Decide whether and how to target national LIHEAP dollars

[MaxGhenis]

Context

PR #688 adds LIHEAP recipient-household calibration targets, but only for counts. It intentionally does not add a national LIHEAP dollar target.

The current CPS/SPM concept is a pooled annual reported energy-assistance amount (spm_unit_energy_subsidy_reported), while LIHEAP administrative dollars are split across heating, cooling, crisis, and weatherization. Recipient-household counts are a cleaner crosswalk than dollars, so counts were added first.

Problem

We still need a documented decision on whether and how to target LIHEAP dollars nationally.

Right now there is no national LIHEAP dollar target in either:

• the unified policy_data.db calibration pipeline, or
• the legacy EnhancedCPS / build_loss_matrix() path

Options

1. Keep counts only.
   This is the cleanest short-run choice and avoids forcing a weak admin-to-survey dollar crosswalk.

2. Add a pooled national LIHEAP dollar target using total direct LIHEAP benefits.
   This is simple, but conceptually weak because CPS reports one pooled amount while admin LIHEAP splits benefits across components.

3. Add a pooled national LIHEAP dollar target using heating assistance only.
   This may align better with what respondents report, but it is still not a direct concept match.

4. Target both recipient counts and recipient average benefit.
   This is effectively count plus dollars, but frames the objective in a way that may be easier to defend analytically.

5. Keep dollars as a validation metric only.
   Calibrate counts, then compare weighted CPS/ECPS LIHEAP dollars against admin totals without forcing the fit.

6. Build separate LIHEAP component modeling first.
   Model or impute heating, cooling, crisis, and weatherization separately, then target each component. This is the most defensible but highest-effort option.

Recommendation

Near term, either:

• keep counts only, or
• treat dollars as a validation metric only

If we do add a single national dollar target before component modeling, heating-assistance dollars are probably the least-bad pooled target.

Acceptance criteria for follow-up work

• decide which LIHEAP dollar concept to use, if any
• document the concept choice and rationale in repo docs or PR notes
• if implemented, add the target to the unified national target pipeline
• if implemented, ensure the legacy national build_loss_matrix() path consumes the same target from policy_data.db

References

• PR [codex] Add LIHEAP targets to local calibration DB #688
• CPS/SPM variable: spm_unit_energy_subsidy_reported
• LIHEAP admin profiles used for counts in [codex] Add LIHEAP targets to local calibration DB #688