A viral chart on recent graduate unemployment is misleading

A recent viral chart from the Federal Reserve Bank of New York on the labor market prospects of recent college graduates has sparked a steady flow of in-depth reporting and thinkpieces over the last few months about the future of STEM education, the effects of AI on labor markets, and even high-skilled immigration.

A segment of the chart, which shows estimates of unemployment rates for college graduates aged 22–27 by field of study, is shown below. Elevated unemployment rates for computer engineering (7.5 percent) and computer science majors (6.1 percent) garnered special attention, feeding into growing concerns that majoring in technical STEM fields like computer science no longer leads to better job prospects.

One problem: Under the hood, the data presented in the chart is misleading.

The data underlying the New York Fed’s chart comes from the 2023 American Community Survey (ACS), filtering for respondents aged 22 to 27 with at least a college degree who are not currently attending school. While the ACS is itself a large survey, the sub-samples highlighted in the chart — young college graduates with very specific majors — are quite small. As a result, the confidence intervals around these estimates are huge.

Take the estimated 7.5 percent unemployment rate for computer engineering majors, for example. Using special weights provided by the ACS, we can calculate the confidence interval on this estimate. We find that we can say with 95 percent confidence that the unemployment rate for new computer engineering graduates is somewhere between four percent and 11 percent. For perspective, the most we can feel comfortable saying is that the job market for those graduates looks like something between the absolute height of the Great Recession and the booming job market of 2019, just before the pandemic. That’s it.

In some of the fields with the highest unemployment rates, the confidence bands are even wider. For recent physics majors, we are 95 percent certain that the unemployment rate is somewhere between two percent and 12 percent. For public policy majors, from zero percent to 13 percent.

In most cases, the confidence intervals around these estimates of recent grad unemployment are so large that it makes no sense to use them to make firm conclusions about the returns to particular majors, let alone use them to make policy decisions.

Using broader categories of degree fields (thus a larger sample), we can modestly reduce the uncertainty around estimates of recent graduate unemployment. The ACS buckets six majors under the category of “computer and information sciences.” For recent graduates who majored in these fields, there was a substantial rise in unemployment between 2022 and 2023, from just under four percent to just under six percent.

But there is still a catch here: the unemployment rate is not always the best indicator of labor market slack. In order to be classified as unemployed, respondents need to be in the labor market actively looking for work. For majors with especially poor job prospects, some workers may give up entirely, dropping out of the labor force and mechanically lowering the measured unemployment rate. On the other end of the spectrum, recent graduates from programs that yield solid job prospects may be less disillusioned, continuing to search for work.

An alternative measure, the employment-to-population ratio (EPOP), skirts this problem, simply measuring the share of recent graduates by major who have a job.

For computer and information science graduates, the employment-to-population ratio tells a much more positive story. A full 90 percent of 22- to 27-year-olds who majored in those fields were employed in 2023. That is down more than one percent from 2022, but well above the entire decade leading up to the Covid-19 pandemic and statistically indistinguishable from 2021.

The job market has certainly weakened for these graduates since 2022 (along with the broader labor market), but as of 2023 it remained very strong by historical standards.¹

Unlike the American Community Survey, which only extends through 2023, more recent data cannot produce estimated unemployment by major. Given the surge in layoffs from Big Tech firms in 2023 and 2024, it is likely that the job prospects for young graduates of computer and information science programs have worsened since the last ACS.

We can instead turn to a more timely, related data source, the Current Population Survey, within which we can identify changes in employment among young graduates in computer occupations.² In this data, job growth in computer occupations for college graduates aged 22–27 flatlined for much of 2023 and 2024, albeit with a possible re-acceleration over the last year.

Yet it is worth drawing out the relative differences between young graduate employment in computer occupations and overall young graduate employment. In the five years leading up to the pandemic (2014–2019), young graduate employment in computer occupations grew four times faster than overall young college graduate employment. In the five years since, growth among both groups has slowed, but computer occupation employment has still grown at a rate four times faster than overall employment. As with the ACS data, CPS estimates for such a small segment of the population are highly uncertain, but they are the most timely data we have.

None of the above suggests that a degree in computer science will continue to be a sound investment long into the future. Participation in some computer science programs has dropped substantially over the last few years, perhaps anticipating future labor market shifts that have yet to show up in the data. At the same time, the labor market has been steadily weakening and may weaken further, leaving more jobless graduates of all degree programs.

Still, news of the much-anticipated implosion of the computer science graduate job market is greatly exaggerated, at least for now.

The Github repository for this analysis can be viewed here.

1. The Census Bureau did not release standard one-year American Community Survey estimates for 2020 because of challenges conducting the survey during the pandemic, but did release experimental data, which is included here. A longer explanation from the Census Bureau can be found here.

2. Computer science and information majors account for about 30 percent of workers in computer occupations. Note that this percentage only concierge workers’ undergraduate degree fields.