Housing Access and Fertility: A Survey Experiment on Housing Timing and Housing Attributes in Brazil

2026-09-11

2026-09-25

(1) Weights on housing attributes (space, neighborhood, tenure stability, ownership, monthly income) recovered from pairwise scenarios, for two responses: whether the couple is more likely to have a first child within three years and whether it has more children over the lifetime.
(2) Expected number of children of the same couple housed now versus only at age 45, in total and by age, and the direct comparison between the two timings.
(3) The premium a house carries over an equal, illiquid windfall, for the same two responses.

Scenario responses are five-point scales from “much more for Option 1” to “much more for Option 2”, with a “não sei” option, recoded from −2 to +2 toward a named attribute (toward the home in the windfall scenario) using the recorded side assignment; “não sei” is dropped. The weights in (1) are the attribute coefficients from the stacked responses, with income as the reference attribute, estimated separately for each of the two responses and pooled over doses and sides. In (2), respondents fill a grid of the couple’s total number of children by age (0 to 5, “5 or more” coded 5, “não sei” dropped); the estimands are the within-respondent difference between the two timings, in total and at each age, the share judging the couple to have more children if housed now, and the difference in the lifetime effect between the age-25 and age-35 arms.

Dose–response in the voucher amount and in the lease length; cash versus voucher; ownership versus a permanent protected lease; the module in which the couple’s ideal number of children is fixed at three; the stated ranking of the attributes; respondents’ own reports of whether housing changed the timing or the number of their children; responses by sub-group, including the sampling segments (consórcio members, other and intending buyers, non-buyers), age, sex, parity and income, and descriptive comparisons of these groups with each other and with population benchmarks. All exploratory.

Online survey (Qualtrics, Brazilian Portuguese) with quota sampling on housing-purchase status. After screening and basic demographics, 15 percent of respondents read a note that research finds housing important for family decisions, 15 percent that it is a minor factor, and 70 percent read nothing. All respondents then see five of ten pairwise scenarios, drawn with even presentation, in which a childless couple aged 30 faces two housing options that differ in exactly one attribute, everything else equal: space, neighborhood, tenure stability (a guaranteed lease of 5 or 10 years), ownership or monthly income (a food voucher of R$500, R$1,000 or R$2,000); one slot shows ownership versus a permanent protected lease with probability one third. The side of each option, the voucher amount and the lease length are randomized at the respondent level; half of the respondents answer an early comprehension check in each scenario; half of those who see ownership versus income also answer a cash version. Respondents are then assigned with equal probability to one of three modules: (1) housing timing, in which a recently married, childless couple aged 25 or 35 receives a paid-off program apartment now or, in a second scenario shown in random order, only at age 45; (2) the same timing scenarios with the couple’s ideal number of children fixed at three; (3) a couple aged 30 wins R$300,000 that either must buy a home or is invested and locked for ten years, with a cash follow-up. Half of the sample also answers a stated ranking of the attributes. The survey ends with the respondent’s own housing and fertility history and demographics; other blocks are exploratory.

Not available

Random draws by the survey software (Qualtrics) at the start of the survey flow, stored in the data export.

Individual respondent.

No

1,500 individuals.

1,500 completes: 350 lottery-selected consórcio buyers, 100 bid-selected buyers, 350 members still paying in, 325 other buyers, 325 intending buyers and 50 non-buyers.

About 750 respondents per pairwise scenario (500 for ownership versus stability, 250 for ownership versus the permanent lease); about 500 per module, and 250 per couple age within modules 1 and 2; framing note 225, 225 and 1,050; early comprehension check 750 and 750.

0.15 scale points for the mean response to one pairwise scenario (about 750 respondents, response standard deviation of about 1.5 on the −2 to +2 scale, 80 percent power, 5 percent two-sided).