Gender, work, and care
In a fictional household, two adults sit down to plan the next month. Their combined paid hours look manageable until they add the hours needed to care for a child, arrange appointments, travel and sleep. One adult earns more per hour, so reducing the other's paid work appears to minimize the immediate loss of income. Yet repeating that decision may affect experience, promotion and future bargaining within the household. A sensible response to today's constraint can become part of tomorrow's unequal position.
This chapter examines paid work, unpaid care and the arrangements connecting them. It separates hours from hourly rewards, group averages from mechanisms and observed choices from the conditions under which those choices are made. The aim is to explain patterns without assuming that every household has the same interests, that all differences reflect discrimination or that a freely stated preference is independent of available options.
Earnings combine several moving parts
For a simple hourly job, earnings equal paid hours multiplied by the hourly rate. Annual earnings also depend on weeks worked and periods outside paid employment. Real compensation can include other components, but even the simple identity shows why an annual earnings difference cannot automatically be interpreted as unequal pay for an identical hour of work.
Suppose one person works thirty hours a week at twenty units per hour for fifty weeks. Annual earnings are 30,000 units. Another works forty hours at the same rate for fifty weeks and earns 40,000. Their hourly rates are equal; their annual earnings differ by 10,000. Explaining the difference requires explaining the hours, not declaring the earnings gap nonexistent.
Now hold hours constant and change the rate. That produces another pathway. Change weeks worked, access to overtime or time spent outside employment, and further pathways appear. A useful decomposition separates these components so that we can investigate them. It does not tell us which component is fair or why it differs.
The population matters here too. A comparison among full-time workers excludes people working part time or outside paid work. A comparison among people with positive earnings excludes those with none. These restrictions may answer a particular question, but they can omit consequential differences in access to employment and time allocation. The definition belongs beside the reported gap.
Occupational composition can produce an aggregate difference
Consider a stipulated workforce with one hundred women and one hundred men. There are two job types, paying twenty and forty units per hour. Within each type, everyone receives the same rate. Seventy women and thirty men hold the lower-paid jobs; thirty women and seventy men hold the higher-paid jobs. These are invented counts, not labor-market statistics.
The women's average hourly rate is twenty-six units: seventy times twenty plus thirty times forty, divided by one hundred. The men's average is thirty-four units. The aggregate difference is eight units, about 23.5 percent of the men's average, despite equal rates within each job type. Composition alone generates the gap in this example.
That result does not establish equal opportunity. We still need to explain access to the two jobs. Training, hiring, schedules, location, preferences and unequal treatment could all matter. If a rule excludes qualified applicants from the higher-paid job, equal pay within each type would not remove the barrier. If the jobs genuinely require different preparation, we can investigate access to that preparation.
Nor does the example show that every real pay gap is compositional. It shows why a single average is insufficient to choose among mechanisms. A real analysis would need appropriate data and a defined question. The purpose of the arithmetic is to prevent an aggregate difference from being mistaken for a direct comparison of identical work, while preserving the importance of explaining occupational sorting.
Unpaid activity is work even when earnings records omit it
Care, household maintenance and coordination can take time without producing a wage payment. A dataset of earnings therefore cannot measure all productive effort or all contributions to a household. Zero earnings do not mean zero work. This distinction is analytical: it identifies a limit of the measure before we begin to evaluate how contributions should be recognized.
Time measurement is not effortless either. A person may prepare food while supervising a child, or remain responsible for responding while doing another activity. Adding every reported activity can exceed twenty-four hours if simultaneous tasks are counted separately. A diary must specify whether it records a primary activity, secondary care, responsibility or some combination.
Predictability matters as well as duration. Two hours of scheduled care may be easier to combine with employment than two hours of interruptions distributed unpredictably across a workday. A time total can conceal fragmentation, coordination and the need to remain available. A study of work and care should select measures that match the proposed burden rather than assume that all hours are interchangeable.
We should also distinguish the person performing a task from the person responsible for ensuring it happens. Making an appointment, remembering a deadline and arranging a replacement can involve different demands from completing the visible task. These activities need explicit definitions if they are to be measured, rather than being added to an explanation whenever an observed pattern needs another label.
A household ledger makes the constraint visible
Here is a deliberately simplified weekday allocation. Each hour appears once; there is no simultaneous activity in this model. The two adults, Avery and Morgan, each have twenty-four hours. Travel and personal tasks are included in the final category. Their names do not stipulate a gender; later comparisons can ask how actual expectations assign these roles.
| Daily activity | Avery | Morgan |
|---|---|---|
| Paid work | 8 hours | 6 hours |
| Unpaid care and household work | 4 hours | 6 hours |
| Sleep | 8 hours | 8 hours |
| Other necessary activities | 4 hours | 4 hours |
| Total | 24 hours | 24 hours |
Suppose Avery earns twenty units per paid hour and Morgan earns twelve. Their daily earnings are 160 and 72, totaling 232. If they divide paid work equally at seven hours each and unpaid work equally at five hours each, the household earns 140 plus 84, or 224. Equalizing these hours reduces current household earnings by eight units under the stated rates.
The calculation identifies an immediate tradeoff. It does not prescribe which arrangement is best. The adults may value equal time, continuity of employment, professional development or relief from a demanding task differently. Their future rates may also respond to hours and experience, which this one-day model holds fixed. A short-run income maximum is not automatically a long-run household maximum.
Consider another option: purchase two hours of replacement care for twenty units in total, allowing Morgan to work eight paid hours while Avery's schedule stays fixed. Morgan's additional earnings are twenty-four units, so the household's current net cash rises by four units. This result depends on available care at the stipulated price, suitable timing and the absence of other costs. It is an accounting illustration, not a claim that such care is generally available.
Choices depend on feasible alternatives
If Morgan reduces paid work in the original arrangement, we should not immediately infer a stronger intrinsic preference for care. The lower hourly rate makes that response less costly in current cash terms. The rate itself may reflect prior experience, job access, qualifications or other processes. A choice can express both personal values and adaptation to unequal opportunities.
The reverse inference is also unwarranted: a constrained choice is not necessarily meaningless to the person making it. Morgan may value caregiving deeply while also wanting a more secure employment path. Sociology can examine the conditions of a choice without treating participants as incapable of understanding their own lives. Preferences, constraints and negotiations can all enter the account.
Household members may not have equal influence over the decision. Control of money, access to independent resources, information and recognized obligations can affect negotiation. A household-level income total conceals these relationships. To understand whose interests prevail, we need evidence about the decision process and the alternatives available to each person.
Repeated decisions can create feedback. Less paid work may reduce opportunities for experience or advancement in a particular workplace; a widening rate difference can then make the same allocation appear even more financially compelling. This is a proposed mechanism, not an inevitable sequence. A workplace with flexible advancement or a change in care arrangements could interrupt it.
Following earnings around a first birth
Kleven, Landais and Søgaard's 2019 study uses Danish administrative records to examine outcomes around parenthood. Its main balanced sample follows first births from 1985–2003, with observations before and after birth, and the article reports a long-run gender earnings penalty of roughly twenty percent associated with children. This is a historical Danish study, not a current estimate for every country or a raw gap between all women and men. Published article.
The event-study design aligns observations by time relative to first birth, using the preceding year as a reference and accounting for age and calendar year. It examines several labor-market margins. Its hours measure uses pension-contribution categories with a cap, a limitation when interpreting changes at higher hours. Longer-run estimates can also include the consequences of subsequent children rather than isolating one birth in all respects.
The design is informative because event time reveals a sequence that a single cross-sectional average can hide. But first birth is not randomly assigned. Interpretation depends on assumptions about the paths outcomes would have followed without the event, with stronger demands for longer-run comparisons. The article's “penalty” is an estimated relative outcome pattern; it is not a fine imposed by a named institution or proof of one exclusive mechanism.
We should therefore resist turning the result directly into a story about biology, discrimination or preferences alone. Several processes can change around parenthood: time demands, hours, employer responses, job choice and household arrangements. Distinguishing them requires additional evidence. A well-designed estimate can establish an important pattern while leaving part of its explanation open.
Reading an event-time comparison
Imagine a separate fictional chart indexed to one hundred in the year before a first birth. One parent's earnings path falls to eighty and later rises to ninety; the other's rises to one hundred and ten. The index expresses change relative to each person's own baseline, not their original earnings levels. If the baselines differed, equal index values would not imply equal amounts of money.
A reference year is also not automatically a counterfactual. Without the birth, earnings might have risen with experience, fallen during a recession or changed for another reason. Comparing the later outcome only with the baseline can mix the event with those other changes. Age and calendar-time adjustments address parts of this problem under their assumptions; they do not make every unobserved difference disappear.
Patterns before the event can help assess whether groups were already moving differently, but a visually flat pre-event pattern is not a guarantee of identification. The available period may be short, estimates imprecise or relevant changes anticipated. We should inspect the design and uncertainty rather than treat a graph's shape as self-validating.
The outcome's definition remains central. An earnings measure that includes zeros captures employment exits differently from a wage measure observed only among employed people. If those remaining employed are selected, their average wage can rise even while the broader group's earnings fall. A chart should identify who remains in the calculation at each point.
Workplace rules connect time to rewards
Consider a fictional firm where promotion requires being available for unscheduled evening meetings. The rule may reward useful responsiveness in some tasks, but it may also exclude workers whose care responsibilities are fixed. To evaluate it, ask what the meetings accomplish, whether the timing is necessary and whether another arrangement could preserve performance.
A policy allowing flexible hours could change access, yet its effect depends on actual use and consequences. If workers formally have the option but expect that using it will damage promotion prospects, the written rule is an incomplete description. Evidence about requests, approvals, assignments and advancement would help establish how the policy operates.
There can also be unequal effects from an apparently equal offer. Two workers may receive the same flexibility while facing different caregiving demands or different authority to rearrange tasks. The question is whether the arrangement makes participation feasible, not simply whether the policy sentence applies to everyone. Outcomes and mechanisms should be measured alongside formal eligibility.
This does not establish that every scheduling difference is unjustified. Some activities genuinely require coordination at a particular time. The analysis should examine necessity and alternatives rather than assume that either managerial convenience or individual preference settles the issue. A concrete task and a credible alternative make the tradeoff assessable.
Evaluating a response requires several outcomes
Suppose the firm introduces predictable schedules. A useful evaluation might examine earnings, retention, task performance and care coordination. Improvement in one outcome does not guarantee improvement in all. A reduction in hours could reflect an unwanted loss of work or a valued release from excessive demands; interpretation requires information about the situation and the participant's aims.
The policy might also shift work to colleagues or change hiring. Those effects belong in the evaluation if the proposed mechanism makes them plausible. Ignoring them could exaggerate benefits; inventing them without evidence could unfairly dismiss a useful reform. The standard is to identify relevant consequences and investigate them, rather than attach a generic list of possible objections to every intervention.
Our household ledger and the Danish study operate at different levels. The ledger demonstrates an exact tradeoff under stipulated assumptions. The study estimates a historical pattern using records and a design. Neither alone tells us the best arrangement for a particular household. Together they show why explaining gendered inequality requires attention to time, rewards, care, selection and the institutions connecting them.
Application
Recalculate the workforce averages and the household's three cash outcomes. Then propose one workplace change and one care-related change, identifying the pathway each would alter and an outcome that could reveal an unintended cost. Write a 200-word interpretation of the Danish study that states its population, event, measure and causal limits.
Check your understanding: If women and men have equal hourly rates within each occupation, must their overall average earnings be equal, and does a parenthood event study identify one exclusive cause of the later gap?
Expected answer: No. Occupational composition, paid hours, weeks worked and employment participation can differ even with equal within-occupation hourly rates. An event study can reveal and estimate a pattern around parenthood under stated assumptions, but it does not by itself separate every mechanism involving care, employment rules, preferences or unequal treatment.