We have moved to a new website. You are being redirected!

In August, we examined crime trends across D.C. neighborhoods and made a few interesting observations about the neighborhoods responsible for D.C.'s crime surge in 2015. With Halloween coming up, we wondered whether certain neighborhoods experience a seasonal spike in crime this time of year. In this post, we use D.C. crime statistics from the District's Crime Incidents dataset from 2011 to 2014 to measure whether there is a Halloween crime bump. The source code for this analysis can be found here.

Does your neighborhood have a Halloween crime spike? Look it up in Figure 2 below.

Methodology 

We calculated the average weekly crime incidents for the week of October 24 to October 31 for the years 2011-2014 (Halloween Week). Then, we calculated the average weekly crime incidents from 2011-2014. We compared the average weekly crime incidents during Halloween Week to the average weekly crime incidents during the period from 2011-2014 to measure whether there is a crime bump during Halloween Week. We estimated the statistical significance of the observations using the z-test. Finally, we excluded neighborhoods with very low crime incidents to avoid distorting the results.

We saw in a previous chart-it series that the crime rate exhibits a seasonal pattern and, therefore, it may be more appropriate to compare crime incidents within clusters of months that exhibit the same seasonal pattern (or, to seasonally adjust the dataset). Using that methodology, we could exclude the low crime winter months, which would increase the average weekly crime rate and, therefore, reduce the magnitude of the Halloween crime bump. The results were not significantly different under either approach so we stuck to using the entire dataset. However, this is by no means a comprehensive study of D.C. crime trends.

Halloween Crime Spike

Figure 1 (yes, we had to use a Halloween color theme) shows the five D.C. neighborhoods with the highest Halloween crime bumps. 'Howard University, Le Droit Park, Cardozo/Shaw' has historically seen the largest increase in crime (about 40.3 percent) during Halloween. The 'Howard University, Le Droit Park, Cardozo/Shaw' observation is statistically significant at the < 0.03 p-value level. In other words, there is a less than 0.3 percent probability that the magnitude of the observed Halloween Week crime bump is by random chance. The statistical significance of the observations for the other neighborhoods is provided in Figure 2 below.

Interestingly, three of the top five neighborhoods were also in chart-it's list of top 10 most improved D.C. neighborhoods in terms of a reduction in violent crime between 2014 and 2015. Hopefully this improvement will carry over into Halloween and will not cause visitors to and resident of these neighborhoods to be complacent.

Figure 1. Neighborhoods With Largest Halloween Crime Spike

Washington, D.C., as a whole, experienced an approximately 11.6 percent crime spike during Halloween Week (with a p-value of 0.22). This is primarily the result of more thefts, robberies, burglaries and thefts from cars, although there is also a large percentage increase in homicides and sex crimes as these crimes are relatively rare to begin with. Bottom line: be aware of your surroundings and don't forget to lock your house and car doors. 

Does your neighborhood have a Halloween crime spike? Look it up in Figure 2.

Figure 2. All Neighborhoods With Measurable Crime Spike

Neighborhood	Halloween Crime Spike	Probability Spike Due To Random Chance
Howard University, Le Droit Park, Cardozo/Shaw	40.3%	0.3%
Capitol View, Marshall Heights, Benning Heigh	39.0%	0.6%
Eastland Gardens, Kenilworth	34.1%	12.8%
Brightwood Park, Crestwood, Petworth	29.4%	1.3%
Mayfair, Hillbrook, Mahaning Heigh	28.4%	6.7%
Edgewood, Bloomingdale, Truxton Circle, Eckington	27.6%	3.1%
Takoma, Brightwood, Manor Park	26.2%	6.1%
Southwest Employment Area, Southwest/Waterfront, Fort McNair, Buzzard Poin	24.2%	10.2%
Columbia Heights, Mt. Pleasant, Pleasant Plains, Park View	21.6%	3.2%
Georgetown, Burleith/Hillanda	20.8%	9.5%
Congress Heights, Bellevue, Washington Highland	20.6%	5.0%
Historic Anacostia	18.8%	18.8%
Twining, Fairlawn, Randle Highlands, Penn Branch, Fort Davis Park, Dupont Park	18.5%	12.8%
Shaw, Logan Circ	18.2%	9.4%
Fairfax Village, Naylor Gardens, Hillcrest, Summit Park	15.5%	22.0%
NoMa, Union Station, Stanton Park, Kingman Park	13.6%	16.5%
Douglass, Shipley Terrac	10.7%	25.2%
Woodridge, Fort Lincoln, Gateway	9.1%	32.7%
Spring Valley, Palisades, Wesley Heights, Foxhall Crescent, Foxhall Village, Georgetown Reservoi	8.9%	39.1%
Lamond Riggs, Queens Chapel, Fort Totten, Pleasant Hi	8.7%	33.9%
Friendship Heights, American University Park, Tenleytown	8.6%	32.9%
Kalorama Heights, Adams Morgan, Lanier Heigh	8.0%	31.6%
Woodland/Fort Stanton, Garfield Heights, Knox Hi	7.2%	35.3%
Dupont Circle, Connecticut Avenue/K S	5.8%	29.7%
Brookland, Brentwood, Langdon	5.5%	35.3%
Ivy City, Arboretum, Trinidad, Carver Langston	4.7%	36.0%
River Terrace, Benning, Greenway, Fort Dupon	4.4%	38.2%
North Cleveland Park, Forest Hills, Van N	3.2%	45.5%
Capitol Hill, Lincoln Park	3.1%	43.0%

We have moved to a new website. You are being redirected!

This is part three of our four-part series exploring data from the DMV area's Capital Bikeshare system. In the first two parts of the series (published in early 2015), we examined rider and bike patterns based on data from 2.9+ million Bikeshare rides taken in 2014. Now that it's October 2015, and Bikeshare has released data for the first half of 2015, today’s post will explore popular routes and stations during the year from July 2014 to June 2015 (about 3.1 million Bikeshare rides). The entire series covers the following topics: 

- Part 1 explores the riding patterns of Bikeshare members in 2014. 

- Part 2 looks at bike activity and workload patterns in 2014.

- Part 3 explores popular routes and stations in during the year ending June 2015.

- Part 4 will explore trends over the period from 2010 to the first half of 2015. 

The Capital Bikeshare system data can be found here. The source code for the data analysis can be found here (for route patterns) and here (for station patterns). 

Routes

First, let's explore popular routes among Bikeshare users. Route popularity varies by member type as shown in Figure 1 below. Casual riders (mostly tourists), not surprisingly, mostly ride around D.C.’s popular monuments, the Lincoln Memorial, Jefferson Memorial and the Smithsonian/National Mall stations. In fact, casual riders overwhelmingly ride around the monuments, resulting in the top five routes by casual riders actually being the top five routes overall. The most popular route among casual users is from Jefferson Dr & 14th St SW to... Jefferson Dr & 14th St SW (a ride around the National Mall).

Registered riders (i.e., D.C. residents) mostly ride outside of the area around the monuments. Tourists could benefit from observing the popular routes among registered users to explore some of the less well-known parts of D.C. that the locals frequent, including Eastern Market and H St Corridor.

Figure 1. Most Popular Bikeshare Routes

Casual Riders	Registered Riders
Jefferson Dr & 14th St SW to Jefferson Dr & 14th St SW	Columbus Circle / Union Station to 8th & F St NE
Lincoln Memorial to Jefferson Memorial	8th & F St NE to Columbus Circle / Union Station
Jefferson Dr & 14th St SW to Lincoln Memorial	Lincoln Park / 13th & E Capitol St NE to Eastern Market Metro / Pennsylvania Ave & 7th St SE
Lincoln Memorial to Jefferson Dr & 14th St SW	Eastern Market Metro / Pennsylvania Ave & 7th St SE t Lincoln Park / 13th & East Capitol St NE
Smithsonian / Jefferson Dr & 12th St SW to Smithsonian / Jefferson Dr & 12th St SW	New Hampshire Ave & T St NW to Massachusetts Ave & DuPont Circle NW

The most popular route may also vary by month (Figure 2 below). Interestingly, the most popular route among registered users is the same in every month, except in January, when the reverse route was the most popular. Why are so many people riding from Union Station to the H St Corridor? It may be the result of limited public transportation to the H St Corridor. H St appears to be a big beneficiary of the Bikeshare system. 

Among casual riders, the most popular route is always some combination of stations around the National Mall.

Figure 2. Most Popular Route by Month

Month	Casual Riders	Registered Riders
July 2014	Lincoln Memorial to Jefferson Dr & 14th St SW	Columbus Circle / Union Station to 8th & F St NE
Aug 2014	Jefferson Dr & 14th St SW to Jefferson Dr & 14th St SW	Columbus Circle / Union Station to 8th & F St NE
Sept 2014	Jefferson Dr & 14th St SW to Lincoln Memorial	Columbus Circle / Union Station to 8th & F St NE
Oct 2014	Lincoln Memorial to Jefferson Memorial	Columbus Circle / Union Station to 8th & F St NE
Nov 2014	Lincoln Memorial to Jefferson Memorial	Columbus Circle / Union Station to 8th & F St NE
Dec 2014	Lincoln Memorial to Jefferson Dr & 14th St SW	Columbus Circle / Union Station to 8th & F St NE
Jan 2015	Jefferson Dr & 14th St SW to Jefferson Dr & 14th St SW	8th & F St NE to Columbus Circle / Union Station
Feb 2015	Jefferson Dr & 14th St SW to Jefferson Dr & 14th St SW	Columbus Circle / Union Station to 8th & F St NE
Mar 2015	Jefferson Dr & 14th St SW to Jefferson Dr & 14th St SW	Columbus Circle / Union Station to 8th & F St NE
Apr 2015	Lincoln Memorial to Jefferson Memorial	Columbus Circle / Union Station to 8th & F St NE
May 2015	Lincoln Memorial to Jefferson Memorial	Columbus Circle / Union Station to 8th & F St NE
June 2015	Jefferson Dr & 14th St SW to Jefferson Dr & 14th St SW	Columbus Circle / Union Station to 8th & F St NE

Measuring distance traveled is a little tricky because we don’t know the exact path taken during each ride. Some rides might follow a direct path from point A (starting station) to point B (end station) but others might start and end at the same station with a long ride in between (as we observed earlier). Others might dock their bike to re-start the 30 minute timer before continuing on to their ultimate destination. Nevertheless, we can still roughly estimate ride distance in two ways. We can either calculate the distance between the starting and ending station (using the haversine formula to estimate distance between two latitude/longitude coordinates), or we can divide the duration of each trip (in minutes) by an assumed average bike speed - about 10 miles per hour (compared to 5.9 miles per hour implied by the first method).

Figure 3 shows the results under each of these two methods. The average ride is between 0.9 and 1.8 miles, depending on our estimation method. The true average is probably somewhere in that range. 

Figure 3. Route Distance

Stations

Now let's take a look at the Bikeshare stations. The number of stations has continued to grow during the year ending June 2015. As shown in Figure 4 below, the system started the year with about 325 active stations and ended it with close to 351, an eight percent increase (or 26 stations). There were 362 unique active stations during the year, about 11 stations more than the 351 that were active in June 2015. It is likely that these 11 stations were either temporarily or permanently shut down, or moved to a different location (and, therefore, changed names) during the year. 

The number of stations increased gradually over the course of the year with November 2014 and April 2015 seeing the largest increase with 10 and six new stations, respectively.

Figure 4. Active Stations by Month

As shown in Figure 5, stations are most densely located between Foggy Bottom, Thomas Circle, Downtown and Chinatown, with scant coverage in certain areas, particularly in the low income neighborhoods east of the Anacostia River. Capital Bikeshare recently announced plans for a major expansion in Washington, DC, partly to service areas with little coverage, but also to better reach casual riders around the National Mall. Most of the new stations are proposed to be built in the area from the National Mall and north to Columbia Heights, with a small to be built proportion east of the Anacostia river.

Figure 5. Station Location Density Map

Figure 6 summarizes the workload distribution across stations based on the number of trips started or ended at that station. The bottom quantile of Bikeshare stations recorded less than 7.2 trips per day during the year. The top quantile recorded more than 80.8 trips per day. We measure station activity using rides per day to adjust for the fact that some stations were not operating during the entire year. It would not be accurate to compare the total number of rides for a station operating for 12 months versus a station operating for just three months. 

Figure 6. Summary of Workload by Station

	Rides Per Day
Minimum:	0.2
25th percentile:	7.2
50th percentile (median):	25.1
75th percentile:	80.8
Maximum:	374.2
Standard Deviation	57.6

Let's take a closer look at the distribution of ride activity by station. Figure 7 shows that half of stations registered less than 25 rides per day, with the remainder spread generally evenly between 25 and 200+ rides. At the extremes, about 15 percent of bike stations logged either less than 3.9 rides or more than 100.9 rides per day in 2014. This demonstrates Bikeshare’s social/political objective to provide a convenient mode of transportation for tourists and locals. A profit maximizing system might avoid many of the low traffic areas. The downside of course is that the system is more expensive as a result.

Figure 7. Distribution of Station Activity

But each ride is not created equal. There also appears to be a difference in the duration of rides from one station to another. The shortest trips originate from stations in the area from Downtown to Adams Morgan (highly efficient Washingtonians, no doubt) and around the neighborhood and business districts in Arlington. On the opposite end, the National Mall area (mostly casual riders) and 'suburban DC' neighborhoods tend to have the longest duration trips. The record, however, goes to the Prince St & Union St station (in Old Town Alexandria), which had by far the longest trips at 37 minutes! Figure 8 shows the median trip duration for each station. Red means longer trips and blue means shorter trips. Click on a station to see its median ride duration.

Figure 8. Median Ride Duration by Station 

We were also interested in the relationship between how busy a station is (in rides per day) and its size (number of docks). In other words, we wanted to know how successful Bikeshare has been at placing the larger stations in the busiest areas and the smaller station in less busy areas. First, let's take a look at station sizes (Figure 9). Bikeshare stations have between seven to 42 docks, with a median size of 15 docks. 

Figure 9. Bike Station Size

	Bike Station Size (# of Docks)
Minimum:	7
25th percentile:	13
50th percentile (median):	15
75th percentile:	19
Maximum:	42
Standard Deviation	5.5

Now let's plot rides per day against station size to measure the relationship between how busy a station is and its size (Figure 10). The correlation is relatively strong at 0.58. In other words, the busier stations tend also to be larger, which makes sense and reflects well on Bikeshare's station placement strategy. But not all busy stations are large, which may lead to congestion (either no bikes or no docks).

Figure 10. Relationship Between Station Size and Station Activity

One rough way to measure station congestion is to calculate the ratio of ‘rides per day’ to ‘station size’ (the higher the ratio, the more congested the station). Figure 11 below shows this ratio for each station (red means more congested and blue means less congested). Click on a station to see its ratio.

Figure 11. Station Congestion Map

Figure 12 shows the stations with the highest ratio of rides per day to station size. The stations with the highest ratio are generally some of the busiest stations in the entire system and, therefore, most likely to be the targets of bike rebalancing by Bikeshare. As a result, we also show the most congested stations excluding the top 50 busiest stations by total number of rides (which are less likely to benefit from rebalancing). A more accurate way to measure station congestion is to collect real time station data, which can then be used to calculate the percent of the time that a station either has too many bikes or not enough bikes. That is an analysis for a future write-up.

Figure 12. Most Congested Stations

All Stations	Excluding Top 50 Stations by Total Ride Count
Lincoln Memorial	3rd & D St SE
Jefferson Dr & 14th St SW	4th and East Capitol St NE
15th & P St NW	24th St & N St NW
25th St & Pennsylvania Ave NW	Eastern Market / 7th St & North Carolina Ave SE
1st & M St NE	3rd St & Pennsylvania Ave SE
14th & R St NW	8th & D St NW
20th St & Florida NW	10th & E St NW
Columbus Circle / Union Station	1st & K St SE