppose) represent arbitral decisions like so. Yellow are union wins, green are mixed results, and blue are employer wins.

Note that, in this representation, both the union win and the mixed outcome category result in the worker being better off than they were before the decision. This suggests that looking at the “employer win” category (blue) is a useful way to get a quick and dirty sense of decision patterns.

The graphic above summarizes all decisions. The literature suggests that different types of disputes (e.g., discipline and termination grievances, salary and benefits grievances, grievances addressing seniority, selection, promotion and layoff) will have different win-loss patterns. I have teased apart the data that we have along these lines in the graphic below. Sorry the images are a bit har to read, the lines (top to bottom) are grievances addressing seniority, selection, promotion and layoff, salary and benefits grievances, discipline and termination grievances, and the overall average.

We do seem to see some interesting differences. Note that, in the discipline and termination decisions, the employer typically bears the onus (at least initially) or proving discipline was warranted. In most others kind of grievances, the union bears the onus of proving the grievance should be upheld.

If the “who you get affects what you gets” hypothesis is correct, we should see differences among the decision patterns of different arbitrators. I have presented below a randomly drawn selection of the early data in this regard (carefully anonymized) with the overall average at the top.

What this suggests is that there appear to be large differences in decision outcomes among arbitrators. Two important caveats are worth keeping in mind. This first is that the number of cases in the dataset to date for each arbitrator varies and is, overall, small. Small samples tend to yield swingy numbers, so we shouldn’t jump to conclusions based on a small sample. These differences may attenuate over time as we add cases (although we’re not seeing that yet in the data)

The second is that the facts of each case almost certainly impact the decision of the arbitrator. Our expectation is that, over many cases, differences between cases should attenuate (i.e., wash out) these case-specific differences. Together, these caveats also suggest that eliminating arbitrators with relatively few recorded decisions from the final dataset is likely appropriate.

When we look at arbitrator records on discipline and termination cases (which seem to be the largest single category of cases), we see similarly large differences among arbitrators. I have not visually presented that data, given the small number of cases for each arbitrator.

-- Bob Barnetson

No comments: