By Jack Willoughby   The Effect of Public Transportation on Crime

I. Introduction

In response to population growth, scarce natural resources, and the growing atmospheric problems caused by greenhouse gasses, many cities have sought to reduce the number of drivers on the road to limit both traffic and fossil fuel consumption. One theoretical means of reducing the number of cars on the road is to expand the use of public transportation. Such an expansion can have unintended negative effects, however, if it results in an increase in the prevalence of crime near the new transit stations or bus stops. The goals of this paper are to present a model of how public transportation affects crime, to demonstrate how theory has been confirmed or challenged by previous studies, and to use data related to Durham’s Bull City Connector to test the model. Analysis of available crime data indicates that the Bull City Connector has had no noticeable effect on the total amount or distribution of crime near its route.

II. Theoretical Model

As first presented by Ihlanfeldt (2003), the effect of public transportation on crime can be spatially modeled and decomposed. The expected value of a crime (π) to the person who is committing the crime is a function of the expected payoff of the crime (w), the cost of committing the crime (c), and the expected punishment of being caught, which is equal to the product of the probability of being caught (p) and the expected penalty conditional on being caught (f). The expected value of committing a crime in neighborhood H is modeled in (1):

Screen Shot 2014-04-09 at 10.00.25 PM

Furthermore, (c) can be decomposed into the accounting costs of committing a crime, (b), the transportation costs to the crime (tc), and the opportunity costs of committing a crime. Accounting costs of committing a crime include the cost of buying a weapon or tool needed for the crime, the cost of buying disposable black clothes or a mask, the cost of paying someone for help, etc. The opportunity costs of committing a crime are the forgone welfare that a criminal could have earned through legitimate activities plus the value of improved mental wellbeing from not being a criminal (g). These forgone earnings are equal to the income from legal employment (e) minus the transportation costs of traveling to a job (tj). The costs of committing a crime are modeled in (2):

Screen Shot 2014-04-09 at 10.03.07 PM

Aggregating (1) and (2), the net expected value of committing a crime is (3):

Screen Shot 2014-04-09 at 10.03.34 PM

Transportation costs can also be broken down into accounting and opportunity costs. The accounting costs are the sun of the monetary costs of transportation, (m), which include the gas needed to drive, depreciation of a motor vehicle, and public transportation fare. The opportunity costs are the time it takes to make the journey, which can be decomposed into the value of time (v) times the amount of time traveled, which equals distance (d) divided by speed (s). This relationship is modeled in (4):

Screen Shot 2014-04-09 at 10.05.36 PM

Subscript (i) indicates the trip being taken. As noted above, subscript (j) will indicate the trip to a job, and subscript (c) will indicate the trip to commit a crime. Incorporating this into (3) yields (5):

Screen Shot 2014-04-09 at 10.06.47 PM

To determine the effect of public transportation on crime, we need to take the partial derivative of (5) with respect to transportation (T) and drop all terms that are not affected by a change in public transportation. The result of this marginalization is modeled in (6):

Screen Shot 2014-04-09 at 10.08.39 PM

Next, we can decompose crimes into two types: crimes committed in outside neighborhoods (O) and crimes committed in the neighborhood in which the criminal is a resident (R). Adding these geographical distinctions as superscripts, we can revise (6) into (7):

Screen Shot 2014-04-09 at 10.09.12 PM

An expansion in public transportation that results in its increased use implies that the marginal effect of the expansion is to decrease the total transportation costs of its riders, but it does not indicate if the mechanism through which it decreases total costs is through monetary or time costs. With one decision (whether or not to use public transportation) and two decision factors (monetary costs and time costs), the relative contributions of effects of public transportation on speed and money are unidentifiable. Theoretically, however, it is conceivable that improvements public transportation decrease both time and monetary costs. Public transportation is often highly subsidized and therefore cheap, and in cities where traffic and parking are time consuming, it is likely also faster than alternative methods of transportation. Given that the addition of new public transportation is both faster and cheaper than previous methods of existing transportation, its marginal effect on crime would be threefold. The addition of public transportation would:

(A) Decrease the transportation costs of committing crimes in distant neighborhoods that contain public transportation stops;

(B) Increase the relative transportation cost of committing crimes in criminals’ own neighborhoods relative to remote neighborhoods; and

(C) Decrease the transportation costs associated with having a legitimate job.
Therefore, assuming that the new public transportation is cheaper and faster than previous modes of transportation, its three theoretical marginal effects should be to decrease crime in areas in which criminals live, increase crime near public transportation stops in neighborhoods not previously populated with criminals, and decrease total crime levels as legitimate employment becomes a more attractive substitute to crime. The latter two effects operate in competing directions, so while crime should certainly decrease in areas currently populated by criminals, it theoretically will change ambiguously in areas not currently populated by criminals, depending on the relative contributions of the effects on crime due to (A) and (C).

III. Previous Research

The theoretical predictions of the effect of public transportation on crime have been tested empirically with mixed results. Plano (1993) found no significant relationship between proximity to rail transit stations and crime using data from the opening of stations in the Baltimore Metro system. Block and Block (2000) found a significant positive relationship between proximity to subway stations and street robberies in both the Bronx borough of New York City and the Northeast Side of Chicago. Liggett et al. (2003) found that the Green Line light rail transit system in Los Angeles had no effect on the overall levels of crime and the spatial distribution of crime in Los Angeles. Ihlandfeldt (2003) analyzed Atlanta’s MARTA system to find that the addition of rail transit stations resulted in the increase of crime in the center city, but did not affect crime levels in the outer limits of the city. Denver Regional Transportation District (2006) analyzed crime patterns near stations of the Central Corridor light rail transit line in Denver and found no evidence of an increase in crime as a result of its inception. SANDAG (2009) studied the expansion of the Green Line transit system in San Diego and found that both crime rates and distribution of crime were unaffected by the expansion of the public transportation system, as well as that residents did not feel more or less safe as a result of the expansion. Taken together, these studies indicate variation in the effects of public transportation, suggesting that specific cities respond differently to its expansion. Additionally, the rare statistical significance suggests that the magnitude of the effect of public transportation on crime is likely small.

IV. Case Study: Durham’s Bull City Connector

A. Introduction
On August 16, 2010, the Bull City Connector (henceforth BCC) began operation in the city

of Durham, NC. In conjunction with Duke University, Durham launched the BCC to more seamlessly connect Duke and downtown Durham. During operating hours, the bus runs every 15-20 minutes, depending on the time of day and day of the week, and it is completely free to the public. It stops 34 times on its loop bounded by Duke’s Central Campus on one end and Durham’s Golden Belt District on the other, passing through Brightleaf Square and downtown Durham along the way. The full route is shown in Appendix 1. Using available data, the effect on crime of this exogenous improvement in public transportation will hereafter be investigated.

B. Data and Empirical Method
First, regions of interest were identified, the boundary maps of which are contained in

Appendix 2. The theoretical model establishes a differential effect on crime of public transportation in neighborhoods in which criminals live versus neighborhoods not inhabited by criminals, so neighborhoods of different home values were studied. The first area studied was “Trinity South,” which is comprised of the neighborhood bordered by N Buchanan Blvd on the west, Urban Ave on the north, N Duke St on the east, and W Main St on the south. This region is directly to the north of the BCC route. Using estimates from, most houses in the region are valued between $300-450K. A control for this region, which is comprised of houses of similar value but is not immediately adjacent to the BCC’s route, is the “Trinity North” neighborhood. This region is to the north of Trinity South, and is bordered by N Buchanan Blvd to the west, W Club Blvd to the north, Ruffin St to the east, and Green St to the south. Next, low wealth neighborhoods were identified, both near and removed from the BCC route. Adjacent to the BCC’s route is the “Holloway” Neighborhood, which is comprised the residential areas surrounding Holloway St bordered by N Roxboro St to the east and N Alston St to the west. Using estimates, a large share of these houses are valued between 50-150K. Of similar value are the houses immediately north of North Carolina Central University, in the “North NCCU” neighborhood. This neighborhood is isolated from the BCC by Rt. 147, and is comprised of the residential areas bordered by Fayetteville St to the west, Linwood Ave and Simmons St to the north, S Alston St to the east, and Dupree St to the south.

Crime data on these four regions from the Durham Police Department Crime Mapper were used in analysis. The Crime Mapper displays spatially identified crimes across Durham for every month from January 2010 to the present. For this analysis, the total number of crimes in each month from January to July in both 2010 and 2011 was recorded for each of the four areas described above. The data from 2010 illustrate crime before the inception of the BCC, while the data from 2011 describe crime immediately after its inception. The analysis was confined to this limited dataset because no information on crimes before 2010 is available on the Crime Mapper.

A difference-in-difference analysis was used to identify the effect of the BCC on crime. The first stage of the difference-in-difference was to subtract the number of crimes in a region in a given month in 2010 from its corresponding month in 2011. Months were compared to corresponding months throughout the analysis to control for any seasonal variation in the amount of crime in Durham. This first difference yields the change in crime from before to after the launch of the BCC in each of the four regions. Next, a second difference is needed to control for any city-wide changes in crime that were occurring in Durham independent of the inception of the BCC. To control for these exogenous changes, changes in crime in areas geographically removed from the BCC were subtracted from the areas of interest adjacent to the BCC route. Furthermore, to attempt to balance the city-wide changes in crime as well as possible, control areas contained houses of similar value to the areas of interest. In keeping neighborhood quality constant, the relationship between holistic changes in crime in Durham and quality of neighborhood will be roughly accounted for.

C. Results
The theoretical model developed above predicts that crime should certainly decrease in areas

currently populated by criminals, and it theoretically will change ambiguously in areas not currently populated by criminals, depending on the relative effects on transportation costs to a job vs. to commit a crime. The BCC meets the assumptions of the model since it is free, so the monetary costs of travel will certainly decrease, and it travels to areas in which parking is often time-consuming, so it plausibly increases the speed with which people can travel. Assuming that criminals are more likely to live in lower value homes, more criminals per capita will reside in the poorer areas than the wealthier areas. Therefore, the theoretical model predicts that crime will increase in the wealthier area adjacent to the BCC, as compared to the control wealthy area. It predicts ambiguous results in the poorer area. The difference-in-difference results are displayed in table 1:

Screen Shot 2014-04-09 at 10.10.09 PM

As displayed in the table, there is no noticeable relationship between crime and the introduction of the BCC. In wealthier residential areas, crime in close proximity to BCC increased more after the launch of the BCC than in areas farther from the BCC in 3 out of 7 months, and in poorer areas it increased in 4 out of 7 months. Furthermore, neither result is statistically significant from zero (p- values = 0.9282 and 0.2378, respectively). Therefore, the theoretical hypotheses of the inception of the BCC are not confirmed by analysis of the available data.

D. Discussion
The lack of observed relationship between the introduction of the BCC and crime rates in

nearby neighborhoods could be due to the small sample of data used in analysis or the reality that the BCC actually does not affect crime in Durham. First, data were only used for 4 regions over 14 total months. The unavailability of data from before January 2010 confined the amount of data accessible for use in the study, but if more data were available, a potentially more significant result could be deduced. More likely, however, is that the BCC does not actually affect crime in Durham. The bus system shuts off at 10pm on Monday-Thursday, and at midnight every other day of the week, and is therefore not operational in the hours of the night when crime may conceivably take place. Also, from end to end, the BCC route is only about 3.5 miles long, so it may not provide criminals access to areas that were previously inaccessible. Duke University and Medical Center are at one end of the route, and I would suspect that people considering working at these employers would find a decrease in transportation costs, which might decrease overall crime as criminals turn to legitimate employment. An unlikely but possible explanation is that both of the regions studied were not home to criminals, and as theory would predict the shift in the distribution of crime toward these areas caused by the BCC was washed out by the decrease in the total number of criminals as former-criminals shifted to legitimate employment. Finally, as Durham’s once-impoverished areas improve, there may not be enough of a discrepancy between the number of criminals residing in the wealthier and poorer areas to establish any differential effect of the addition of the BCC.

V. Conclusion and Extensions

Many public officials believe that expanding public transportation is a solution to the existing problems of air pollution and excessive automobile traffic that current transportation methods have precipitated. Residents of currently peaceful neighborhoods counter this sentiment by arguing that the expansion of public transportation will provide criminals access to their homes. Theoretically, expansion of public transportation that results in a decrease in transportation costs should result in the shift of crime from neighborhoods in which criminals live to other neighborhoods, and also a shift from crime to legitimate employment as the travel costs associated with having a job decrease. Empirically, this relationship has rarely been found in the past, either due to a flaw in its theoretical development or a lack of available data. To test the application to Durham, data from the introduction of the Bull City Connector were collected and analyzed, but no relationship was found between the proximity to the Bull City Connector and change in crime rates after its inception.

The consideration of how changes in a city affect crime is a subject that has not been extensively researched, largely because isolating spatial effects on crime requires unique natural experiments and geo-coded data. As a result, there are many extensions for future research, even in Durham. For example, how did the construction of the Durham Bulls Park affect surrounding crime? How will the construction of Durham’s first skyscraper affect crime? How has the gentrification of Durham affected the spatial distribution of crime? Before policy decisions are made in the future, government officials should consider how their actions might change the incentives that prompt criminals to commit crimes. Only after assessing the externalities on crime, among other factors, can one truly value the effect of a change in Durham.

Appendix 1: Bull City Connector Route (photo taken from Google Maps)

Screen Shot 2014-04-09 at 10.13.20 PM

Appendix 2: Areas of Data Collection (photos taken from Google Maps)

Screen Shot 2014-04-09 at 10.15.49 PM

Appendix 3: Data

Data are on the number of crimes committed in the corresponding area and time period. Crimes reported are arson, assault, burglary, homicide, larceny, motor vehicle theft, robbery, and rape. Regions are defined in Appendix 2.

Screen Shot 2014-04-09 at 10.16.24 PM


Block, R., & Block, C. “The Bronx and Chicago: Street robbery in the environs of rapid transit stations.” In V.

Goldsmith, P. McGuire, J. Mollenkopf, & T. Ross (Eds.), Analyzing crime patterns: Frontiers of practice. Thousand
Oaks, CA: SAGE Publications, Inc. (2000): 137-153.
Denver Regional Transportation District. “Technical Memorandum: Neighborhood vs. Station Crime Myths and Facts.”
Durham Police Department. “Crime Mapper Online Software.” Durham, 2010-2011.
GoTriangle. “Bull City Connector: First Year Report, August 2010-2011.”
Ihlanfeldt, Keith R. “Rail Transit and Neighborhood Crime: The Case of Atlanta, Georgia.” Southern Economic Journal 70.2
Liggett, Robin, Anastasia Loukaitou-Sideris, and Hiroyuki Iseki. “Journeys to crime: assessing the effects of a light rail
line on crime in the neighborhoods.” (2003).
Plano, Stephen L. “Transit-Generated Crime: Perception Versus Reality–A Sociogeographic Study Of Neighborhoods
Adjacent To Section B Of Baltimore Metro.” Transportation Research Record 1402 (1993).
SANDAG. “Understanding Transit’s Impact on Public Safety.” (2009).