Pre-registration
The US COVID labour-market shock 2020-2024 produced a sharp asymmetric reallocation — leisure/hospitality and brick-and-mortar retail collapsed in 2020 then over-recovered nominal wages relative to trend, while professional/information sectors saw remote-work entrenchment with persistently elevated WFH share and lower CBD office utilisation through 2024.
Falsification criterion — what would disprove this
This hypothesis is considered falsified if:
Multi-metric checklist on US 2020Q1-2024Q4 sectoral data. The hypothesis is SUPPORTED if at least 4 of 5 metrics meet their thresholds; REFUTED if 2 or fewer meet; PARTIAL otherwise. Each metric below is independently sourced (BLS CES, BLS CPS, FRED) and probes a different layer (employment level, wage growth, hours, WFH share, sector-specific JOLTS quits).
formal test & threshold
test: us_post_covid_sectoral_reallocation_2020_2024_multi_metric threshold: SUPPORTED if >=4/5 metrics meet thresholds; REFUTED if <=2/5.
Method
- Template
multi_metric_checklist- Clustering
none- Sample
- 1 countries · 2019 – 2024
- Evidence type
- canonical_case_multi_metric
Asymmetric-reallocation pattern requires multi-metric design rather than single coefficient. Each metric independently sourced and probes a different layer of the post-COVID adjustment.
Data
| Variable | Source | Transform |
|---|---|---|
leisure_hospitality_employment outcome | fred:USLAHtier 1 | level |
information_sector_employment outcome | fred:USINFOtier 1 | level |
leisure_hospitality_ahe outcome | fred:CES7000000003tier 1 | pct_change_yoy |
aggregate_private_ahe outcome | fred:CES0500000003tier 1 | pct_change_yoy |
total_nonfarm_payrolls outcome | fred:PAYEMStier 1 | level |
covid_recession_indicator treatment | constructed:NBER recession 2020-02 to 2020-04, post-recession 2020-05+tier 5 | indicator |
cpi_headline control | fred:CPIAUCSLtier 1 | pct_change_yoy |
real_pce control | fred:PCEC96tier 1 | log |
● ready · ● pending · ● reconstruct-needed
Detailed result card
Result card — post_covid_labour_reallocation_us_2020_2024
Verdict: inconclusive (data gaps)
Reason: 1 metrics met, 3 pending; 3 more need resolution
Pre-registered rule: SUPPORT if >= 4 of 5 metrics met; REFUTE if <= 2 met (impossible to hit support).
Counts: 1 MET · 1 NOT_MET · 1 PENDING_DATA · 2 PENDING_EVAL
Primary country: USA
Metric-by-metric
| # | Metric | Status | Observed | Threshold | Notes |
|---|---|:---:|---:|---|---|
| 1 | leisure_hospitality_employment_collapse_2020 | NOT_MET | 0 (2020) [peak_to_trough_pct_decline] | >35% decline peak-to-trough | |
| 2 | leisure_hospitality_wage_overrecovery_2024 | MET | 112 (2024) [pct_increase_from_baseline] | >5pp cumulative excess over aggregate private AHE 2019-12 to 2024-12 | |
| 3 | information_sector_employment_recovery | PENDING_EVAL | 3.06e+03 (2022) [max_in_window_fallback] | by 2024-12 employment >= 2019-12 level (full recovery or above) | threshold expression unparseable by regex |
| 4 | wfh_share_persistence_2024 | PENDING_DATA | | >25% of full-time employees still hybrid or fully remote in 2024 | No usable vintage for: bls:american_time_use_survey_telework |
| 5 | retail_brick_mortar_employment_shortfall | PENDING_EVAL | 3.24e+03 (2024) [max_in_window_fallback] | by 2024-12 employment <= 95% of 2019-12 level | threshold expression unparseable by regex |
Claim
The US COVID labour-market shock 2020-2024 produced a sharp asymmetric reallocation — leisure/hospitality and brick-and-mortar retail collapsed in 2020 then over-recovered nominal wages relative to trend, while professional/information sectors saw remote-work entrenchment with persistently elevated WFH share and lower CBD office utilisation through 2024.
Interpretation
Verdict is inconclusive (data gaps) — 1 metric(s) cannot be evaluated because the underlying data source is not yet in the vintages pipeline, and 2 metric(s) have data but a threshold expression the auto-evaluator does not recognise (complex conditions, discrete event counts, cross-country gaps). Close these gaps then re-run.
Steelman live concerns
See hypotheses/steelman/post_covid_labour_reallocation_us_2020_2024.md for the strongest opposing arguments. Canonical-case multi-metric evidence is a pattern match, not a causal identification — the result card should be read as 'outcome trajectory matches the predicted pattern to degree X' rather than 'policy P caused the outcome'.
Provenance
Vintages pinned in manifest.yaml. Full per-metric diagnostics in diagnostics.json. Machine-readable results in metric_results.parquet.
Strongest opposing argument
Every hypothesis ships with its charitable opposing argument. The framework earns credibility by handling objections at their strongest, not weakest.
Notes
2020s natural experiment — pandemic shock + structural WFH adoption. Multi-metric to capture asymmetric sectoral pattern; single-coefficient test would mask the reallocation story.