DWP workplace pension participation: the first pensions claims in the UK scorecard (#98)

app/public/data/lanes.json

@@ -357,6 +357,17 @@
    "running": false,
    "updated": "2026-08-21",
    "note": "5 model claims, 7 Chronicle facts (6 statistical + 1 non-simulated uprated EU-SILC survey input), 6 derived ratios dispositioned; 2 concept-mismatch attachments"
+  },
+  {
+   "id": "dwp-pension-participation",
+   "source": "DWP",
+   "area": "workplace pension participation",
+   "mode": 1,
+   "country": "UK",
+   "stage": "ingested",
+   "running": false,
+   "updated": "2026-08-25",
+   "note": "1377 claims \u2014 the first external validation of any pensions quantity in the UK scorecard (participation among auto-enrolment ELIGIBLE employees, GB, ASHE-derived)"
   }
  ]
 }

data/lanes.json

@@ -357,6 +357,17 @@
    "running": false,
    "updated": "2026-08-21",
    "note": "5 model claims, 7 Chronicle facts (6 statistical + 1 non-simulated uprated EU-SILC survey input), 6 derived ratios dispositioned; 2 concept-mismatch attachments"
+  },
+  {
+   "id": "dwp-pension-participation",
+   "source": "DWP",
+   "area": "workplace pension participation",
+   "mode": 1,
+   "country": "UK",
+   "stage": "ingested",
+   "running": false,
+   "updated": "2026-08-25",
+   "note": "1377 claims \u2014 the first external validation of any pensions quantity in the UK scorecard (participation among auto-enrolment ELIGIBLE employees, GB, ASHE-derived)"
   }
  ]
 }

scorecard_db/build_db.py

@@ -19,6 +19,7 @@
     campaign_us  staged day-1/day-2 campaign results (claim matching)
     uk_externals five UK primary-source families + Chronicle staging
     uk_deductions FRR family
+    dwp_pensions  workplace pension participation (ASHE, GB)
     produce_uk + campaign_uk  archive-resolved UK reckoner attaches
     be_jrc      JRC EUROMOD-BE model claims + honest demo attachments;
                 final so its 2026-08-21 lane update cannot be regressed by
@@ -47,6 +48,7 @@
     ingest_solo,
     ingest_uk_deductions,
     ingest_uk_externals,
+    ingest_dwp_pensions,
     ingest_urban,
     produce_campaign_uk,
 )
@@ -94,6 +96,9 @@ def build(db_path: Path) -> dict:
         ("campaign_us", lambda: ingest_campaign.ingest(db_path)),
         ("uk_externals", lambda: ingest_uk_externals.ingest(db_path)),
         ("uk_deductions", lambda: ingest_uk_deductions.ingest(db_path)),
+        # DWP workplace pension participation (#98): the first external
+        # validation of any pensions quantity in the UK scorecard.
+        ("dwp_pensions", lambda: ingest_dwp_pensions.ingest(db_path)),
         ("produce_uk", lambda: produce_campaign_uk.produce(db_path)),
         (
             "campaign_uk",

scorecard_db/ingest_dwp_pensions.py

@@ -0,0 +1,198 @@
+"""Ingest DWP workplace pension participation (#98).
+
+PolicyEngine-UK models pensions — contributions, their relief, the age
+limit, the salary-sacrifice interaction — and until now not one of the
+15,858 UK external claims said anything about pensions. This is the
+external side: 1,377 participation rates by earnings band, age band and
+region, split public/private/overall, 2009-2025.
+
+Two things ride on every claim because leaving them in a methodology
+note would make the rows uninterpretable:
+
+  denominator   The rate is a share of employees ELIGIBLE for automatic
+                enrolment — an earnings-trigger and age-range definition
+                that has moved over the series. A participation rate
+                whose denominator is unstated says nothing.
+  survey_axis   DWP derives these from ONS ASHE, an employer survey of
+                jobs; the certified PE-UK world is FRS-based. A
+                PE-vs-DWP gap is a survey-population difference before
+                it is an engine question — the same axis the LPC lane
+                (#88) carries.
+
+Geography is GB, not UK: ASHE excludes Northern Ireland, and the
+registry keeps the two apart for the same reason #91 keeps IFS's
+coverage-restricted analyses apart.
+
+    PYTHONPATH=. python -m scorecard_db.ingest_dwp_pensions data/scorecard.db
+"""
+
+from __future__ import annotations
+
+import json
+from pathlib import Path
+
+from .db import LANE_SQL, SCORES_SQL, ScorecardDB
+from .harvest import REPO, finish, require_fields
+from .models import ExternalScore, Metric, TimeBasis, UnitConcept
+from .relationships import uk_relationship
+from .uk_aliases import canon
+
+EXTERNALS = REPO / "data" / "externals"
+
+ADAPTER_SOURCE = "dwp-pension-participation"
+SOURCE = "dwp_pensions"
+LANE_ID = "dwp-pension-participation"
+LANE_UPDATED = "2026-08-25"
+FEED_UPDATED = "2026-08-19"
+LANE_FEED_META = {
+    "source": "DWP",
+    "area": "workplace pension participation",
+    "mode": 1,
+    "country": "UK",
+}
+
+PUBLICATION = {
+    "title": "DWP Workplace pension participation and savings trends: 2009 to 2025",
+    "url": (
+        "https://www.gov.uk/government/statistics/"
+        "workplace-pension-participation-and-savings-trends-2009-to-2025"
+    ),
+    "date": "2026-07-30",
+    "country": "UK",
+}
+
+_KNOWN_FIELDS = frozenset(
+    {
+        "axis",
+        "country",
+        "denominator",
+        "edition",
+        "geography",
+        "metric",
+        "period",
+        "program",
+        "sector",
+        "source",
+        "source_column",
+        "status",
+        "subgroup",
+        "survey_axis",
+        "unit_concept",
+        "value",
+    }
+)
+
+_METRICS = {
+    "participation_rate": (Metric.PARTICIPATION_RATE, UnitConcept.SHARE, "share")
+}
+# the adapter's axis name -> the uk_aliases axis its values are closed on
+_AXES = {"earnings_band": "earnings_band", "age_band": "age_band", "region": "region"}
+
+
+def _load() -> list[dict]:
+    path = EXTERNALS / f"{ADAPTER_SOURCE}.json"
+    if not path.exists():
+        raise FileNotFoundError(
+            f"{path} missing — run sources/dwp-pension-participation/adapter.py first"
+        )
+    rows = json.loads(path.read_text())
+    for row in rows:
+        require_fields(row, _KNOWN_FIELDS, SOURCE)
+        if row["source"] != ADAPTER_SOURCE:
+            raise ValueError(f"{SOURCE}: row source {row['source']!r} is wrong")
+        if row["country"] != "UK":
+            raise ValueError(f"{SOURCE}: row country {row['country']!r} is not 'UK'")
+    return rows
+
+
+def stage() -> tuple[list[ExternalScore], dict]:
+    scores: list[ExternalScore] = []
+    counts: dict[str, int] = {}
+    for row in _load():
+        metric, unit, value_kind = _METRICS[row["metric"]]
+        canon(SOURCE, "unit", row["unit_concept"])
+        axis = row["axis"]
+        if axis not in _AXES:
+            raise ValueError(f"{SOURCE}: unregistered axis {axis!r}")
+        value = float(row["value"])
+        if not 0.0 <= value <= 1.0:
+            raise ValueError(
+                f"{SOURCE}: participation rate {value} is not a share — the "
+                "publication gives fractions, and a percent slipping in here "
+                "would be a hundredfold error nobody could see downstream"
+            )
+        cond = {
+            "country": "UK",
+            # GB, not UK: ASHE excludes Northern Ireland
+            "geography": canon(SOURCE, "geography", row["geography"]),
+            "program": canon(SOURCE, "program", row["program"]),
+            "sector": canon(SOURCE, "sector", row["sector"]),
+            # which axis this row cuts on, so an earnings band and an age
+            # band never collide on one subgroup key
+            "axis": axis,
+            "subgroup": canon(SOURCE, _AXES[axis], row["subgroup"]),
+            # the two facts that make the row interpretable at all
+            "denominator": row["denominator"],
+            "survey_axis": row["survey_axis"],
+            "edition": row["edition"],
+        }
+        scores.append(
+            ExternalScore(
+                source=SOURCE,
+                metric=metric,
+                unit_concept=unit,
+                period=int(row["period"]),
+                # ASHE April reference year, carried as a calendar year
+                time_basis=TimeBasis.ANNUAL,
+                value=value,
+                conditions=cond,
+                calibration_relationship=uk_relationship(SOURCE, metric)[0],
+                source_model="dwp_ashe",
+                source_column=row["source_column"],
+                publication=PUBLICATION,
+                value_kind=value_kind,
+                status=row["status"],
+            )
+        )
+        counts[axis] = counts.get(axis, 0) + 1
+    return finish(scores, SOURCE), counts
+
+
+_EXPECTED = {"earnings_band": 357, "age_band": 459, "region": 561}
+
+
+def ingest(db_path: Path) -> dict:
+    scores, counts = stage()
+    if counts != _EXPECTED:
+        raise ValueError(f"claim accounting drifted: {counts} != {_EXPECTED}")
+    db = ScorecardDB(db_path)
+    rows = [ScorecardDB.score_row(s) for s in scores]
+
+    from .baselines import register_baselines_txn
+    from .ingest_harvest import sync_lane_feed
+
+    with db.conn:
+        db.conn.execute("DELETE FROM external_scores WHERE source = ?", (SOURCE,))
+        db.conn.executemany(SCORES_SQL, rows)
+        register_baselines_txn(db)
+        detail = (
+            f"{len(rows)} claims — the first external validation of any "
+            "pensions quantity in the UK scorecard (participation among "
+            "auto-enrolment ELIGIBLE employees, GB, ASHE-derived)"
+        )
+        db.conn.execute(LANE_SQL, (LANE_ID, "ingested", detail, LANE_UPDATED))
+    sync_lane_feed(
+        db,
+        REPO / "data" / "lanes.json",
+        FEED_UPDATED,
+        lanes={LANE_ID: LANE_FEED_META},
+    )
+    db.close()
+    return {"claims": len(rows), "by_axis": dict(sorted(counts.items()))}
+
+
+if __name__ == "__main__":
+    import sys
+
+    out = Path(sys.argv[1] if len(sys.argv) > 1 else "data/scorecard.db")
+    print(json.dumps(ingest(out), indent=1))

scorecard_db/relationships.py

@@ -222,6 +222,20 @@ def effective_relationship(program, metric):
     "nothing in pe-uk-data consumes them.",
 )
 
+# DWP workplace pension participation (#98). ASHE-derived, like the LPC
+# lane, so the survey difference is the first divergence axis.
+_DWP_PENSIONS_HELD = (
+    CR.HELD_OUT,
+    "DWP's workplace pension participation estimates are derived from ONS "
+    "ASHE, an employer survey of jobs, where the certified "
+    "policyengine-uk world is FRS-based; no pe-uk-data target and no "
+    "policyengine-uk parameter is fitted to a participation rate "
+    "(consumption surfaces read 2026-08-25 at the certified pins). Note "
+    "the engine models pension contributions and their relief but has no "
+    "pension commencement lump sum at all, so the pensions question this "
+    "scorecard can ask is narrower than the pensions system.",
+)
+
 _UKMOD_HELD = (
     CR.HELD_OUT,
     "UKMOD is a peer microsimulation, not a calibration source; no PE UK "
@@ -276,6 +290,8 @@ def uk_relationship(source, metric, program=None, kind=None):
         if program in OBR_CONSUMED_WELFARE_PROGRAMS:
             return _OBR_CONSUMED
         return _OBR_UNCONSUMED
+    if source == "dwp_pensions":
+        return _DWP_PENSIONS_HELD
     if source == "ukmod":
         return _UKMOD_HELD
     if source == "hm_treasury":

scorecard_db/uk_aliases.py

@@ -55,6 +55,21 @@ def _alias(source: str, axis: str, source_value: str, canonical: str) -> None:
         ("dwp_takeup:housing_benefit_pensioners", "obr:housing_benefit_on_jsa"),
         ("dwp_takeup:housing_benefit_pensioners", "ukmod:housing_benefit"),
         ("dwp_takeup:benefit_units", "ukmod:families"),
+        # DWP's auto-enrolment ELIGIBILITY earnings bands are defined by
+        # the enrolment trigger, not by an income distribution, so they
+        # are not UKMOD's quantiles nor anyone's deciles.
+        ("dwp_pensions:gbp_10k_20k", "ukmod:q1"),
+        ("dwp_pensions:gbp_70k_plus", "ukmod:q5"),
+        # DEFERRED, deliberately: DWP's age bands are also not LPC's
+        # minimum-wage age bands (one is an auto-enrolment range, the
+        # other a wage-rate category), but the `lpc` vocabulary is
+        # registered on the #93 branch and not here. Asserting a pair
+        # against a source nobody has registered is the same overclaiming
+        # the review caught on #92 — the ledger would read as if the
+        # distinction had been checked when there is nothing on the other
+        # side to check against. The pairs go in once #93 lands; until
+        # then the distinction lives in prose, in
+        # sources/dwp-pension-participation/source.json.
     }
 )
 
@@ -299,6 +314,48 @@ def _alias(source: str, axis: str, source_value: str, canonical: str) -> None:
     _identity(_src, "unit", ["gbp", "households"])
 
 
+# --- DWP workplace pension participation (#98) -------------------------------
+# Geography is GB, not UK: ASHE excludes Northern Ireland, and the two
+# are kept apart for the same reason #91 keeps IFS's coverage-restricted
+# analyses apart from the UK. The earnings and age bands are the
+# publication's own and are closed per source — DWP's auto-enrolment
+# eligibility bands are not UKMOD's income groups and not LPC's age
+# bands, even where the numbers coincide.
+_identity("dwp_pensions", "geography", ["GB"])
+_identity("dwp_pensions", "program", ["workplace_pension"])
+_identity("dwp_pensions", "sector", ["public", "private", "all"])
+_identity(
+    "dwp_pensions",
+    "earnings_band",
+    [
+        "gbp_10k_20k",
+        "gbp_20k_30k",
+        "gbp_30k_40k",
+        "gbp_40k_50k",
+        "gbp_50k_60k",
+        "gbp_60k_70k",
+        "gbp_70k_plus",
+    ],
+)
+_identity(
+    "dwp_pensions",
+    "age_band",
+    [
+        "age_22_25",
+        "age_26_30",
+        "age_31_35",
+        "age_36_40",
+        "age_41_45",
+        "age_46_50",
+        "age_51_55",
+        "age_56_60",
+        "age_61_65",
+    ],
+)
+_identity("dwp_pensions", "region", _UK_REGIONS)
+_identity("dwp_pensions", "unit", ["share"])
+
+
 def canon(source: str, axis: str, value: str) -> str:
     """Canonical value for (source, axis, source_value); unknown raises."""
     try: