You run a lab in Rome where students come to play for money. Before each session you tick two things on the screen they will see: what they must do to compete, and what they can win. Then you press Run, watch the room fill, and read the two bars that come out, one for the men and one for the women. Eight sessions, then you write back.
Find out, in eight lab sessions, whether women are less competitive than men, and write back.
Find out, in eight lab sessions, whether women are less competitive than men, and write back.
Each session: 24 students, in groups of four, two men and two women. The students never change between sessions. Only what you put on their screen does.
Six ways to set the screen, eight sessions. One session of 24 is noisy; repeat the ones you want to believe.
Tick both boxes above; the screen fills in as you do.
Σ sums · € price · the boxes are the prizes · ■ men ■ women, how hard they went for the top
Your eight sessions are in the notebook below. Write back one sentence. The notebook will be read against it.
You wrote back:
Three economists ran the price apparatus in Rome: 168 students, a flat firm and a steep one, 32 rounds each. The women bid as high as the men; for the top of the flat firm a few points higher. The women got the top job more often than their share. Nobody earned more for being at the top: the middle jobs paid better. The famous gap comes from labs where competing means a race at sums; this one measures something else, the willingness to pay for rank, and finds no gap in it.
Their own sentence, near enough: neither way of measuring is the better one; they measure different things, and what you infer about women and competition depends on which you chose.
Find out, in eight lab sessions, whether women are less competitive than men, and write back.
The lab. Each session, 24 students come in and sit in groups of four: two men, two women. They play once for the lab's money, ECU (10 ECU is a euro). The students are the same kind of people every session. Only their screen changes, and you write it.
Your two choices. In the Apparatus window, tick what the students must do to compete: solve sums (whoever solves the most of their four wins; a student may also sit out for a sure amount) or name a price (each student writes the most they would pay for a job; the highest offer gets it and pays the second-highest). Then tick what they can win: one winner (350, the rest 50), a flat ladder (350, 200, 200, 50) or a steep ladder (450, 150, 150, 50). Both boxes start empty every session. The window below, What the students will see, fills in as you tick, so you can read their screen before you run it.
Run. Press Run session. The room fills, the clock runs, and a Result window opens with two bars: how hard the men went for the top, and how hard the women did. Under sums, the bar is the share who chose to compete. Under price, it is what they offered for the top job, out of every 100 it is worth. Then: who got the six top jobs, what the two groups took home, and a line on what this apparatus actually measured. When the apparatus is the one the paper ran, the paper's own result is shown in grey underneath. Press Write it in the notebook to keep the row.
The notebook. One row per session: Σ for sums, € for price, the prize boxes, the two bars small, and the result in words. There are six ways to set the screen and eight sessions. One session of 24 is noisy; a result worth believing is one you have seen twice, and when you run an apparatus again the result window says what the sessions say together.
Write back. After the eighth session, pick one sentence for the journalist. The game reads your notebook against it and says whether your own sessions back what you sent, then shows what the paper found.
The paper. Andrej Angelovski, Jordi Brandts and Werner Güth (2026), Bidding for better jobs: an experiment on gender differences in competitiveness without a real-effort task, Theory and Decision, doi:10.1007/s11238-026-10132-9. Open access. Game built from this paper; the paper is the authority where the two disagree.
Most lab evidence that women compete less comes from one kind of experiment: a few minutes of a task, winner takes all. The authors ask whether that is competing as a career knows it, where getting ahead means spending money, time and effort to secure a rank. So they made competing a sealed bid: 168 students in Rome bid, 32 times each, for a top job and two middle jobs in a flat firm (350 / 200 / 200, default 50) and a steep one (450 / 150 / 150, default 50), first inside groups of four and then across eight. The bids of women and men did not differ significantly; women bid a few points higher for the top of the flat firm. Women ended up in the top job more often than their share. Nobody earned more for being at the top: the middle jobs paid better. Their conclusion is not that women are more competitive, but that what you infer about gender and competition depends on how you measure competing, and that neither way is the right one.
You are the experimenter. The paper's whole point is a design decision its authors made, so the game hands you that decision as a switch: what competing means on the students' screen (a race at sums, or a price for the job) and what they can win (one winner, a flat ladder, a steep ladder). Eight sessions of 24 students; two bars each; a notebook; a sentence back to the journalist. The students never change between sessions, so whatever moves in the bars was moved by the apparatus.
The price apparatus is reconstructed from the paper. Each student's bid for a rung is the rational benchmark (the salary less the 50 everyone gets anyway) plus a deviation drawn to match the paper's Figure 4 and Table 1 (bids for the top about 40 ECU under the benchmark, women about 14 higher in the flat firm; bids for the middle near the benchmark), with the bunching at the benchmark the paper's appendix describes. The paper's own allocation rules then run: the top job to the highest bidder at the second-highest bid, the middle jobs to the next two at the third's. The reconstruction reproduces, to within the tolerances in the calibration file, the paper's bid deviations, the share of top jobs won by women, the earnings by position and the absence of a gender difference in earnings.
The sums apparatus is reconstructed from the literature the paper argues with, as the paper's introduction and section 2 describe it: Niederle and Vesterlund's 2007 lab, where about three in four men and one in three women chose the tournament, men and women performed alike, and the gap went with beliefs about one's own rank and attitudes to risk. Each student here has an ability, a private view of their standing (men's more flattering), an aversion to a gamble (women's stronger), and a taste for the top that does not differ by gender, which is the paper's finding. A student competes when the prize they expect under their own view, less their aversion times the spread, beats the sure alternative. None of these numbers is the paper's data.
Two cells are ours alone. The paper never ran a sums contest with its prize ladders; those cells cross the two switches so the player can test either on its own. In this model a ladder barely moves the gap, because a view of one's own rank still decides who competes. And the one-job price cell is a second-price auction for a single job, with no gender difference, as the auction literature the paper cites reports for that rule.
Mechanism kind: reconstructed (price apparatus, from the paper); reconstructed from the paper's literature (sums apparatus); toy (the two crossed cells and the one-job price cell). Every number is in the config block at the top of the mechanism, tagged with its source, and the calibration check is in the repository.
Fonts: Liberation Sans (Red Hat, SIL Open Font License), the free Arial, because that is what the paper's screens were set in. A beta, built from the paper. Builder's review only; the author finishes it.