NYC 311 Complaints — Manhattan
Manhattan Dog Waste Analysis
2026-05-14
Manhattan Has a Dog Waste Problem.
Here are Ways to Reduce it.
We analyzed 28 months of 311 complaints to find out where the problem is at its worst, why it keeps getting worse, and what the city can actually do about it.
Dog Waste Is a Public Health Problem — Not Just an Aesthetic One
Dog waste on city streets carries serious pathogens, including Escherichia coli, Salmonella, roundworm (Toxocara canis), hookworm, and Giardia, disease-causing organisms that can survive in soil for months or even years. Young children and those with weakened immune systems face the greatest risk from direct contact with contaminated ground. When it rains, fecal matter washes into storm drains and flows into the East and Hudson Rivers, adding to contamination in local waterways.
The consequences extend beyond direct contact with waste. Dog waste left on the street may serve as a food source for rats. This report finds a correlated pattern in the 311 data: community districts with high rates of dog waste complaints also consistently show high rates of rat sighting complaints.
Manhattanites clearly feel this. The share of 311 activity made up by dog waste complaints has been rising gradually since mid-2022. This past winter was the worst on record for dog waste complaints.
So We Decided to Find Out What’s Actually Going On.
You can’t fix what you can’t measure. That’s why we collaborated with BetaNYC to pull every 311 complaint filed in Manhattan from January 2024 through April 2026, a total of 1,704,389 complaints, and combined it with a six-year historical view extending back to January 2020.
We mapped where complaints cluster, tracked how they rise and fall with the seasons, compared all 12 Manhattan community districts against each other, and layered in city infrastructure data: litter baskets, bag dispensers, dog runs, and Business Improvement Districts. We also looked at the Bronx, where a different pattern in the dispenser data opens a window into whether free bags actually change behavior.
The result is likely the most detailed block-level picture of Manhattan’s dog waste problem that has ever been assembled in one place and shared with the public.
Want to See How Your Block Shapes Up?
BetaNYC’s interactive map shows the blocks with the three highest counts of dog-waste service requests from January 1, 2024 to 2026 year to date per Community District. The map is zoomable and pannable, with toggleable layers for litter baskets and privately-owned public spaces (POPs), whose litter baskets are not part of the DSNY system. Community district boundaries and dog runs are also shown on the map. The statistical analysis in this report draws on the same underlying 311 dataset. BetaNYC’s original analysis can be viewed here.
What This Data Actually Measures
Calling 311 to report dog waste requires a specific kind of motivation. A person has to notice the problem, feel enough frustration or civic concern to act on it, know that 311 exists and how to use it, and believe the call will make a difference. In practice, most people who step around dog waste on the sidewalk do not report it. The decision to call is shaped by how bothered someone is, how familiar they are with city services, and how much they trust that reporting will lead to anything. These factors vary widely across neighborhoods, demographics, and individual temperament.
This means that a high volume of dog waste complaints in a given area may reflect not only more waste on the ground, but also a more engaged or more frustrated resident population that is more likely to reach for their phone. Conversely, low complaint rates in some areas may not mean those streets are cleaner. They may mean residents there are less likely to report. The 311 dataset is the best available proxy for where dog waste is a problem, but it can also be read, in part, as a map of who reports.
There is a substantial body of academic research examining whether 311 complaint data tracks real-world conditions or primarily reflects reporting behavior, and the answer, as best as researchers have been able to determine, is some combination of both. Unpacking that question for Manhattan dog waste specifically would require independent on-the-ground measurement that falls outside the scope of this study. What this report can say is that the patterns in the data are consistent, geographically coherent, and corroborated by infrastructure and demographic factors that independently predict where complaints are likely to be high. They should be read as strong signals, with the understanding that the underlying reality in any given neighborhood may be more or less acute than the complaint count alone suggests.
Here’s What the Data Shows
Manhattan filed 1,704,389 311 complaints between January 2024 and April 2026. Of these, 1,694 (0.10%) concerned dog waste. That figure is almost certainly an undercount: the city routes these complaints through two separate agencies, and not every incident gets reported to 311 at all.
(0.10% of all 311)
(February 2026)
(Feb–Mar peak vs. Oct–Nov trough)
(0.24% of its 311 calls)
(est. 200 installed 2018)
Key Findings
- How many people actually call. Of 1,704,389 Manhattan 311 complaints filed over this period, 1,694 concerned dog waste — 0.10% of all activity. That’s a small fraction overall, but a persistent one that shows up consistently across every neighborhood and every season. And it’s almost certainly an undercount: most people who step around a pile on the sidewalk don’t reach for their phone.
- Where in Manhattan the problem is worst. CD 12 (Washington Heights and Inwood) stands out sharply, with dog waste making up about 0.24% of all its 311 activity — 17.3× higher than the best-performing district. Statistical testing confirms this is a genuine outlier — the gap is too large to be explained by the district’s overall complaint volume or by chance. In fact, the top 10 blocks in CD 12 together represent roughly 25% of the district’s total complaint volume. The north–south divide holds across the full borough.
- The seasonal pattern. The February–March peak and October–November trough recur across all six years of data. The 11.4× figure for this period reflects the February 2026 spike. February 2026 was the highest-proportion dog waste complaint month in six years, roughly double that of February 2025. A single day that month set the all-time record.
- Litter baskets. Blocks with few or no public litter baskets show significantly higher dog waste complaint rates. When there’s nowhere nearby to throw it away, some residents leave it on the street. This held up even after accounting for gaps in the basket inventory data near parks.
- Bag dispensers. Of approximately 200 waste bag dispenser units installed in Manhattan in 2018, only 25 now appear in the NYC Parks dispenser dataset, which lists only currently active units; decommissioned ones are removed from the public record entirely, so these are the ones we can confirm are still standing. The program has effectively collapsed outside CD 3. CD 12, the highest-complaint district in the borough, has no working dispensers at all.
- BIDs and supplemental cleaning. Manhattan’s ~25 Business Improvement Districts fund street cleaning on top of baseline city services, but BID-covered blocks do not show consistently lower dog waste complaint rates once neighborhood effects are accounted for. Supplemental cleaning alone doesn’t move the needle — the data points toward infrastructure investment as the more effective solution.
- Dog waste and rats. Areas with high rates of dog waste complaints also tend to have high rates of rat sighting complaints. This connection holds at both the community district level and the neighborhood level, and actually grows stronger once the single most extreme dog waste outlier is set aside.
Top Recommendations
- Targeted basket installation on specific high-complaint, low-coverage blocks in the districts where the data shows persistent concentration.
- Waste bag dispenser expansion into uptown Manhattan, where coverage is effectively nonexistent despite the area generating the borough’s highest complaint volumes.
- Front-loaded enforcement and outreach timed to begin in November or December, ahead of the late-winter peak, rather than ramping up only after complaints have already risen.
The Analysis
Complaint Categories & Volume in Context
The table below groups all 311 complaints by type. Because 311 callers
and the agencies that process complaints do not always use a single
consistent label for the same underlying issue, the dog waste category
is defined by a combination of two fields: complaint_type
and descriptor, rather than a single field alone.
Dog waste complaints combine two types that together capture the full signal: complaints filed as Dirty Condition / Dog Waste (handled by the Department of Sanitation, typically on sidewalks and streets) and complaints filed as Animal in a Park / Animal Waste (handled by the Department of Parks and Recreation). Without combining both, the true volume of dog waste complaints would be understated.
This combined categorization provides a more accurate picture of issue volume than relying on any single label, but it is still incomplete. Not all dog waste incidents are reported to 311, and some may be filed under complaint types not captured here.
| Category | Complaints | % of Total |
|---|---|---|
| Other | 1,690,371 | 99.18% |
| Rat Sighting | 12,324 | 0.72% |
| Dog Waste | 1,694 | 0.10% |
Dog waste complaints represent approximately 0.10% of all Manhattan 311 requests, a small fraction of overall activity that nonetheless carries disproportionate weight for residents’ daily quality of life.
The table below shows the top 20 complaint types by raw count. It is
useful for context, but two features make it an incomplete picture on
its own: (1) Similar issues appear as separate rows.
Dog waste complaints, for example, are split across at least two
complaint_type values, so neither row alone reflects the
full signal. (2) Broad categories dilute the signal. A
complaint type like Dirty Condition encompasses many different
underlying issues — dog waste, illegal dumping, overflowing litter
baskets, and more — so its count overstates the prevalence of any one
specific problem within it.
| Complaint Type | Count | % of Total |
|---|---|---|
| HEAT/HOT WATER | 180,692 | 10.60% |
| Illegal Parking | 159,020 | 9.33% |
| Noise - Residential | 151,224 | 8.87% |
| Noise - Street/Sidewalk | 115,287 | 6.76% |
| Encampment | 72,976 | 4.28% |
| UNSANITARY CONDITION | 56,998 | 3.34% |
| Homeless Person Assistance | 55,494 | 3.26% |
| Noise - Commercial | 50,035 | 2.94% |
| Noise | 49,082 | 2.88% |
| Vendor Enforcement | 39,082 | 2.29% |
| PLUMBING | 36,562 | 2.15% |
| Street Condition | 33,157 | 1.95% |
| Noise - Helicopter | 31,456 | 1.85% |
| PAINT/PLASTER | 31,256 | 1.83% |
| Dirty Condition | 29,354 | 1.72% |
| For Hire Vehicle Complaint | 28,445 | 1.67% |
| Water System | 25,742 | 1.51% |
| Noise - Vehicle | 24,093 | 1.41% |
| DOOR/WINDOW | 23,772 | 1.39% |
| Traffic Signal Condition | 21,378 | 1.25% |
Key finding: Every number in this report is almost certainly an undercount. Most people who encounter dog waste on the sidewalk don’t report it. The ones who do are a self-selected group: more frustrated, more familiar with city services, more likely to believe reporting will lead to something. That means the true scale of this problem is larger than what the 311 data captures, and the neighborhoods that look worst in this analysis may actually be underrepresenting the gap. The best thing a resident can do to help fix this (beyond cleaning up after their own dog) is call 311 when they see a problem. More reports means better data, and better data means more targeted resources.
If you see dog waste on your block: call 311. Every additional report helps close the gap between what the 311 data shows and what is actually on the ground. Future iterations of this analysis should also explore ways to measure the problem that don’t rely solely on 311, including field audits, observation studies, or community surveys that could help verify where the problem is genuinely worst.
Trends Over Time
Monthly dog waste complaint volumes from January 2024 through April 2026 show clear seasonal patterns. The first chart below tracks raw monthly counts. The second shows dog waste complaints as a share of all Manhattan 311 activity, with a smoothed trend line overlaid. The trend line uses a method called LOESS (locally weighted regression), which fits a flexible curve through the data to show the underlying direction while filtering out the noise of individual months rising and falling.
Historical seasonal trend (Jan 2020 – present)
The charts above cover January 2024 through April 2026 — the window for which full individual complaint records are available. To place that period in longer context, the chart below extends the view back to January 2020, using monthly totals drawn from the same NYC Open Data source. This six-year picture reveals how the 2025–26 winter compares against prior years, and whether the trend in dog waste complaints has been growing, flat, or declining over time.
February 2026 is the highest-proportion month in six years. The COVID-19 period (shaded) is visible as a suppressed baseline: restrictions reduced overall 311 activity but appear to have reduced dog waste complaints proportionally less, leaving a modest upward blip in the share. The smoothed trend line shows a gradual upward drift in the dog waste share since mid-2022, punctuated by the February 2026 spike.
February 2026 stands as the most extreme single month in the data: complaint volume ran approximately double the same month in 2025, with one day in particular recording the highest single-day total in the dataset. The spike is visible as the dark-red bar in the chart above, well clear of the LOESS trend line.
Notably, this spike coincides with significant winter weather. It is worth studying whether complaint surges in colder months are partly explained by the dynamics of snow: waste that accumulates beneath it becomes newly visible during the thaw, and residents may also be less motivated to clean up after their dogs when the street is already buried. Future analyses could examine whether spikes cluster in areas with less consistent snow removal or at corners where snow tends to pile, which would have implications for how the city coordinates sanitation and snow operations.
Comparing the December–April windows across years shows that the elevation of complaints in January through March is a consistent seasonal pattern, not a 2026 anomaly. The 2025–26 winter tracked above the prior year in nearly every month, with February as the most pronounced gap.
February 2026 was the worst month on record — roughly double the same month in 2025. One day in February 2026 set the all-time single-day record for dog waste complaints in Manhattan. The 2025–26 winter tracked above every prior year in nearly every month from December onward.
Key finding: Complaints are strongly seasonal, peaking in late winter and early spring (February–March) and bottoming out in fall (October–November), a typical swing of roughly 2.9×. February 2026 drove the measured ratio for this period to 11.4× and broke the six-year record by a wide margin. Enforcement and outreach resources should be scheduled to ramp up in November or December, ahead of the late-winter peak.
Schedule enforcement campaigns, outreach, and basket servicing to ramp up in November or December, before the late-winter spike, not after. Use the six-year smoothed trend to monitor whether the dog waste problem is growing relative to all other 311 activity, and calibrate resource levels accordingly.
Community District Analysis
To compare dog waste complaint activity across Manhattan’s 12 community districts, the chart below shows dog waste complaints as a proportion of each district’s total 311 complaints rather than as a raw count. Using proportions controls for the fact that some community districts simply generate more 311 activity overall. A district with more total complaints would appear to have more dog waste complaints even if residents there are no more likely to report them.
The connected dot chart below orders all 12 community boards from south to north, making the uptown–downtown divide visible at a glance. Each point is color-coded by its rank-based grade relative to other Manhattan districts.
The table below shows the raw numbers behind each community board’s grade: total 311 complaints, dog waste complaints, and the proportion used to determine the ranking.
| Community Board | Total Complaints | Dog Waste Complaints | Grade |
|---|---|---|---|
| 1 | 80,045 | 36 | A |
| 2 | 114,695 | 35 | A+ |
| 3 | 147,357 | 81 | A |
| 4 | 130,660 | 86 | B |
| 5 | 134,765 | 19 | A+ |
| 6 | 90,199 | 54 | B |
| 7 | 155,691 | 129 | C |
| 8 | 113,665 | 113 | C |
| 9 | 131,903 | 97 | B |
| 10 | 193,220 | 209 | C |
| 11 | 116,335 | 144 | C |
| 12 | 256,141 | 626 | D |
The map below shows the same data geographically, with each community board shaded by its grade. Hover over any district to see its dog waste proportion and ranking.
Regression-based anomaly detection
A linear regression draws a line through the data representing what each district’s dog waste complaint count would be, given its total 311 volume, if all districts behaved typically. Districts that fall above that line generate more dog waste complaints than their overall 311 activity would predict. The gap between what a district actually generated and what the model predicted is called the residual: a larger residual means a genuine dog waste concentration, not just a high-volume district.
| Community Board | Total Complaints | Dog Waste (actual) | Dog Waste (expected) | Residual | vs. Expected |
|---|---|---|---|---|---|
| 12 | 256,141 | 626 | 490 | 136 | Above |
| 1 | 80,045 | 36 | -41 | 77 | Above |
| 11 | 116,335 | 144 | 68 | 76 | Above |
| 6 | 90,199 | 54 | -11 | 65 | Above |
| 8 | 113,665 | 113 | 60 | 53 | Above |
| 9 | 131,903 | 97 | 115 | -18 | Below |
| 4 | 130,660 | 86 | 111 | -25 | Below |
| 2 | 114,695 | 35 | 63 | -28 | Below |
| 7 | 155,691 | 129 | 187 | -58 | Below |
| 3 | 147,357 | 81 | 162 | -81 | Below |
| 10 | 193,220 | 209 | 300 | -91 | Below |
| 5 | 134,765 | 19 | 124 | -105 | Below |
Adding 95% confidence interval bands goes one step further: these bands show the range of outcomes we’d expect from normal variation alone. Districts that fall outside the bands are outliers too large to be explained by chance. Their excess is a genuine signal, not noise. Under this framework, CD 12 (Inwood / Washington Heights) sits above the upper band. CDs 2, 3, and 5 fall below the lower band. The remaining districts sit within the expected range.
Dog waste index: CD-level proportion vs. city-wide baseline
For additional context, each community district’s dog waste complaint proportion can also be compared against the five-borough city-wide rate. An index above 1.0 means the district has a higher share of dog waste complaints than the city-wide average, though the more striking gaps are within Manhattan itself.
Index = (CD dog waste proportion) ÷ (city-wide dog waste proportion)
| Community Board | CB Proportion | City-wide Proportion | Index |
|---|---|---|---|
| 1 | 0.04% | 0.09% | 0.48 |
| 2 | 0.03% | 0.09% | 0.32 |
| 3 | 0.05% | 0.09% | 0.58 |
| 4 | 0.07% | 0.09% | 0.70 |
| 5 | 0.01% | 0.09% | 0.15 |
| 6 | 0.06% | 0.09% | 0.64 |
| 7 | 0.08% | 0.09% | 0.88 |
| 8 | 0.10% | 0.09% | 1.06 |
| 9 | 0.07% | 0.09% | 0.78 |
| 10 | 0.11% | 0.09% | 1.15 |
| 11 | 0.12% | 0.09% | 1.31 |
| 12 | 0.24% | 0.09% | 2.60 |
Block-level hotspot analysis
The community board comparisons above treat each district as a single unit. This section zooms in on CD 12 — the borough’s highest-complaint district — to ask where within that district the problem is concentrated. If complaints cluster on a small number of specific blocks, targeted action there would have an outsized effect on district-wide numbers.
Within CD 12, complaints are not evenly distributed across the district. Matching each complaint to its nearest street block reveals a sharp concentration: Riverside Drive between West 158th and 160th Streets accounts for 360 complaints between 2020 and 2026 (per BetaNYC’s six-year extract; the main statistical analysis in this report covers January 2024–April 2026), and the top 10 blocks in the district together represent approximately 25% of all its complaint volume. This concentration suggests that a small number of targeted infrastructure improvements — additional baskets or, given that CD 12 currently has no waste bag dispensers at all, first-ever dispenser installations on specific blocks — could have an outsized effect on district-wide numbers.
Key finding: 1 district, CD 12, earns a D, and the regression confirms it generates more complaints than its overall 311 activity would predict. This is a genuine outlier, not an artifact of high overall volume. CD 12 is statistically significant above the upper confidence band; CDs 2, 3, and 5 are significantly below.
Target infrastructure investment and community outreach at C- and D-grade districts first. Use the regression residual rankings to sequence intervention across all 12 districts, moving from the largest excess downward.
Infrastructure Analysis: Litter Baskets
Two kinds of infrastructure are relevant so that residents have a way to clean up after their dogs: public litter baskets and waste bag dispensers. Both are lacking in the neighborhoods that need them most.
Basket density vs. dog waste proportion
The two maps below share the same grid. On the left, each cell is shaded by the number of DSNY litter baskets it contains. On the right, each cell is shaded by its dog waste complaint proportion; cells in the low-basket-density quartile are outlined in blue so you can visually compare the two layers.
| Basket density tier | Grid cells | Total complaints | Dog waste complaints | Dog waste % | Index vs. Manhattan avg |
|---|---|---|---|---|---|
| No baskets | 107 | 34,264 | 65 | 0.19% | 1.94 |
| Low density (bottom quartile) | 153 | 273,335 | 555 | 0.20% | 2.08 |
| Medium / high density | 428 | 1,375,033 | 1,025 | 0.07% | 0.76 |
An index above 1.0 means that tier has a higher dog waste complaint proportion than the Manhattan-wide average; below 1.0 means lower. The low-density threshold is set at the bottom quartile of basket counts — that is, the point below which the least-served 25% of cells fall — across all grid cells that contain at least one basket (≤ 8 baskets per cell).
To test whether the difference between low-coverage and well-served areas is statistically significant, a one-sided permutation test is used. “No baskets” and “Low density” cells are collapsed into a single low-coverage group and compared against “Medium / high density” cells. The observed difference in dog waste proportions (low-coverage minus medium/high) is compared against 10,000 permutations of the group labels; the one-sided p-value is the share of permutations that produced a difference at least as large as observed.
The observed dog waste proportion in low-coverage cells is 0.20% vs. 0.07% in medium/high-density cells — a difference of 0.13% percentage points. The one-sided permutation p-value is < 0.01 (10^{4} permutations).
Sensitivity check: excluding park-adjacent cells
The basket inventory is known to under-record baskets inside parks (only 357 “Parks Basket” entries for all of Manhattan). Cells that fall inside or immediately border a park polygon may therefore appear as “low basket density” for a spurious reason — not because the street network is poorly served, but because park baskets are missing from the data. This section repeats the entire analysis after dropping those cells.
Park boundaries come from the NYC
Parks Properties dataset, downloaded once and cached. The following
typecategory values are treated as parks: Flagship Park,
Community Park, Neighborhood Park, Nature Area, Historic House Park,
Recreational Field/Courts, and Parkway. This covers Central Park,
Riverside Park, Fort Tryon Park, Morningside Park, Marcus Garvey Park,
and all other named parks of meaningful size; small triangles, plazas,
community gardens, and playgrounds are excluded from the exclusion. Each
grid cell whose center falls inside any park polygon is dropped. The
basket-density quartile threshold from the main analysis is reused so
the “low density” definition is directly comparable.
| Basket density tier | Grid cells | Total complaints | Dog waste complaints | Dog waste % | Index vs. Manhattan avg |
|---|---|---|---|---|---|
| No baskets | 72 | 27,070 | 61 | 0.23% | 2.31 |
| Low density (bottom quartile) | 119 | 244,478 | 518 | 0.21% | 2.17 |
| Medium / high density | 413 | 1,338,543 | 1,001 | 0.07% | 0.77 |
The observed dog waste proportion in low-coverage cells (parks excluded) is 0.21% vs. 0.07% in medium/high-density cells — a difference of 0.14% percentage points. The one-sided permutation p-value is < 0.01 (10^{4} permutations).
Key finding: Low litter-basket coverage is associated with meaningfully higher dog waste complaint rates, and this gap holds even after removing park-adjacent cells where basket data is known to be incomplete. Prioritize basket installation and servicing in low-coverage grid cells that also rank high on dog waste complaint proportion. A small number of targeted additions in the most underserved areas is likely to have a disproportionate effect on complaint volume.
A fuller audit of basket placement on the highest-complaint blocks identified in the hotspot analysis would help produce a prioritized installation list. Basket installation is a well-understood, low-cost operation.
Infrastructure Analysis: Bag Dispensers
In 2018, NYC Parks announced the installation of approximately 1,000 canine waste bag dispensers across all five boroughs to mark the 40th anniversary of the city’s Pooper Scooper law, providing free, on-street units intended to give residents an easy way to pick up after their dogs. To assess how many survive today, this analysis draws on the NYC Parks dispenser dataset, which lists only currently active units; decommissioned ones are removed from the public record entirely and leave no trace. The units counted here are the ones we can confirm are still standing. Anything that’s been removed simply disappears from the data.
New Yorkers shouldn’t have to wait for another anniversary to get more bag dispensers.
| Borough | ~Installed 2018 | Active today | Retained |
|---|---|---|---|
| Bronx | 200 | 192 | 96% |
| Manhattan | 200 | 25 | 12% |
| Brooklyn | 250 | 2 | 1% |
| Queens | 150 | 7 | 5% |
| Staten Island | 150 | 17 | 11% |
The Bronx has maintained its network with near-completeness (~96%) while every other borough has lost the vast majority of its units. Brooklyn retains fewer than 1%. The cause of this divergence is not documented in any public dataset; it likely reflects differences in park maintenance budgets, restocking contracts, or vandalism rates across boroughs.
Within Manhattan, what little dispenser coverage exists is concentrated in Community District 3 (Lower East Side / Chinatown); uptown Manhattan, including CD 12, has effectively no dispenser presence at all.
Bronx: dispenser density and complaint rates
Because attrition makes cross-borough comparison unreliable, the following analysis is confined to the Bronx, the only borough where there is enough variation in dispenser coverage across community districts to test whether dispensers are associated with lower complaint rates.
The Spearman correlation between active dispenser count and dog waste complaint proportion across Bronx community districts is ρ = -0.59 (p < 0.05, n = 12 CDs). Spearman correlation is a measure of how strongly two things move together (here, whether districts with more dispensers tend to have fewer complaints) on a scale from −1 (perfectly opposite) to +1 (perfectly aligned). It ranks the data first rather than using raw values, which makes it less sensitive to extreme outliers. Because dispensers were originally placed in areas with higher existing dog waste pressure, the raw correlation likely understates any protective effect: high-complaint CDs received more dispensers precisely because of their complaint history. A before/after design would require installation date records, which are absent from the public dataset. The figure and correlation coefficient should be read as descriptive, not causal.
Key finding: Manhattan’s dispenser network is effectively non-existent north of CD 3, and Bronx data suggest density matters. CD 12 — the highest-complaint district in the borough — has no dispensers at all. The complete absence of coverage across the entire uptown corridor, despite it generating the borough’s highest complaint volumes, is both an equity gap and a missed low-cost intervention that should be addressed in the next procurement cycle.
Work with Parks and DSNY to install more dispensers in uptown Manhattan, where coverage is effectively absent. Address the gap in the next procurement cycle.
Dog Waste × Rat Co-occurrence
A note on the rat sighting category: For this section, rat sightings are defined as Rodent / Rat Sighting only. Other rodent-related complaint types are excluded.
Community district level
Each dot below represents one of Manhattan’s 12 community districts. The horizontal axis shows each district’s dog waste complaint proportion; the vertical axis shows its rat sighting proportion. Districts in the upper-right corner are high on both.
The Pearson correlation between the two proportions is r = 0.29 (p = 0.364, n = 12 districts). Pearson correlation measures how strongly two things move together, on a scale from −1 to +1. A value of +1 would mean a perfect match — every district that ranks higher on dog waste also ranks proportionally higher on rats. A value of 0 means no relationship at all. A negative value would mean the two move in opposite directions. An r of 0.29 indicates a weak positive relationship: community districts with a higher share of dog waste complaints tend to also have a higher share of rat complaints. The p-value indicates how likely it would be to see a relationship this strong by chance alone; values below 0.05 are the conventional threshold for calling a result statistically meaningful.
Sensitivity check — excluding CD 12. Community District 12 (Washington Heights / Inwood) sits far to the right of all other boards on the dog waste axis, making it a leverage point that can pull the correlation in its direction. Removing it and re-running the same test gives r = 0.73 (p = 0.011, n = 11). The correlation strengthens to 0.73, suggesting CD 12 was actually dampening the overall pattern — its unusually high dog waste rate is not accompanied by a proportionally high rat rate, which pulls the line flatter. Either way, with only 11 or 12 data points the result should be read as directional evidence rather than a firm conclusion.
Neighborhood level
Manhattan has 32 residential Neighborhood Tabulation Areas (NTAs), a geographic unit sized between community districts and individual blocks in scale, giving this analysis roughly 2.7 times as many data points and more statistical power to detect a real pattern.
The Pearson correlation at the neighborhood level is r = 0.23 (p = 0.209, n = 32 NTAs). This analysis has more statistical power than the community board comparison above — roughly 2.7 times as many data points — so the result carries more weight.
An r of 0.23 reflects a weak positive association: neighborhoods with a higher share of dog waste complaints also tend to show a higher share of rat complaints relative to their overall 311 activity. The p-value of p = 0.209 indicates this result does not clear the conventional 0.05 significance threshold, so the pattern should be interpreted cautiously.
Sensitivity check — excluding Washington Heights (South). This NTA sits far to the right on the dog waste axis — its proportion is roughly three times the next highest neighborhood — making it a high-leverage point. Removing it gives r = 0.59 (p < 0.001, n = 31). Excluding the outlier actually strengthens the correlation to 0.59. Washington Heights (South) has a very high dog waste rate but a relatively modest rat sighting rate, so including it was flattening the overall pattern. The underlying relationship across the rest of Manhattan appears tighter than the full-sample figure suggests. This is a useful reminder that with a moderate number of geographic units, a single unusual neighborhood can meaningfully shift a Pearson correlation.
One important caveat: correlation does not imply causation. Dog waste on streets can attract rats by providing a food source, which may explain part of this pattern — but both complaints also tend to concentrate in denser, more active residential neighborhoods, so some or all of the association could reflect shared underlying factors (population density, park access, reporting culture) rather than a direct link between the two problems.
Geographic distribution
The two maps below show the same proportions spatially. Each NTA polygon is shaded from low (light) to high (dark). NTAs with no geocoded complaints appear in grey.
Key finding: Dog waste and rat complaints co-occur across neighborhoods — districts and NTAs with elevated dog waste rates tend also to have elevated rat sighting rates. The signal strengthens once the largest outlier is set aside, suggesting the pattern is real and widespread rather than driven by a single extreme case.
Treat high-dog-waste areas as rat-risk areas and coordinate interventions accordingly. Coordinate dog waste enforcement and basket installation with pest control operations in high-complaint neighborhoods. Consider dual-issue community messaging (addressing both problems at once) in areas that rank high on both complaint types. Reducing dog waste in the highest-concentration blocks may have downstream benefits for rat activity.
Business Improvement Districts
Business Improvement Districts (BIDs) are self-taxing geographic zones where property owners pay a supplemental assessment to fund services beyond baseline city provision, including additional street cleaning and sanitation crews. If BID-funded supplemental cleaning has a meaningful effect on dog waste accumulation or complaint rates, BID-covered blocks should show lower dog waste complaint proportions than comparable non-BID areas in the same neighborhood. Conversely, BIDs often cover high-pedestrian-traffic commercial corridors where dog walking is common, which could inflate complaint rates regardless of cleaning intensity. This section tests those competing hypotheses.
BID vs. non-BID complaint proportions
Across all geocoded complaints, the dog waste proportion is 0.02% inside BID territory and 0.12% outside, a raw difference of 0.09% percentage points. That raw comparison is hard to interpret on its own, because BIDs tend to be located in busy commercial areas where dog walking is common regardless of how much cleaning happens.
The chart below tries to correct for that by comparing BID and non-BID areas within the same neighborhood, so the two sides of the comparison are drawn from similar surroundings. Each point represents one NTA that contains both BID-covered and non-BID complaints.
| NTA | BID complaints | Non-BID complaints | BID dog waste % | Non-BID dog waste % | Difference (BID − non-BID) |
|---|---|---|---|---|---|
| Washington Heights (South) | 707 | 96,141 | 0.00% | 0.45% | -0.45 pp |
| Chinatown-Two Bridges | 16,342 | 12,243 | 0.03% | 0.16% | -0.13 pp |
| Upper East Side-Lenox Hill-Roosevelt Island | 1,312 | 36,614 | 0.00% | 0.11% | -0.11 pp |
| Upper West Side-Manhattan Valley | 9,543 | 26,188 | 0.05% | 0.15% | -0.10 pp |
| Washington Heights (North) | 8,835 | 91,650 | 0.06% | 0.14% | -0.09 pp |
| Harlem (South) | 3,939 | 71,882 | 0.03% | 0.11% | -0.08 pp |
| Tribeca-Civic Center | 5,043 | 21,439 | 0.00% | 0.07% | -0.07 pp |
| Morningside Heights | 1,118 | 20,343 | 0.00% | 0.07% | -0.07 pp |
| Chelsea-Hudson Yards | 24,545 | 44,743 | 0.02% | 0.09% | -0.06 pp |
| Financial District-Battery Park City | 43,172 | 10,951 | 0.02% | 0.08% | -0.06 pp |
| Murray Hill-Kips Bay | 5,266 | 36,987 | 0.02% | 0.07% | -0.05 pp |
| Upper West Side (Central) | 4,420 | 63,098 | 0.02% | 0.08% | -0.05 pp |
| Hell’s Kitchen | 9,062 | 47,582 | 0.02% | 0.07% | -0.05 pp |
| West Village | 14,726 | 22,429 | 0.02% | 0.07% | -0.05 pp |
| East Village | 5,507 | 68,928 | 0.00% | 0.05% | -0.05 pp |
| Upper West Side-Lincoln Square | 12,399 | 40,976 | 0.03% | 0.08% | -0.04 pp |
| East Midtown-Turtle Bay | 11,607 | 17,413 | 0.03% | 0.07% | -0.04 pp |
| Upper East Side-Carnegie Hill | 9,021 | 26,490 | 0.01% | 0.05% | -0.03 pp |
| Greenwich Village | 16,709 | 22,902 | 0.01% | 0.03% | -0.02 pp |
| Gramercy | 7,430 | 18,987 | 0.04% | 0.05% | -0.01 pp |
| Midtown South-Flatiron-Union Square | 36,096 | 12,796 | 0.01% | 0.02% | -0.01 pp |
| Lower East Side | 9,365 | 37,126 | 0.05% | 0.05% | +0.00 pp |
| SoHo-Little Italy-Hudson Square | 14,558 | 24,031 | 0.02% | 0.01% | +0.01 pp |
| Midtown-Times Square | 58,225 | 21,172 | 0.02% | 0.00% | +0.01 pp |
| East Harlem (North) | 294 | 71,595 | 0.34% | 0.12% | +0.22 pp |
Individual BID rankings
Each BID’s dog waste complaint proportion is shown below, ranked from highest to lowest. The dashed vertical line marks the Manhattan-wide average. Red bars are above average; blue are below. BIDs with fewer than 50 total 311 complaints are excluded.
| BID | Total complaints | Dog waste complaints | Dog waste % | Index vs. Manhattan |
|---|---|---|---|---|
| Washington Heights BID | 9,542 | 5 | 0.05% | 0.54 |
| Fifth Avenue Association BID | 8,332 | 4 | 0.05% | 0.49 |
| Lower East Side BID | 11,038 | 5 | 0.05% | 0.46 |
| Columbus Amsterdam BID | 11,413 | 5 | 0.04% | 0.45 |
| Bryant Park BID | 2,483 | 1 | 0.04% | 0.41 |
| SoHo Broadway BID | 5,118 | 2 | 0.04% | 0.40 |
| Lincoln Square BID | 10,350 | 4 | 0.04% | 0.40 |
| 125th Street BID | 5,351 | 2 | 0.04% | 0.38 |
| Hudson Yards/Hell’s Kitchen | 15,090 | 5 | 0.03% | 0.34 |
| East Mid-Manhattan BID | 9,962 | 3 | 0.03% | 0.31 |
| West Village BID | 10,125 | 3 | 0.03% | 0.30 |
| Chinatown | 21,643 | 6 | 0.03% | 0.28 |
| Garment District | 10,852 | 3 | 0.03% | 0.28 |
| Downtown Alliance BID | 47,150 | 11 | 0.02% | 0.24 |
| Columbus Avenue BID | 4,599 | 1 | 0.02% | 0.22 |
| Times Square BID | 32,922 | 5 | 0.02% | 0.16 |
| Union Square Partnership BID | 13,387 | 2 | 0.01% | 0.15 |
| Village Alliance BID | 13,502 | 2 | 0.01% | 0.15 |
| Grand Central Partnership | 19,561 | 2 | 0.01% | 0.10 |
| Flatiron/23rd Street Partnership | 29,397 | 3 | 0.01% | 0.10 |
| 34th Street BID | 16,814 | 0 | 0.00% | 0.00 |
| 47th Street BID | 292 | 0 | 0.00% | 0.00 |
| Hudson Square | 4,608 | 0 | 0.00% | 0.00 |
| Madison Avenue BID | 6,645 | 0 | 0.00% | 0.00 |
| Meatpacking BID | 4,974 | 0 | 0.00% | 0.00 |
| NoHo BID | 4,609 | 0 | 0.00% | 0.00 |
Regression: dog waste vs. total complaint volume by BID
The same regression method used for community boards above is applied at the BID level: dog waste complaints as a function of total 311 complaints. BIDs above the fitted line generate more dog waste complaints than their overall 311 activity would predict; those below generate fewer. Dashed segments show each BID’s residual.
BID complaint rate map
Each BID polygon is shaded by its dog waste complaint proportion. NTA outlines provide orientation. BIDs below the 50-complaint threshold appear in grey.
Key finding: BID-funded supplemental cleaning shows a measurable association with lower complaint rates. BID-covered blocks show a dog waste complaint proportion of 0.02%, compared to 0.12% on non-BID blocks. Within NTAs, controlling for neighborhood character, the median gap is -0.05 pp (25 NTAs with sufficient complaints in both zones). The within-NTA comparison is the more meaningful test: BIDs tend to occupy commercial corridors with higher foot traffic, so the global gap alone overstates any cleaning effect. That the gap persists within neighborhoods is an encouraging signal.
Share the block-level hotspot data from this analysis with the Manhattan BID Association and individual district managers. Complaint rates within BID boundaries reflect many factors outside any district's control, including foot traffic, dog density, and the mix of residential and commercial uses, so variation across BIDs is not a straightforward measure of cleaning effectiveness. What the hotspot data can help with is identifying specific blocks where infrastructure gaps and elevated complaints overlap, giving district managers a concrete starting point for directing resources toward basket installations or dispenser coverage where it is likely to make the most difference.
Thanks for reading.
Everything you wanted to know about dog waste but were afraid to ask — unless you’re a data nerd, in which case, read on.
Data & Methodology
Software and tools
This report was produced entirely in R, using R Markdown to combine analysis code and narrative in a single reproducible document. The rendered output is a self-contained HTML file. Key packages used:
- Data access and wrangling:
httr(API requests),jsonlite(JSON parsing),dplyr,readr,lubridate - Visualization:
ggplot2,plotly(interactive charts),patchwork(multi-panel layouts),ggrepel(non-overlapping labels),scales - Tables and output:
knitr
All raw data files are cached locally on first download; subsequent runs of the document skip network calls if the files already exist.
BetaNYC produced a parallel interactive map and block-level analysis using the same underlying 311 dataset. That work, including the block-level complaint counts and infrastructure overlays, is available at betanyc.github.io/mbpo-dog-poop.
Data sources
NYC 311 Service Requests The primary dataset is the
NYC
311 Service Requests from 2010 to Present, accessed via the Socrata
Open Data API. Full individual records (one row per complaint) were
downloaded for Manhattan, January 2024 through April 2026, in monthly
batches. A supplementary aggregate query — fetching only monthly totals
rather than individual records — extends the historical view back to
January 2020 for the six-year trend chart. A separate Bronx extract
covering the same January 2024–April 2026 window supports the dispenser
density analysis. Fields used: created_date,
complaint_type, descriptor,
community_board, latitude,
longitude, incident_address,
agency.
DSNY Litter Baskets The DSNY
litter basket inventory provides the location and type of every
public litter basket maintained by the Department of Sanitation. Records
were filtered to a bounding box covering Manhattan. Fields used:
basketid, baskettype, point
(coordinates).
NYC Parks Waste Bag Dispensers The NYC Parks dog waste bag dispenser dataset lists currently active dispenser units by borough and community board. Only active units appear in the dataset; decommissioned units are not retained. The Manhattan extract was filtered to community boards 1–12 and further refined using the NTA point-in-polygon test to exclude units in adjacent boroughs and Roosevelt Island that fall within the bounding box. A separate Bronx extract aggregated active units by community board for the cross-sectional dispenser analysis.
NYC Parks Dog Runs The NYC Parks dog run locations dataset provides the name and polygon geometry of official off-leash dog run areas across Manhattan. Centroids were computed for map display.
NYC Parks Properties (park boundaries) The NYC Parks Properties dataset provides polygon boundaries for all NYC Parks land. For the sensitivity analysis, the following park types were used to define exclusion zones: Flagship Park, Community Park, Neighborhood Park, Nature Area, Historic House Park, Recreational Field/Courts, and Parkway. This covers Central Park, Riverside Park, Fort Tryon Park, Morningside Park, Marcus Garvey Park, and all other named parks of meaningful size, while excluding small triangles, plazas, and playgrounds.
NYC Community District Boundaries Community board polygon geometries were downloaded from the NYC Community Districts layer (NYC ArcGIS Open Data), filtered to Manhattan (BoroCD 101–112).
NYC Neighborhood Tabulation Areas (NTAs) NTA polygon geometries were downloaded from the NYC Neighborhood Tabulation Areas 2020 layer (NYC ArcGIS Open Data), filtered to Manhattan residential NTAs (BoroCode = 1, NTAType = 0). NTAs are geographic units defined by the NYC Department of City Planning to represent stable, recognizable neighborhoods; they sit between community boards and individual blocks in scale.
Business Improvement District (BID) Boundaries BID polygon geometries were downloaded from the NYC BIDs layer (NYC ArcGIS Open Data) and filtered to Manhattan by bounding box.
Complaint categorization
Dog waste complaints are identified using a two-field match on
complaint_type and descriptor, because
reporting responsibility is divided between two agencies depending on
location:
- Dirty Condition / Dog Waste — filed with the Department of Sanitation (DSNY), typically for waste on sidewalks and streets
- Animal in a Park / Animal Waste — filed with the Department of Parks and Recreation, typically for waste in parks and at curbside
Neither type alone captures the full volume. This report combines both into a single “Dog Waste” category throughout. Rat sightings are defined as Rodent / Rat Sighting only; other rodent complaint subtypes (signs of rodents, rat burrows) are excluded. All other complaint types are grouped as “Other.”
Geographic assignment
Complaints with valid latitude/longitude coordinates were assigned to NTAs and BID zones using a pure-R ray-casting point-in-polygon algorithm. Each point is tested against NTA or BID polygon rings using a vectorized crossing-number method with bounding-box pre-filtering to reduce computation. Points are assigned to the first matching polygon; unmatched points (those falling outside all NTA boundaries, typically in waterways or at the borough edge) are excluded from NTA-level analyses only.
Community board assignment uses the community_board
field from the 311 records directly, parsed to extract the numeric board
identifier (1–12). Records with missing or out-of-range community board
values are excluded from board-level analyses.
Spatial grid analysis (basket coverage)
Manhattan was divided into a regular grid of rectangular cells (0.003 decimal degrees per side), approximately 333 m north–south and 250 m east–west at Manhattan’s latitude. Each 311 complaint with valid coordinates was assigned to a cell; each litter basket was similarly assigned. Cells with fewer than 10 total 311 complaints were excluded to avoid instability from very low complaint volumes.
Cells were then classified into three basket density tiers:
- No baskets: zero baskets recorded in the cell
- Low density (bottom quartile): basket count at or below the 25th percentile among cells that have at least one basket
- Medium / high density: all remaining cells
The dog waste complaint proportion was computed for each tier and compared. The low-density threshold was derived from the basket distribution and reused as-is for the park-exclusion sensitivity check to keep the tier definitions directly comparable.
Statistical methods
Proportions rather than raw counts. Throughout this report, dog waste complaint volumes are expressed as a proportion of total 311 activity — dog waste complaints divided by all complaints from the same geographic unit or time period. This controls for the fact that some community boards, neighborhoods, and months simply generate more 311 calls overall. A board with twice the population would be expected to generate roughly twice as many complaints of every type; comparing proportions levels that playing field.
Linear regression (community board and BID anomaly detection). Ordinary least squares regression was used to model each unit’s dog waste complaint count as a function of its total 311 complaint volume. The fitted line represents the expected dog waste count for a unit of that overall size. Residuals (observed minus expected) identify units generating more or fewer dog waste complaints than their size would predict. A 95% confidence band was added to formally classify residuals as statistically significant: units outside the band are flagged as outliers unlikely to arise from chance variation.
Permutation test (basket density vs. complaint rates). To test whether the difference in dog waste complaint proportions between low-coverage and well-served grid cells is statistically significant, a one-sided permutation test was used. The “No baskets” and “Low density” tiers were collapsed into a single low-coverage group. The observed difference in dog waste proportions (low-coverage minus medium/high) was then compared against 10,000 random permutations of the group labels across cells. The one-sided p-value is the proportion of permutations that produced a difference at least as large as the observed value. This non-parametric approach makes no assumptions about the shape of the underlying distributions.
LOESS smoothing (temporal trend). The six-year share chart uses locally weighted regression (LOESS) with a span of 0.3 to fit a smooth trend curve through the monthly data. LOESS fits a separate weighted regression at each point along the time axis, giving more weight to nearby months and less to distant ones, producing a flexible curve that captures gradual directional shifts without overfitting to individual months.
Pearson correlation (dog waste and rat co-occurrence). The Pearson correlation coefficient (r) measures the strength and direction of the linear relationship between dog waste complaint proportions and rat sighting complaint proportions across community boards and NTAs. Values range from −1 (perfect negative relationship) to +1 (perfect positive relationship). P-values test whether the observed correlation is large enough to be unlikely under the null hypothesis of no relationship. With 12 community boards, statistical power is limited; the NTA-level analysis (32 units) carries more weight. Sensitivity checks removing the single largest outlier are reported alongside the full-sample results.
Spearman correlation (Bronx dispenser analysis). The Spearman correlation (ρ) between active dispenser count and dog waste complaint proportion across Bronx community boards ranks both variables before computing the correlation, making it more robust to extreme values than Pearson. This was used in preference to Pearson because one or two community boards may have outlier dispenser counts.
City-wide index. Each community board’s dog waste proportion was divided by the city-wide dog waste proportion (computed across all five boroughs via a single aggregate Socrata query) to produce an index. An index value of 1.0 means the board matches the five-borough average; values above 1.0 indicate a higher-than-average share of dog waste complaints.
Within-NTA BID comparison. To isolate any BID cleaning effect from the neighborhood-character confound, dog waste complaint proportions were computed separately for BID-covered and non-BID areas within each NTA that had at least 50 complaints in both zones. The within-NTA difference (BID proportion minus non-BID proportion) was computed for each qualifying NTA, and the median of those differences is reported as the within-neighborhood BID effect.
Known limitations
- Undercount by design. The combined dog waste category understates actual incidents — not all incidents are reported to 311, and the two-field definition may miss edge cases.
- Reporting bias. 311 complaint rates reflect both the prevalence of the problem and residents’ propensity to report it. High complaint rates in some neighborhoods may partly reflect higher civic engagement rather than more waste on the ground. This is discussed more fully in the "What This Data Actually Measures" section above.
- Missing coordinates. Complaints without valid latitude/longitude values are excluded from all spatial analyses (NTA assignment, grid analysis, BID assignment). The exclusion rate is noted in the output but is not expected to be geographically systematic.
- Basket inventory completeness. The DSNY basket dataset is known to under-record baskets located inside parks (only a small number of “Parks Basket” entries appear for all of Manhattan). The park-exclusion sensitivity check is designed to test whether this gap drives the basket-density finding; results are robust to that exclusion.
- Dispenser attrition data. The NYC Parks dispenser dataset retains only currently active units. There is no public record of installation dates or decommissioning dates, making it impossible to reconstruct the dispenser network at any prior point in time. The 2018 installation figures used in the attrition table are drawn from a 2018 NYC Parks press release and are approximate.
- Cross-sectional design. All analyses in this report are cross-sectional snapshots — they compare places at a single point in time, not before and after an intervention. Correlations between infrastructure and complaint rates are descriptive and do not establish causation. A before/after or randomized design would be needed to confirm causal effects.