We look here at whether any noticable trends existed in time, if there are any seasonal effect, whether pit bulls bite less people over time, and whether borough-level effects exist. The figure below looks at the number of dog bites over time across all boroughs, broken down by breed type.
There is a moderate overall increase in the mean number of bites per month by non-pitbull breeds. However, for pit bulls, the trend in number of bites per month is more stable. In fact, there’s evidence of a slight decrease in pit bull bites, on average. Additionally, fluctation is very evident within any given year, most likely corresponding to seasonal effects, with overall dog bites rising during the summer months and falling during the winter months.
The figure below shows these bites trends by borough. In Manhattan, we see the largest mean difference in pit bull vs. non-pit bull bites per month; trends over time are similar between breed types. Conversely, a smaller mean difference between breeds was observed in Staten Island, with a nearly identical trends over time. Brooklyn and Queens both demonstrate a widening gap over time, with the pit bull bites decreasing and non-pit bull bites increasing notably. Interestingly, there is an inverse trend between pit bulls and other breeds; pit bull bites increase over time while non-pit bull bites decreased.
After visually inspecting the temporal dynamics of pit bull bites in time, we decided to formally test the hypothesis that the number of pit bull dog bites decreased from 2015-2018 in NYC. Motivated by our exploratory visualizations above, we also decided it was necessary to control for seasonal and borough-level effects. We fit a linear regression where our final model took the form
\[ Number \ of \ Bites_i \sim \beta_0 + \beta_1 Time_i + \beta_2 Pitbull_i + \beta_3 Pitbull_i:Time_i + \beta_4 Season_i + \beta_5 Borough_i + \varepsilon_i,\] for the \(i^{th}\) month. Results are below, and model assumptions are checked after the discussion.
| term | estimate | p.value |
|---|---|---|
| (Intercept) | 19.237 | 0.000 |
| time | 2.038 | 0.005 |
| breedPit Bull | -12.423 | 0.000 |
| seasonSpring | 6.886 | 0.000 |
| seasonSummer | 11.016 | 0.000 |
| seasonFall | 5.490 | 0.000 |
| boroughBrooklyn | 5.946 | 0.000 |
| boroughManhattan | 5.041 | 0.000 |
| boroughQueens | 9.392 | 0.000 |
| boroughStaten Island | -8.919 | 0.000 |
| time:breedPit Bull | -1.912 | 0.059 |
After adjusting for season and borough, we see an almost negligible increase in the expected number of pit bull bites for a given month (increase of 0.126 bites per additional month). Our final regression results describe a baseline mean number of bites of 19.24 in the Bronx, in the winter, for non-pit bulls - from here there is a moderate expected mean increase in bites per month overall (2.03 bites). This estimate changes based on the borough, season, and breed.
Ultimately, pit bull bites increased almost negligibly compared to other breeds (at the 10% significant level) in NYC from 2015-2018, adjusting for significant seasonal and borough level effects. Our initial intuitions that seasonal fluctuation and borough significantly affect the number of dog bites over time are also supported. Unfortunately, we cannot conclude whether media campaigns to destigmatize and rehabilitate pit bulls in recent years have directly caused the relative decline in pit bull aggression and citation which we observed. Still, we have identified a meaningful difference in pit bull bites over time compared to other breeds.