Event–Crime Association Dashboard

Interactive coefficient translator for the Houston zip code–day study

Scenario calculator

Translate a change in planned-event count into the model-implied conditional association in expected reported crime. All manuscript presets use the verified PPML log-link specification.

Selected model

Event count: 0 → 1

Model 1 — All crime: ln(Event + 1)

Estimated association
+2.67%
Transformed event multiplier2.000
Selected coefficientβ = 0.038
Expected-crime multiplier1.027
Associated expected incident difference
+0.088
Illustrative comparison expected count
3.381
μ0 × expected-crime multiplier
Transformation check
Uses (E1 + 1)/(E0 + 1), never E1/E0.

Calculation trace

Interpretation guardrail

Outcome modelPPML log link
Event exposureln(Event + 1)
Claim typeConditional association
1

Select a model

Choose a coefficient directly verified against Tables 2–3 of the manuscript.

2

Set event counts

The calculator transforms both counts with the regression model's log-plus-one exposure.

3

Read the association

Results describe expected reported crime conditionally; they are neither causal effects nor forecasts.

ZIP-code spatial map

Explore observed crime and planned-event patterns across the 211 ZIP codes in the balanced 2019–2024 panel. This module is descriptive; it does not estimate ZIP-specific effects or imply that higher mapped values were caused by planned events.

All years · 2019–2024
Houston study ZIP-code choroplethZIP Code Tabulation Areas colored by the selected observed metric, with all study-area ZIP boundaries retained.
Legend

Verified coefficient explorer

Browse every coefficient currently supported by the calculator. The explorer is descriptive and does not expand the empirical scope beyond the manuscript presets.

Manuscript estimates
Robust standard errors clustered by zip code.
PresetβSEMeanSource
Descriptive outcome means
Houston zip code–day panel, N = 462,512.
OutcomeObservationsMeanStd. dev.

About the translation

The dashboard converts published coefficients into transparent scenario comparisons without re-estimating the model or changing the paper's substantive claims.

How the calculator works

Planned-event exposure enters the PPML model as ln(Event + 1). The outcome remains a non-negative crime count under a log-link conditional mean. This distinction matters: the result is an association in expected reported crime, not a percentage change in Crime + 1.

  • Event counts must be non-negative whole numbers.
  • The baseline and comparison counts are transformed separately.
  • Reported robust standard errors generate coefficient-based 95% intervals.
  • The incident difference applies the relative association to an illustrative baseline mean.

Exact formulas

ratio = (E₁ + 1) / (E₀ + 1)

crime multiplier = ratioβ

association (%) = 100 × (ratioβ − 1)

incident difference = μ₀ × (ratioβ − 1)

Interpretation boundaries

The tool is interpretive rather than predictive. It does not estimate causal effects, forecast realized crime, evaluate intervention effectiveness, or prescribe police deployment, private-security staffing, or resource allocation. The reported coefficients already reflect observed public- and private-safety responses embedded in the study period.