Commensuration

Yale Law School ranks 1st. The University of New Mexico ranks 99th. One runs nine law journals, the other trains more Native American lawyers than anywhere else in the country. Both are remarkable, but on the list every difference between them collapses into a single number: 98. The concept is commensuration: pressing things of different natures, things that cannot be directly compared, into numbers on a single ruler. The act of "pressing into one number" is itself power: it does not describe the world, it decides what counts in the world and what becomes invisible from then on. The reactivity machine you met in R01 has this as its first part: you need a shared ruler before there is a rank that can be gamed.

One story lights up this layer of power better than any other, and it comes from Espeland's own fieldwork. The U.S. Bureau of Reclamation wanted to build the Orme Dam at the confluence of two rivers, which would flood the ancestral sacred land of the Yavapai. The technocrats produced a cost-benefit analysis, trying to put the sacred site and "another piece of land plus a compensation payment" on the same ruler. The Yavapai refused to price the sacred site. This was not a bargaining posture: certain relationships are constituted precisely by the act of refusing to put a price on them, and the philosopher Raz named this constitutive incommensurability. You do not price your own child, not because the price would be too high, but because not pricing is part of what that relationship is. Cost-benefit analysis putting a price on "what a statistical life is worth" is the same logic wearing a different face. Wherever this logic spreads, refusing to put a price on something is the only thing left that can stop it. So declaring something incommensurable is not irrational stubbornness in this framework, it is a legitimate political act that marks a moral boundary. Conversely, commensuration is not always oppressive: the excluded fighting to "be counted" (getting their losses into the compensation formula, getting their group into the statistics) is an equally common political demand. The same equivalence machine can flatten differences or hand out admission tickets, depending on who holds it and against whom it runs.

The paper that turned that story into theory is Espeland and Stevens's 1998 programmatic article in the Annual Review of Sociology. They named the counterintuitive part: commensuration is not the discovery of similarities that already exist between things, it actively creates comparability. It first assumes that things which were not comparable belong to the same kind, then places precise quantified intervals between them. Their phrasing: "Commensuration transforms qualities into quantities, difference into magnitude." They went further and characterised it as a mode of power, because it decides what is relevant to what and what becomes invisible. Prices, votes, cost-benefit ratios, standardised test scores and rankings are all concrete forms of the same process. The paper has been cited nearly three thousand times: when three thousand studies pick up a concept, what it hit is not one industry's quirk but a general-purpose part of modern life. The paper's real ambition was to establish these forms as a single independent object of sociological study, and to give four sets of variables that must each be asked about: motives, technical form, effects, resistance. Those four questions are the skeleton of the module below, and they work directly as your own design checklist.

Back to the ranking scene in R01. In the 2007 ranking study, commensuration takes three concrete forms. First, simplification: the entire complexity of a law school is pressed into one number, which renders a great deal of hard-to-quantify information "irrelevant." 173 deans once co-signed a letter warning applicants that "these ranking systems are inherently flawed," and the letter changed not a single position. Nearly nine in ten deans said together that the ruler is broken, and the ruler went on working: its power does not reside in the assent of the measured. Second, it unifies and differentiates at once: it first anchors all law schools as one kind, then uses precise ranks to place each school in a unique relation of above and below to every other, and it also turns "this year's self" and "past years' selves" into a frame-by-frame photo series. In the paper's words, only Yale can still credibly claim to be "the best." Third, it provokes reflection about the relation between numbers and reality, and different people take completely different attitudes to the same ranking. Applicants who know nothing read the ranking as objective quality and ignore its method; the administrators forced to manufacture the numbers themselves are the most likely to become sceptics. One interviewee's exact words: what we are forced to do is shameful, we are a Catholic school, we teach honesty, and yet we have to count non-legal jobs as employment like everyone else. The closer you are to the ruler, the less you believe it.

Commensuration is the cognitive infrastructure of reactivity. For reactivity to happen, the evaluated must first be able to see their own relative position on a shared ruler: without commensuration there is no number to game and no rank to chase. The division of labour in the 2007 paper is that commensuration changes cognition and the self-fulfilling prophecy changes expectations, and only the two stacked together produce the machine in R01. Dig one layer further and classification comes before commensuration. You saw in R15 how the nineteenth-century statistical wave invented the "normal person." Bowker and Star's Sorting Things Out (1999) adds the reminder that behind any indicator stands a naturalised classification infrastructure: classification has to cut a heterogeneous world into comparable units before commensuration can lay scale marks between the units. A mature classification system gets black-boxed, and people forget how much political and ethical struggle it once involved. Espeland's 2013 "Commensuration and Cognition" connects this layer to cultural cognitive sociology: commensuration is fundamentally a reorganisation of classification and attention, and it happens before anyone consciously weighs anything, which makes it more efficient than persuasion. You do not need to win the argument about the value of what New Mexico does, you only need to publish a table showing it 99th. In 2026 Smidt simply called commensuration the "equivalence machine" of liberal modernity, reopening the normative argument about which things should be protected from the ruler. This line scaled up to the state is R05 Legibility, and scaled down to individual identity is R06 Looping effects.

The concept later grew several corrections, each drawing a boundary around the sweeping claim that "all indicators are dangerous." First, Nelken in 2021 used pandemic-era country rankings to argue for distinguishing "comparison" from "commensuration": many so-called indicators are only loose weak comparisons and do not meet the strong condition of "sharing one cardinal metric." If that holds, reactivity may be fiercest only under genuine strong commensuration. Second, Van Bommel and colleagues in 2023 gave empirical evidence in sustainability reporting for commensuration crowding out morality: once values are converted into scores, what people discuss shifts from "should we" to "how many points." Third, Gerdin and Englund in 2019 turned the camera on the public responses of the evaluated, cataloguing the discursive tactics academics use to contest commensuration: denying comparability, restating qualitative value, exposing methodological flaws. Commensuration is not a one-way accomplished fact, it is a continuously contested process, and the later accounting literature simply calls it commensuration work, a form of labour that has to be continuously invested. Fourth, Berman and Hirschman's 2018 general critique applies here too: the word quantification covers indicators, rankings, algorithms and big data, far too many heterogeneous phenomena, and an over-broad concept is itself a disease of this field.

One open question: classical commensuration assumes people can read the ruler. A deep model's score is an unreadable black box, so might this produce a new "black-box commensuration," where the evaluated do not even get the chance to reverse-engineer and can only comply blindly? The D1 report lists this as the open question on this line.

The one-line takeaway: pressing incomparable things into one number is not mainly an act of measurement, it is a decision about what becomes invisible.