AI was supposed to hit new grads hard. So far, unemployment data says otherwise.
AI was supposed to hit new grads hard. So far, unemployment data says otherwise.
人工智能本应重创应届毕业生，但目前的失业数据却显示并非如此。
Last month, we shared word of a Stanford University study that found entry-level employment in so-called “AI-impacted” occupations lagging well behind that in other fields. Now, a new working paper from economics researchers at Munich’s CESifo finds the opposite, arguing point blank that “there is no evidence of any significant, widespread displacement or reduction in hiring of recent college graduates in absolute or relative levels.”
上个月，我们分享了斯坦福大学的一项研究，该研究发现，在所谓的“受人工智能影响”的职业中，入门级就业岗位远远落后于其他领域。然而，慕尼黑 CESifo 经济研究人员的一份最新工作论文得出了相反的结论，明确指出：“没有证据表明近期大学毕业生的就业受到任何显著、广泛的替代或招聘减少，无论是在绝对水平还是相对水平上。”
In “The Early Impacts of AI on Employment Among Recent College Graduates,” researchers Robert Fairlie and Jane Wu said they decided to focus on recent graduates “because changes in labor demand may first appear through reductions in hiring.” As AI gets good enough to at least perform the “relatively standardized tasks” in many entry-level office jobs, they argue, firms could reduce new hiring for simpler roles rather than laying off more experienced long-term employees.
在《人工智能对应届大学毕业生就业的早期影响》一文中，研究人员罗伯特·费尔利（Robert Fairlie）和简·吴（Jane Wu）表示，他们决定关注应届毕业生，因为“劳动力需求的变化可能首先表现为招聘的减少”。他们认为，随着人工智能的发展，足以胜任许多入门级办公室工作中“相对标准化的任务”，企业可能会减少对简单岗位的招聘，而不是解雇经验更丰富的长期员工。
There’s some reason to believe 2026’s graduating job seekers might be more at risk of AI displacement than those graduating just a year or two prior. The CESifo researchers point to a recent sharp increase in the number of firms “replacing a large number of employee tasks with AI” in a Census survey, as well as broad increases in AI spending per employee and ChatGPT Enterprise token use in the last 2026 months.
有理由相信，2026 年的求职毕业生可能比前一两年毕业的学生面临更大的被人工智能替代的风险。CESifo 的研究人员指出，最近的一项人口普查调查显示，越来越多的企业“用人工智能取代了大量员工任务”，此外，过去 12 个月里，每位员工的人工智能支出和 ChatGPT 企业版的使用量也出现了广泛增长。
Anecdotally, some major names also think that this is the year AI is finally capable enough to start replacing the jobs of some recent graduates. Venture capitalist Marc Andreessen said earlier this year that “AI literally until December [2025] was not actually good enough to do any of the jobs that they’re actually cutting.” And BlackRock CEO Larry Fink said in March that “the speed at which AI is changing” led him to worry that “when this year’s college graduates enter the workforce, we could see the highest unemployment rate among them in years—even without a recession.”
从轶事来看，一些知名人士也认为，今年人工智能终于具备了开始取代部分应届毕业生工作的能力。风险投资家马克·安德森（Marc Andreessen）今年早些时候表示：“直到 2025 年 12 月，人工智能实际上还不足以胜任他们目前正在削减的任何工作。”贝莱德首席执行官拉里·芬克（Larry Fink）在 3 月份也表示，“人工智能变革的速度”让他担心，“当今年的大学毕业生进入职场时，我们可能会看到他们中多年来最高的失业率——即使没有经济衰退。”
Nothing to see here (yet)
目前尚无异常
To determine if those kinds of worries were valid, the CESifo researchers looked at detailed microdata from the US Census’ Current Population Survey to determine unemployment trends among recent college graduates (i.e., Bachelor’s degree recipients 22 to 25 years old who aren’t pursuing higher degrees). Since these unemployment numbers predictably spike as fresh graduates enter the job market in the summer months, the researchers looked at year-over-year and seasonal trends going back to 2022 (which is both the year employment returned to pre-pandemic levels and the year ChatGPT was released).
为了确定这些担忧是否成立，CESifo 的研究人员查看了美国人口普查局“当前人口调查”的详细微观数据，以确定应届大学毕业生（即 22 至 25 岁、未攻读更高学位的学士学位获得者）的失业趋势。由于随着应届毕业生在夏季进入就业市场，这些失业数据通常会出现预期的激增，研究人员查看了自 2022 年以来的同比和季节性趋势（2022 年既是就业恢复到疫情前水平的一年，也是 ChatGPT 发布的一年）。
At a base level, the summer unemployment rate for those young college graduates in 2026 (7.3 percent) was well within the range seen in previous years (from 6.3 percent in 2022 to 7.8 percent in 2024). The 2026 numbers were similarly unremarkable when expanded to include graduates who told the CPS survey they “want a job” even though they weren’t actively looking for one (and thus don’t officially count as part of the official “unemployed” labor force).
从基础层面来看，2026 年这些年轻大学毕业生的夏季失业率（7.3%）完全在往年的范围内（2022 年为 6.3%，2024 年为 7.8%）。当将范围扩大到包括那些在调查中表示“想要工作”但并未积极寻找工作（因此在官方统计中不计入“失业”劳动力）的毕业生时，2026 年的数据同样表现平平。
To test the robustness of these findings, the researchers created statistical tests to compare the recent college graduates both with non-college graduates in the same age range and with older college graduates (aged 30 to 49). They also broke down employment by potential “AI exposure” based on a 2023 study of which job roles AI systems were best equipped for. In almost all of the comparisons, any trend differences between the groups in the 2022 to 2026 time frame studied were not statistically significant.
为了测试这些发现的稳健性，研究人员进行了统计测试，将应届大学毕业生与同年龄段的非大学毕业生以及年龄较大的大学毕业生（30 至 49 岁）进行了比较。他们还根据 2023 年关于人工智能系统最适合哪些工作角色的研究，按潜在的“人工智能暴露度”对就业情况进行了细分。在几乎所有的比较中，在 2022 年至 2026 年的研究时间范围内，各组之间的任何趋势差异在统计学上均不显著。
Overall, the data “tell a consistent story in which unemployment among recent college graduates in summer 2026 was not unusually high relative to earlier summers” across comparison groups, the researchers wrote.
研究人员写道，总的来说，这些数据“讲述了一个一致的故事，即 2026 年夏季应届大学毕业生的失业率相对于往年夏季而言并不异常高”。
So how does this analysis square with the recent Stanford study that found an almost completely opposite result? Well, Stanford’s study was based on payroll data from HR firm ADP, which covers a decent cross-section of the economy but might miss some elements of a wider Census survey. The ADP data also looks at the total supply of jobs in various fields, while the unemployment rate being studied here also takes into account the aggregate demand for those jobs. That demand could easily shift even if the supply of jobs in certain fields starts contracting due to AI.
那么，这项分析如何与斯坦福大学最近得出的几乎完全相反的研究结果相吻合呢？斯坦福大学的研究基于人力资源公司 ADP 的工资单数据，该数据涵盖了经济中相当一部分横截面，但可能遗漏了更广泛的人口普查调查中的某些要素。ADP 数据还着眼于各领域的工作总供给，而此处研究的失业率还考虑了这些工作的总需求。即使某些领域的工作供给因人工智能而开始萎缩，这种需求也可能轻易发生转移。
Overall, the CESifo researchers conclude that the Summer 2026 unemployment data serves as a “useful first test” of how accelerating AI usage is—or, as the data show, is not—impacting the current US job market. But current trends do not imply future performance, of course, and the researchers warn that “if the intensity of AI use in the workplace continues to increase, the graduating classes of 2027 and later might be more affected than the class of 2026, and additional years of data will be needed to hone in on whether effects emerge as workplace use of AI deepens.”
总的来说，CESifo 的研究人员得出结论，2026 年夏季的失业数据是衡量人工智能使用加速如何影响（或者如数据所示，并未影响）当前美国就业市场的一个“有用的初步测试”。当然，当前的趋势并不代表未来的表现，研究人员警告说：“如果工作场所中人工智能的使用强度继续增加，2027 年及以后的毕业生可能比 2026 届毕业生受到更大的影响，我们需要更多年份的数据来确定随着人工智能在工作场所的深入应用，是否会出现相关影响。”