Title: Observational Surveys: Implementation and Analysis
1Observational Surveys Implementation and Analysis
2Observational Surveys - Speakers
- William W. Stenzel, D.Sc.
- Associate Director, Center for Public Safety
(Management Consulting) - Roy E. Lucke
- Director, Research and Development for the Center
for Public Safety
3Notebook Materials
- A hardcopy of all of the PowerPoint slides for
this session (Observational Surveys) can be
found in the seminar notebook.
4Why Use Observational Surveys?
- For the Illinois Traffic Stop Statistics Study,
disparities are going to be calculated by
comparing - the racial composition of traffic stops
- the racial composition of the driver population.
- The racial composition of the driver population
is going to estimated by using adjusted census
data at the city and county level.
5Why Use Observational Surveys?
- What options does an agency have if there is
concern that the adjusted census data will not
provide accurate information about the racial
composition of the driver population in its
jurisdiction? - One option is to obtain a better estimate of the
racial composition of the driver population with
the use of observational surveys.
6Center for Public Safety Experience with
Observational Surveys
- The Center for Public Safety has conducted three
observational surveys for agencies in Illinois - Highland Park (Sept-Oct 2001 Stenzel)
- Hinsdale (May 2004 Lucke)
- Schaumburg (August 2004 Lucke)
7Observational Survey Topics
- Observational Surveys
- Topic 1 Data Collection
- the nuts and bolts of how to conduct a survey
- Topic 2 Data Summarization
- putting the survey data into a format suitable
for review and analysis - Topic 3 Data Analysis
- comparing and assessing the survey and traffic
stop data
8Observational Surveys - Topic 1
- Topic 1 Data Collection
- The nuts and bolts of how to conduct a survey
9Conducting Observational Studies
-
- Once the decision is made to do an observation
study, there are three major tasks - Determining what data to collect
- Identifying data collection sites
- Recruiting and training observers
10Conducting Observational Studies
- In addition to the three major tasks, other steps
include - Scheduling data collection
- Equipping the data collectors
- Developing forms, data entry and data analysis
procedures
11Preliminary Information
- It is only possible to obtain good driver
demographic information from stopped vehicles - Observations can only be made at intersections
with traffic signals or stop signs - Efforts to observe drivers on controlled-access
roadways were not successful
12Site Selection
- Primary Criteria
- Conduct observations at or near intersections
that are among the agencys high traffic stop
locations. - Agency should try to identify locations for all
stops, not just where citations are issued - Also identify times of day and days of week for
stops so observation times can be matched as well
as possible
13Site Selection, Continued
- Site must provide a good view of stopped vehicles
- Steep shoulders may raise observers too high
- Sweeping right turn lanes might keep observers
too far from lanes to see - No limit on number of lanes observers need only
check lanes they can clearly see - Site must be safe for observers
- There must be a shoulder or sidewalk curbs are
desirable - Observers must be free from potential harassment
14Survey Sessions
- Session is a 2 or 3 hour observation period
- Sessions should be distributed across all days of
the week, according to traffic stop information - Sessions can be done
- Mornings
- Afternoons
- Early evenings
- Again, dependent on stop information and
available daylight - Surveys should be done in both directions of
travel
15The Survey Team
- Three individuals are needed for each session
- Observer
- Recorder
- Counter
16The Survey Team, Continued
- Team members can be recruited from a number of
possible sources. - Agency volunteers or auxiliaries
- College students (e.g., criminal justice
students) - Crossing guards
- Temporary labor pools
- Etc.
17The Survey Team, Continued
- Training must be provided to survey team members
- Classroom instruction covering the nature of the
project and what they will be expected to do - Practice sessions under guidance of project
leaders - Team members must be scheduled in groups of
threes at dates and times identified for surveys.
- Have substitutes available
- Project leaders should oversee all observation
sessions
18Sample Agenda for Observer Training
- Agenda
- Review Agenda
- Complete Forms
- Driver Survey
- Vehicle Counter
- Field Work
- Schedule/Signups
19Survey Team Equipment
- Safety vests
- Traffic counting devices (or digital cameras)
- Clipboards and pencils
- Rain gear (ponchos, umbrellas, writing pouches)
20Data Items to be Recorded
- Each agency must decide what data items they
believe are important to capture. Candidates
items include - Driver race/ethnicity
- Driver gender
- Driver age
- Number of passengers in vehicle
- Driver residency
- Type of vehicle
21Data Collection
- Paper check mark form
- Scantron form
- Palm or other hand-held device
- Tablet-type personal computer
- Each session should be stored as a separate file,
either in a physical packet or data file
22Sample Data Collection Form
23Number of Observations
- The number of data collection sessions can be
affected by - Traffic volume
- Roadway configuration (number of lanes)
- Stop signs or traffic signals
- Number of data items to be collected
- General observations
- Higher capture rates (75 - 100 of drivers) at
stop signs, but usually lower traffic volumes - Lower capture rates (20 - 75) at signalized
intersection depending on volume, number of
lanes, and signal timing
24Observation Limitations
- Can be done only during daylight hours
- Glare from windows (or window tinting) can affect
observation - Weather (rain, snow, excessive heat)
- Subjective decisions be observers
- Cost of doing surveys (labor intensive activity)
25Topic 2 Data Summarization
- Data Summarization
- Recordkeeping
- Data Entry
- Data Base Software
26Topic 2 Data Summarization
- Recordkeeping - additional information that
should/can be added to each observation (record) - Location (should)
- Day of the week (should)
- Time of day (should)
- Direction of traffic (optional)
- Data collectors (optional)
27Topic 2 Data Summarization
- Data Entry - getting the data into an electronic
format - The old fashion way keying the data in
- Use machine-readable data collection forms (e.g.,
Scantron) - Download from a file created at the time of data
collection (e.g., from a Palm Pilot or a PC
tablet)
28Topic 2 Data Summarization
- Data Base Software - a computer program that can
be used to - Manipulate the data (i.e., sort and filter)
- Display the data (i.e., print summary tables and
charts) - Describe the data (i.e., compute various
descriptive attributes) - number of observations
- Average value
- Minimum and maximum values
- Examples (Access, EXCEL)
29Topic 2 Data Summarization
30Observational Surveys - Topic 3
- Topic 3 Data Analysis
- Comparing and assessing the survey and
- traffic stop data
31Topic 3 Data Analysis
- Data analysis consists of comparing two sets of
data - Traffic stop data
- Driver survey data
- And addressing the question Are differences
between the two sets of data important?
32Data Analysis A Sample Comparison
5
Two data sets
- Question Are the differences in the percentages
between the traffic stop data and the driver
survey data in the each racial category
important? - Are differences due only to natural variation, or
- Are differences due to the some outside influence
on the officers decision about whom to stop
(e.g., race)?
33Statistical Benchmarking
- Statistical benchmarking consists of
- Comparing two sets of data
- Encounter data the racial composition of drivers
in traffic stops, and - Survey data the racial composition of drivers
who are potential participants in a traffic stop - A procedure for assessing the significance of
differences in the percentages between the two
data sets
and
34Statistical Benchmarking
- Highland Park
- Statistical benchmarks were used to assess the
importance of the differences in the percentages
in the driver survey and traffic stop data. - The benchmarks were determined using a
statistical procedure called confidence
intervals.
35Confidence Interval Example
- Example Is the difference between the two
percentages for Hispanics (i.e., 24.0 and 18.6)
important? - One way to address this is to determine a range
of values (i.e., a confidence interval) for the
expected number of traffic stops involving
Hispanic drivers.
36Confidence Interval Example
- Example The confidence interval for the number
of Hispanics in the traffic stop data shown above
is - 64, 96.
- This interval can be interpreted as follows
- If the decision about who to stop is not
influenced by race, then the expected number of
Hispanics stopped, due to normal variation,
should fall between 64 and 96.
37Confidence Intervals Example
- The upper and lower limits for the confidence
interval can be interpreted as statistical
benchmarks for the number of Hispanics stopped. - The limits are determined based on
- Total number of traffic stops (408)
- Estimated number of Hispanics in the driver
population - Selected confidence level
38Confidence Interval Example
1
- Example The statistical benchmarks for the
expected number of Hispanics, 64, 96, is based
on a confidence level of 95. - The 95 confidence level means the margin of
error is 5. - A 5 margin of error means that there is 5
chance that even with normal statistical
variation the number of Hispanics stopped could
fall below 64 or above 96.
39Statistical Benchmark Example
2
- The upper and lower benchmarks for each racial
category are shown at the bottom of the table. - These benchmarks are compared with the actual
number of encounters in each racial category. - Except for Hispanics, the actual number of stops
within each category falls within the benchmark
limits.
40Statistical Benchmark Example
- The number of Hispanics stopped in this example,
98, is outside the statistical benchmarks of 64,
96. - THIS DOES NOT PROVE RACIAL PROFILING.
- It indicates that further investigation is needed
to determine what special circumstances might be
present that are influencing the number of
Hispanics that are stopped.
41Why Use 95?
- Use of 95 for the confidence interval is a
conservative approach that assumes that racially
motivated policing is not occurring unless there
is significant evidence to the contrary. - Justification for a conservative approach is
appropriate in view of the many uncertainties
associated with the data - Difficulty in identifying race
- Different driver behaviors by race
- Different driver behaviors by gender and age
- Unknown mix of drivers by gender and age by race
42How Can I Use Statistical Benchmarks?
- The benchmarking procedure described is based on
statistical procedure called the two-sample test
for proportions. - Its use requires a basic understanding of applied
statistics. (Note Statistical Benchmarks for
Police Traffic Stops in seminar notebook.) - To help departments that may want to use
statistical benchmarking based on this procedure,
the Center for Public Program has put an
easy-to-use spreadsheet on its website that can
be used to find statistical benchmarks.
43Statistical Benchmarking Spreadsheet
44Statistical Benchmark Spreadsheet
- The statistical benchmarking spreadsheet can be
found on the website for the Center for Public
Safety - www.northwestern.edu/nucps
- Select Links
- Select Racial Profiling
- At bottom of page under Recent Articles find
- Benchmarking Spreadsheet
45Contact Information
- William Stenzel
- 847/491-8995
- wstenzel_at_northwestern.edu
- Roy Lucke
- 847/491-3469
- rlucke_at_northwestern.edu