The Job Switch Is Not the Ladder
The subject of this series is the U.S. gendered job ladder: how men and women move between employers, occupations, and employment states, and why those moves can produce different wage growth. The opening focus is a deceptively simple question: if two groups change jobs at similar rates, why might their pay ladders still diverge? Understanding that distinction gives readers a better way to interpret earnings statistics, workplace choices, and policy changes that improve some parts of a job while changing others.
An earnings table shows a level, not a route
The Bureau of Labor Statistics offers a useful starting point. Its current chart reports median usual weekly earnings for full-time wage and salary workers by broad occupation and sex. In the second quarter of 2026, the all-occupation median was $1,251: $1,380 for men and $1,131 for women. The broad categories also sit at very different levels. In management, professional, and related occupations, the figures were $1,928 for men and $1,476 for women. In service occupations, they were $920 and $730.
Those numbers establish the wage dimension of the problem. Occupations are not interchangeable containers: the typical weekly earnings associated with one broad group can be far above those associated with another. The difference between men's and women's figures also shows why occupation matters when we discuss the gender pay gap. Where people work affects the pay they can receive, and the distribution of men and women across occupations is part of the story.
Even these figures need to be read at the right level of aggregation. A broad occupation groups jobs with different tasks, credentials, schedules, and employers. Its median summarizes a distribution; it does not describe every worker in that group or reveal the spread between lower- and higher-paid jobs. The chart is also a second-quarter, not-seasonally-adjusted snapshot for full-time wage and salary workers. It leaves out the history that placed those workers there, including their earlier pay and whether they arrived through a direct employer change or after an interruption.
That limitation is not a defect in the BLS measure. A level statistic is doing the job it was designed to do: locating the typical wage in a category at a point in time. The mistake would be asking it to answer a movement question. To understand a ladder, we need to compare starting positions with destinations and observe what happens between them.
But the table answers a narrower question than its visual simplicity might suggest. It describes a current level for a defined group of full-time workers. It does not tell us how long a worker has been in an occupation, what they earned before arriving there, whether they recently changed employers, what kind of offer they considered, or what noncash features the job provides. It is a map of positions, not a record of routes between them.
That difference matters because a person can reach a new job with a higher weekly wage, a better schedule, or both. Another person can make a move that improves flexibility but leaves pay nearly unchanged. If we look only at the destination's wage, those moves may appear equally disappointing. If we look only at whether an employer changed, they may appear equally successful. Neither summary captures the full decision.
A simple thought experiment makes the distinction concrete. Two workers can each leave an employer in the same month. One moves from a lower-paid occupation to a higher-paid one but accepts a rigid schedule; the other moves to a role with more predictable hours while remaining near the previous wage. The count records one switch for each. It does not record that the first move may raise cash earnings more, while the second may raise the value the worker attaches to the job. This example does not claim that any particular worker makes either choice. It shows why the outcome must be measured along more than one axis.
The same issue appears when we compare groups. An observed difference in current earnings may reflect occupational sorting, different starting positions, different destinations, or changes that occur after a move. A current gap alone cannot tell us which part is responsible. The relevant question is not simply who is in a higher-paid category today, but whether comparable moves produce comparable opportunities to progress.
The puzzle begins when switching is counted
The Chicago Fed working paper “Job Search, Job Amenities, and the Gender Pay Gap” examines this problem using the Current Population Survey and the Survey of Consumer Expectations, with supporting evidence from the Survey of Income and Program Participation. Its central finding is not that women switch jobs less often. Across the three data sources, men and women have nearly identical job-to-job transition rates, including after controlling for observable characteristics and occupation.
At first glance, that result might seem to imply similar opportunities to climb. A job-to-job transition means moving directly from one employer to another. It is an important kind of mobility, but it is still a count. It tells us that a move happened, not what the worker moved toward or what the move returned.
The wording matters. The paper reports nearly identical rates and no statistically significant gender difference under its stated definitions; it does not establish that every subgroup, occupation, or type of offer has the same experience. Nor does it say that switching is unimportant. Job-to-job mobility is one route through which workers can find better matches and obtain wage growth. The point is narrower and more useful: the frequency of that route is not enough to measure its payoff.
The surrounding flows are different. Women in the paper's CPS sample are more likely than men to leave employment for nonemployment and less likely to re-enter employment from nonemployment. Their employment-to-population ratio is 66.5%, compared with 79.1% for men. Even among direct job-to-job moves, the destinations differ: roughly one quarter of women's such transitions lead to part-time work, compared with about one tenth of men's.
So the same headline rate can contain different routes. One group may move between full-time employers while another makes more moves into part-time work. One group may remain continuously employed while another is more likely to experience an interruption and later re-enter. These differences affect both immediate weekly earnings and the position from which a later gain is measured.
The contrast becomes sharper when the researchers account for the worker's prior pay. In year-over-year regressions, the female coefficient on log hourly-wage growth is −0.110 after controlling for prior wages, and the corresponding coefficient for log weekly-earnings growth is −0.141 after controlling for prior weekly earnings. In plain terms, conditional on where workers started, women's measured pay growth was about 11 log points lower for hourly wages and 14.1 log points lower for weekly earnings. Without the prior-pay control, the growth gap is close to zero. Starting position changes what a comparison means.
This is why “same rate of switching” and “same return to switching” must remain separate. The first describes frequency. The second depends on the destination, the worker's starting rung, the continuity of employment, and the way the new job combines pay with other features. A mobility statistic can be accurate while still being incomplete.
The pay-growth estimates also require careful reading. They are log changes from one period to another, not dollar increases that can be applied to every worker. They are regression comparisons after accounting for prior pay and other listed characteristics, not a controlled experiment in which researchers assigned workers to different employers. The result is still important: once the comparison starts from a similar measured position, the gap in growth is no longer hidden by differences in where people began.
The level of measurement matters too. The paper finds a larger gender gap in individual wage growth than in occupation-level wage growth when it studies occupation switches: about −11 log points versus −4.9 log points. An occupation average can therefore show a meaningful pattern while missing movement within an occupation or other parts of an individual wage history. Reading the two measures together prevents a broad category from standing in for a complete career.
What counts as climbing?
The phrase “job ladder” often carries an unstated assumption: higher rungs are simply jobs with higher wages. That assumption is convenient, but it is too narrow for the evidence here. The paper models jobs as differing in productivity and nonwage amenities as well as wages. Amenities can include flexibility, work pace, scheduling, commuting conditions, or the possibility of working remotely. A job can therefore be more attractive in one dimension and less advantageous in another.
Productivity and amenities are not interchangeable labels. Productivity describes what the match between a worker and a firm produces; an amenity describes a feature of the work that affects its value to the worker but is not simply cash pay. The model lets firms choose amenity provision at a cost and lets workers value that provision differently. Wages are negotiated around the value created by the match, while an outside offer can change the worker's bargaining position. These links mean that the same cash wage can sit inside jobs with different total value, and that two offers with similar wages can generate different choices.
This does not mean that every wage difference is fully compensated by a benefit that a table fails to count. It means that the choice facing a worker is not reducible to a single number. A worker may accept a lower wage for a job that better fits desired hours or family responsibilities. A firm may offer useful flexibility in a role with less room for wage growth. An outside offer may improve a worker's total job value even when its cash wage is not the largest available.
That is also why “better job” needs a comparator. Better pay is one judgment; better fit, greater stability, or more usable time are others. The paper does not collapse these judgments into a universal ranking. Instead, it asks how different valuations and constraints change the jobs workers accept, the occupations in which they sort, and the gains available from later offers. The framework leaves room for trade-offs rather than assuming that one dimension secretly measures all the others.
The paper's evidence points in that direction without settling every reason behind the choices. Women report lower reservation wages and a preference for fewer work hours, while also placing more weight on several nonwage features. In the survey, women are more likely to say they would reject a relocation, a longer commute, or a required increase in hours at any offered wage. In job-satisfaction regressions, men's satisfaction is more closely associated with wages, while quality-of-life and childcare benefits have a stronger association with women's satisfaction. These are group-level patterns in stated preferences and conditional relationships, not proof of fixed traits or a complete account of social constraints.
That broader definition also clarifies the role of occupations. The paper's occupation-level amenity index combines measures of physical demands, pace, flexibility, and teleworkability. It finds that the female share of occupations rises with the amenity index and falls with occupation-level wages after accounting for both dimensions. When women switch occupations, they gain somewhat more in measured amenities but substantially less in occupation-level wages, conditional on their earlier position. The result is a steeper amenity ladder alongside a flatter wage ladder.
The opening lesson is therefore not that wages do not matter. They matter greatly, and the BLS figures show why wage levels deserve attention. The lesson is that wage levels, switching rates, destinations, and total job value answer different questions. A careful account of the gendered job ladder must keep all of them in view while avoiding a leap from observed group differences to claims about innate ability.
The next step is to examine the evidence more closely: what the equal job-to-job rate establishes, what it hides, and why prior pay and measurement level change the apparent return to a move. Only after that distinction is clear can we ask how amenities, search, separation, and bargaining send workers toward different rungs.