The Stack Overflow Developer Survey 2024 (n=65,437) found that 66% of professional developers hold a BA/BS or MA/MS degree, while only 49% report having learned to code at school — a 17-point gap that captures the dominant pattern of the modern tech labour market: most developers learn online (82% report using online resources, 85% among professionals) but a substantial majority still ultimately obtain a formal credential. 24% have no degree at all. Combined with the HBS Institute for Business in Global Society finding (February 2025, dataset of 11,300+ job postings 2014-2023) that fewer than 1 in 700 new hires benefited from corporate skills-based hiring reforms, this paints a picture in which the no-degree path remains structurally penalised in hiring even where employers publicly claim otherwise. The 34% action-side regret proxy combines the no-degree share with the behavioural signal that many self-taught entrants later return to formal education — a recoverable but costly correction.
On the inaction side, the Federal Reserve’s 2023 SHED survey (n=11,400) found that 32% of bachelor’s-or-higher holders do not feel their education’s benefits outweighed the costs, and the field-specific regret rate for computer and information sciences graduates is 31%. Combined with the Stack Overflow signal that degree-holders frequently report learning their actual job skills online rather than at school, the 31% inaction-side regret rate captures the share of degree-path developers who experience their formal education as disconnected from the work it was meant to prepare them for. Neither side is a direct self-reported regret measure; both are constructed proxies, and the entry is published with proxy_only: true.
The 0.03 regret_delta places this entry near balance, well within proxy-estimate noise. Read it as “roughly equal regret on both sides” rather than as a confident action-dominates verdict. The decision is strongly domain-limited: it applies primarily to software development, design, marketing, content production, and adjacent fields where portfolios and demonstrated work outweigh credentials. In licensed professions — medicine, law, accounting, nursing, teaching, engineering — the self-taught path is functionally impossible because state licensing requires accredited-programme completion, and the comparison disappears. In skilled trades, apprenticeship and certification consistently beat formal degree on wage, time, and debt measures. The Gilovich long-term inaction-regret pattern would predict the inaction side should dominate at long horizons; that the action side leads modestly in this estimate reflects the visible, concrete, recoverable nature of credential gaps — self-taught developers can return to school — while degree-holders’ tuition and opportunity costs are sunk and irrecoverable.







