Compute President and top-10 compensation CAGRs over a shared ten-year window

The top-10 CAGR was hardcoded to the last five years while the President
CAGR used the full range, so the two cards covered different periods.
Also fix start_comp, which returned the end value.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01X9jG4CUFcH6n4vFvgtwndY

src/admin_analytics/dashboard/pages/about.py

@@ -32,8 +32,8 @@ reports detailed compensation for officers, directors, trustees, and key
 employees of tax-exempt organizations. The University of Delaware (EIN
 516000297) and UD Research Foundation (EIN 516017306) are the filing entities.
 Titles are normalized to canonical roles (President, Provost, VP Finance, etc.)
-using pattern matching. CAGR is computed as compound annual growth rate from
-first to last available year.
+using pattern matching. President and top-10 CAGRs are compound annual growth
+rates over the most recent ten tax years of data.
 
 **Administrative Cost Ratios** use IPEDS finance survey data. "Institutional
 support" is the IPEDS functional expense category that most closely

src/admin_analytics/dashboard/queries.py

@@ -8,6 +8,9 @@ import polars as pl
 from admin_analytics.config import UD_UNITID
 from admin_analytics.irs990.titles import normalize_title
 
+# Compensation CAGRs cover the most recent ten tax years of Schedule J data
+CAGR_WINDOW_YEARS = 10
+
 # Shared CTE for CPI adjustment
 _CPI_CTE = """
 WITH annual_cpi AS (
@@ -125,19 +128,18 @@ def query_aggregate_comp(conn: duckdb.DuckDBPyConnection) -> pl.DataFrame:
     """).pl()
 
 
-def query_aggregate_comp_cagr(conn: duckdb.DuckDBPyConnection) -> dict | None:
-    """CAGR of aggregate Schedule J compensation over the last 5 years of data."""
-    df = query_aggregate_comp(conn)
+def _windowed_cagr(df: pl.DataFrame, value_col: str) -> dict | None:
+    """CAGR over the most recent CAGR_WINDOW_YEARS tax years of ``df``."""
+    if df.height == 0:
+        return None
+    end_year = df["tax_year"][-1]
+    df = df.filter(pl.col("tax_year") > end_year - CAGR_WINDOW_YEARS)
     if df.height < 2:
         return None
 
-    # Use last 5 years of available data
-    df = df.tail(min(5, df.height))
-
     start_year = df["tax_year"][0]
-    end_year = df["tax_year"][-1]
-    start_comp = float(df["total_comp"][0])
-    end_comp = float(df["total_comp"][-1])
+    start_comp = float(df[value_col][0])
+    end_comp = float(df[value_col][-1])
     n_years = end_year - start_year
 
     if n_years <= 0 or start_comp <= 0:
@@ -148,13 +150,18 @@ def query_aggregate_comp_cagr(conn: duckdb.DuckDBPyConnection) -> dict | None:
         "cagr_pct": round(cagr, 1),
         "start_year": start_year,
         "end_year": end_year,
-        "start_comp": int(end_comp),
+        "start_comp": int(start_comp),
         "end_comp": int(end_comp),
     }
 
 
+def query_aggregate_comp_cagr(conn: duckdb.DuckDBPyConnection) -> dict | None:
+    """CAGR of top-10 Schedule J compensation over the last ten tax years."""
+    return _windowed_cagr(query_aggregate_comp(conn), "total_comp")
+
+
 def query_comp_cagr(conn: duckdb.DuckDBPyConnection) -> dict | None:
-    """Annualized growth rate (CAGR) of President compensation.
+    """Annualized growth rate (CAGR) of President compensation over the last ten tax years.
 
     Tracks the President role specifically using title normalization.
     Returns dict with cagr_pct, start_year, end_year, start_comp, end_comp,
@@ -183,27 +190,7 @@ def query_comp_cagr(conn: duckdb.DuckDBPyConnection) -> dict | None:
         .sort("tax_year")
     )
 
-    if df.height < 2:
-        return None
-
-    start_year = df["tax_year"][0]
-    end_year = df["tax_year"][-1]
-    start_comp = df["top_comp"][0]
-    end_comp = df["top_comp"][-1]
-    n_years = end_year - start_year
-
-    if n_years <= 0 or start_comp <= 0:
-        return None
-
-    cagr = ((end_comp / start_comp) ** (1.0 / n_years) - 1) * 100
-
-    return {
-        "cagr_pct": round(cagr, 1),
-        "start_year": start_year,
-        "end_year": end_year,
-        "start_comp": start_comp,
-        "end_comp": end_comp,
-    }
+    return _windowed_cagr(df, "top_comp")
 
 
 def query_top_earners(

tests/test_dashboard_queries.py

@@ -7,6 +7,8 @@ from admin_analytics.dashboard.queries import (
     query_admin_per_student,
     query_admin_faculty_ratio,
     query_top_earners,
+    query_comp_cagr,
+    query_aggregate_comp_cagr,
     query_comp_by_role,
     query_comp_vs_cpi,
     query_staff_composition,
@@ -129,6 +131,23 @@ class TestTopEarners:
         assert df_empty.height == 0
 
 
+class TestCompCagr:
+    def test_uses_last_ten_tax_years(self, db_conn):
+        # 12 years of data; the window should start at 2015, not 2013
+        for year in range(2013, 2025):
+            comp = 1_000_000 if year < 2015 else 500_000 * 2 ** ((year - 2015) / 9)
+            db_conn.execute(
+                "INSERT INTO raw_990_schedule_j (tax_year, person_name, title, total_compensation) "
+                "VALUES (?, 'P', 'PRESIDENT', ?)",
+                [year, round(comp)],
+            )
+        for result in (query_comp_cagr(db_conn), query_aggregate_comp_cagr(db_conn)):
+            assert (result["start_year"], result["end_year"]) == (2015, 2024)
+            assert result["start_comp"] == 500_000
+            assert result["end_comp"] == 1_000_000
+            assert result["cagr_pct"] == 8.0
+
+
 class TestCompByRole:
     def test_groups_by_role(self, db_conn):
         _seed_990(db_conn)