Add structural mortgage interest data support

changelog.d/structural-mortgage-interest.added.md

@@ -0,0 +1 @@
+Convert imputed deductible mortgage interest into structural mortgage balance, interest, and origination-year inputs when the installed `policyengine-us` supports federal MID cap modeling, while preserving total current-law interest deductions via residual investment interest inputs.

policyengine_us_data/datasets/cps/extended_cps.py

@@ -6,16 +6,27 @@
 import pandas as pd
 from policyengine_core.data import Dataset
 
-from policyengine_us_data.datasets.cps.cps import *  # noqa: F403
-from policyengine_us_data.datasets.puf import *  # noqa: F403
+from policyengine_us_data.datasets.cps.cps import CPS, CPS_2024, CPS_2024_Full
+from policyengine_us_data.datasets.puf import PUF, PUF_2024
 from policyengine_us_data.storage import STORAGE_FOLDER
+from policyengine_us_data.utils.mortgage_interest import (
+    STRUCTURAL_MORTGAGE_VARIABLES,
+    convert_mortgage_interest_to_structural_inputs,
+    impute_tax_unit_mortgage_balance_hints,
+)
+from policyengine_us_data.utils.policyengine import has_policyengine_us_variables
 from policyengine_us_data.utils.retirement_limits import (
     get_retirement_limits,
     get_se_pension_limits,
 )
 
 logger = logging.getLogger(__name__)
 
+
+def _supports_structural_mortgage_inputs() -> bool:
+    return has_policyengine_us_variables(*STRUCTURAL_MORTGAGE_VARIABLES)
+
+
 # CPS-only variables that should be QRF-imputed for the PUF clone half
 # instead of naively duplicated from the CPS donor. These are
 # income-correlated variables that exist only in the CPS; demographics,
@@ -445,6 +456,15 @@ def generate(self):
         )
 
         new_data = self._rename_imputed_to_inputs(new_data)
+        if _supports_structural_mortgage_inputs():
+            new_data = impute_tax_unit_mortgage_balance_hints(
+                new_data,
+                self.time_period,
+            )
+            new_data = convert_mortgage_interest_to_structural_inputs(
+                new_data,
+                self.time_period,
+            )
         new_data = self._drop_formula_variables(new_data)
         self.save_dataset(new_data)
 
@@ -472,11 +492,17 @@ def _rename_imputed_to_inputs(cls, data):
     # due to entity shape mismatch.
     _KEEP_FORMULA_VARS = {
         "person_id",
-        "interest_deduction",
         "self_employed_pension_contribution_ald",
         "self_employed_health_insurance_ald",
     }
 
+    @classmethod
+    def _keep_formula_vars(cls):
+        keep = set(cls._KEEP_FORMULA_VARS)
+        if not _supports_structural_mortgage_inputs():
+            keep.add("interest_deduction")
+        return keep
+
     # QRF imputes formula-level variables (e.g. taxable_pension_income)
     # but we must store them under leaf input names so
     # _drop_formula_variables doesn't discard them. The engine then
@@ -526,7 +552,7 @@ def _drop_formula_variables(cls, data):
             if (hasattr(var, "formulas") and len(var.formulas) > 0)
             or getattr(var, "adds", None)
             or getattr(var, "subtracts", None)
-        } - cls._KEEP_FORMULA_VARS
+        } - cls._keep_formula_vars()
         dropped = sorted(set(data.keys()) & formula_vars)
         if dropped:
             logger.info(

policyengine_us_data/datasets/puf/puf.py

@@ -12,6 +12,11 @@
 from policyengine_us_data.datasets.puf.disaggregate_puf import (
     disaggregate_aggregate_records,
 )
+from policyengine_us_data.utils.mortgage_interest import (
+    STRUCTURAL_MORTGAGE_VARIABLES,
+    convert_mortgage_interest_to_structural_inputs,
+)
+from policyengine_us_data.utils.policyengine import has_policyengine_us_variables
 from policyengine_us_data.utils.uprating import (
     create_policyengine_uprating_factors_table,
 )
@@ -643,6 +648,18 @@ def generate(self):
                 self.holder[key] = np.array(self.holder[key]).astype(float)
                 assert not np.isnan(self.holder[key]).any(), f"{key} has NaNs."
 
+        holder_tp = {
+            variable: {self.time_period: values}
+            for variable, values in self.holder.items()
+        }
+        if has_policyengine_us_variables(*STRUCTURAL_MORTGAGE_VARIABLES):
+            holder_tp = convert_mortgage_interest_to_structural_inputs(
+                holder_tp,
+                self.time_period,
+            )
+        self.holder = {
+            variable: values[self.time_period] for variable, values in holder_tp.items()
+        }
         self.save_dataset(self.holder)
 
     def add_tax_unit(self, row, tax_unit_id):