Publish near-term unemployment outlook from the Okun satellite

economy/index.html

@@ -127,7 +127,7 @@ <h2>What has happened, and what the model sees next.</h2>
         <article class="economy-stat">
           <p class="economy-stat-label mono">REAL-TIME FORECAST RECORD</p>
           <p class="economy-stat-value">1 scored</p>
-          <p class="economy-stat-change">1 archived round</p>
+          <p class="economy-stat-change">2 archived rounds</p>
           <dl>
             <div><dt>Observation</dt><dd>current scorecard</dd></div>
             <div><dt>Vintage</dt><dd>generated scorecard</dd></div>

forecasts/index.html

@@ -76,12 +76,12 @@ <h2>One round in, one result.</h2>
     <div class="prose">
 <!-- scorecard-status:begin -->
       <div class="forecast-summary" aria-label="Current forecast record">
-        <article><strong>1</strong><span>Archived round</span></article>
+        <article><strong>2</strong><span>Archived rounds</span></article>
         <article><strong>1</strong><span>Scored period</span></article>
         <article><strong>0.12pp</strong><span>Latest absolute error</span></article>
       </div>
       <p class="forecast-next">
-        <strong>Next due:</strong> 2026Q2 (UK real GDP, y/y), which lands when the ONS publishes it.
+        <strong>Next due:</strong> 2026Q2 (UK real GDP, y/y, unemployment), which lands when the ONS publishes it.
       </p>
 <!-- scorecard-status:end -->
       <p>
@@ -153,6 +153,16 @@ <h2>What was claimed, and when.</h2>
           <a class="mono" href="/forecasts/rounds/2026-07-21/boe-svar.json" download>Download artifact · <code>2026-07-21/boe-svar.json</code> →</a>
           <a class="mono" href="https://github.com/PolicyEngine/macro/blob/main/forecasts/rounds/2026-07-21/boe-svar.json">Open on GitHub with commit history →</a>
         </article>
+        <article class="forecast-round">
+          <div><span class="mono">2026-07-28</span><strong>okun-unemployment</strong></div>
+          <dl>
+            <div><dt>Data edge</dt><dd>2026Q1</dd></div>
+            <div><dt>Periods</dt><dd>4</dd></div>
+            <div><dt>Scored</dt><dd>0</dd></div>
+          </dl>
+          <a class="mono" href="/forecasts/rounds/2026-07-28/okun-unemployment.json" download>Download artifact · <code>2026-07-28/okun-unemployment.json</code> →</a>
+          <a class="mono" href="https://github.com/PolicyEngine/macro/blob/main/forecasts/rounds/2026-07-28/okun-unemployment.json">Open on GitHub with commit history →</a>
+        </article>
       </div>
 <!-- scorecard-rounds:end -->
       <p>

forecasts/outturns.json

@@ -23,6 +23,12 @@
       "publisher": "ONS",
       "dataset": "D7G7 (CPI 12-month rate), quarterly average",
       "typical_release_lag": "~3 weeks after month end"
+    },
+    "unemployment": {
+      "series": "UK unemployment rate (aged 16+, SA), level",
+      "publisher": "ONS",
+      "dataset": "MGSX, quarterly",
+      "typical_release_lag": "~7 weeks after quarter end (LFS)"
     }
   },
   "observations": [
@@ -35,4 +41,4 @@
       "release_updated": "2026-07-21T23:00:00.000Z"
     }
   ]
-}
+}

forecasts/rounds/2026-07-28/okun-unemployment.json

@@ -0,0 +1,72 @@
+{
+  "schema_version": 1,
+  "round_id": "2026-07-28",
+  "archived_utc": "2026-07-28T08:27:48Z",
+  "model": {
+    "name": "okun-unemployment",
+    "description": "Okun's-law satellite: deterministic mapping of the boe-svar UK GDP growth forecast to the ONS unemployment rate (MGSX). Fit du_t = alpha + beta*g_t + gamma*du_{t-1} by OLS on 1992Q1-2025Q1 with 2020Q1-2021Q2 dummied out (furlough decoupled unemployment from GDP): alpha=0.0611, beta=-0.0474, gamma=0.1871, sigma=0.172pp, n=123, R2=0.368.",
+    "source_pipeline": "PolicyEngine/boe-var-model src/boe_var/okun.py at main (PR #8, merge 302f7ec)",
+    "generated": "2026-07-28"
+  },
+  "information_set": {
+    "data_edge": "2026Q1",
+    "forecast_start": "2026Q2",
+    "first_period": "2026Q2",
+    "last_period": "2027Q1",
+    "gdp_input": "archived round forecasts/rounds/2026-07-21/boe-svar.json (data edge 2026Q1)",
+    "unemployment_input": "ONS MGSX through 2026Q1 (u=5.0, du=-0.2), vintage 2026-07-26"
+  },
+  "units": "percent, unemployment rate level",
+  "variables": [
+    "unemployment"
+  ],
+  "caveats": [
+    "Horizon capped at 4 quarters: the rolling-origin validation (boe-var-model results/okun_validation.json) shows the end-to-end SVAR->Okun forecast beats a random walk at horizons 1-4 ex-COVID (RMSE ratios 0.82-0.99) and is clearly worse beyond 4 quarters.",
+    "The improvement over no-change is not statistically significant at 73 overlapping origins (Diebold-Mariano p >= 0.14).",
+    "Bands carry GDP-forecast uncertainty only, mapped through the deterministic Okun equation; Okun residual uncertainty (sigma 0.17pp per quarter) is NOT sampled, so the stated bands understate total uncertainty."
+  ],
+  "benchmarks": {
+    "random_walk": null,
+    "drift": null,
+    "ar1": null,
+    "official": null
+  },
+  "forecast": {
+    "2026Q2": {
+      "unemployment": {
+        "median": 4.9714,
+        "lo68": 4.9447,
+        "hi68": 4.9976,
+        "lo90": 4.9304,
+        "hi90": 5.0146
+      }
+    },
+    "2026Q3": {
+      "unemployment": {
+        "median": 4.9626,
+        "lo68": 4.8897,
+        "hi68": 5.0333,
+        "lo90": 4.8507,
+        "hi90": 5.0828
+      }
+    },
+    "2026Q4": {
+      "unemployment": {
+        "median": 4.9405,
+        "lo68": 4.8137,
+        "hi68": 5.0744,
+        "lo90": 4.7354,
+        "hi90": 5.1609
+      }
+    },
+    "2027Q1": {
+      "unemployment": {
+        "median": 4.9332,
+        "lo68": 4.7351,
+        "hi68": 5.1395,
+        "lo90": 4.611,
+        "hi90": 5.2693
+      }
+    }
+  }
+}

forecasts/scorecard.json

@@ -1,12 +1,13 @@
 {
   "_comment": "Generated by forecasts/score.py \u2014 do not edit by hand.",
-  "rounds": 1,
+  "rounds": 2,
   "periods_scored": 1,
   "next_period_to_score": "2026Q2",
   "pending_detail": {
     "period": "2026Q2",
     "variables": [
-      "gdp"
+      "gdp",
+      "unemployment"
     ]
   },
   "status": "1 forecast period(s) scored",
@@ -41,6 +42,17 @@
           "coverage_90": 1.0
         }
       }
+    },
+    {
+      "round_id": "2026-07-28",
+      "model": "okun-unemployment",
+      "data_edge": "2026Q1",
+      "archived_utc": "2026-07-28T08:27:48Z",
+      "path": "forecasts/rounds/2026-07-28/okun-unemployment.json",
+      "periods_forecast": 4,
+      "periods_scored": 0,
+      "entries": [],
+      "by_variable": {}
     }
   ]
 }

index.html

@@ -100,7 +100,7 @@ <h2>Latest outturns beside the model's near-term view.</h2>
           <tbody>
           <tr><th scope="row">Real GDP growth, y/y</th><td>0.9% <span class="mono">2026Q1</span></td><td>1.1% <span class="mono">2026Q2</span> · 68% range 0.6%–1.7%</td></tr>
           <tr><th scope="row">CPI inflation, y/y</th><td>2.8% <span class="mono">2026Q2</span></td><td>2.7% <span class="mono">2026Q3</span> · 68% range 1.9%–3.5%</td></tr>
-          <tr><th scope="row">Unemployment rate</th><td>5.0% <span class="mono">2026Q1</span></td><td>not forecast by the suite yet</td></tr>
+          <tr><th scope="row">Unemployment rate</th><td>5.0% <span class="mono">2026Q1</span></td><td>5.0% <span class="mono">2026Q2</span> · 68% range 4.9%–5.0% · <a href="/forecasts/">Okun satellite</a></td></tr>
           </tbody>
         </table>
       </div>

validation/index.html

@@ -638,6 +638,24 @@ <h2>Against a Bank of England paper, and against outturns.</h2>
         inflation, and the single frozen-edge result must not be read as broad
         GDP forecast superiority.
       </p>
+      <p>
+        <strong>Unemployment satellite.</strong> The unemployment rate is not
+        one of the eight VAR variables; the published unemployment outlook is
+        an Okun's-law satellite — a small regression of the quarterly change
+        in the ONS unemployment rate on year-on-year GDP growth (fit
+        1992Q1&ndash;2025Q1 with the furlough quarters 2020Q1&ndash;2021Q2
+        dummied out, since they decouple unemployment from GDP and attenuate
+        the coefficient about threefold). In a 73-origin rolling test fed by
+        the VAR's own GDP forecasts, it beats a no-change benchmark at
+        horizons 1&ndash;4 excluding furlough targets (relative RMSE
+        0.82&ndash;0.99) and is clearly worse beyond four quarters, so the
+        published path is capped at four quarters. The improvement is not
+        statistically significant at 73 overlapping origins, and the bands
+        carry GDP-forecast uncertainty only. Specification, code, and the
+        full skill table:
+        <a href="https://github.com/PolicyEngine/boe-var-model/blob/main/results/okun_validation.json">okun_validation.json</a>
+        in the model repository.
+      </p>
       <figure class="vfig">
         <svg class="vchart" data-chart="svar-fan" viewBox="0 0 760 340" role="img" aria-labelledby="svar-fan-t svar-fan-d">
 <title id="svar-fan-t">boe-svar: out-of-sample forecast fan from the frozen 2024Q2 edge vs ONS outturns</title>