I was on strike a couple of weeks ago, for the span of three days. Not just me, of course; the faculty at l'Université Laval is unionised.

The whole notion of faculty unions seems a bit odd to me, and it's not just the incongruous image of seeing tenured professors whose salaries put them in the top 5% of the income distribution walking picket lines. When it comes to the day-to-day functioning of the university, professors are the bosses: we decide what gets taught, we decide who teaches it, and we decide what research gets done. And the decision-making process behind those decisions is invariably consensual: more than anything else, the functioning of a typical department resembles that of an anarcho-syndicalist commune.

So any advantage professors might see in unionising must come from the perceived rents they can obtain from collective bargaining. But where would those rents come from? Arthur Hosios and Aloysius Siow address this question in their CJE article "Unions without rents: The curious economics of faculty unions". It turns out that these rents are extracted from high-quality scholars and redistributed to their lower-quality colleagues. Here's the abstract from the ungated version:

Since the faculty is the de facto residual claimant of the university, any earnings gain must come from an increase in university revenue. A model is developed to explain the fact that faculty unions have a negligible positive (and oftentimes negative) effect on average faculty wages. A union that aims to promote the interests of the median faculty member negotiates wage redistributions and more onerous working conditions (e.g., greater teaching loads), despite the negative impact these changes have on research output. Union formation may sometimes appear to raise average wages, but only because the benchmark non-unionized university fails to maximize average wages when scholars derive non-pecuniary benefits from research and teaching. Using Canadian data, we estimate that there is no gain in faculty earnings or university revenue after a public university unionizes. Nevertheless, there is evidence for earnings redistribution within a unionized faculty, from high- to low-paying academic disciplines and from assistant to full professors, and for a decrease in research output.

And here are a few paragraphs from the introduction outlining their argument:

Public universities and the overwhelming majority of private universities are operated as non-profit organizations. Setting aside for-profit universities, there are no shareholders to retain university earnings; after accounting for fixed costs and support staff, university revenue is distributed to faculty. As a first approximation, then, the faculty itself is the residual claimant of the university. From this perspective, a university is effectively a partnership that sells the research and teaching services of its members, and divides the resulting revenue among the faculty. Of course, once we reject the view of a university as a profit-maximizing firm, and of faculty members as simply another input to production, we necessarily reject the conventional view of a faculty union as an indifference map confronting a demand curve or iso-profit line.

In this paper, we develop a model of union formation in higher education. We assume that non-unionized universities determine compensation rules and workload to maximize the welfare of the representative faculty member, whereas unionized shops aim to maximize the welfare of the median faculty voter. In this situation, there is no gain or rent for the faculty of a given university from unionization. The view that rents are absent for individual non-profit universities, whether public or private, does not deny that faculty at several public universities and colleges in a given jurisdiction may choose to organize together to enhance their bargaining position vis-à-vis government. Indeed, this is why many public-sector workers (including grade school and high-school teachers) organize. Of course, acknowledging that there is a gain from collusive behavior does not explain why the faculties at individual institutions choose to unionize.

The overall absence of rents appears to be consistent with the data. While the union wage advantage across all industries in North America ranges from 10 to 20 percent, the situation in higher education is very different. Recent estimates of the union wage gain in Canada and the US range from marginally positive to strictly negative. However, these estimates are based on samples that include public institutions in statewide systems that bargain collectively, and these systems include large numbers of junior colleges and comprehensive universities. Many of the larger reported wage gains are specifically associated with junior colleges and comprehensive universities.

The average faculty wage gain following unionization, net of union dues, appears quite modest, and yet we observe that some faculties choose to become unionized  The obvious question is, why? … [W]e describe a model of a university, with a heterogeneous faculty, in which relatively low-ability scholars may benefit from union formation if compensation schemes are subsequently negotiated that redistribute income away from their high-ability colleagues. Ability is used here as a proxy for a scholar’s marginal value product, reflecting the combined effects of effort, productivity and discipline-specific relative output prices. Though it may be difficult to directly test whether redistribution is the primary aim of those in favor of unionization, as claimed here, it is clear that redistribution is a consequence. There is ample U.S. evidence that full professors gain more than other ranks following unionization, and that unionization is associated with less wage dispersion across fields…

A union redistributes wage income from high- to low-ability scholars in our model by imposing compensation rules that are less sensitive to performance than those at non-unionized universities. This is consistent with evidence that research output (publications) has a smaller impact on wages at universities with faculty unions… In turn, the resulting attenuation of research incentives has two effects: first, research output per scholar falls following unionization…; and second, per capita university income falls to the extent that university income is sensitive to research output. Since per capita university income is the average faculty wage in our model of an income-sharing partnership, union formation is expected to lower average wages. Some, but not all, U.S. studies report a negative union wage gain.

Our model thus predicts that faculty unions negotiate compensation schemes that discourage research and, as a result, that average faculty wages will be reduced as research output falls. While the first prediction seems to be robust, the second depends on two simplifying assumptions: first, the only income-generating activity in which scholars engage is research and, second, the non-pecuniary benefits of research are nil. In a multi-tasking ve
rsion of our model, in which scholars engage in research and teaching and enjoy non-pecuniary benefits from both, union formation can cause average wages to increase. When unions negotiate compensation and working conditions, including teaching loads, unions again promote less performance pay, which discourages research, but they also favor heavy teacher loads that enhance university income. The key insight her  is that non-unionized universities choose relatively light teaching loads for their faculty, below the level that maximizes per capita income, because their faculties receive nonpecuniary benefits from research, and teaching raises the marginal disutility of supplying research effort. Recognizing that non-unionized universities fail to maximize per capita income, the alternative compensation-teaching package negotiated by a faculty union can lead to a wage increase.

Union formation at Canadian universities is shown to have a very small positive effect on average faculty wages, but no effect on university income. We find evidence of income redistribution towards senior faculty, and of less wage dispersion across disciplines, following union formation. Information on the rank-discipline composition of union executive boards is presented that is consistent with the underlying voting model developed elsewhere in the paper. Finally, we show that research output, but not quality, decreases when the faculty is unionized. We find no evidence in our data that unions are able to raise wages by forcing a reallocation of non-wage portions of the university budget to faculty. It is possible that a faculty union causes universities to reduce their use of part-time labor. For the same total number of students, fewer part-time instructors causes full-time teaching loads and wage incomes to rise. Unfortunately, our data describe full-time faculty alone.

We extend our model … to explain why universities that participate in more competitive markets for faculty are less likely to be unionized. As individual income is positively related to performance pay, the re-distributive gains for low-ability scholars from forming a union and reducing performance pay are, in part, offset by the potential departure of more-able colleagues. If a university’s more-able faculty members are mobile, a faculty union will opt for more performance pay than otherwise. Competition from other universities for able scholars thus reduces the net benefit of forming a union, which, in part, explains why research universities (which are relatively well-endowed with more-able researchers) are less likely to unionized.