Programmer and developer employment 50% below May 2024
Will the latest BLS OEWS release show combined computer programmer and software developer employment 50% or more below the May 2024 level?
Chance by Q4 2031
2%
Cumulative probability
- Latest forecast
| Year | Q1 | Q2 | Q3 | Q4 |
|---|---|---|---|---|
| 2026 | 0% | |||
| 2027 | 0% | <1% | <1% | <1% |
| 2028 | <1% | <1% | <1% | <1% |
| 2029 | <1% | <1% | <1% | <1% |
| 2030 | <1% | <1% | <1% | <1% |
| 2031 | 1% | 2% | 2% | 2% |
Model reasoning
For this to resolve yes, U.S. programmer and software developer jobs in the government's annual occupational survey would have to halve, to about 882,000. The latest count, for May 2025, was 1.78 million, slightly above the May 2024 baseline. No large occupation has ever halved within six years in this survey, and the survey pools three years of data, which slows any decline. Only the releases from 2028 to 2031 count. The thing that would most change my view is a sudden, economy-wide collapse in tech-industry employment caused by AI.
Where things stand
Baseline and threshold. I downloaded the BLS May 2024 national file (oesm24nat.zip; national_M2024_dl.xlsx is dated 2025-03-11, so it is the original release). Computer Programmers (15-1251) were 109,870 and Software Developers (15-1252) were 1,654,440, so B = 1,764,310 and the YES threshold is ≤ 882,155.
Latest release. The May 2025 estimates came out on 15 May 2026 (USDL-26-0725), about six weeks late because of the shutdown. The national file (oesm25nat.zip) shows 92,230 programmers and 1,687,890 developers. That makes a combined 1,780,120, which is 0.9% above B. YES would need a further fall of about 50.4%.
Which releases count. Under the observation-period rule, the May 2026 estimates (due around spring 2027) do not qualify, because their reference period ends before 16 Sep 2026. In practice it doesn't matter: they will not be anywhere near half. That leaves four qualifying releases in the window:
- May 2027 estimates, out around spring 2028
- May 2028 estimates, out around spring 2029
- May 2029 estimates, out around spring 2030
- May 2030 estimates, out around spring 2031
Recent releases have come out between 31 March and 15 May, so I put about 30% of each year's chance in Q1 and 70% in Q2. Nothing arrives in Q3 or Q4.
Why the survey method slows any fall. Each OEWS estimate pools six survey rounds collected over three years (Handbook of Methods, https://www.bls.gov/opub/hom/oews/calculation.htm). Firms that responded and haven't changed much keep the job mix they reported up to three years earlier. Only total employment at each workplace is refreshed to the current period. So a mid-2028 collapse would show up only partly in the spring-2029 and spring-2030 releases.
Base rate (computed)
I built a panel from the BLS national files for 2000–2025 (oesNNnat.zip / oesmNNnat.zip at bls.gov/oes/special-requests/). Within periods where the occupation codes were consistent, I looked at every occupation with at least 500,000 jobs over horizons of one to six years. None of 4,016 cases halved; for occupations with at least 1 million jobs, none of 1,867 did. The worst fall for programmers plus software engineers combined was −8.6% (2000→2002, the dot-com bust). In the 2008→2010 recession it was −4.6%. The base rate before AI is essentially zero.
Current evidence (as of Oct 2026)
- Industry payrolls (BLS monthly employment survey, pulled through the BLS API):
- Computer systems design: 2,355.4k in Sep 2026, −1.3% on the year and −3.5% since May 2024 (series CES6054150001).
- Software publishers: 652.9k in Aug 2026, +1.5% since May 2024 (CES5051320001).
- Information sector: −6.5% since May 2024.
- Household survey: software developers were 2.254M in 2025, up from 2.085M in 2022 (cpsaat11).
- Job postings: Indeed's software-development postings were up about 15% in June 2026 from early 2025 (Business Insider, Aug 2026).
- Company cuts: some firms made large AI-justified cuts, such as Block (−40%) and WiseTech (about −30%). But Meta dropped its plan to shrink teams by up to 60% (Reuters/TechSpot, 26 Aug 2026), and laid-off engineers are being re-hired elsewhere.
- Forecasters: Metaculus Labor Hub professionals (July 2026) put software-developer employment at −8.1% by 2030 and −13.8% by 2035. The question page shows about −7.4% for 2030 with a range of roughly −17% to +6%. A Manifold market prices a decline of more than 15% from 2023 to 2028 at 42%. A 50% drop is far out in all of these tails.
Paths to YES and to NO
- YES path. In practice the only route is a very fast AI scenario. Software engineering would have to be fully automated by about 2027, with adoption across the economy fast enough that actual employment falls well over 50% by 2029. That is what it would take for a three-year pooled estimate to cross the threshold.
- Size of that path. I put the chance of transformative AI by early 2028 at roughly 10%. Metaculus's median for "weakly general" AI is 2028; for strong AGI it is 2033. Given that, I put the chance of an OEWS reading at or below the threshold by the May 2030 estimates at about 10–20%. Errors and other paths add very little.
- NO paths. Gradual change is the main one. Two others push the same way:
- The 2028 revision of the occupation codes is still unsettled: its second Federal Register notice was still TBD on 30 Sep 2026. If developers are merged with excluded occupations, there is no comparable total, so the question cannot resolve YES.
- BLS budget pressure could cut or delay the survey.
Chance each release crosses the threshold (if none has before): 0.05% for the 2028 release, 0.2% for 2029, 0.45% for 2030 and 0.9% for 2031, plus a negligible background. Cumulative: about 1.6% by end-2031.
Strongest case against my estimate. AI capability is moving fast in 2026: frontier models have been withheld, the government has mandated release delays, and leading figures predict end-to-end automation of software work. If an intelligence explosion happens in 2027, past base rates may not apply. That is why I'm not below 1%. But the lag built into the survey method, together with stable data so far, keeps me well under 5%.
Resolution criteria
Question. By the last day of each quarter, has the latest BLS OEWS release shown combined computer programmer and software developer employment 50% or more below the May 2024 level?
Definitions.
- BLS OEWS release: the U.S. Bureau of Labor Statistics’ annual Occupational Employment and Wage Statistics national cross-industry employment estimates, covering all ownership types combined.
- Combined employment: the sum of the published employment point estimates (
TOT_EMP) for Computer Programmers (2018 SOC 15-1251) and Software Developers (2018 SOC 15-1252). These are OEWS job counts, not employment projections, job postings, full-time equivalents, or counts including the self-employed. Software Quality Assurance Analysts and Testers (15-1253), web developers, and other computer occupations are excluded. - May 2024 level (B): the sum of those two rows in BLS’s May 2024 national estimates, file
national_M2024_dl.xlsxin oesm24nat.zip, using the version publicly available on the opening date. This baseline is fixed; subsequent baseline revisions do not change it. - Latest release: at each qualifying publication date, the release for the most recent OEWS reference period then available. An official correction to that release can qualify; a revision to an older, superseded reference period cannot newly trigger YES.
- 50% or more below: combined employment is less than or equal to 0.50 × B, calculated using the published employment estimates without intermediate rounding or adjustments for sampling error.
- Classification changes: renamed or renumbered occupations count only where BLS’s official classification documentation preserves the same scope. If either occupation is split, sum its exhaustive successor categories without double counting. If occupations are merged with excluded occupations, use only a BLS-published breakdown or comparable estimate that isolates the original combined scope; do not allocate employment using assumed shares.
Resolves YES if a qualifying BLS publication shows the latest release’s combined employment at or below the threshold. The Resolution window and Observation period conventions apply: the publication must be on or after 2026-09-16, and the estimate’s reference period must end after that date. The May 2024 baseline is used only for comparison. Once a qualifying publication meets the threshold, YES persists for every later quarter-end, even if employment subsequently rises or BLS corrects the estimate. No attribution to AI is required.
Resolves NO for a quarter if no qualifying publication has met the threshold by that quarter-end. Declines in only one occupation do not qualify unless the combined total meets the threshold. Forecasts, private estimates, subnational or industry-specific figures, and releases only repeating or revising pre-opening observations do not qualify. Missing, suppressed, or discontinued estimates are not treated as zero; an unavailable comparable combined total cannot establish YES.
Resolving source. BLS only: the annual national OEWS tables and downloadable national files linked from OEWS Tables, including the national data interface at data.bls.gov/oes/, and official BLS corrections. Use the national download if presentations disagree, unless BLS explicitly identifies a corrected figure. This is a narrower source requirement than Public confirmation (A2). Resolve on the qualifying publication date under Resolution date (A3), not the estimate’s reference date.