No, AI is Not Taking Your Job...Yet

Recent college grads are having a hard time finding jobs, but ChatGPT isn’t the reason.

Hey, everyone! My name is Jordan, and I have a Ph.D. in Economics from the University of Pennsylvania. I want to ask important questions and answer them meaningfully using the hard work economists have put into their research. If you learn anything interesting from this, please like and subscribe!

News reports suggest that recent college graduates have faced difficulties due to factors like the growing use of artificial intelligence. For example, data from the NY Fed shows that unemployment rates for recent graduates have increased compared to older graduates since the economy began recovering from the pandemic in 2021. AI technology became widely popular around 2022, which might initially seem connected to these higher unemployment rates.

Artificial intelligence (AI) is a technology that allows computers to perform tasks usually done by people, like understanding documents, recognizing images, or making decisions. AI research began decades ago, but major advances only started around the 2000’s. In 2012, deep learning, which is a method where computers learn from large amounts of data, led to breakthroughs like facial recognition and self-driving cars. By 2022, ChatGPT was introduced, which was a breakthrough in what is known as Gen AI; Gen AI uses logic that allows users to ask detailed questions and receive detailed output such as code, images, and writing. This technology is advancing at a quick rate.

But AI is not responsible for the rising college graduate unemployment rate.

To understand how much recent college graduates are being impacted, I compare unemployment rates between recent graduates and mid-career adults, grouped by education and gender; I divide the recent graduate unemployment rate by the mid-career worker unemployment rate. This comparison helps us understand how recent college graduates are faring relative to older graduates with similar education levels. The figure below shows that recent college graduates have increasingly higher unemployment rates compared to older college graduates.

There’s also a smaller increase among recent high school graduates compared to older high school graduates, but it is not as strong. This trend has been ongoing since 2010 (way before Gen AI and ChatGPT was introduced) but became more noticeable after the economy started recovering from the pandemic. Among these groups, men with bachelor’s degrees have experienced the greatest increase in relative unemployment rates.

To better understand unemployment, it helps to look at why people become unemployed. Are they getting laid off or leaving jobs? Are they entering the workforce and beginning to apply for jobs? Or are they just stuck being unemployed? After looking at all three, I found that it is the latter case that has been happening in recent years; recent college graduates are increasingly experiencing longer unemployment spells compared to older college graduates. This issue has become especially pronounced for men, as shown in orange in the graph below. Male college graduates are now over 2.5 times more likely than their older counterparts to remain unemployed from one month to the next.

This trend has been steadily getting worse since 2010. An important follow-up question is whether recent graduates are also spending longer periods unemployed overall. Does our earlier observation hold true here as well? As I show below, the answer is yes — recent graduates are indeed facing longer stretches of unemployment.

What might be causing this trend? While it’s hard to say how much each factor matters, I’ll explore several possible explanations using economics research—including the role of automation and job loss due to AI.

Supply of Bachelor’s degrees has outpaced demand

The number of people earning Bachelor’s degrees has grown steadily, especially compared to other types of higher education. But has the demand for these degrees grown just as quickly? There are a few signs that suggest maybe not. First, around the year 2000, the extra pay college graduates earn over high school graduates (the college wage premium) stopped rising. This suggests the number of college graduates matches the number employers want for the same wage since 2000. Second, one study found about half of recent college graduates start out in jobs below their skill level, though this improves somewhat over their careers.

Lastly, another study noted that since 2000, jobs requiring advanced thinking skills, such as managers, engineers, lawyers, or programmers, have become relatively less common, which may have led to the second point of underemployment; college graduates may look for a job longer if they think they will become underemployed in a job that pays less and requires lower skills. However, it's also possible that the average skill level among college graduates has declined over time as well, but this argument is not as supported.

Graph from educationdata.org

Computer science majors have recently seen higher unemployment rates, raising concerns that AI could reduce demand for these jobs. While some skills may become less relevant, others could grow in value — I’ll touch on that later in the post. Another possible explanation is a mismatch between the number of students choosing computer science and the number of available jobs. I explore that in the graphs below:

Here, we see a clear cycle: demand for software developers — a common job for computer science majors — spiked in 2022. Many entry-level roles were likely filled that year, so fewer are available now. Meanwhile, the number of computer science graduates has continued to rise. Since students often choose their majors based on trends from a few years earlier, we may see a drop in computer science majors in the coming years as the job market adjusts.

Overall, this may point to a deeper shift in the job market that’s leading to longer unemployment for young college graduates. If fewer jobs are being posted while more students are graduating, it makes sense that new grads might have to wait longer to land the jobs they expected. Still, this doesn’t fully explain why the trend started accelerating after 2016.

The hiring process has grown longer for entry-level positions

One possible reason is that it now takes longer to get hired for entry-level jobs. Imagine you're a recent grad applying to several jobs. In a competitive market, each application might involve weeks of interviews, only to end in rejection. Without much experience, you may face multiple rejections, and the time spent in each hiring process adds up, leading to a longer period of unemployment. More experienced workers can signal their expertise from on-the-job learning more effectively.

Time to hire is one way to measure the interview length, but it could be indicative of difficulties finding someone skilled enough for the position rather than a process in place, especially when unemployment rates are low. After the Great Recession, one study found that recruiting intensity didn’t increase enough to keep up with vacancy postings, which means the lack of intensity could have led to increased time to hire.

To see how much the time to hire has increased over the years, I have included the graph below. It shows a strong increase from 2010 to 2017 — right where we see new graduates getting impacted by unemployment duration the most. During this time, businesses were growing and posting more jobs, but when unemployment rates are low, the marginal gain from effort may be small.

Graph from DICE

There isn’t much information on how much recruiting effort differs between entry-level jobs, so this potential explanation should be considered in future economics research. However, the trend for recent college graduates could very well depend on this cyclical change.

Automation and Gen AI could be replacing early college careers

The final possible explanation I consider is generative AI and automation. The concern is that Gen AI could replace common entry-level jobs, especially those held by recent college grads. Fears about automation aren’t new — robots have long been used in manufacturing to take over routine tasks. If automation were wiping out jobs across the board, we might expect unemployment to be much higher than the usual 4% seen outside of recessions and pandemics, but that hasn’t happened. While manufacturing jobs have declined since the 1980s, other sectors like hospitality and professional services have continued to grow.

One paper breaks down the impact of new technology into two main effects: displacement and productivity. The displacement effect happens when technology fully replaces jobs, like robotic arms in car factories, but there’s also the productivity effect, which can offset that loss by making a business more productive and allowing it to expand. For example, self-checkouts may reduce the need for some workers, but if the store becomes more profitable, it might expand and hire even more people overall. Economists call this kind of technology “labor complementary” because it can ultimately lead to more hiring, not less. The opposite would be “labor substituting.”

It’s difficult to say exactly how much job loss AI will cause, but we do know the job market is likely to look very different in the next five years. The IMF estimates that about 25% of jobs in advanced economies are highly exposed to AI. A 2013 study predicted that administrative and sales jobs, like cashiers and telemarketers, were most at risk. However, more recent research suggests the productivity effect may be stronger than the displacement effect, showing that AI tools like ChatGPT have so far boosted efficiency more than they’ve replaced workers.

The widespread adoption of the computer and Internet created a spectrum of new jobs that could not have been imagined in the 1980’s, like software developers and content creators. The widespread adoption of machine learning, which is the basis of social media algorithms, created jobs devoted to optimizing it, like data scientists and analysts. And, with the widespread adoption of Gen AI and AI chatbots, we will see a shift from software developers and data scientists to something new — something devoted to developing big ideas through AI.

Conclusion

In summary, the recent rise in unemployment for recent graduates relies on a trend of relatively longer unemployment spells happening since 2010. The evolution of this trend is to be determined, but it is not obvious that it is being driven by AI adoption — it could very well be driven by supply, demand, and vacancy search effort by businesses. The adoption of AI will create a new set of jobs that could make other jobs less valuable at a slower rate than we anticipate; automation of certain jobs takes time, and graduates will adapt. We will see skills among the future graduates and current graduates adjust to these new times, just as we have seen them do with the adoption of the computer, the adoption of the assembly line, and the adoption of the plow.