Eric Cai
All research

Local Law 97 · Independent empirical research

Above or Below the Line?

Did the first binding year produce new concentration immediately below the reconstructed emissions threshold?

Local Law 97 creates an emissions threshold, but the public benchmarking data do not provide a single audited compliance variable. I reconstruct statutory emissions and an LL84-based counterfactual emissions limit, define the eligible annual analysis samples, and test whether 2024 added new mass immediately below the reconstructed line.

Ongoing · updated September 2026

CompletedThreshold-distribution analysis

In progressEmissions-response analysis

PlannedDenominator-response analysis

No detectable increase in near-threshold mass appeared in 2024 relative to the 2022–2023 pre-binding anchor.

Reconstructing the line

E

Reconstructed statutory emissions

Built from reported energy inputs and applicable statutory emissions factors.

L

LL84-based counterfactual emissions limit

Built from reported property-use areas and the applicable ESPM intensity limits used in the frozen analysis.

R = (E - L) / L0% = reconstructed line
Not official audited DOB filing data. Not a universal representation of every alternative methodology available to every mixed-use property.

The public axis expresses R as percent above or below the reconstructed limit: negative values are below the line; positive values are above it.

Sample construction

Emissions completeness, limit completeness, and structural eligibility are parallel checks within linked observations; their intersection defines the primary sample.

Sample constructionProperty ID × Calendar Year
102,484Benchmarking property-year observations
94,694Linked covered-building observations

7,790 excluded: no link to the defined covered-building universe.

Three parallel checks within linked observations
84,206

Emissions-reconstructable linked observations

10,488 excluded

Reconstructed emissions incomplete under the specified fuel-mapping rules

91,111

Limit-reconstructable linked observations

3,583 excluded

Primary LL84-based counterfactual limit incomplete under the specified property-use rules

58,963

Structurally eligible observations

35,731 excluded

Does not satisfy the strict one-to-one identity definition or fixed Article 320 scope

Intersection of all three checks
Primary threshold-analysis observations56,836

25,053 unique Property IDs

202219,631
202318,693
202418,512

Within the structurally eligible observations, 2,127 were excluded: emissions or primary counterfactual-limit reconstruction is incomplete.

The result represents strict one-to-one LL84 reporting units in fixed Article 320 scope with complete E and L reconstruction; it does not represent the full covered-building universe.

2024 showed no new concentration at the line

Near the reconstructed lineBin mass · % of annual sample

2022

n = 19,631
2022: distribution around the reconstructed limitBars show stored full-year bin mass within minus 10 to plus 10 percent of the reconstructed limit. The reconstructed line is zero percent. The left and right primary five-percent windows are shaded. The vertical scale is identical for all three years.0%0.5%1%Reconstructed line−10%−5%0%+5%+10%

2023

n = 18,693
2023: distribution around the reconstructed limitBars show stored full-year bin mass within minus 10 to plus 10 percent of the reconstructed limit. The reconstructed line is zero percent. The left and right primary five-percent windows are shaded. The vertical scale is identical for all three years.0%0.5%1%Reconstructed line−10%−5%0%+5%+10%

2024

n = 18,512
2024: distribution around the reconstructed limitBars show stored full-year bin mass within minus 10 to plus 10 percent of the reconstructed limit. The reconstructed line is zero percent. The left and right primary five-percent windows are shaded. The vertical scale is identical for all three years.0%0.5%1%Reconstructed line−10%−5%0%+5%+10%

Percent above or below the reconstructed limit

Primary windows: −5% to 0% and 0% to +5%Same axes across years · no smoothing

2024 minus the anchor

Difference in stored bin mass.
2022–2023 anchor, equally weighted.

2024 minus the equally weighted 2022–2023 anchorStored bin-mass differences in percentage points. These descriptive bars do not have bin-level confidence intervals or significance tests.-0.20.0+0.2−10%−5%0%+5%+10%Percentage points
No confidence bands or significance claims for individual bins. Uncertainty belongs to the aggregate contrast.

Each year uses its own eligible annual analysis sample. Year-to-year distribution changes can reflect both within-property changes and changes in sample composition.

Change in near-threshold concentration

-0.210 percentage points

95% CI: [-0.585, +0.151] percentage pointsp = 0.2622

2024 versus the equally weighted 2022–2023 pre-binding anchor

No detectable increase in near-threshold mass appeared in 2024 relative to the 2022–2023 pre-binding anchor. Both anchor years are post-enactment and are not untreated controls.

The share of each annual analysis sample below the reconstructed line was 86.9% in 2022, 90.6% in 2023, and 90.9% in 2024.

Metric details

Near-threshold left-minus-right mass (A_t):

  • 2022: 0.647 percentage points
  • 2023: 0.551 percentage points
  • 2022–2023 anchor: 0.599 percentage points
  • 2024: 0.389 percentage points

Current and planned work

In progress — emissions response

The current analysis asks whether properties facing greater pre-existing threshold exposure experienced different changes in reconstructed statutory emissions from 2023 to 2024.

This is a reduced-form association analysis. It is not presented as a causal regression-kink design.

Planned — denominator response

A later analysis will examine whether changes in reported building characteristics materially alter reconstructed threshold exposure. No result from that analysis is shown yet.

Boundaries

  • The reconstruction is not official audited property-level DOB filing data; mixed-use properties and alternative regulatory methodologies require additional caution.
  • The annual distribution is not a balanced panel; year-to-year shares do not show that the same properties moved below the line.
  • No near-threshold concentration does not imply no behavioral response through other margins.
  • Current evidence does not establish manipulation, strategic reporting, or a causal policy effect.

Aggregate data

The public site may offer the sanitized aggregate near-threshold series and sample-lineage table used to render the figures. Source code and research repositories remain private.

Near-threshold seriesCSV Sample-lineage tableCSV

Other featured research

From Policy Text to Traceable Classification