Add V inflation flag to full model yaml

[jvendries]

cc: #461
Closes:

What changed? Why?

The apply_inflation_to_V flag has been added to the 2025 USA Cornerstone full model configuration. This enables inflation adjustment to be applied to the V matrix as part of the model's calculation pipeline.

Testing

Verify that running the 2025 USA Cornerstone full model with this configuration applies inflation adjustments to V as expected and that model outputs reflect the change.

[jvendries]

• Add V inflation flag to full model yaml #460 👈 (View in Graphite)
• Refactor disagg calls into cornerstone_disagg_pipeline.py #453
• main

This stack of pull requests is managed by Graphite. Learn more about stacking.

[WesIngwersen]

Test is fixed but I question now the intent of derive_cornerstone_q_and_vnorm_for_year() when vnorm is not changing from 2017.

[jvendries]

@WesIngwersen This PR adds apply_inflation_to_V flag to 2025_usa_cornerstone_full_model.yaml, but the new default model config is 2025_usa_cornerstone_v0_3. A few questions on this:

1. Is this flag meant to be added to the production pipeline?
2. If so, are the effects described below the intended outcome of applying the flag to the production pipeline?

Effects of setting apply_inflation_to_V: True on 2025_usa_cornerstone_v0_3 config

What the flag does

Inflates the Make matrix V from 2017 chain dollars to model_base_year (2024) by scaling each industry row with the industry price index. That flows into Vnorm and into V-norm-weighted commodity price ratios used in v0.3’s final A/q inflation step.

v0.3 pipeline context (unchanged by the flag)

• model_base_year: 2024, usa_ghg_data_year: 2024
• A/q: CEDA fallback — scale 2017 → 2022, then commodity PI 2017 → 2024 (adjust_summary_A_and_q_dollar_year: True)
• B: (E / x) @ Vnorm with authoritative 2024 x (use_E_data_year_for_x_in_B: True)
• Waste disaggregation: on

Today (flag off): E, x, and final A/q are in 2024 dollars, but Vnorm and commodity-PI weights use 2017 Make-table structure.

---

Impact by output

Output	Changes?	Mechanism
B (direct EFs, D)	Yes — likely largest impact	Vnorm moves from 2017 → 2024 supplier mix; E and x stay 2024
A (Adom, Aimp)	Yes	Final step uses commodity PI, whose weights come from Vnorm
scaled q (published q)	Yes	Same commodity PI path as A
L, M, N	Yes	Downstream of A and B
E	No	Flowsa GHG at 2024
x (B denominator)	No	Authoritative BEA gross output at 2024
y vectors	No	Separate industry PI path
Published V sheet	No	Export uses uninflated derive_cornerstone_V()

v0.3 is more affected than full_model for A/q, because v0.3 uses commodity PI (V-norm-weighted) while full_model uses industry PI for that step.

---

Dollar-year coherence

Component	Flag off (v0.3 today)	Flag on
E, x	2024	2024
Vnorm in B	2017 structure	2024 structure
Commodity PI weights	2017 supplier mix	2024 supplier mix
Final A, q	2024	2024

The flag aligns IO structure in B and commodity pricing with the 2024 model year, instead of mapping 2024 E and x through a 2017 Make table.

---

What would not change

• GHG allocation (E from Flowsa)
• BEA authoritative x for B
• Summary-table structural scaling (2017 → 2022 ratios)
• Other v0.3 flags (CEDA fallback, UMD GHG, waste disagg, update_inflation_factors)
• Exported V (still 2017 $)

---

Operational consequences

• Snapshot tests would fail (B, A, q, L, M, N all move)
• EF diagnostics vs CEDA/USEEIO baselines would shift
• Treat as a methodology change — needs snapshot regen, diagnostics review, and likely a version bump

---

Bottom line

Adding apply_inflation_to_V: True to v0.3 would change B, A, q, and supply-chain matrices (L, M, N) by using 2024 Vnorm and 2024-weighted commodity PI, while leaving E, x, y, and exported V unchanged. It improves dollar-year consistency between emissions/output data and the Make-table structure used in B and A/q pricing, at the cost of a full numerical model shift.

[WesIngwersen]

Thanks @jvendries for that AI-assisted summary. Yes I was looking into it. @bl-young if this is a v0.3 candidate how can we best add a model spec with it to see the effects? (or @jvendries feel free to answer)

[jvendries]

@bl-young if this is a v0.3 candidate how can we best add a model spec with it to see the effects? (or @jvendries feel free to answer)

For testing purposes I would assume it's just a matter of creating a new yaml config based on the v0.3 config and add the apply_inflation_to_V=True flag at the end of the file. @bl-young please correct me if I'm wrong

[bl-young]

For testing purposes I would assume it's just a matter of creating a new yaml config based on the v0.3 config and add the apply_inflation_to_V=True flag at the end of the file.

Yes that seems right. You are getting a conflict now since we changed the name of the model config file in #486

[WesIngwersen]

Ok i vote for adding that new config and then doing diagnostics on it to view it as a potentially v0.3 addition.

[jvendries]

Ok i vote for adding that new config and then doing diagnostics on it to view it as a potentially v0.3 addition.

You mean before it's pulled to main, right?
I can tackle this early next week if that's alright timing wise.