[codex] Add congressional district outcome percentages

changelog.d/codex-congressional-district-outcome-pct.added

@@ -0,0 +1 @@
+Add winner, loser, and no-change percentages to the congressional district impact output.

src/policyengine/outputs/congressional_district_impact.py

@@ -16,7 +16,8 @@ class CongressionalDistrictImpact(Output):
 
     Groups households by congressional_district_geoid (integer SSDD format
     where SS = state FIPS, DD = district number) and computes weighted
-    average and relative household income changes per district.
+    average and relative household income changes per district, plus the
+    district-level shares of people who are winners, losers, or unchanged.
     """
 
     model_config = ConfigDict(arbitrary_types_allowed=True)
@@ -36,6 +37,11 @@ def run(self) -> None:
         baseline_income = baseline_hh["household_net_income"].values
         reform_income = reform_hh["household_net_income"].values
         weights = baseline_hh["household_weight"].values
+        household_count_people = (
+            baseline_hh["household_count_people"].values
+            if "household_count_people" in baseline_hh.columns
+            else np.ones_like(weights)
+        )
 
         # Only include valid geoids (positive integers)
         unique_geoids = np.unique(geoids[geoids > 0])
@@ -50,6 +56,7 @@ def run(self) -> None:
 
             b_inc = baseline_income[mask]
             r_inc = reform_income[mask]
+            people_weights = household_count_people[mask] * w
 
             weighted_baseline = float((b_inc * w).sum())
             weighted_reform = float((r_inc * w).sum())
@@ -60,6 +67,27 @@ def run(self) -> None:
                 if weighted_baseline != 0
                 else 0.0
             )
+            capped_baseline = np.maximum(b_inc, 1.0)
+            income_change = (r_inc - b_inc) / capped_baseline
+            people_total = float(people_weights.sum())
+
+            if people_total == 0:
+                winner_percentage = 0.0
+                loser_percentage = 0.0
+                no_change_percentage = 1.0
+            else:
+                winner_percentage = float(
+                    people_weights[income_change > 1e-3].sum() / people_total
+                )
+                loser_percentage = float(
+                    people_weights[income_change <= -1e-3].sum() / people_total
+                )
+                no_change_percentage = float(
+                    people_weights[
+                        (income_change > -1e-3) & (income_change <= 1e-3)
+                    ].sum()
+                    / people_total
+                )
 
             geoid_int = int(geoid)
             state_fips = geoid_int // 100
@@ -72,6 +100,9 @@ def run(self) -> None:
                     "district_number": district_number,
                     "average_household_income_change": float(avg_change),
                     "relative_household_income_change": float(rel_change),
+                    "winner_percentage": winner_percentage,
+                    "loser_percentage": loser_percentage,
+                    "no_change_percentage": no_change_percentage,
                     "population": total_weight,
                 }
             )

tests/test_congressional_district_impact.py

@@ -29,13 +29,15 @@ def test_basic_district_grouping():
             "congressional_district_geoid": [101, 101, 202],
             "household_net_income": [50000.0, 60000.0, 40000.0],
             "household_weight": [1.0, 1.0, 1.0],
+            "household_count_people": [2.0, 2.0, 2.0],
         }
     )
     reform = _make_sim(
         {
             "congressional_district_geoid": [101, 101, 202],
             "household_net_income": [52000.0, 62000.0, 42000.0],
             "household_weight": [1.0, 1.0, 1.0],
+            "household_count_people": [2.0, 2.0, 2.0],
         }
     )
 
@@ -52,12 +54,16 @@ def test_basic_district_grouping():
     assert d101["district_number"] == 1
     assert abs(d101["average_household_income_change"] - 2000.0) < 1e-6
     assert d101["population"] == 2.0
+    assert d101["winner_percentage"] == 1.0
+    assert d101["loser_percentage"] == 0.0
+    assert d101["no_change_percentage"] == 0.0
 
     # District 202: baseline = 40000, reform = 42000, change = 2000
     d202 = by_geoid[202]
     assert d202["state_fips"] == 2
     assert d202["district_number"] == 2
     assert abs(d202["average_household_income_change"] - 2000.0) < 1e-6
+    assert d202["winner_percentage"] == 1.0
 
 
 def test_negative_geoid_excluded():
@@ -67,13 +73,15 @@ def test_negative_geoid_excluded():
             "congressional_district_geoid": [-1, 0, 101],
             "household_net_income": [50000.0, 50000.0, 50000.0],
             "household_weight": [1.0, 1.0, 1.0],
+            "household_count_people": [1.0, 1.0, 1.0],
         }
     )
     reform = _make_sim(
         {
             "congressional_district_geoid": [-1, 0, 101],
             "household_net_income": [55000.0, 55000.0, 55000.0],
             "household_weight": [1.0, 1.0, 1.0],
+            "household_count_people": [1.0, 1.0, 1.0],
         }
     )
 
@@ -90,13 +98,15 @@ def test_zero_weight_district_skipped():
             "congressional_district_geoid": [101, 202],
             "household_net_income": [50000.0, 50000.0],
             "household_weight": [1.0, 0.0],
+            "household_count_people": [1.0, 1.0],
         }
     )
     reform = _make_sim(
         {
             "congressional_district_geoid": [101, 202],
             "household_net_income": [55000.0, 55000.0],
             "household_weight": [1.0, 0.0],
+            "household_count_people": [1.0, 1.0],
         }
     )
 
@@ -113,13 +123,15 @@ def test_weighted_average_change():
             "congressional_district_geoid": [101, 101],
             "household_net_income": [40000.0, 80000.0],
             "household_weight": [3.0, 1.0],
+            "household_count_people": [1.0, 1.0],
         }
     )
     reform = _make_sim(
         {
             "congressional_district_geoid": [101, 101],
             "household_net_income": [42000.0, 82000.0],
             "household_weight": [3.0, 1.0],
+            "household_count_people": [1.0, 1.0],
         }
     )
 
@@ -129,3 +141,83 @@ def test_weighted_average_change():
     # Weighted avg change: (3*2000 + 1*2000) / 4 = 2000
     assert abs(d["average_household_income_change"] - 2000.0) < 1e-6
     assert d["population"] == 4.0
+    assert d["winner_percentage"] == 1.0
+
+
+def test_district_winner_loser_percentages_are_people_weighted():
+    """Winner and loser shares should use household_count_people * weight."""
+    baseline = _make_sim(
+        {
+            "congressional_district_geoid": [101, 101, 101],
+            "household_net_income": [1000.0, 1000.0, 1000.0],
+            "household_weight": [1.0, 1.0, 1.0],
+            "household_count_people": [1.0, 3.0, 2.0],
+        }
+    )
+    reform = _make_sim(
+        {
+            "congressional_district_geoid": [101, 101, 101],
+            "household_net_income": [1100.0, 900.0, 1000.0],
+            "household_weight": [1.0, 1.0, 1.0],
+            "household_count_people": [1.0, 3.0, 2.0],
+        }
+    )
+
+    impact = compute_us_congressional_district_impacts(baseline, reform)
+
+    district = impact.district_results[0]
+    assert district["winner_percentage"] == 1 / 6
+    assert district["loser_percentage"] == 3 / 6
+    assert district["no_change_percentage"] == 2 / 6
+
+
+def test_district_small_changes_use_no_change_threshold():
+    """Changes within +/-0.1% should be counted as no change."""
+    baseline = _make_sim(
+        {
+            "congressional_district_geoid": [101, 101, 101],
+            "household_net_income": [1000.0, 1000.0, 1000.0],
+            "household_weight": [1.0, 1.0, 1.0],
+            "household_count_people": [1.0, 1.0, 1.0],
+        }
+    )
+    reform = _make_sim(
+        {
+            "congressional_district_geoid": [101, 101, 101],
+            "household_net_income": [1001.0, 999.5, 990.0],
+            "household_weight": [1.0, 1.0, 1.0],
+            "household_count_people": [1.0, 1.0, 1.0],
+        }
+    )
+
+    impact = compute_us_congressional_district_impacts(baseline, reform)
+
+    district = impact.district_results[0]
+    assert district["winner_percentage"] == 0.0
+    assert district["loser_percentage"] == 1 / 3
+    assert district["no_change_percentage"] == 2 / 3
+
+
+def test_district_outcome_percentages_fall_back_to_household_weights():
+    """Missing household_count_people should default to household-weighted shares."""
+    baseline = _make_sim(
+        {
+            "congressional_district_geoid": [101, 101, 101],
+            "household_net_income": [1000.0, 1000.0, 1000.0],
+            "household_weight": [1.0, 2.0, 1.0],
+        }
+    )
+    reform = _make_sim(
+        {
+            "congressional_district_geoid": [101, 101, 101],
+            "household_net_income": [1100.0, 900.0, 1000.0],
+            "household_weight": [1.0, 2.0, 1.0],
+        }
+    )
+
+    impact = compute_us_congressional_district_impacts(baseline, reform)
+
+    district = impact.district_results[0]
+    assert district["winner_percentage"] == 1 / 4
+    assert district["loser_percentage"] == 2 / 4
+    assert district["no_change_percentage"] == 1 / 4