Informality and Motherhood Penalties over the Life Cycle

Karolina Andrade, Diogo Britto, Lucas Finamor, Alexandre A. Fonseca, Fernando Mattar, Breno Sampaio and Gabriel Ulyssea — “Informality and Motherhood Penalties over the Life Cycle,” presented by Finamor at the NBER Summer Institute, Gender in the Economy, on 29 July 2026. No discussant; the room asked questions throughout. Source note: no public draft exists yet, so this is written from the recording and the conference slide deck. Every number below is one the speaker showed or said.

Here is a thing that is true about the Brazilian formal labor market and that turns out to explain a great deal.

If you are a Brazilian woman with at most a high school education, and you work in the formal sector — registered contract, social security contributions, unemployment insurance, maternity leave, the whole protective apparatus — the probability that you work part time is 2.5 percent. Not “low.” Effectively zero. The formal sector does not sell that product.

If you work informally, it is 24.3 percent. If you are self-employed, 38.7 percent. And that is with part time generously defined as anything under thirty hours.

Same story on the other flexibility margins. About half of self-employed women work from home. Commuting times are far lower in both non-formal sectors, even conditioning on commuting at all.

So the formal job in Brazil is a bundle, and it is sold as a bundle. You cannot buy the social security without the rigid schedule. And what a woman needs after a birth is on the other side of that line.

The identity problem, solved by four datasets

To measure any of this you need to know who is whose mother. Brazil does not hand you that. The authors build it out of four administrative sources in sequence: the civil registry (child’s ID and mother’s name), tax returns from 2006 to 2019 (both IDs, but only for the top of the income distribution, since most Brazilians don’t file), the Cadastro Único welfare registry (which covers the bottom, and carries both parents’ names), and the school census from 2008 to 2017 (names but no IDs). Mothers and fathers are matched separately, and each birth takes whichever link is available.

The result is roughly 40 million births matched to parents, joined to RAIS — the matched employer-employee register covering the universe of formal contracts — and to the quarterly PNAD household survey. The presenter’s line, which he attributed to a joke with colleagues the previous day, is that this is Scandinavian data in a developing country. The joke undersells it slightly: no Scandinavian register has to tell you whether the job exists.

Forty percent of the penalty happens before the birth

The design is a difference-in-differences with matching: treated parents had their first child in 2012 or 2013, controls had theirs four years later in 2016 or 2017, matched on gender, birth cohort, municipality, education and prior-year formal-sector and welfare-registry status. Sample restricted to at most high school. Outcome: probability of formal employment. And — this is the part that does the work — everything is at monthly resolution.

At monthly resolution something appears that annual data cannot see. The mothers’ line does not break at the birth. It breaks about nine months earlier.

Roughly forty percent of the total motherhood penalty has already materialized by the time the child arrives. Which has an immediate methodological implication the speaker was happy to state: if you are estimating child penalties on annual data with year −1 as the reference, you are normalizing on a year that already contains a chunk of the treatment. Use −2.

Split the sample by formality status twelve months before the birth and the pre-birth effect resolves into two completely different stories. Women who were formally employed show no divergence at all during pregnancy — and then, four dots after the birth, they fall off a cliff. Four dots is four months, which is the length of Brazilian maternity leave (job protection runs to six, which is why the drop isn’t perfectly sharp at four). Women who were not formally employed show the opposite: control women keep flowing into formal jobs month after month, and the pregnant women’s line goes completely flat. The pipeline into formal employment closes the moment they conceive.

Asked what the flat line means — discrimination, search effort, morning sickness, anticipation of discrimination — the presenter declined to pick, repeatedly and I think correctly. “It’s all mixing together here.” The data cannot separate them, the model lumps them into one parameter, and someone should write the paper that pulls them apart.

(A nice detail surfaced in the Q&A: someone guessed there must be a twelve-month tenure requirement to qualify for maternity leave. There isn’t. Employed in the last month of pregnancy and you are entitled, which — as the speaker observed drily — is precisely why getting hired while visibly pregnant is difficult.)

The model, and the thing that turns out not to matter

The structural piece is a life-cycle model in semesters. Each period a woman chooses whether to try to conceive — try, because conception is uncertain — and then makes three nested labor decisions: participate or not; if participating, formal, informal or self-employed, subject to search frictions and offer arrival; and conditional on working, full or part time, where the part-time option is only available if the offer carries it. Experience accumulates separately in formal and non-formal sectors and is rewarded differently in each. Marital status is exogenous — you get married and divorced at random, which the speaker introduced as “in a very romantic way.”

Children hit through three channels. A participation cost in utility that depends on the age of the youngest child. A childcare time cost that subtracts from leisure and — crucially — varies by sector. And a search cost that depresses offer arrival for pregnant women and mothers of infants.

The estimates, by SMM, matching the event studies as moments:

Participating in the labor force with a baby aged nought to six months is worth a 13.3 percent drop in both consumption and leisure; by age three that falls to 2.1 percent. The childcare time cost, if working formally, is 20.8 hours a week with an infant and 9.8 hours at age three — and it is 82 percent smaller if working informally, 88 percent smaller if self-employed. Search costs: arrival rates are 94 percent lower when pregnant, 25.6 percent lower with a baby under twelve months. The 94 percent is the flat line, expressed as a parameter.

Now the counterfactuals, and the finding I did not expect. Give formal jobs the part-time availability of self-employment — always an option — and the penalty falls a bit on the intensive margin and barely at all on participation. Give formal jobs the commuting times of informal work: also helps a little. What actually kills the intensive-margin penalty is equalizing the childcare cost parameter — the residual bucket holding everything about flexibility that isn’t hours or commute. Being able to shift your schedule, skip the late meeting, not work the evening.

Which is a slightly awkward result, because it says the binding constraint is the one the model measures as a residual. The speaker was straightforward about this: they did not know a priori what it would be, and it turns out to be neither of the two things everyone measures. It is worth saying that a residual that large is also where any misspecification would end up, and the paper’s persuasiveness rests on the residual being flexibility rather than something else the three sectors differ in.

The trap

The last three minutes contained the paper’s real point, delivered at a run. Suppose you run a formalization program — tighter monitoring, higher cost of operating informally — calibrated to move five percentage points from non-formal to formal employment with no change in participation. This is a policy many governments want and several have tried.

Long-run welfare goes up about one percent. And the participation of mothers with young babies goes down.

Which follows directly, once you see the bundle. You have pushed women into the sector that does not sell the product they need. The intertemporal trade-off is the whole title: informality is a bad job that pays worse, accumulates the wrong kind of experience and carries no insurance, and it is also the only place a Brazilian mother can buy a schedule. Formalize the market and you have taken away the bad option without supplying the good one.

(Nobody in Brasília is going to design a formalization program around whether a woman can skip a four o’clock meeting. But the paper’s arithmetic says that is the margin, and that the protective apparatus — registered contract, social security, four months of leave, six months of job protection — is a package that, for the women it was written for, arrives fused to a schedule they cannot work.)