TL;DR
- About 3,300 trick-or-treat-age kids live in Watertown, and on Halloween night they will concentrate on a handful of neighborhoods: the north-side “state avenues” (Colorado, Michigan, Indiana and California Avenues North) lead the forecast.
- The reason is the apartment effect: apartments and multi-unit buildings hold about 42% of the city’s homes but only about 1% of its candy-giving front doors, so a lot of kids walk to the house streets.
- Halloween falls on a Saturday this year. Expect an earlier start and more families travelling across neighborhoods. Last year’s rain and 40 mph gusts made it the worst Halloween evening since at least 2000; one year in three is wet.
The map
Switch between where kids are likely to go and where they live. Blocks with fewer than five homes are not shown.
What changed since 2024
In October 2024 we published a first version of this analysis: 1,103 property addresses grouped into H3 hexagons, with “spillover” to neighbouring blocks within 225 metres. It found the dense blocks but had no estimate of who actually lives there, and it treated an apartment complex as a single address.
This year’s model rebuilds every step:
- Who lives where. Census counts of children by age (ACS 2020–2024) are applied to every home at each neighborhood’s real rate of kids per household. Apartment units are estimated from building size and floors, then calibrated so the city total matches the Census: 13,430 modelled homes against 13,426 counted.
- Where the candy is. Houses with their own front door hand out candy; apartment buildings mostly don’t.
- How kids walk. Instead of straight-line circles, visits travel along Watertown’s actual street and sidewalk network, favouring streets with more front doors and avoiding busy roads.
The busiest streets
The forecast concentrates on the north side. Colorado Avenue North stands out, followed by a cluster of streets with similar scores. The index is relative: it compares streets with each other, not a head count.
Data
| Street | Traffic index |
|---|---|
| Colorado Avenue North | 25.9 |
| North Michigan Avenue | 18.4 |
| Columbia Street | 16.7 |
| McClelland Street | 16.6 |
| Pleasant Street North | 16.1 |
| Weldon Drive | 16.1 |
| Indiana Avenue North | 16.1 |
| California Avenue North | 16 |
| Kieff Drive | 13.6 |
| Colorado Avenue | 13.6 |
The apartment effect
This is the finding that explains the map. Apartments and multi-unit buildings are home to about 42% of Watertown’s households:
Data
| Home type | Housing units |
|---|---|
| Single-family homes | 5,177 |
| Two- and three-family | 2,665 |
| Apartments & multi-unit | 5,588 |
But they account for only about 1% of the doors that hand out candy:
Data
| Home type | Doors |
|---|---|
| Single-family homes | 5,176 |
| Two- and three-family | 1,955 |
| Apartments & multi-unit | 104 |
Kids from apartment neighborhoods don’t stay home; they walk to the nearest streets lined with houses. That is why the busiest blocks are often next to, not inside, the places where most children live. Switch the map between its two layers to see it.
How far kids walk
A typical trick-or-treat visit happens about 630 metres from home, and almost a quarter happen more than a kilometre away, largely kids heading from apartment areas to the house streets.
Data
| Walking distance | Share of visits |
|---|---|
| Under 250 m | 18.1% |
| 250–500 m | 21.5% |
| 500–750 m | 21.2% |
| 750 m–1 km | 16.3% |
| Over 1 km | 22.9% |
Who is out there
The Census tracts covering Watertown are home to roughly 4,100 children aged 3 to 14. The model weights them by how likely they are to trick-or-treat: younger children go with parents on shorter routes, and only some teenagers still go.
Data
| Age | Children |
|---|---|
| 3–4 | 698 |
| 5 | 472 |
| 6–8 | 1,085 |
| 9–11 | 1,083 |
| 12–14 | 756 |
Halloween on a Saturday
October 31 is a Saturday in 2026. No school the next morning usually means an earlier start, longer evenings and more families driving to the neighborhoods they consider the best. The busiest streets on this map are likely to be busier still.
What the weather usually does
Last year’s Halloween was the worst for trick-or-treaters since at least 2000. At Watertown International Airport the evening of October 31, 2025 brought steady rain, sustained winds of about 26 mph with gusts to 43 mph, and a temperature of 40°F that felt like 28°F. It got worse right as it got dark: between 6 and 7 p.m. the rain picked up and the gusts hit 40 mph, and it stayed that way all evening.
Data
| Year | Peak wind (mph) |
|---|---|
| 2025 | 43 |
| 2024 | 32 |
| 2023 | 6 |
| 2022 | 6 |
| 2021 | 30 |
| 2020 | 17 |
| 2019 | 26 |
| 2018 | 22 |
| 2017 | 31 |
| 2016 | 6 |
| 2015 | 5 |
| 2014 | 13 |
| 2013 | 29 |
| 2012 | 6 |
| 2011 | 6 |
| 2010 | 10 |
| 2009 | 33 |
| 2008 | 10 |
| 2007 | 30 |
| 2006 | 24 |
| 2005 | 12 |
| 2004 | 21 |
| 2003 | 7 |
| 2002 | 8 |
| 2001 | 17 |
| 2000 | 7 |
Across the 26 Halloween evenings since 2000 (4 to 9 p.m.):
- Rain or drizzle fell on 8 of them, about one year in three.
- The typical 6 p.m. temperature is around 50°F, but in 6 years it felt like freezing or colder at some point in the evening.
- The sun sets at 5:54 p.m. on October 31 in Watertown, and it is dark by about 6:25. Daylight saving time ends the next morning, so this is the last late sunset of the year.
A reliable forecast for a specific evening only exists about a week ahead. We will update this post with the National Weather Service forecast in the days before Halloween.
Other places to trick-or-treat this year
Several organizations host daytime or early-evening events. Check with the organizer before you go, since details can change.
- Boo at the Zoo, Zoo New York at Thompson Park: October 17–18 and 24–25 (zoo calendar).
- Fort Drum Trunk or Treat, Magrath Sports Complex ball fields: Friday, October 23, 5–8 p.m. Free and open to the general public (Fort Drum MWR).
- VFW Post 1400 Trunk-or-Treat, 231 Bellew Ave: Saturday, October 24, 1–4 p.m. (as listed by local media).
- SPCA of Northern New York Trunk-or-Treat, 25056 Water St: Saturday, October 31, 2–5 p.m. (as listed by local media), which ends before most neighborhood trick-or-treating begins.
What it means for local businesses
Halloween demand in Watertown is not spread evenly; it gathers in a few neighborhoods on one evening. For local shops, cafés, food trucks and anyone selling treats, that suggests a few practical moves: stock up and stay open later if you are near the north-side or east-side clusters, show up where the crowds already are (trunk-or-treats and the busiest streets), and reach the households that hand out candy in those neighborhoods in the week before. Some neighborhood house streets will be handing out treats to hundreds of visiting kids; those households are buying in bulk.
This is the kind of question WBA’s data work answers for any local business: where is the demand, when does it peak, and who is closest to it.
Method and limits
- Weather: hourly observations from Watertown International Airport (station ART) for October 31, 2000–2025, via the Iowa Environmental Mesonet; sunset from standard solar calculations.
- Data: American Community Survey 2020–2024 five-year estimates (tables B09001, B11001, B25001) at the census-tract level; City of Watertown parcel records (property type and lot shape only, no owner information); OpenStreetMap streets, sidewalks and buildings.
- Kids: ages 3–4 weighted 0.8, 5–11 weighted 1.0, 12–14 weighted 0.5, applied at each tract’s rate per household and the city’s 88% occupancy rate.
- Walking: visits are spread across reachable front doors with a gravity model: attraction falls off with walking distance (400 m scale), with a 1.6 km limit; doors on arterial roads count 40%.
- Privacy: results are shown for H3 resolution-10 hexagons (about a block) and only where there are at least five homes.
- Limits: this is a model, not a count. Tract-level Census data is coarser than individual blocks, OpenStreetMap is incomplete in places, and weather, decorations and neighborhood traditions are not in the data. Treat the map as a guide to where it will be busy, not a guarantee.
The 2024 version of this analysis is on LinkedIn.