What counts as a remote-enabled career move?
The U.S. study asks a precise version of the remote-work question: what is associated with entering a job whose employer-side opportunity structure makes remote work possible? This chapter follows that job-entry lifecycle—from the worker’s previous position, through the matching of employer postings, to wage, seniority, geographic, and organizational outcomes. The aim is to understand what the comparison measures and how much confidence its weighting and robustness checks can reasonably provide.
That precision matters because “remote worker” would be too broad a label for the design. The study does not observe a random group of people being assigned to work from home, nor does it directly record how many days each entrant works remotely. It identifies whether the destination opportunity was advertised as remote-capable at the time of hiring, then compares the subsequent transition with a carefully adjusted set of fully on-site entries.
From a job history to a remote-eligibility measure
The unit of analysis is a job-entry event: the start of a new position after a previous employment spell for the same individual. This unit turns a career history into a sequence of transitions. Instead of asking whether one worker earns more than another at a single moment, the study asks what changes between the old job and the new one. The previous job supplies the baseline for wage, seniority, occupation, employer, industry, and metropolitan location; the new job supplies the destination against which mobility is judged.
That transition-based unit is important for two reasons. First, it makes “upward” relative to the worker’s own prior position rather than a comparison of people at different career stages. A move from an entry-level role to a junior role is evaluated differently from a move by someone who was already a director. Second, it makes the timing of remote eligibility explicit. The relevant question is what kind of opportunity the worker entered, not whether they later described their lifestyle as remote.
The study links two large data systems. Longitudinal employment histories provide job spells, predicted wages, seniority levels, occupations, employers, and locations. Employer vacancy records provide information about the work arrangements attached to posted opportunities. The researchers connect a job entry to postings from the destination employer that match the employer, occupation, country, and metropolitan area in a three-month window before entry. The window is intended to capture the opportunity structure facing the entrant shortly before the transition rather than a policy that may have changed long afterward.
Within each matched employer–occupation–location–month cell, the study calculates the share of postings offering any form of remote work. The categories include fully remote, territory remote, temporary remote, and partial-office arrangements. A transition is classified as remote-eligible when at least half of the matched postings offer one of those forms. A transition below that threshold is treated as non-remote-eligible and belongs to the comparison group.
The threshold is a practical classification rule, not a claim about the worker’s exact schedule. A destination with a majority of remote-capable postings signals an employer-side opportunity structure in which remote work is available for that kind of role and place. It does not say that every posting is fully remote, that the worker negotiated the same arrangement, or that the arrangement lasted throughout employment. The measure is better understood as “entered a job market where remote eligibility was part of the available design” than as “worked remotely for a known fraction of time.”
The resulting U.S. sample contains approximately 48 million job-entry events with defined remote-work eligibility between 2020 and 2024, including roughly one million transitions into remote-eligible roles. The scale allows the study to examine differences across prior wages, occupations, managerial status, industries, metropolitan conditions, and time. It also brings a selection issue: the people whose job histories can be linked to postings, and whose wages and characteristics are observed, are not necessarily a complete picture of every worker who changed jobs.
The lifecycle can be summarized as follows:
The sequence keeps three objects separate. The worker history establishes a transition. The postings describe the destination opportunity. The analysis compares outcomes after trying to make the treated and comparison transitions similar on characteristics measured before entry. Confusing these objects would turn a supply-side eligibility measure into a claim about realized remote behavior.
What the comparison is designed to mean
Workers who enter remote-eligible and fully on-site jobs may differ before the transition. They may have different wages, seniority, occupations, industries, previous employers, locations, or demographic characteristics. If those differences were ignored, a higher wage change among remote-eligible entrants could simply reflect the fact that they were already on a different career trajectory.
The study addresses this concern with inverse-probability weighting. First, it estimates the probability that a transition enters a remote-eligible job given observed pre-entry characteristics. Non-remote transitions are then reweighted so that their observed distribution resembles the distribution among remote-eligible transitions. In intuitive terms, a fully on-site transition that looks much like a remote-eligible transition receives more influence in the comparison than one that comes from a very different part of the labor market.
The method targets a comparison among observed types of transitions; it does not manufacture a randomized experiment. The weighting procedure can balance the variables included in the model, but it cannot balance characteristics that were not measured or measured poorly. A worker’s unobserved ability, urgency to move, quality of negotiation, or relationship with a hiring manager may affect both the type of job they enter and their later outcome. The weighted comparison is therefore a way to reduce obvious differences, not a guarantee that the two paths would otherwise have been identical.
Imagine two transitions by workers with similar prior wages, seniority, occupation, employer size, location, and entry year. One enters a destination whose matched postings are mostly remote-capable; the other enters a fully on-site destination. The study’s comparison asks whether their next-job outcomes differ after the broader sample has been reweighted to make such similarities common. It does not ask whether the first worker would have received the same offer, negotiated the same terms, or behaved the same way if remote eligibility had been removed. Those are counterfactuals the data cannot observe directly.
The weighting also changes how a reader should interpret the size of the result. A coefficient is not the average wage increase for all workers who might someday work remotely. It is an estimate attached to the transitions represented in the treated group and to the comparison constructed from observed covariates. Workers with no plausible on-site counterpart may contribute little to the effective comparison, while workers in occupations and places where both arrangements are possible carry more of the identifying variation. That is a strength for like-for-like interpretation, but it narrows the population to which the estimate can safely be generalized.
The regression stage adds controls and fixed effects for factors such as previous wages, previous employer size and prestige, seniority, ethnicity, occupation, industry, metropolitan area, and entry year. These adjustments serve different purposes. Prior wage and seniority locate the worker on a career ladder. Occupation and industry account for different baseline possibilities for remote work and wage growth. Metropolitan controls account for local labor-market conditions. Entry-year effects absorb broad changes in hiring conditions, including the unusual evolution of remote work after 2020.
The outcomes are deliberately plural. Wage growth is the log change between the destination job and the immediately preceding job. An upward seniority move records whether the new role has a higher standardized seniority level. A cross-metropolitan move records a change in the metropolitan or non-metropolitan area attached to the old and new jobs. Destination employer size and prestige describe the organizational route taken by the transition. Each outcome answers a different part of the career question.
A higher wage change indicates a larger immediate step in pay relative to the prior job, but it does not tell us why the step occurred or whether it persists. A seniority move indicates a change in rank, but job titles can mean different things across employers. A cross-metropolitan change indicates that the transition reached beyond the previous local labor market, but it does not tell us whether the worker moved home. Employer size and prestige indicate what kind of organization received the worker, but neither is a complete measure of job quality.
The study therefore treats mobility as a pattern across outcomes rather than as one master score. If remote-eligible entry were associated only with a higher wage level, selection would remain a strong explanation. If it were associated with higher wage growth, more upward seniority moves, and more cross-metropolitan transitions, the result would be more consistent with a change in the structure of job matching. Even then, the interpretation would be an association under the study’s design unless the remaining alternatives were convincingly addressed.
The robustness checks make that interpretation more demanding. The researchers expand the posting-matching window from three months to six and obtain similar results. They use a stricter threshold, requiring more than 0.8 rather than 0.5 of matched postings to offer remote work; the signs remain similar. They also compare remote and non-remote entries within the same individual when the person experiences both types of transition. That comparison absorbs stable individual characteristics that are difficult to measure directly, although it does not remove changes in the person’s circumstances between job entries.
A further test restricts matched transitions to switches within the same occupation. This matters because a worker may move into a higher-paying occupation that is also more likely to offer remote work. Holding occupation constant reduces that particular route for confounding. The study also adds occupation-by-entry-year fixed effects in the full sample, absorbing occupation-specific changes over time in demand, pay norms, and remote feasibility. The estimates are attenuated in stricter comparisons but remain positive for the main outcomes reported in the paper.
These checks strengthen the claim that the observed pattern is not merely an artifact of one threshold, one matching window, or a simple difference between different kinds of workers. They do not prove that eligibility is exogenous. The study still relies on career histories built from public professional profiles, predicted position-level wages, employer postings, and inferred demographic characteristics. Posting coverage may be incomplete. Occupational upgrading may still be related to a time-varying factor that also affects remote eligibility. The analysis does not observe long-run mentoring, tacit knowledge transfer, or network formation inside the destination organization.
There is also a conceptual limit to the treatment itself. Employer-side eligibility is a condition of access, not a dose of remote work. Two workers may enter jobs classified in the same way but use different schedules, receive different support, or spend different amounts of time in the office. One may value the flexibility but work mostly on site; another may use it to remain in a distant community. The estimated difference therefore belongs to the opportunity associated with the job entry, not to a simple formula relating days at home to career returns.
This distinction also protects the later comparison with French wage evidence. The U.S. study is designed to observe whether a certain kind of job transition is associated with a different trajectory. It is not designed to answer whether people who work from home earn more because remote work increases their productivity. That second question needs a different outcome, a different treatment measure, and stronger attention to prior worker characteristics.
For now, the main methodological lesson is straightforward: a career-mobility estimate is only as clear as its transition definition. Once the destination opportunity, comparison group, and outcome are named precisely, the first study can be read for what it actually supports. The next step is to put those outcomes together and ask whether they form a coherent, distributionally meaningful pattern.