Sunday
Aug032014

Analysis of Spikes in Red Light Camera Tickets

The city has acknowledged the Tribune's analysis and is now reviewing over 9,000 tickets.  There are other stories that say this number will climb as high as 16,000 tickets.  The city has also stated they will more carefully monitor their RLC contractor and post the ticket violations daily for each intersection.  This transparency would aid in ensuring the integrity of the red light camera tickets.

I decided to do my own analysis of the camera spikes.  I used a statistical technique that looks for outliers in a time series.  I ran this on both a daily summation and a weekly summation of the tickets.  The results were revealing: 

  • Over 100 intersections had a spike  
  • At least 20,000 to 40,000 tickets are affected by the spikes
  • Spikes are largely pre-2010.  I found that 2007 and 2008 accounted for 68% of all the spikes.  By 2010, the amount of spikes dropped considerably.  This suggests that whatever the cause of the spikes was, it was largely corrected several years ago.
  • Unlike the Tribune, a spike was not found at Western and Touhy

I have attached to this post, a complete spreadsheet with the results.  

Some caveats:  

 

  • The analysis looks for spikes, this means values that are outside the normal range.  I tried to filter out the low spikes, but if the number of tickets is too low, that is also considered a spike.  
  •  I ran the analysis on both a weekly and daily basis.  At some intersections, the pattern of traffic would result in spikes on a weekly basis, because Saturday could be the busiest travel day.  I did try to control for this (as well as seasonal changes), but you may see this.  
  • The dates provided are approximate with a few days of the spike.
  • UPDATE 8/8/2014 - A small amount of the spikes were the initial values, I updated the analysis to remove those spikes

 

So go find some spikes!  Take a look at the data, see my RLC site, and analyze using a daily or weekly period.  You can adjust the date range to zoom in on specific dates.

Let me know what you find

Sunday
Jul272014

Red Light Camera Analysis

Red Light Camera Analysis offers the ability to see data on accidents that occurred within a 100 feet of a red light camera between 2009 to 2012. The ticket data includes tickets from 2007 to March 2014. Ticket data does not exist for all intersections. The analysis is automated with over 300 intersections, so please let us know if something doesn’t look right.

My goal was to get something up quickly.  Eventually, I will add more features and refine it.  Please drop me a line and let me know if you have any suggestions.  This is just one of many related MiningChi projects on data visualization and statistics.


Wednesday
Jul232014

Its not traffic flow causing the ticketing spikes

To account for the spikes, the city has argued that it could be due to increased traffic flow.  One way to see if traffic flow increases is to look at accidents.  After all, as the flow increases, so does the likelihood of accidents.  So I took a look at the traffic accidents for the first two intersections mentioned in the Tribune story.  I looked at overall accidents and graphed the data by week to make it easier to view.

At the Halsted & 119th intersection, a significant spike occurred between April 29, 2011 and June 19, 2011.  A cursory look at the accident data doesn’t show a similar spike.

 

At the Kimball, Lincoln and McCormick intersection, a significant spike occurred between Dec. 30, 2011 and Jan. 10, 2012.  A cursory look at the accident data doesn’t show a similar spike.

 

I will look at the other intersections this weekend.  But it doesn’t look like traffic flow increases is the source of the spikes.

 

Wednesday
Jul232014

Do more red light camera tickets reduce accidents?

I have created a graphic here that explains the relationship between traffic tickets and accidents at red light camera intersections for Chicago.  The explanation of the graphic is as follows:

  If the purpose of red light cameras is to ticket risky behavior that leads to accidents, then you would expect intersections that give out lots of tickets to have lots of accidents.  After all, more people are breaking the law, so you would expect that as law breaking increases, so would accidents.

    A scatter plot is often used to investigate relationships between two variables. In the first figure (the inset left box), you can see a scatter plot of the number of tickets versus accidents.  Its quite clear there is no relationship between tickets and accidents.  Whether an intersection gives out lots of tickets or few tickets has no relationship to the number of accidents.


    You can see the relationship in second figure between tickets and accidents for red light camera intersections in Chicago.  The data is sorted by the number of tickets given at an intersection.  Please click on the links to see the full graphic on Tableau's website.  This makes it clear that there is no obvious relationship between tickets and accidents. The third figure provides the same data in a map view.


    

The caveats here are that we are measuring whether the overall number of tickets affects accidents.  I only have the aggregated ticket data for four years for these intersections, so I can't analyze trends in the ticket data.  A more nuanced study could look at tickets and accidents over time to see if there was a relationship.  This analysis also does not address whether the presence of a red light camera drops accident rates, that issue is addressed elsewhere.

Data: Accident data from IDOT for 2009-2012, Aggregated ticket data between 2009-2012. A few of the red light cameras were actually put into place in 2009 and 2010.  For these, there would be a short time period where accidents occurred but no tickets could be given out.  Accident counts were determined by looking at total accidents that occurred within a 100 feet of the red light camera.

Credits: Rajiv Shah & Aaron Moore (GIS work) & Barnet Fagel (for acquiring the ticket data from the city)

Tuesday
Jul222014

Spikes in Red Light Camera Ticketing

The Tribune conducted an analysis of 4 million tickets issued since 2007 and found a troubling pattern.  Read the article, it is very well done.

A 10-month Tribune investigation documented more than 13,000 questionable tickets at 12 intersections that experienced the most striking spikes; similar patterns emerged at dozens of other intersections responsible for tens of thousands more tickets. Among the key findings:

Cameras that for years generated just a few tickets daily suddenly caught dozens of drivers a day. One camera near the United Center rocketed from generating one ticket per day to 56 per day for a two-week period last summer before mysteriously dropping back to normal.

Tickets for so-called rolling right turns on red shot up during some of the most dramatic spikes, suggesting an unannounced change in enforcement. One North Side camera generated only a dozen tickets for rolling rights out of 100 total tickets in the entire second half of 2011. Then, over a 12-day spike, it spewed 563 tickets — 560 of them for rolling rights.

Many of the spikes were marked by periods immediately before or after when no tickets were issued — downtimes suggesting human intervention that should have been documented. City officials said they cannot explain the absence of such records.

As you can surmise, there has been outrage (again) over these "deviations".  

I am frankly astonished that this happened. I would have expected RedFlex and the city to monitor violations at intersections.  They should have found these spikes and addressed them right away.  After all, isn't the city reviewing ticketing data to help determined camera placement and effectiveness?  That these spikes continued for such a long period speaks to how the cameras are treated as revenue producing devices and not safety devices.  As long as the revenue flowed, no one in the city or at RedFlex questioned the operation of the cameras. 
Monday
Jul142014

Surveillance in Chicago: Growing, but for what purpose?

I recently wrote a draft chapter covering the history of surveillance in Chicago over the last ten years.  I have presented this numerous times and have finally taken the time to get my thoughts on paper.  It summarizes and brings together a lot of trends in Chicago, such as CLEAR, blue light cameras, red light cameras, heat list, license plate recognition, and the massive fiber optic infrastructure the city owns.
Here is the abstract: 
Chicago is now one of the leading cities in the world in using surveillance technology. This growth began in 2003 when cameras were used to fight street level crime. Since then, Chicago’s camera network has grown to over 25,000 units, along with the use of a variety of new technologies. This chapter explains this growth and how it affects policing.
Feel free to send me comments, this is still a work in progress
Monday
Jul072014

iPhones, CTA, & Crime

A couple of months ago, I wrote about possible reasons why crime on the CTA dropped in the winter of 2014.

Bloomberg recently analyzed smart phones, activatation locks and crime:

Mobile phone technology companies have faced pressure from public officials over the past year to add mechanisms for disabling the devices if they’re lost or stolen to help curb resale potential. More than 30 percent of robberies in major cities involve mobile phones, with smartphones often targeted because of their high value, according to the Federal Communications Commission.

Following Apple Inc. (AAPL)’s release of a kill switch in September, thefts of iPhones in some cities “plummeted,” said New York Attorney General Eric Schneiderman, who helped start a coalition of law enforcement officials and consumer and public safety activists that has prodded the industry to add the theft-deterrence measures.

The Secure Our Smartphones Initiative group said in a report today that robberies involving Apple products in New York dropped 19 percent in the first five months of 2014 compared with the same period last year. In San Francisco and London, robberies involving Apple products dropped 38 percent and 24 percent, respectively, according to Schneiderman’s office.

“Today, the smartphone industry acknowledges that its wonderful products have been driving an international crime wave,” Schneiderman said in a news conference in Manhattan today. “That change in attitude has opened the door to great possibilities.”

Since i suspect smart phones are more of a target for robberies occuring on the CTA then in the city generally, this can partially explain the large drop in robberies on the CTA this winter.


Thursday
Jun122014

First Chicago robber caught via facial recognition gets 22 years

Ars Technica reports that the robber that was caught by facial recognition has now been sentenced.   

The Chicago Sun-Times reported that the Chicago Police Department acquired the technology via a $5.4 million federal grant.

In February 2013, Pierre Martin robbed a man at gunpoint while on a Chicago Transit Authority (CTA) train. After taking the man’s phone, Martin jumped off the train. However, his image was captured by CTA surveillance cameras and was then compared to the Chicago Police Department’s database of 4.5 million criminal booking images. Martin, who already had priors, had a mugshot in the database. He was later positively identified by witnesses.

At trial, Martin also admitted to committing a similar robbery also on the Pink Line in January 2013—his face was captured during both robberies.

Saturday
May102014

CTA Winter 2014 Crime Drop

CTA officials recently credited new security cameras for starting 2014 with fewer robberies and thefts on its buses, trains and L platforms.  The CTA claims a 38% drop in robberies and thefts dropping 23%.  On its face, this sounds impressive and a validation of the cameras.  After all, the idea that cameras can reduce robberies and thefts is sound.  But before we get carried away, lets dig a little deeper.

CAMERAS: The CTA has been adding cameras for the last five years.  I think the count is around 5,000 cameras for the CTA.  They just didn't start rolling out in 2014, so it doesn't make sense to attribute the drop solely due to cameras.

LONG TERM TRENDS: Lets consider the history of crime on the CTA.  A short review of CTA stories on my blog finds plenty of stories about CTA crime going up.  A key factor in the robberies and thefts is the shift to smart phones and mobile devices in the last five years.

A Tribune analysis in 2012 found that between 2009 and 2011 there was a surge in thefts and robberies. They also found that arrests were made in less than 4 percent of thefts and about 15 percent of robberies.  In January 2014, NBC Chicago found that thefts on the CTA have gone up more than 68% percent over the past five years.  So CTA is a magnet for robberies.

Its safe to say over the last 10 years, thefts and robberies have probably increased on the CTA as opposed to overall crime. 

CRIME DATA:  Interestingly, both of the studies by the media are critical of the data provided by the CTA.  CTA data consistently seems to always to underplay the extent of crime.  This is very disappointing and has become a trend in Chicago.  Its why I am forced now to carefully scrutinize the data provided by the city.

CTA DATA:  This analysis focuses on the totals for thefts and robberies the first three months of 2014 as compared to 2013. The Tribune found faults (again) with how the CTA counts thefts and robberies.  The Tribune’s data instead showed (actually higher) drops in crime with robberies falling 43% to 112 and thefts falling 20% to 423.  For comparison, in the city overall, robberies fell 24% and thefts fell 17%.  Comparing the CTA to Chicago, we see that the fall in thefts was similar to the general crime trend, however, there is still a large difference in robberies.  So clearly, some factor has dropped the robberies more on the CTA. 

ANALYSIS:  Just based on this data, its not clear why robberies dropped much faster.  The cameras have been installed for many years, so unless the police just started using them in the past three months, they shouldn’t be a factor.  So what else could it be?

Weather - We know the weather was extremely cold in 2014, so could that have lead to a drop in robberies?  Perhaps people kept their mobile devices in their pockets and robbers were less tempted?

Regression to the Mean - We know that robberies have grown much faster than the general crime trend, so could they be returning to their overall average?

Increased police presence/surveillance - Could this drop be due to a focus by police on robberies?

To answer these questions we will have to continue to watch the robbery numbers.  If the robbery numbers go up in summer or during a mild winter next year, then weather appears to be a factor.  If the robbery numbers continue to drop during the summer, then it could be due to increased police presence/surveillance.

In New York, robberies were down 50% for their transit system (MTA) in March 2014 compared to March 2013.  I didn’t check any other cities, but this could point to a factor such as weather or declining interest in smart phones as a larger global trend affecting robberies. 

Wednesday
May072014

Critique of Chicago's DOT studies on red light cameras

The city maintains an page with background on the red light cameras. In it, the city claims:

Analyses of Chicago’s red-light equipped intersections conducted by CDOT found that dangerous angle crashes were reduced by an average of nearly 30% when a high angle crash rate intersection was equipped with red-light cameras.  Rear-end crashes were found to increase on average at red-light camera equipped intersections.  Rear-end crashes are more likely to result in minor injuries or property damage.  The safety goal of the red-light program remains focused on reducing the most dangerous crashes.  Rear-end crashes tend to decrease in frequency as driver behavior changes over time to comply with the red-light traffic laws.

The two different analyses were done in September 2010 and August 2011.  

The first study looks at 50 intersections where cameras were installed between 2006 and 2008. The study then compares the accidents which occured two years before the cameras were installed and two years after the cameras were installed. This study finds accidents dropped 9.8%.

The second study looks at 106 intersections where cameras were installed between 2006 to 2008. The study then compares the accident rate in 2005 compared to the accident rate in 2010 for the intersections. This study finds accidents dropped 8.22%.  

Lets look at the second study first.  It appears the red light cameras are making a difference. However, if we think about it, how do we know accidents dropped because of the red light cameras. What if there was another factor, say people driving less, that could account for the drop? To control for this, we need to know what the general trend was for accidents in Chicago.  In 2005 there was 119,133 accidents and by 2010 this dropped to 80,922 acccidents, so reduction of 32%!  Once we take this into account, we realize that actually intersections with red light cameras became more dangerous relative to the rest of the city.

So at this point, it should be obvious that either the red light cameras dramatically increase accidents or there is something wrong with the data.  In this case, its a problem with the data.  Starting in 2009, Illinois changed the definition of an accident from $500 to $1500 in damage. By increasing the dollar amount, this is going to reduce what qualifies as an accident.  In fact, between 2008 and 2009, accicents dropped 36%, which is largely due to the change in the definition of an accident.

The flaws with the first study should also become obvious now.  Since it relies on data from 2009 and 2010, that data is going to be significantly biased towards a reduction in accidents.  As a result, you can’t compare the raw numbers after 2009 to prior years.  This is why my study did not include 2009 data.  I am currently working on another study, but that will only use data after 2009.

The bottom line is the Chicago's DOT analysis is useless.  Its comparing apples to oranges, since the definition of accidents changed during the time period of the analysis.  If the city really believes this is an accurate assessment of the red light cameras, then it shows red light cameras are actually a significant danger because accidents have only fallen 8%, while accidents in the city dropped 36%.