“Acting Wife” or “Acting Female”? Marriage Market Signaling Across Religious Boundaries

Participants attend a one-hour career discussion workshop in four-person groups. The intervention is the randomly assigned composition of the group: whether a woman's three peers are all female, include male peers of her own religion, or include male peers of a different religion. All other elements of the workshop — the ice-breaking activity, the career questionnaire, the facilitated career discussion, and the post-discussion survey — are identical across groups. Participants are informed that group composition is randomized but are not told that it varies by gender and religion.

2027-02-01

2027-02-28

Two indices of expressed career ambition, constructed from binary job choices made during the workshop: (i) a two-item index based on a salary–hours trade-off and a promotion–travel trade-off, adapted from Bursztyn, Fujiwara and Pallais (2017); and (ii) an extended three-item index that additionally includes a salary–commute trade-off.

For each of the three ambition items, the respondent's choice is coded one if she selects the more ambitious option and zero otherwise. Each item is then standardized relative to the mean and standard deviation among women assigned to the female-only reference condition. The two-item index is the average of the standardized salary–hours and promotion–travel items; the extended index averages all three standardized items. Both are analysed as primary outcomes, adjusting for a privately elicited baseline career-ambition measure using ANCOVA.

(i) Differences between privately expressed baseline career ambition and ambition expressed during the workshop; (ii) the accuracy of women's beliefs about men's preferences over a future partner's career ambition; (iii) heterogeneity in the primary outcomes by a privately elicited pre-treatment belief about the marriage-market penalty to female career focus; (iv) two placebo job choices; and (v) the corresponding audience effects among male participants.

The private ambition benchmark is the share of six baseline career-ambition statements the respondent endorses, on the same 0–1 scale as the binary workshop choices; the secondary outcome is the difference between the workshop choice average and this benchmark. Belief accuracy compares each woman's privately stated belief about men's partner-ambition preferences with the mean preference men reported privately at baseline, reporting both the signed gap and absolute prediction error. The marriage-market-penalty belief is a five-point agreement item, standardized within the female sample and interacted with treatment assignment. The two placebo choices involve social impact versus coworker interaction, and salary versus subsidized meals.

The study is a lab-in-the-field experiment with 440 third and final-year undergraduates at the Faculty of Economics and Business of a large public university in Indonesia, drawn equally from four demographic groups: Muslim men, Muslim women, Christian men and Christian women. Participants first complete a private online baseline survey. They are then invited to a career discussion workshop held over five rounds on separate days within a single week, each round comprising 88 participants in 22 four-person groups.

There are six group types: all-female, all-male, same-religion mixed-gender, and different-religion mixed-gender, with the number of men in mixed groups varying from one to three. The same 22 configurations are implemented in every round. For women, the all-female groups are the reference condition; the contrast between same-religion and different-religion male peers isolates marriage-market signaling from a general male-audience effect, exploiting Indonesia's near-absolute religious endogamy. Men serve as a placebo group.

The study is funded for five workshop rounds. Should additional funding be confirmed before data collection begins, the same 22 group configurations would be implemented in seven or eight rounds, scaling all cell sizes proportionally. The design, hypotheses and analysis plan are unchanged in that case, and the registration would be updated before data collection.

Not available

Randomization done in office by a computer. Within each workshop round, a computer program randomly permutes participants within each of the four demographic groups (Muslim men, Muslim women, Christian men, Christian women) and matches them to the demographic slots defined by the 22 pre-specified group configurations. The four permutations are executed together as a single allocation for the round.

Individuals are randomly assigned to four-person discussion groups. The group's composition defines the treatment condition, so the discussion group is the unit of treatment and of clustering. Randomization is blocked by workshop round and constrained by the demographic composition required by the 22 group configurations.

Yes

110 four-person discussion groups (22 groups in each of five workshop rounds). If additional funding is confirmed, 154 groups (seven rounds) or 176 groups (eight rounds).

440 undergraduate students (110 Muslim men, 110 Muslim women, 110 Christian men, 110 Christian women). If additional funding is confirmed, 616 students (seven rounds) or 704 students (eight rounds).

Women (220 total): 25 all-female groups, 100 women (reference condition); 30 same-religion mixed-gender groups, 60 women; 30 different-religion mixed-gender groups, 60 women.

Men (220 total, placebo group): 25 all-male groups, 100 men (reference condition); 60 men in same-religion mixed-gender groups; 60 men in different-religion mixed-gender groups.

At 80% power and a 5% two-sided test, without covariate adjustment: 0.46 standard deviations of the career-ambition index for the comparison of 60 women with same-religion male peers against 100 women in all-female groups, the same 0.46 SD for the corresponding different-religion comparison, and 0.51 SD for the direct comparison of the two 60-woman male-peer conditions. Adjusting for privately elicited baseline ambition via ANCOVA, and assuming a baseline–endline correlation of 0.4 to 0.5, these fall to 0.42–0.40 SD and 0.47–0.44 SD respectively. Under the contingent expanded design, the unadjusted MDEs scale by the square root of 5/R in the number of rounds R, falling to 0.39 SD and 0.43 SD at seven rounds and 0.36 SD and 0.40 SD at eight. For comparison, Bursztyn, Fujiwara and Pallais (2017) report effects of 0.56 SD and 0.77 SD.