Generative AI and Early-Career Hiring: Counts, Entry Rates, and the Employment Base

The story of how new technology changes the workforce is often told as a sudden shift, a moment when machines take over and people are left behind. But the reality is usually more subtle, unfolding in the quiet mechanics of hiring and job stability long before the headlines arrive. Economists study these shifts by looking at two different ways of measuring work: the raw number of people hired, and the rate at which those hires happen relative to the size of the existing workforce. Imagine a factory floor where the total number of workers is shrinking; even if the factory hires the same number of new people as before, the rate of hiring relative to the total staff might look different. Understanding the difference between a simple count of new workers and the mathematical rate of hiring is crucial, because a drop in the number of new hires does not always mean employers have stopped looking for talent. It might simply mean the pool of experienced workers they are hiring from has changed size or composition. This distinction becomes especially important when examining how artificial intelligence, specifically the new generation of tools that can write and reason like humans, is affecting young people trying to start their careers.

A recent study by researcher Yi Wu at The University of Texas at Austin investigates exactly this question. The paper asks whether the rise of generative AI has caused a specific drop in the hiring of early-career workers, defined as those between the ages of 22 and 34, compared to older, established workers. To find the answer, the researcher did not rely on surveys or guesses about what companies were thinking. Instead, they turned to a massive, detailed government dataset called the Quarterly Workforce Indicators. This data tracks real hiring events across thousands of industries in the United States, distinguishing between "all new hires" and "stable new hires." A stable hire is a specific type of job record: it is counted only when a new employee has stayed in the job long enough to be considered a full-time, steady part of the workforce for a whole quarter. By focusing on these stable hires, the study looks at the jobs that actually stick, rather than short-term or trial positions.

The researcher compared industries that are highly exposed to language-based artificial intelligence with those that are less exposed. The exposure was measured by how much the typical tasks in an industry rely on skills that AI tools can now perform, such as writing, coding, or analyzing text. The study found that after 2022, when these AI tools became widely available, the number of stable new hires for young workers fell significantly more in the high-exposure industries than in the low-exposure ones. This drop was not just a small fluctuation; it was a clear, measurable decline. However, the study also found that this drop was much more pronounced for stable hires than for all hires, suggesting that the AI impact is specifically affecting the jobs that lead to long-term employment.

Here is where the story takes a surprising turn. When the researcher broke down the numbers to see why the count of young hires was falling, they discovered that the drop was not driven by a sudden, statistically significant change in how often companies were hiring young people relative to their current staff. Instead, the decline was almost entirely located in the size of the workforce itself. The study separated the hiring count into two parts: the hiring rate (how many new people are hired per unit of existing staff) and the employment base (the total number of stable workers already on the payroll). The data showed that the hiring rate for young workers did not exhibit a statistically reliable decline. In fact, the estimated coefficient for the rate was positive but imprecise, and its sign varied depending on how the data was transformed, meaning the evidence for a distinct drop in the entry rate is less firmly supported. The entire decline in the number of young hires came from the fact that the pool of stable, experienced workers in these industries was shrinking or changing in a way that reduced the denominator.

This finding challenges a common assumption. Many might expect that if AI is replacing human work, companies would simply stop hiring young people at the same rate they used to. The study suggests a different mechanism. The decline in young hires appears to be a consequence of a longer-term trend in the composition of the workforce that was already happening before the pandemic, and which continued afterward. The "employment base" for young workers in AI-exposed industries was already trending downward, and the new AI tools did not necessarily cause a sudden stop in hiring activity. Rather, the hiring shortfall is concentrated in the shrinking pool of stable jobs available to be filled. The study explicitly notes that this pattern was visible in the data before the widespread adoption of the most famous AI chatbots, suggesting that the forces at play are deeper and older than the recent technological boom.

The research also looked at whether this trend was a sudden break from the past or a continuation of an older pattern. By comparing the post-2023 data to the trends seen between 2017 and 2019, the study found that the decline in hiring counts did accelerate slightly after 2023. However, even this acceleration was driven primarily by changes in the employment base, not by a sudden collapse in the hiring rate. The study is careful to state that while the numbers show a clear correlation between AI exposure and fewer young hires, it does not prove that AI is the sole cause. Other factors, such as the economic recovery from the pandemic or other industry changes, could be contributing. Yet, the evidence suggests that the drop in young hires is not simply a result of companies hiring fewer people per existing worker; the rate of hiring remained relatively steady, while the total number of stable positions available for young people decreased.

One of the most important takeaways from the paper is the distinction between a count and a rate. If you only look at the total number of young people getting hired, it looks like a disaster for their career prospects in AI-heavy fields. But if you look at the rate at which they are hired relative to the size of the workforce, the picture is much less dramatic. The study concludes that the employment base deserves as much attention as the worker flows when we try to understand how technology changes the labor market. The decline in young hires is real, but it is concentrated in the denominator—the size of the stable workforce—rather than in the numerator, which represents the hiring activity itself. This means that the story of AI and early-career workers is not just about companies stopping their hiring; it is about a complex shift in the structure of the workforce that predates the current AI wave and continues to shape who gets hired and who does not.

The study does not claim to have solved the mystery of how AI will affect the future of work, nor does it suggest that the situation is hopeless. It simply provides a clearer, more precise map of what is happening right now. By separating the hiring rate from the employment base, the research reveals that the drop in young hires is not a sudden, sharp rejection of young talent by employers. Instead, it is a reflection of a broader, slower-moving change in the composition of the workforce in industries where AI is most relevant. The findings suggest that the challenges facing early-career workers are deeply rooted in the structure of these industries, and that understanding the size and stability of the workforce is just as important as counting the new hires. In the end, the paper offers a quiet but powerful correction to the narrative: the problem is not necessarily that companies have stopped hiring young people, but that the landscape in which they hire has fundamentally shifted, and that shift began long before the first AI chatbot went online.