Reporting and Analysis
Benchmarks
Finding an appropriate benchmark against which to compare traffic citations or warnings is one of the most difficult parts of the data collection effort. By themselves, the data collected by officers about the demographics of traffic stops say very little. For example, if a jurisdiction discovers that 45% of its traffic stops on a particular highway are of Hispanic drivers, that number by itself does not reveal very much. Instead, the agency would want compare the proportion of stops to an appropriate benchmark or base rate that reflects the demographics of the driving population. There is no clear consensus, however, about the most statistically sound population against which to compare traffic stops. In Racially Biased Policing: A Principled Response, the Police Executive Research Forum recently noted that the creation of an accurate benchmark is at best a "very challenging endeavor" (2001).
With this in mind, there are a number of benchmarks that can be used to analyze the data, and different benchmarks seem better suited to different jurisdictional situations. Departments may choose to compare traffic stop demographics against external benchmarks or internal benchmarks. Additionally, departments may choose to conduct a demographic analysis of search data as one measure of racial differences in post-stop activity.
External Benchmarks
To date there is little agreement about which type of external benchmark is most appropriate for the analysis of traffic stop data. The following sections on population data, modified population estimate, and traffic surveys provide a brief overview of some of the methodologies currently being used to construct external benchmarks for traffic stop data.
Population Data
Some jurisdictions have sought to use residential population data, broken down by race, to estimate the racial percentages of persons using the jurisdiction's roads (see: San Diego Police Department, 2001; San Jose Police Department, 2001). This methodology may be proper when used appropriately. Census data is more appropriate for jurisdictions that experience low traffic flow from outside the local area.
When using census data it is important to first ensure that the population data are sufficiently current. The 2000 Census data is obviously the most appropriate measure of population demographics. Second, it is important to only use residential data for that portion of the population that is of driving age. Because the age demographics for different racial groups may vary, it is vital that the residential benchmark includes only individuals who are of legal driving age. Other available data concerning a person's access to vehicles reported by race may help jurisdictions refine the residential population benchmark even further. Finally, residential population benchmarks are least appropriate for examining the racial demographics of individuals stopped by the police who reside outside that particular jurisdiction. In order to most effectively utilize census population data the analyst must be able to partition out all stops for individuals who reside outside of the local jurisdiction, comparing only residential stops to residential population data. As a result, it is important to have some indication of driver residence in the data collection protocol if one plans to use census data during analysis.
Modified Population Estimate
Residential population data can also be modified to capture information about the racial demographics of surrounding communities. To compute an adjusted racial demographic, analysts would first divide the population (16 and over) of each racial group in the surrounding communities by their distance (in miles) from the target community. This weighted population data would result in an adjusted demographic composition for the jurisdiction that can be compared to the demographics of traffic stop data.
Preliminary analysis of traffic stop data has utilized both the residential census data and the modified residential population estimates as comparative benchmarks to construct an "index of disparity." It is important to note that such an index may be useful to assess the magnitude of differences between citation demographics and census and/or modified census estimates for each jurisdiction, but it does not provide a definitive measure of profiling. Other intervening variables such as time of day, location or shift might explain the existence of such racial disparities.
Traffic Surveys
Census data alone is an inappropriate or at best limited measurement tool for some agencies because they experience a heavy volume of commuter traffic from drivers who do not reside in the local jurisdiction. Alternatively, some analysts have utilized traffic surveys to determine the racial makeup of individuals, and in some instances violators, on interstate roadways. To survey the population of individuals traveling on a given roadway, analysts have used both stationary surveys where teams of observers are standing on street corners noting the race of drivers at intersections and rolling surveys where teams of observers travel on the highways noting the race of drivers as they pass the test vehicle. While stationary and rolling road surveys are becoming a more acceptable method of assessing driving populations, they can be both costly and time consuming, particularly for studies involving multiple agencies.
Other analysts have tried to construct measures of the demographics of a driving population using existing information on traffic accidents. In these studies analysts have compared traffic stop demographics to the existing demographic information from traffic accident data (see: Washington State Police, 2001; San Diego Police Department, 2001).
Internal Benchmarks
Although the construction of an external benchmark is a difficult task, there are a variety of internal analyses that are more easily developed and produce very useful information. The first set of internal analysis that could be conducted would be to use the stop demographics of individual officers or groups of officers as the baseline against which to compare stops. For example, a jurisdiction could compare traffic stop data for the same unit (or the same officer) over time, or could compare that data for several units (or several individual officers) that patrol the same area of similar areas. This requires the data collection system to collect officer level information, or at the least it must collect unit level information.
Information about officer or unit behavior is virtually invisible in most jurisdictions at this time. By collecting such information a department can identify the typical pattern of behavior of its officers and discern if outliers exist. This kind of analysis could be done for individual officers or for individual neighborhoods. Data collection could determine for example that the typical officer stops ten cars per shift and issues four citations. Once this information is known, the behavior of all officers can be evaluated in terms of this measure. If an officer is stopping fifty cars in a shift, she may be an officer who is working very hard in an area or she may be causing increased community resentment in a particular neighborhood.
A final benefit of the collection of officer specific or unit specific information is the ability to track changes over time. If departments were regularly collecting information on the characteristics of traffic stops, they would be able to detect trends in the use of this law enforcement tactic. If for example, the number of stops decreased drastically in one section of a community, the data would alert officials to this change. Similarly, if in one area there was a marked increase in stops of Asian motorists, a department could investigate to determine the cause of this increase. In addition, having this data would allow departments to respond more quickly to complaints from community groups of racial profiling.
Analysis of Searches
Analyses may also be performed to determine demographic differences in the activities that occur once a car is stopped and the driver is issued a citation (e.g. searches). No external benchmarks are needed for this analysis because the analyst has the whole universe of cars that are cited against which to compare the post-stop activity. For example, searches may be analyzed to determine if minority drivers who are cited are disproportionately searched as compared to white drivers who are cited.
In order to understand the relationship between search/citation patterns and search outcomes, it might be useful for local jurisdictions to be able to connect traffic stop data collection information to routine search forms or arrest sheets. If a large number of black motorists are stopped and searched but the subsequent searches of black motorists yields much lower hit rates than those searches conducted on individual of other racial groups a department would be able to flag areas of concern and follow them more closely. Additionally, connecting citation data with search information can provide agencies with important information about both the quality and the quantity of contraband found to determine any racial disparities in search patterns.
