NYC comparator lab
Monthly data through mid-2026

Are the “Mamdani effects” actually unusual?

Test prominent claims against contemporaneous peer cities—or swap in a peer as a placebo. Every tab uses an inclusive comparator panel by default, preserves the long pre-period, and rebuilds the estimate when you change the focus or comparison group.

15outcomes with comparison panels
57places available as a focus
6+ yearsminimum history
Homicide rate

Did NYC’s monthly homicide rate change unusually after the new administration took office?

30 homicide-reporting cities · Jan 2018–Jun 2026
Difference-in-changes estimate

NYC relative to comparator cities

29 of 29
Latest estimate · Jun 2026−0.0495% CI −0.50 to +0.41 per 100k
Post-event months outside zero0 / 6Pointwise 95% intervals
Historical window8+ yearsJan 2018 to Jun 2026
Estimator
Pretrend
Chart window
Parallel-trends diagnostic · 12 monthsNo clear linear pretrendLinear slope with Newey–West uncertainty; diagnostics have limited power.
Monthly slope−0.005 per 100k95% CI −0.020 to +0.009
Implied 12-month drift−0.07 per 100kPre-event linear component
Pretrend fitR² 2%Share explained by a line
Window sensitivity6m · no clear slope12m · no clear slope24m · clear slope
NYC relative to comparator cities · conventional DiDNYC relative to selected comparators, with the pre-event reference month fixed at zero and pointwise 95 percent confidence intervals.-1.8-1.2-0.50.10.8Mamdani takes officet−1 = 0-48-36-24-120Months relative to Jan 2026NYC minus comparator change · homicides per 100k/month
Conventional estimate 95% CI Mamdani takes officeNegative = New York’s homicide rate fell more than the selected panel.
Placebo explorer

Swap the focus city

Keep the NYC event date fixed, but treat another place as the focal city and compare it with every eligible peer.

Primary analysis15 of 15 outcomes available
What this tab estimates

Context first, attribution second

Conventional DiD. Rebase New York and every peer to Dec 2025, then subtract the equal-weight peer change from New York’s change.

Pretrend adjustment. Choose 6, 12, or 24 pre-event months; 12 is the default. The fitted drift excludes the normalized baseline and preserves t−1 = 0. Short windows are more local but lower-powered.

Uncertainty. Conventional bands use cross-peer variation plus Poisson count variance. Adjusted bands approximately add Newey–West slope uncertainty in quadrature; all intervals are pointwise.

Policy timing

All 29 other cities with complete homicide data enter the inclusive default. Six cities in CCJ’s broader report lack a usable homicide series.