The 360: How will artificial intelligence reshape the job market

There is a lot of anxiety out there, and while influencers and experts speculate, UCLA economist Till von Wachter and other scholars from the California Policy Lab recognized that clear, up-to-date data is needed to facilitate informed conversations on the topic.

Their new tool, the California AI-Unemployment Tracker, is a great start to gathering verifiable data that addresses this inquiry.

We recently caught up with von Wachter and asked him some of our initial questions:

How long might it take for the California economy to feel the impact of AI on the labor market?

The timing of AI's potential impact on job loss can be hard to predict. To help spot early warning signs of AI-related job losses, we built the California AI-Unemployment Tracker together with the California Employment Development Department. Available to the public online, the tracker allows following month by month whether the classic leading indicator of labor market stress - new claims for unemployment insurance - are shifting toward workers in occupations that are more exposed to AI.

So far, we do not see a statewide surge in unemployment claims, consistent with comparable findings at the national level. However, there have been some increases for certain groups (such as college-educated workers in highly AI-exposed occupations) and regions (highly exposed workers in the SF Bay Area) after the release of ChatGPT 3.5 at the end of 2022 - a moment that is often treated as the beginning of widespread use of AI. These are exactly the groups where we would have expected to see initial effects. It is reassuring, however, that there has not been an upward trend in claims for these groups recently.

How should people straight out of college prepare for the AI economy?

Several recent studies find that employment among workers aged roughly 22 to 25 in white-collar occupations highly exposed to AI has grown more slowly or declined since the introduction of ChatGPT. This matters because there is overwhelming evidence that entering the labor market in a weak labor market can lead to earnings losses and other difficulties that persist for a decade or more.

There are several common pieces of advice to young individuals starting to work in a difficult labor market. One is to be flexible - for example, the desired type of work may not be available in the desired location, or vice versa. Another recommendation is to get real work experience early, including through routes people may not always think of initially, such as registered apprenticeships, national service programs like AmeriCorps and the California Service Corps or internships with employers. Finally, in the current environment, it might be a good idea to invest in skills that are complementary to AI, meaning skills that become more valuable as AI improves. This could include being able to identify when AI could be used to streamline work processes (and knowing a subject well enough to catch when AI gets it wrong), as well as critical thinking, which assesses problems worth working on, and developing soft skills, which could include managing teams or advising clients.

For policymakers who are focused on this issue, it's important to keep in mind that many reemployment and training services are provided through the unemployment insurance system. Unfortunately, this may exclude younger workers because unemployment insurance eligibility is contingent on having prior work experience. Building clearer entry points for young workers strikes me as a promising direction.

What should policymakers pay attention to in order to prepare the workforce for changes in the economy catalyzed by AI?

The important infrastructure needed to address potential AI-induced labor market pressures already exists: systems that flag layoffs early; programs that help prevent them; unemployment benefits; and programs that help with finding a new job or retraining. The work now is making sure that these components are modernized to prepare for AI-related job loss. This includes making sure benefits are getting to eligible people, coverage is expanded where needed and ensuring that training programs are offering the types of training that align with the jobs that employers are hiring for.

There are three things I would prioritize. First, expand the Work Sharing program (called Short-Time Compensation in other states), which lets a business cut hours instead of laying people off. This is a win-win program for employers and employees, but it also could reach many more employers than it currently does. Second, I'd suggest more focus on understanding and measuring which workers are missing out on unemployment insurance benefits, and why. Third, there is always room to improve our data infrastructure around unemployment, which would allow for better, more real-time analysis on who is losing work. That would allow for a more informed, timely response.

One last point. Even if there are labor effects from AI, and even if they initially land on high earners first, the evidence from past recessions is overwhelming that a broader downturn hits more vulnerable workers hardest. Recessions slow hiring across the board, and when jobs are scarce, workers with more education tend to have an easier time getting and keeping them, which pushes everyone else further down the ladder.

What should journalists be writing about now to help readers understand AI and its potential future impacts on the labor market and jobs?

Widespread, painful job losses are not inevitable. How the impact of AI on workers plays out depends a great deal on what businesses and policymakers do now, so how to prepare U.S. labor market institutions is itself a story that is worth following. Another important aspect is to better understand how AI is really impacting workers and businesses. The strongest reporting draws on several angles at once: unemployment claims, job postings, what companies say about themselves and what workers report. That is also a good argument for building faster and better data. With so much speculation about the future of work, our aim with the California AI-Unemployment tracker is to provide facts rather than fear, and reporting that separates what we know from what we are guessing.

Von Wachter is professor of economics at UCLA, faculty director of the California Policy Lab's UCLA site and director of the Federal Statistical Research Data Center. His research examines how labor market conditions and institutions affect the well-being of workers and their families.

His research has been supported by the National Science Foundation, the National Institute on Aging, the Social Security Administration, the Smith Richardson Foundation, the Sloan Foundation, the Russell Sage Foundation, the Hilton Foundation and Arnold Ventures.

Von Wachter has been an expert witness in numerous testimonies before committees of U.S. Congress and has provided expert assistance to the City and County of Los Angeles, the state of California, the U.S. Department of Labor, the Canadian Labor Ministry, the OECD, the United Nations and the International Monetary Fund.