The First Rung Is a Different Remote-Work Problem
Remote and hybrid work are the subject of this series, with a particular focus on the relationship among distributed teams, employers’ training choices, and young college graduates trying to enter remote-capable jobs. The question is not simply whether an employee can finish today’s assignment away from an office. It is whether a team can create enough of the small, repeated teaching interactions that turn an inexperienced hire into a capable contributor. That distinction gives readers a practical way to judge remote-work claims: separate what happens to current output, to an employee’s decision to stay, and to the development of people who are still learning the work.
Remote work has changed the shape of a job
Working from home was not invented by the pandemic. It was, however, transformed by it. A Bureau of Labor Statistics analysis follows the National Longitudinal Survey of Youth 1979 cohort—Americans born between 1957 and 1964—across more than four decades. In 1988, members of this cohort spent an average of 2.0% of their work time at home. By 2020, that share had reached 8.7% among all employed workers. Among workers who reported at least some telework, the average share rose from 11.2% to 25.3% over the same comparison.
The change was not just a smooth increase. Full-time telework rose sharply from 3.9% of employed cohort members in 2018 to 15.3% in 2020. After the pandemic shock eased, full-time remote work declined, but intermediate arrangements continued to grow. By 2022, the average share of work time spent at home among people with some telework had fallen to 21.7%—below the 2020 peak, but still well above the earlier pattern. What had been occasional or highly concentrated in a small group had become, for many remote-capable workers, a regular mixture of home and office.
The transitions in the data make the distinction more concrete. Among workers who reported no telework in 2018, 81% were still in that category in 2020. At the other end, 81.5% of those who were already working from home full time remained full-time teleworkers. The least intensive arrangements were less stable: 37.4% of people who worked from home for two hours or less in 2018 reported no telework in 2020. By 2022, many people who had been working from home nearly full time or hardly at all had moved into intermediate categories. Remote work was therefore not one switch that moved the whole labor market from office to home. It was a set of job designs with different degrees of continuity, different amounts of shared time, and different opportunities for a beginner to observe how work is done.
That difference matters for training. An occasional home day may change where someone completes routine work without changing the team’s basic learning rhythm. A schedule with several predictable office days may preserve frequent observation while giving employees quiet time and relief from commuting. A near-full-time remote arrangement asks the team to reproduce almost all of its feedback, explanation, and relationship-building through deliberate channels. Treating these arrangements as equivalent makes both the evidence and the policy argument less precise.
That history matters because a work arrangement is also a learning environment. The office, the home, and the digital channel between them determine not only where a task is performed but how easily someone can see a colleague’s reasoning, ask a question, receive a correction, and try again. Those opportunities may be abundant, scheduled, and visible—or they may be scarce and dependent on a beginner’s willingness to interrupt someone who is not physically nearby.
Remote work is also not equally available to everyone. The BLS cohort data show that education, occupation, race and ethnicity, sex, and region are associated with different degrees of telework access and intensity. Over time, workers with bachelor’s or master’s degrees became more represented among full-time teleworkers, while high-school graduates remained more represented in the no-telework group. In the 2020–22 period, the no-telework group also included higher shares of Black and Hispanic workers than the full-time-telework group. These are descriptions of who occupied different kinds of jobs, not evidence that a demographic group is inherently more suited to working at home.
That distinction is important for entry-level work. Remote access can be a benefit negotiated by an established employee, but it can also be a property of the occupation and team to which a newcomer is admitted. A new graduate may be applying into a remote-capable field without having the experience, bargaining power, or social knowledge that lets an established colleague shape the arrangement. If the job is designed around independent execution, distance may be a welcome feature. If it is designed around learning through shared work, the same distance may change who is considered affordable to hire.
The BLS findings do not tell us exactly why these differences persist; the article points to occupational structure and regional conditions among possible explanations. They do, however, caution against speaking of remote work as if it were a single employee choice. A software engineer, a hospital technician, and a retail worker do not face the same menu of choices. Nor do a senior specialist and a first-time hire, even when both are technically eligible to work from home. The relevant unit of analysis is often the job’s learning system: what is visible, who supplies correction, and how much of that support is available before a worker has learned to ask for it.
Productivity is three different questions
The public argument often asks whether remote work is productive, as if productivity were one observable thing. For an organization, at least three questions are hiding inside that word.
The first is current performance: Does an employee complete useful work at an acceptable standard? The second is retention: Does the employee want to remain, and does the organization avoid the cost of replacing them? The third is development: Is the employee acquiring the judgment, context, and habits needed to handle more difficult work independently? The first two can improve while the third becomes harder to produce. A person who already knows the job may work effectively from home; a person who has never learned the job may need a very different support system.
A useful model keeps these outcomes separate:
This is why a favorable performance result cannot settle the development question. In a randomized trial at Trip.com, 1,612 university-graduate employees were given either the option to work from home two days a week while spending three days in the office, or an office-only schedule. The hybrid option reduced quits and improved job satisfaction. Over the following review periods, the researchers found no meaningful damage to performance grades, promotions, or computer engineers’ lines of code. That is valuable evidence against the claim that a moderate hybrid schedule automatically makes established employees less productive.
But the trial excluded interns and rookies in probation because on-site learning and mentoring were considered especially important for them. The result therefore answers a narrower question than the headline debate usually acknowledges: how does a particular hybrid option affect an existing, university-educated workforce in one technology company? It does not show that the same arrangement has no effect on the process of learning a first job.
Why the first job is the test
The distinction becomes consequential when we look at young college graduates. The New York Fed reports that their unemployment rate averaged 3.1% in 2017–19 and rose to 3.7% in 2022–25. Over the same comparison, unemployment among more experienced college graduates dipped from 1.9% to 1.8%. The gap is striking not because every young person is affected in the same way, but because the increase was concentrated in occupations that can more readily be performed remotely. In non-remotable occupations, the relative increase for young graduates rose during 2020 and then moved back toward its earlier level.
The stakes extend beyond the months spent looking for a first job. The New York Fed notes that entering the labor market in a weak economy can leave lasting marks on earnings and career progression. That does not mean every delayed hire experiences permanent damage, but it changes the cost of getting the diagnosis wrong. If a firm temporarily substitutes experienced workers for beginners, the firm may protect near-term execution. Across many firms, however, that same choice can reduce the number of places where beginners acquire the experience later employers require. A shortage of first chances can become a shortage of qualified candidates rather than evidence that young workers have nothing to contribute.
There are three different evidence questions inside this pattern. The BLS data tell us that remote work became widespread and varied, but they do not identify its causal effect on today’s young workers. The New York Fed compares age groups and occupation categories and then studies one firm’s feedback and hiring patterns, which makes the proposed mechanism plausible but leaves room for sorting and other shocks. The Trip.com experiment gives stronger causal evidence for one moderate hybrid schedule, but its established graduate employees are not the people who were excluded from the experiment. Keeping these questions separate is not a technical nicety. It is what prevents a result about one outcome or population from being used to answer a different one.
The authors estimate that remote work can explain 64% of the increase in unemployment among young college graduates between the pre-pandemic period and 2022–24. That is a back-of-the-envelope estimate, not the result of randomly assigning workers to remote jobs. Occupation choice, labor demand, worker sorting, and other pandemic-era changes could also matter. The timing likewise does not prove that generative AI is irrelevant; it does show that the youth-unemployment increase began before the rapid diffusion of generative AI, and the age difference persisted when occupation-level AI exposure was held constant.
The most useful interpretation is narrower and more demanding. Remote work may have changed the economics of entry-level hiring. If a beginner requires frequent feedback and mentorship, and a distributed team makes those interactions more difficult or costly, an employer may respond by hiring someone who already has the needed experience. That choice can look efficient from the firm’s short-term perspective while depriving beginners of the very first opportunity through which they would acquire experience.
This is the problem the series will examine. It is not a case against flexibility, nor a claim that an office is automatically a school. It is a question about whether the learning system surrounding a job is strong enough for the people who have not yet learned it. The answer depends on the intensity of remote work, the interdependence of the tasks, the employee’s seniority, and whether a team deliberately replaces the informal feedback that proximity makes easier. Before deciding whether remote work damages productivity, we need to understand what proximity is doing in the first place.