One third and twenty-two percentage points are the same finding said two ways. Among respondents who recalled being likely or very likely to stay in academia half a year earlier, 66% became 44% by March 2025 [3], and 22 points off a 66% base is a 33% relative decline [1]. So the third who changed their minds is a third of the academically inclined, not a third of everyone who answered [1]. A further recalled baseline of 93% appears in the reported figures without its March counterpart in the text available [5].
The response rate matters more than the topline share. Raffaella Sadun's team at Harvard Business School, with colleagues at MIT and Monash University, drew from a frame of 2,466 US biomedical labs [7][8], contacted more than 13,000 doctoral students and postdoctoral fellows [9], and received 916 replies from 770 labs at 134 institutions [11]. That is a response rate of about 7% at the most [2]. Only 31% of the labs in the frame produced even one respondent [3], and participating labs averaged roughly 1.2 respondents each [5], so this is a thin scatter across many labs rather than a picture of any lab. Non-response, more than recall, is the live problem: a trainee whose plans had just been upended has more reason to answer six questions about upended plans than one whose plans held.
The recall design, each question asked once for now and once for six months ago [10], is the vulnerability the authors flagged themselves [12], and their check on it is better than most fast-turnaround surveys manage. About a quarter of respondents, roughly 229 people [4], had already answered the PhD satisfaction question in the original lab management study before the end of 2024, ahead of the January 2025 change of administration, and comparing those contemporaneous answers with the recalled ones showed no sign of recall bias, according to the study [13]. That validation covers only the satisfaction item, the one where 17% fewer respondents called themselves satisfied or neutral than six months earlier [6]. The two career-intent items, staying in academia and staying in the country, have no contemporaneous baseline described.
None of this measures anyone leaving. It measures how trainees rated their own likelihoods during a month when funding and international-student policies were being tightened and then blocked in court, at least initially [15]. Sadun says the anxiety arrived unprompted, in interviews that were supposed to be about lab management [14], and she is candid that the instrument is imperfect [17]. Phys.org reports the study's reading that the campaign itself, plus the absence of a final legal resolution, is what fed the uncertainty [16]. For anyone recruiting out of this pool, the usable signal is the speed rather than the level: a large share of these trainees revised a career plan inside six months, and the survey offers no data on whether that revision will be the last one they make.
