Highway Safety: Further Opportunities Exist to Improve Data on	 
Crashes Involving Commercial Motor Vehicles (18-NOV-05, 	 
GAO-06-102).							 
                                                                 
Large trucks make up 3 percent of the nation's registered	 
vehicles, but they were involved in 11 percent of all fatal	 
crashes in 2003. To reduce the fatality rate, the Federal Motor  
Carrier Safety Administration (FMCSA) sets national safety goals 
and works in partnership with states to reach them. Crash data	 
collected by states and submitted to FMCSA is key to these	 
efforts, and to be fully useful, this data must be complete,	 
timely, accurate, and collected in a consistent manner. GAO	 
addressed (1) what is known about the quality of commercial motor
vehicle crash data, and what states are doing to improve it, and 
(2) the results of FMCSA's efforts to help states make		 
improvements.							 
-------------------------Indexing Terms------------------------- 
REPORTNUM:   GAO-06-102 					        
    ACCNO:   A41780						        
  TITLE:     Highway Safety: Further Opportunities Exist to Improve   
Data on Crashes Involving Commercial Motor Vehicles		 
     DATE:   11/18/2005 
  SUBJECT:   Data collection					 
	     Data integrity					 
	     Federal/state relations				 
	     Highway safety					 
	     Motor carriers					 
	     Motor vehicles					 
	     Safety regulation					 
	     Safety standards					 
	     Traffic accidents					 
	     Transportation safety				 
	     FMCSA Safety Data Improvement Program		 

******************************************************************
** This file contains an ASCII representation of the text of a  **
** GAO Product.                                                 **
**                                                              **
** No attempt has been made to display graphic images, although **
** figure captions are reproduced.  Tables are included, but    **
** may not resemble those in the printed version.               **
**                                                              **
** Please see the PDF (Portable Document Format) file, when     **
** available, for a complete electronic file of the printed     **
** document's contents.                                         **
**                                                              **
******************************************************************
GAO-06-102

     

     * Report to Congressional Committees
          * November 2005
     * HIGHWAY SAFETY
          * Further Opportunities Exist to Improve Data on Crashes Involving
            Commercial Motor Vehicles
     * Contents
          * Results in Brief
          * Background
          * States Continue to Have Problems with CMV Crash Data Quality, but
            Are Pursuing a Variety of Improvement Efforts
               * Challenges Remain in Meeting Data Quality Standards
          * Data Quality Problems Often Reflect Difficulties in Collection
            and Processing of Crash Reports
               * Misinterpretation of Criteria and Definitions by Officers
                 Filling Out Crash Reports
               * Crash Report Processing Issues
               * States Are Pursuing a Variety of Efforts to Address Data
                 Quality Issues
               * Data Analysis to Identify Problems
               * Reduction of Crash Database Backlogs
               * Using Electronic Systems to Expedite CMV Crash Data
               * Improving Law Enforcement Training
          * FMCSA's Efforts Have Contributed to CMV Data Quality Improvements
          * FMCSA's SaDIP Has Supported State Efforts to Improve Data Quality
               * SaDIP has Yielded Several Positive Results
               * FMCSA Has Made Improvements in the Administration of SaDIP;
                 However, Management of Grant Awards Raises Concerns
          * Data Quality Map Has Spurred Improvements, but Limitations Curb
            Map's Continued Usefulness
               * The State Safety Data Quality Map is an Incentive for States
                 to Make CMV Crash Data Improvements
               * The Data Quality Map is Important in Addressing Data Quality
               * The State Safety Data Quality Map Has Limitations that May
                 Affect its Effectiveness as a Tool for Monitoring and
                 Measuring CMV Crash Data Quality
          * Conclusions
          * Recommendations for Executive Action
          * Agency Comments and Our Evaluation
     * Objectives, Scope and Methodology
     * State Safety Data Quality Map Limitations
          * Completeness: Overall Completeness is Based on Fatal Crashes Only
          * Timeliness: Timeliness is Not Based on All Reported Crashes
          * Accuracy: Accuracy is Based on Only One Variable
     * FMCSA Reportable Crash, CMV, and Carrier Identification Visor Cards
     * SaDIP Grant and Cooperative Agreement Distribution by State
     * SaDIP Case Study States
          * Georgia
          * SaDIP Projects
          * Crash Data Collection/Reporting Issues
          * State/Federal Coordination Issues
          * Data Quality Map
          * Minnesota
          * SaDIP Projects
          * Crash Data Collection/Reporting Issues
          * State/Federal Coordination Issues
          * Data Quality Map
          * North Carolina
          * SaDIP Projects
          * Crash Data Collection/Reporting Issues
          * State/Federal Coordination Issues
          * Data Quality Map
          * Ohio
          * SaDIP Projects
          * Crash Data Collection/Reporting Issues
          * State/Federal Coordination Issues
          * Data Quality Map
          * Oklahoma
          * SaDIP Projects
          * Crash Collection/Reporting Issues
          * State/Federal Coordination Issues
          * Data Quality Map
          * Washington
          * SaDIP Projects
          * Crash Collection/Reporting Issues
          * State/Federal Coordination Issues
          * Data Quality Map
     * GAO Contact and Staff Acknowledgments

Report to Congressional Committees

November 2005

HIGHWAY SAFETY

Further Opportunities Exist to Improve Data on Crashes Involving
Commercial Motor Vehicles

Contents

Tables

Figures

November 18, 2005Letter

The Honorable Christopher S. Bond Chairman The Honorable Patty Murray
Ranking Minority Member Subcommittee on Transportation, Treasury, the
Judiciary, Housing and Urban Development, and Related Agencies Committee
on Appropriations United States Senate

The Honorable Joe Knollenberg Chairman The Honorable John W. Olver Ranking
Minority Member Subcommittee on Transportation, Treasury, and Housing and
Urban Development, the Judiciary, District of Columbia Committee on
Appropriations House of Representatives

Large trucks make up only 3 percent of the nation's registered vehicles,
but they were involved in 11 percent of all fatal crashes in 2003, the
last year for which complete data is available. That year, large trucks
were involved in more than 430,000 crashes, killing approximately 5,000
people.  1 In 1999, Congress established the Federal Motor Carrier Safety
Administration (FMCSA) within the Department of Transportation2 and
mandated it with reducing crashes, injuries, and fatalities involving
large trucks and buses. Currently, FMCSA has the goal of reducing
commercial motor vehicle crash fatalities to 1.65 fatalities per 100
million miles of travel by 2008. As of fiscal year 2003, the commercial
motor vehicle fatality rate was 2.3 fatalities per 100 million miles
traveled, the lowest recorded since the Department of Transportation
initiated tracking in 1975-but still 40 percent above the 2008 goal.

FMCSA works in partnership with states to reach commercial motor vehicle
safety goals. States are the gatekeepers for the collection and reporting
of commercial motor vehicle crash information. They receive crash reports
completed by law enforcement personnel in local jurisdictions, compile
them, and then submit crash reports to FMCSA. At the federal level, FMCSA
manages a database which provides data that is used in rating motor
carriers according to various safety indicators. Based on this rating,
motor carriers are selected for safety inspections and reviews as part of
FMCSA's enforcement efforts. While the data collected is primarily for
federal use, states use the information to assist overall crash safety
efforts and in setting commercial motor vehicle safety goals for
themselves. Because the data is used in both federal and state
decision-making on a variety of safety-related issues, it is important
that it adequately meets data quality standards.

To be useful to both federal and state decision-makers, crash data must be
complete, timely, accurate, and collected in a consistent manner. However,
there have been concerns about the quality of the information FMCSA and
the states use to direct their efforts. Beginning with the Motor Carrier
Safety Improvement Act of 1999, Congress directed the Department of
Transportation to improve the collection and analysis of data on
commercial motor vehicle crashes. This resulted in the creation of a
commercial motor vehicle data improvement program.3 The program was
reauthorized in 2005.4 Since 2002, about $21 million has been awarded to
states to improve their crash data quality, but data quality problems have
persisted.5 In February 2004, the Department of Transportation Inspector
General released a report discussing limitations of the commercial motor
vehicle crash data.6

In a Senate report accompanying the fiscal year 2005 appropriation for the
Department of Transportation,7 Congress asked that we review FMCSA's
program for helping states improve their commercial motor vehicle crash
data. The report directed us to describe the benefits obtained through the
program, identify what can be done to improve the effectiveness of the
program, and address concerns regarding crash data raised in the February
2004 Department of Transportation Inspector General's report. Accordingly,
this report examines (1) what is known about the quality of commercial
motor vehicle crash data and what states are doing to improve it, and (2)
the results of FMCSA's efforts to facilitate the improvement of the
quality of commercial motor vehicle crash data sent to the federal
government.

To describe the quality of commercial motor vehicle crash data, we
reviewed a number of sources, including data reported by FMCSA and
existing studies on the quality of commercial motor vehicle crash data. We
interviewed officials from FMCSA, contractors that develop FMCSA crash
data tools, and commercial vehicle industry researchers and public
interest organizations to gain their perspective on commercial motor
vehicle crash data quality. To provide information on states' efforts to
improve commercial motor vehicle crash data, we reviewed grant
documentation for 34 states that participated in FMCSA's Safety Data
Improvement Program (SaDIP) as of September 2005.8 We also conducted case
studies in six states that participated in the program.9 States were
chosen based on a wide variety of factors including crash data quality and
level of crash reporting. Additionally, we conducted phone interviews with
states that did not participate, or are no longer participating, in the
program.10 To provide results of FMCSA efforts to facilitate the
improvement of commercial motor vehicle crash data quality, we conducted
interviews with officials from FMCSA on the administration and management
of the SaDIP program. We also analyzed the guidance and support FMCSA
provided to states and assessed FMCSA's role in coordinating commercial
motor vehicle data quality initiatives. We reviewed studies conducted by
the University of Michigan Transportation Research Institute and
determined that the methodologies used in assessing the quality of the
data states submit to FMCSA were sound and that the studies provided
sufficiently reliable results for our purposes. Through site visits, a
review of grant applications, interviews with relevant stakeholders and
experts, and these studies, we were able to determine shortcomings in the
reliability of FMCSA's commercial motor vehicle crash data. However, we
determined that the data was sufficiently reliable for the purpose of case
study selection. Our work was conducted from February 2005 through
November 2005 in accordance with generally accepted government auditing
standards. See appendix I for more details regarding our objectives,
scope, and methodology.

Results in Brief

Overall, commercial motor vehicle crash data does not yet meet general
data quality standards of completeness, timeliness, accuracy, and
consistency. For example, according to FMCSA, as of fiscal year 2004
nearly one-third of commercial motor vehicle crashes that states are
required to report to the federal government were not reported, and those
that were reported were not always accurate, timely, or consistent. Data
quality problems most often stem from errors or omissions either by law
enforcement officers at the scene of a crash or in the processing of crash
reports to a state level database. To address data quality problems, a
number of states are undertaking four major types of data improvement
activities:

o Analyzing existing data to identify where problems are and to develop
plans for addressing them;

o Reducing backlogs of data that have not been entered into state-level
databases in order to create more complete state crash files, through
steps such as hiring contract employees;

o Developing and implementing electronic data systems for collecting and
processing crash information in a more timely, accurate, and consistent
manner; and

o Providing training, such as educating law enforcement officers on the
definitions and criteria for commercial motor vehicle crashes, to create
more accurate and consistent data.

These state efforts are resulting in some progress. Based on analysis of
FMCSA data, a great number of crashes are being reported to FMCSA.
Overall, the total number of commercial motor vehicle crashes being
reported to FMCSA has increased by 59 percent between fiscal year 2000 and
fiscal year 2004, while the length of time it takes states to report these
crashes to FMCSA has decreased as well.

FMCSA has two main efforts to support states in improving their reporting
of commercial motor vehicle crash information-a commercial vehicle crash
data improvement program and a data quality rating system-and both appear
to be beneficial. Through the data improvement program, FMCSA has provided
nearly $21 million in discretionary grants to 34 states between 2002 and
2005. These grants have ranged from $2,000 to $2 million and have helped
states conduct all four data improvement activities previously described.
The six states in our case studies generally improved their data quality,
mainly through projects funded in whole or in part through the grant
program. While we did not find problems with FMCSA's oversight of the
program, we did note that FMCSA does not have formal guidelines for
awarding funds to states. Because these grants are discretionary, and
because more states are expected to participate in the program in the
future, having formal guidelines and systems in place would likely assist
in prioritizing states' requests and ensuring consistency in awarding
funds. FMCSA's second major effort, the State Safety Data Quality map, has
proven to be an important tool for states to use in improving their crash
data as well. This map, created by FMCSA and the Volpe National
Transportation Systems Center, is a color-coded display depicting the
overall data quality for each state in one of three rating
categories-"good" (green), "fair" (yellow), or "poor" (red). According to
both FMCSA and state officials, the map and the underlying rating system
serve as an incentive for states to make improvements to their crash data.
Despite the map's utility thus far, we identified potential problems both
in the methodology used for developing ratings and the risk of drawing
erroneous conclusions from the map. One example of a problem with the
current methodology is that the overall ratings combine information about
crashes with information stemming from FMCSA's inspections of motor
carriers. Combining ratings for both crash and inspection data quality
tends to make it difficult to determine how states are doing specifically
with their crash data. In addition, some states with a "good" rating in
completeness are not reporting all crashes to FMCSA. Rating states as
"good" when in fact they have problems may discourage states from
continuing to devote attention and resources to areas needing improvement
and possibly misdirect program efforts. FMCSA is aware of many of the
limitations of the map, but has not yet developed and implemented a formal
plan to improve it. Addressing these limitations will strengthen the data
quality map as a tool for improving commercial motor vehicle crash data.

To ensure that FMCSA is able to target limited funds as effectively as
possible, we are recommending that FMCSA create specific guidelines and
criteria for awarding commercial motor vehicle crash data improvement
funding. We are also recommending that FMCSA develop a plan to improve the
data quality map, including assessing the methodology for developing
ratings, providing results in greater detail, and documenting any
limitations associated with the map. These enhancements will provide users
with a more useful tool to view the condition and progress made in states'
commercial motor vehicle crash data.

Background

FMCSA was established as a separate administration within the U.S.
Department of Transportation (DOT) on January 1, 2000, pursuant to the
Motor Carrier Safety Improvement Act of 1999. FMCSA issues and enforces
the federal motor carrier safety regulations that govern many aspects of
specified commercial trucking and bus operations, including the interstate
operation and maintenance of commercial motor vehicles (CMV). Regulations
promulgated by FMCSA specify requirements that must be met by drivers of
these vehicles. FMCSA conducts compliance reviews11 of truck and bus
companies, and performs safety audits of new entrants12 into the industry.
In addition, FMCSA trains inspectors to conduct safety audits, while
states and local authorities are responsible for conducting the
inspections and submitting the results to FMCSA.13 This partnership
between FMCSA and the states annually results in about 3 million truck and
bus inspections, 7,000 to 13,000 compliance reviews, and more than 19,000
new-entrant safety audits.

CMV crash data is key to FMCSA's efforts. CMV crash data is collected by
local law enforcement, sent to the state, and then processed and uploaded
by the state into FMCSA's data system. FMCSA maintains a database
management system in each state so they can submit crash reports into
FMCSA's central data system, the Motor Carrier Management Information
System (MCMIS). FMCSA uses the information in its Motor Carrier Safety
Status Measurement System, also known as SafeStat, to target carriers for
compliance reviews to ensure that they are following safety regulations.14
SafeStat uses a variety of data to rank carrier safety, but it places the
heaviest weight on crash data.

Federal and state data quality guidelines call for CMV crash data to meet
four basic quality standards:15

o Completeness: To support adequate decision-making for identifying
problems and developing appropriate countermeasures, data should be
collected for all reportable CMV crashes in the state, and data on all
appropriate crash variables such as the carrier's identification number,
should be submitted to FMCSA.

o Timeliness: To make decisions about current safety problems, identify
trends, or target carriers that pose immediate threats to highway safety,
CMV crash data should be available for state and federal analytical
purposes within a useful timeframe.

o Accuracy: To adequately assess CMV crash problems and target the
appropriate carriers for enforcement, all data within reportable CMV crash
records should be accurate and reliable.

o Consistency: To target carriers nationwide and to compare state results,
CMV crash data should be collected uniformly using the same standards and
applying the same criteria across jurisdictions.

FMCSA has provided more specific guidelines and criteria for meeting each
of these standards. On timeliness, for example, FMCSA calls for states to
submit all CMV crash data to MCMIS within 90 days of when the crash
occurs. In order to facilitate complete reporting of CMV crashes, FMCSA
recommends data elements, such as the identity of the carrier, vehicle
configuration, cargo body type, etc., as the minimum information to be

collected in order to have complete information on CMV crashes.16 FMCSA
has also created criteria to assist in consistent reporting of crash
information. These published criteria are used for identifying reportable
CMV crashes to be submitted to MCMIS (see fig. 1).

Figure 1: Criteria for Selecting Truck and Bus Crashes to Report to FMCSA

Note: Exceptions include crashes that involve: 1) a personally-owned truck
or passenger vehicle meant for personal use only as the sole vehicle
meeting the criteria above, or 2) a driver with a disease condition (e.g.,
stroke, heart attack, diabetic coma or epileptic seizure) and no other
injury or damage occurs, or 3) deliberate intent (suicide, self-inflicted
injury, homicide, etc.), with no unintentional injury or damage. Of the
430,000 CMV crashes that occur each year, about 150,000 meet reporting
criteria.

With the Motor Carrier Safety Improvement Act of 1999, Congress
established a program to improve the collection and analysis of CMV crash
data.17 This resulted in the creation of the Commercial Motor Vehicle
Analysis Reporting System (CVARS). CVARS, which is now known as SaDIP was
originally intended to be a standalone data collection system. After
determining that a separate system would duplicate existing efforts,
however, FMCSA decided to use SaDIP as a federal funding tool to support
state efforts to collect and report CMV crash data.18

Besides changes in scope, SaDIP has also changed greatly in how the
program is administered. Since fiscal year 2001, FMCSA has received
funding from the Congress to implement this program. As a new agency,
FMCSA did not have the appropriate contracting infrastructure in place to
administer the funds. Therefore it transferred funds and the
administrative duties of the program to the National Highway Traffic
Safety Administration (NHTSA), which awarded several grants to states in
fiscal year 2002. NHTSA also provided some of the funding to the General
Services Administration (GSA) to enter into cooperative agreements with
several states to fund multi-year data improvement projects. In fiscal
year 2003, however, FMCSA assumed responsibility for all oversight of
SaDIP funding.

FMCSA provides SaDIP funds to states through two different methods. States
can receive 1-year grants to fund specific projects, or they can enter
into multi-year cooperative agreements in order to fund multiple efforts
that are necessary to identify and reduce data quality problems. States
can receive both of these types of funds from FMCSA and can apply for
grant funding multiple times. SaDIP funding is discretionary in nature,
allowing states to request the amount of funding they need to conduct data
improvement projects. Since 2002, SaDIP funding has been provided to 34
states, and as of September 2005, awards ranged from $2,000 to $2 million
totaling approximately $21 million. (See app. IV for funding distribution
by state.)

In February 2004, the Department of Transportation Inspector General
issued a report on SafeStat, the system FMCSA uses to target its
compliance reviews.19 The report identified a number of problems with
SafeStat, much of which stemmed from the quality of the crash data being
used in calculating carrier ratings. The State Safety Data Quality map,
discussed later in the report, was developed in response to the
recommendations in the 2004 report.

States Continue to Have Problems with CMV Crash Data Quality, but Are
Pursuing a Variety of Improvement Efforts

When measured against generally accepted standards for completeness,
timeliness, accuracy, and consistency, states continue to have challenges
with the quality of their CMV crash data. Many of these challenges are
based in the collection and processing of data at the local level, and
often reflect broader crash data quality problems for all types of vehicle
crashes. To address remaining limitations, and improve the completeness,
timeliness, accuracy, and consistency of the data, states are undertaking
four main types of efforts: analyzing data to identify problem areas,
eliminating backlogs of data not yet entered, creating electronic systems
to expedite data entry, and training law enforcement officers in ways to
improve the data they submit.

Challenges Remain in Meeting Data Quality Standards

The completeness, timeliness, accuracy, and consistency of CMV crash data
is currently not meeting generally accepted data quality standards. Table
1 provides examples of some of the overall problems we (and others) have
identified. Appendix V, which contains summaries of the six states we
visited, provides more specific examples.

Table 1: Examples of Problems with Commercial Vehicle Crash Data Quality

                                        

Completeness    As of 2004, FMCSA estimates that about one-third of the    
                reportable CMV crashes are not being submitted to their data  
                   system-the Motor Carrier Management Information System     
                (MCMIS).a In addition, FMCSA estimates 20 percent of nonfatal 
                  crashes are not being reported. Further, studies indicate   
                that even those crashes reported to MCMIS often have missing  
                                           data.b                             
Timeliness   The average length of time from when a crash occurs to when   
                the crash data is uploaded to MCMIS is 99 days-9 days over    
                the required time limit. While this is not far from the goal, 
                there is substantial variation in timeliness among states.    
                For example, our analysis of state CMV crash data shows       
                timeliness ranges from 13 days to 339 days.c                  
Accuracy     FMCSA assesses accuracy by determining the number and         
                percentage of interstate crashes uploaded to MCMIS without    
                enough information to determine a carrier's DOT number (known 
                as a non-match). As of fiscal year 2004, 15 percent of CMV    
                crash records in MCMIS can not be matched to a carrier's DOT  
                number.d                                                      
Consistency  According to an analysis by FMCSA, 33 of 50 states have crash 
                reports that do not adequately follow the criteria for        
                reporting commercial motor vehicle crashes to FMCSA.e         

aMCMIS contains information on the safety fitness of commercial motor
carriers and hazardous material shippers subject to the Federal Motor
Carrier Safety Regulations and the Hazardous Materials Regulations. States
upload their crash information to MCMIS via FMCSA's SafetyNet.

bSource: FMCSA

cSource: Analysis of MCMIS data from fiscal year 2000 through fiscal year
2004.

dSource: Analysis of fiscal year 2004 MCMIS data.

eSource: Comparison of State Crash Reports with SafetyNet Selection
Criteria and Key Data Secondary Elements Chart, FMCSA, August 12, 2005.

Data Quality Problems Often Reflect Difficulties in Collection and
Processing of Crash Reports

CMV crash data quality problems often stem from issues that occur when
data is initially collected at the scene of a crash and later processed
through the state. We reviewed reports on crash data quality including
individual state reviews conducted by the University of Michigan
Transportation Research Institute (UMTRI).20 We also discussed these
matters with state officials. We identified two key causes of poor data
quality: (1) problems in interpreting how to fill out crash reports at the
scene and (2) crash report processing issues ranging from competing
priorities at the local level to complex routing of crash reports.

Misinterpretation of Criteria and Definitions by Officers Filling Out
Crash Reports

According to studies and our discussions with state officials, data
collected at the scene of a crash can be flawed because of law enforcement
misinterpretation of reporting criteria and definitions.21
Misinterpretation can occur for several reasons, such as infrequent
opportunities for officers to receive training on filling out crash
reports or unfamiliarity resulting from infrequent occurrences of CMV
crashes in an officer's jurisdiction. Below are common problems with
properly reporting CMV crashes:  22

Identifying reportable crashes. While crashes that result in a fatality
are easily identifiable as a reportable crashes, tow-away or injury
crashes are more difficult to identify. For example, UMTRI's review of
eight states showed that five of those states experienced problems with
reporting crashes that did not involve a fatality. According to UMTRI,
this is likely due to a lack of understanding of criteria for reporting
CMV crashes to FMCSA (see fig. 1).

Identifying reportable commercial motor vehicles. FMCSA is responsible for
enforcing safety regulations for interstate carriers, but collects crash
information on a variety of CMVs that meet certain criteria. According to
UMTRI reports, states had problems identifying smaller trucks and they
were less likely to be reported because law enforcement officials are less
likely to recognize them as qualifying CMVs. Further, law enforcement
officials may be confused about reporting interstate and intrastate
carriers to FMCSA. UMTRI's review of 8 states showed that five of the
states encountered problems in identifying appropriate vehicle type for
reporting to FMCSA (see fig. 1).

Crash Report Processing Issues

Several other issues relate to the processing of crash reports. These tend
to fall into the following three main categories:

Competing priorities at the officer level. Data collection is complicated
because at the scene of a crash, an officer's first priority is to ensure
the safety of those involved; data collection is often a lesser concern.
Only after the situation has been stabilized can the officer fill out the
crash report. Competing priorities can make it difficult for law
enforcement to adequately complete a crash report. According to a report
by NHTSA, crash reporting is often a low priority when traffic safety data
is not perceived as relevant to the work of law enforcement. This was
confirmed in states we visited. For example, one state official with whom
we spoke said that submitting crash data is often dependent on how much
priority the local police chief places on data quality.

Manual crash report forms. Typically, law enforcement officers complete
handwritten crash reports, which are then manually submitted for data
processing at a local or state agency. According to an FMCSA official,
nearly all states use manual crash reporting forms to some extent.
Completing crash reports by hand and manually processing the information
has the potential to lead to errors in the data. Data quality may be
further compromised by the use of a supplemental CMV crash report. States
may have a general crash report for use in all crashes and a supplemental
CMV crash report to be completed if a CMV is involved. For example, four
of the six case study states we reviewed used a supplemental form to
report a CMV accident. When data has to be captured in a separate form,
law enforcement officers may not always complete the form, and sometimes
it may be separated from the original crash report or even lost. UMTRI
found in one state that use of the supplemental crash report may be
leading to CMV crashes not being submitted to FMCSA.

Complex processing. Even with a correctly completed crash report that is
properly identified as a CMV crash, states may have a complicated process
in place to route CMV crash data to FMCSA and this may contribute to
lengthy delays. UMTRI state assessments found that processing issues in
some states led to delays and resulted in incomplete submission of crash
data. One official told us that all law enforcement agencies in his state
send crash reports to the state patrol office where the report is scanned
into electronic format. The electronic version of the report is then sent
to the state department of transportation where additional location data
is entered into the report. From here, the report enters the state
database and then is periodically uploaded to FMCSA. This process can take
significant time, especially when the original crash report is missing
information.

States Are Pursuing a Variety of Efforts to Address Data Quality Issues

States are undertaking a variety of projects in order to improve the
completeness, timeliness, accuracy, and consistency of CMV crash data.
These projects encompass four general types of efforts: research and
analysis, reduction of reporting backlogs, electronic processing of crash
records, and training. Figure 2 shows state participation in each of these
four efforts. These projects directly affect the completeness, timeliness,
accuracy, and consistency of the states' CMV crash data.

Figure 2: State Participation in CMV Crash Data Improvement Efforts from
Fiscal Year 2002 through Fiscal Year 2005

Note: Data available for the 34 states that have participated in FMCSA's
data improvement program. Several states are participating in more than
one effort.

Data Analysis to Identify Problems

Eleven of the 34 states have research projects to evaluate crash data
collection and reporting practices.23 Assessing data collection and
reporting processes allows states to determine how well their data meet
each of the data quality standards of completeness, timeliness, accuracy,
and consistency, providing them with a useful tool for determining where
they are going to concentrate their improvement efforts.

Out of 34 states for which we have data, 11 have conducted research to
identify data problems.  During our case studies, we found examples of
states conducting data analysis to identify their CMV crash data
collection and reporting problems. (See app. V for more information on our
case study states.) For example, one state conducted an analysis in order
to determine why approximately half of the eligible crashes in the state
crash file were not being reported to FMCSA. The analysis found that the
state's crash file was not in the correct format to be correctly
transferred to FMCSA. In addition, frequent errors in the state or zip
code fields, and incomplete or inaccurate data, were leading to problems.
A federally funded study of the state's crash data found approximately 50
percent of the state's crash reports for one year (2003) were not reported
to the federal crash file due to problems at the state level coding and
preparing crash data for transfer to FMCSA. Another state conducted a
traffic records assessment that found the major deficiency in the state's
crash file was the lack of timely data. It also identified the lack of
effective quality controls during data entry and non-reporting of eligible
crashes as deficiencies. Following the state's assessment, researchers
made recommendations to help the state prioritize projects, including
planning for the eventual implementation of electronic crash data
collection.

Reduction of Crash Database Backlogs

Fourteen of the 34 states have projects to reduce the backlog of crash
reports that need to be entered into state crash files. In most states,
until crashes are included in the state file, they are not reported to
FMCSA. Hence, backlog reduction is essential for creating a complete state
crash file, and with providing FMCSA with complete data about CMV crashes.
A complete data file is critical for performing crash and trend analysis,
for making informed policy decisions, and for gaining a better
understanding of CMV safety issues.

Of the 34 states for which we have data, 14 are participating in backlog
reduction projects, including four of our six case study states. In these
four states, temporary employees were hired to eliminate data backlogs.
For example, in one state, contract employees were working to reduce a
backlog of 600,000 crash reports. Data entry work began in July 2004 and
state officials expect it to be completed by September 2005. Another state
used temporary employees to eliminate a backlog of approximately 2 million
crash records. This effort began in 2002, and as of June 2005, the data
backlog has been completely eliminated and the state is now processing
crash reports within 3 days of receipt.

Using Electronic Systems to Expedite CMV Crash Data

Twenty-seven of the 34 states have projects to collect or input CMV crash
data through electronic systems. Electronic reporting allows for more
accurate and timely transfer of crash data to state and federal crash
files. Electronic transfer of crash data can reduce duplication and data
entry error, because paper-based data collection systems are subject to
human error and time delays. While these projects enhance the quality of
CMV crash data, often they are large in scope and require law enforcement
agencies to purchase hardware. Uniform electronic crash reporting systems
are heavily dependent on acceptance from all stakeholders and often some
jurisdictions may already have their own systems.

Of the 34 states for which we have data, 27 are participating in
electronic crash record projects, including all six of the case study
states. For example, one state has developed software that facilitates a
nightly transfer of CMV crash records from the Division of Driver and
Vehicle Services to the state patrol agency. The state patrol agency then
submits the crash reports to FMCSA. The state is also working to provide
computer hardware and Internet access so that state patrol officers can
complete crash reports online. Another state is developing electronic
crash reporting capabilities using handheld computers, which will
eliminate handwritten crash reports, and the need to manually enter crash
report data into an electronic system.

Improving Law Enforcement Training

States are providing training to law enforcement officers in CMV crash
data collection. Training for law enforcement officers on how to correctly
identify a CMV and how to accurately complete a police accident report can
help to improve the completeness, accuracy, and consistency of CMV crash
data since many of the mistakes in reporting are made at the crash

site.24 Training is also an opportunity to highlight the link between data
collection and its end use in planning and prevention.

Seventeen of the 34 states have projects aimed at providing law
enforcement training initiatives, and all six of our case study states had
some form of CMV training. In addition, all case study states we visited
recognized problems with the quality of CMV crash data resulting from
issues at the collection point. For example, one state provides on-going
CMV training to teach officers to properly identify and report commercial
vehicle crashes. This state also tries to show law enforcement officers
how the data is used in planning and prevention in order to show the
significance of good data. However, state officials told us that there has
not been a lot of participation in the training from local law enforcement
agencies. Another state is planning to develop a training video and visor
cards25 for statewide dissemination. The visor cards will assist officers
in CMV identification and provide information that will help police
properly complete all the necessary crash reporting documentation.

These state efforts are resulting in some progress. An analysis of FMCSA
data shows that a greater number of crashes are being reported to FMCSA
since 2000. Overall the total number of commercial motor vehicle crashes
being reported to FMCSA has increased by 59 percent between fiscal year
2000 and fiscal year 2004, while the length of time it takes states to
report these crashes to FMCSA has decreased as well.

FMCSA's Efforts Have Contributed to CMV Data Quality Improvements

FMCSA's efforts appear to improve CMV crash data quality and have had a
positive impact on state crash data. Specifically two efforts, the Safety
Data Improvement Program (SaDIP) and the State Safety Data Quality map,
have contributed to the changes. SaDIP funding has allowed states to
improve data completeness, timeliness, accuracy, and consistency by
supporting the implementation of new activities or increasing the scope
and timeliness of existing projects. While FMCSA has provided adequate
administration and oversight of SaDIP, the management of the program with
regard to awarding grants raises some concerns. There are no formal
guidelines in place for awarding funding to the states, and while this has
not yet presented a problem, it may in the future. Funding continues to be
made available, and more states continue to request funds for new
projects. The State Safety Data Quality map is an evaluation tool that
provides ratings for states' crash and inspection data quality, and
displays the ratings so that a state's performance relative to other
states is apparent. The map, which is publicized on FMCSA's Analysis and
Information (A&I) Online Web site, allows state officials and the public
to view the status of states'

CMV crash and inspection data quality.26 The map is being used as an
indicator of states' progress in improving the completeness, timeliness,
and accuracy27 of crash data and by virtue of its public nature it is an
incentive for states to improve their crash information. However, the
methodology used for the ratings has limitations that may hinder the map's
effectiveness in monitoring the status of CMV crash data quality.

FMCSA's SaDIP Has Supported State Efforts to Improve Data Quality

As of July 2005, FMCSA had provided about $21 million to 34 states in
order to assist in improving the quality of CMV crash data.28 Several
states have received funding more than once, and in fiscal year 2005,
FMCSA made available to states almost twice the amount of money than it
had at the beginning of the program ($3.9 million in fiscal year 2000
versus $7.3 million in fiscal year 2005). Additionally, recently passed
transportation legislation authorizes a total of $11 million to be used
for SaDIP funding for

fiscal years 2006-2009.29 The states that have participated in SaDIP
account for about two-thirds of all CMV crashes occurring between April 1,
2004, and March 31, 2005, reported to FMCSA's MCMIS database, and 70
percent of the 2003 fatalities reported to MCMIS. FMCSA's goal is to
provide funding for projects that will have the largest impact on
improving state data describing CMV crashes. FMCSA also encourages states
that are participating in data improvement projects that are already
funded to apply for SaDIP funds in order to provide additional assistance
to these larger efforts. (See table 2 for the annual distribution of SaDIP
funds and app. IV for fund distribution by state.)

Table 2: Annual Distribution of SaDIP funds

                                        

     Fiscal year         Funds budgeted for SaDIP     Funds awarded to states 
2000                                $3,865,000                          $0 
2001                                 2,748,000                           0 
2002                                 5,462,614                   5,207,014 
2003                                 4,967,500                     854,732 
2004                                 4,913,990                   5,449,467 
2005                                 7,340,800                   9,180,181 
Total                              $29,297,904                 $20,691,394 

Source: FMCSA.

Note: These data are current as of September 2005. In addition to funds
provided directly to states, FMCSA provided $1,200,000 to contractors to
assist states in improving data quality, bringing the total amount spent
on state data improvement efforts to $21,891,394.

SaDIP has Yielded Several Positive Results

States have used the SaDIP funds to conduct a variety of projects,
including those discussed earlier in this report. These funds have
benefited states in several different ways, including increased focus on
CMV data quality and advancement and expansion of ongoing broader data
quality projects.

SaDIP increased national attention to the problems associated with the
quality of CMV crash data reported to the federal government. Several
state officials we interviewed stated that they have noticed an increase
in the amount of focus given to CMV crash data issues. For example,
presentations made by FMCSA at several national conferences and workshops
have highlighted the importance of data quality, and informed states of
the various types of assistance, such as training and funding, that are
available to them. During our case studies, we were told the following:

o One state official said that by providing funds to be used for specific
purposes, SaDIP had the effect of focusing attention on data quality
improvement. SaDIP has sustained a high level of interest in data quality
and has been a catalyst to improving traffic records coordination across
state agencies.

o Another state official said SaDIP helped to improve communication
between the state patrol and the state's department of vehicle services.
Because grants were provided to both agencies, analysts at the agencies
are working together and have a better understanding of each other's data
needs and share access to their respective databases.

In half of our case study states, SaDIP has also allowed states to
expedite data improvement projects that were already planned for
implementation.

o One state official told us that the state had been considering a plan to
improve traffic safety data, including CMV crash data, and SaDIP funds
provided it with the means to implement this plan. The long-term funding
provided by the SaDIP cooperative agreement to support a full time
employee was also a crucial element in gaining support from state
decision-makers.

o One state official said the SaDIP grant made it significantly easier to
prioritize and expedite a data project that the state was already
considering implementing. The additional funding allowed the state to
specifically address CMV crash data and use other funding to continue to
address broader crash data issues.

Finally, SaDIP has allowed states to increase the scope of ongoing
projects or develop new initiatives. Two of our six case study states have
used SaDIP funds in addition to other resources to develop comprehensive
and long-term data quality initiatives to address the completeness,
timeliness, accuracy, and consistency of the CMV crash data they report to
FMCSA.

o Officials in one state told us that SaDIP funding supports elements of
the state's electronic traffic information processing initiative. SaDIP
funding has allowed the state to resolve a considerable backlog in crash
records, creating a more complete crash file, and work has begun to
develop a new electronic crash report that will be used-first by the state
police-and then by local law enforcement in the rest of the state. Without
resolution of this backlog, the state was unable to fully implement the
CMV data component of its electronic traffic information processing
initiative. Doing this allowed the state to have more timely, accurate,
and consistent crash data.

o SaDIP funding has helped another state develop an electronic crash
report and traffic records system. As previously stated, this will allow
the state to have more timely, accurate, and consistent crash data. The
state's goal is to develop the new system by 2008 and the state is
evaluating and using multiple sources of funding to achieve this goal.
Early in the process, the state made a strategic decision to target all
crash data, not just commercial vehicle data. While the state uses SaDIP
to develop electronic crash reporting software, SaDIP funding has also
been leveraged to help the state accomplish its data quality improvement
goals more quickly. For example, the new crash reporting capabilities
developed through the SaDIP grant allow for much better crash analysis and
targeting of safety efforts along major commercial vehicle corridors in
the state.

FMCSA Has Made Improvements in the Administration of SaDIP; However,
Management of Grant Awards Raises Concerns

SaDIP has evolved over time and FMCSA has made several efforts to improve
the administration of the program. For example, beginning in fiscal year
2006, FMCSA will be implementing a new application package for states to
use in applying for SaDIP. While SaDIP had some preexisting application
requirements for states, this application package provides a uniform
application for states to use, and it is the first time states are going
to be required to submit quantifiable project objectives and program
measures. This will allow FMCSA to begin to measure the effectiveness of
state improvement efforts. Additionally, SaDIP will be posted on
Grants.gov ( http://www.grants.gov ) in 2006.30 This federal website
provides a single source for grant applicants to electronically find and
apply for federal funds. FMCSA will also be updating its State Program
Manager's manual to include guidelines for the roles and responsibilities
of state program managers in administering SaDIP. Finally, FMCSA will be
issuing funds at designated times during the year. Currently, funds are
awarded to states on a rolling basis, but beginning with the
implementation of the new application package the funds will be awarded on
specific dates to all applicants. This is expected to improve the
program's organization and FMCSA's ability to keep track of grant progress
as the number of program participants increases. Throughout these
administrative changes, FMCSA has maintained sufficient oversight of the
states participating in SaDIP. FMCSA has contracted with a company that is
responsible for monitoring SaDIP participant states, and ensuring that
they are submitting quarterly progress reports containing sufficient
detail to FMCSA. This contractor also has regular conversations with the
FMCSA Division Administrators in the SaDIP states, and it maintains copies
of SaDIP-related paperwork.

While these efforts are positive steps, we have concerns with FMCSA's lack
of guidelines for awarding funds to states. FMCSA has not yet established
formal guidelines for determining how much money a state should receive,
or if the state should receive the funds in the form of a grant or a
cooperative agreement. Since the beginning of the program, funds awarded
to states have ranged from $2,000 to $2 million for projects ranging from
specific activities to broader efforts that span three or more years. (See
app. IV for funds awarded by state.) These awards have taken the form of
both grants and cooperative agreements between the state and the federal
government. Currently these decisions are made on a state-by-state basis,
informally by a small review panel. The state's application, data quality
history, discussions with the state's FMCSA division administrator, and
any other relevant and available data on the state are consulted when
making funding awards.

While the absence of guidelines has not proven to be problematic to date,
having formal guidelines will better ensure consistency and discipline in
managing the grant program among states, particularly as states' needs
become more targeted or more states decide to participate in the program.
It will also add integrity to the grant management system and assist in
providing better administration and oversight of SaDIP projects. For
example, such guidelines would likely allow FMCSA to better assess and
prioritize states' funding requests, including more formally considering
whether proposed activities adequately address problems states identify in
the proposal, whether the amount of funding requested is appropriate for
the proposed activities, and whether multiple-year versus one-year funding
is appropriate. Further, formal guidelines would provide a more structured
framework to evaluate the effectiveness of different SaDIP project
activities and assist in guiding future state improvement efforts.

Data Quality Map Has Spurred Improvements, but Limitations Curb Map's
Continued Usefulness

The State Safety Data Quality map has encouraged states to improve their
CMV data quality, but limitations exist that may hinder the map's
usefulness as a tool for monitoring and measuring CMV crash data quality.
The State Safety Data Quality map was created by FMCSA and the Volpe
National Transportation Systems Center principally to provide context for
both crash and inspection data used in the SafeStat system; however, it
has

evolved into a tool to evaluate state-reported crash and inspection
data.31 It is based on a system that rates crash and inspection data
quality-completeness, timeliness, and accuracy-as "good," "fair," or
"poor."32 The map has proven to be a major incentive for states to
initiate CMV crash data improvements, for gaining support in implementing
these improvements, and as a tool to monitor CMV crash data quality.
However, we have identified some important limitations that can affect the
data quality map's future usefulness.

The State Safety Data Quality Map is an Incentive for States to Make CMV
Crash Data Improvements

Since the State Safety Data Quality map is accessible by the public and
presents data quality ratings in a simplified form, the map is a motivator
for states to improve their CMV crash data. According to officials at
FMCSA, they said the map has been very influential in encouraging states
to improve their CMV crash data. Corroboratory comments came from the
state level, where many of the officials in case study states provided
anecdotes of how the data quality map served as an impetus for initiating
improvements and gaining support.33 Here are examples of how the map
assisted states in initiating or expediting improvements in CMV crash
information.

o One state includes the data quality map ratings in weekly status reports
to agency heads and in reports to the governor. An official in this state
said that once the data quality map was shown to the governor, it raised
the importance of improving CMV crash data. This state now posts the data
quality map on state agency Web sites and uses it as a tool to compare its
data quality efforts to the rest of the nation.

o One state official told us that when agency leaders understand that
their state has a "poor" rating, they are likely to make data improvement
a priority and focus attention and resources on the issue.

o Another official said that the map helped officials to "see the light."
In this state, the data quality map helped initiate a process that led to
improved communication and coordination for data quality at the highest
levels for all safety and security projects within the state.

The Data Quality Map is Important in Addressing Data Quality

FMCSA and states we reviewed use the State Safety Data Quality map to
measure states' CMV data quality and progress being made in their data
improvement efforts. Our review indicated that although the map was
created as a tool for providing context to FMCSA's SafeStat system, both
FMCSA and the states were using it specifically for monitoring the status
of CMV crash data quality. State officials were also using it to identify
data quality problems and assist states in targeting state improvement
efforts. Here are examples from the states we contacted:

o In one state an official reported that it was very helpful for law
enforcement officials and department heads to instantly recognize their
state's data quality status on the map.

o In another state, officials said they knew they had a data quality
problem, but were unable to identify the specific issue. The data quality
map indicated that the state's accuracy was poor, and the problem was
specifically with matching carriers in the state database with crashes in
MCMIS.

o In another state, officials said the data quality map focused attention
on the state's data quality spending and results, and it helped state
officials re-prioritize spending.

Both FMCSA and state officials in many of the states we spoke with
recognized the map as an important tool in measuring progress in their
crash data improvement efforts. Further, FMCSA officials said the map was
used as an assessment of a state's success in the SaDIP program. While
officials also said they reviewed data in the MCMIS system to monitor
state progress, an important measure of success in the program was a
state's status in the State Safety Data Quality map and the underlying CMV
crash measures.

The State Safety Data Quality Map Has Limitations that May Affect its
Effectiveness as a Tool for Monitoring and Measuring CMV Crash Data
Quality

Given the importance that FMCSA and the states attach to the data quality
map, it appears to be a good first step in monitoring states' data
quality. However, we found limitations with many features of the map that
diminish its usefulness as an effective tool for specifically monitoring
and measuring state progress in their improvement efforts. Specifically,
we identified several limitations with the methodology used to develop the
data quality map's ratings, as well as with the measures themselves that
should be addressed. Below are some of the key limitations. A more
detailed discussion of limitations is located in appendix II.

o The data quality map measures used to calculate the completeness,
timeliness, and accuracy of CMV crash data falls short of providing a
complete measure of CMV crash data quality. While each data quality
measure has some limitations, one key measure-completeness-has a number of
difficulties. Most importantly, the completeness measure is limited in the
data it is assessing. The completeness measure only evaluates fatal CMV
crashes, which represents about 3 percent of all reportable CMV crashes.34
In addition, the completeness measure does not assess the completeness of
the information contained within the crash report.

The completeness measure's methodology relies solely on comparing CMV
fatal crash data in MCMIS to data states submit to the Fatality Analysis

Reporting System (FARS).35 This particular approach provides additional
limitations to the completeness measure. First, there are some
definitional differences between FARS and MCMIS data that can account for
about 4 percent of the crash records. Secondly, FARS data is not timely.
For example, the June 2005 map relies on 2003 FARS data; thus the
completeness measure does not reflect the current status of CMV crash
data. In addition, since FARS data is only released once a year, quarterly
issuances of the map do not necessarily reflect changes in the number of
fatal crashes. Table 3 below describes in more detail the limitations for
each crash data quality measure.

Table 3: Comparison of Data Quality Standards and State Safety Data
Quality Map Measures

                                        

Data quality  State Safety                      Limitations 
     standard    Data Quality                                  
                  Map measure                                  
Completeness All reportable  Percentage of                  Includes only  
                CMV crashes in  Fatal Crash                    reported,      
                the state and   Records Reported               fatal crashes; 
                data on all                                    reflects only  
                appropriate     Compares the                   about 3        
                crash variables number of large                percent of all 
                are submitted   trucks in crashes              CMV crashes    
                to FMCSA.       involving a                    required to be 
                                fatality in MCMIS              reported to    
                FMCSA           versus those in                FMCSA.         
                recommends 20   the FARS.b                                    
                CMV crash data                                 Bases          
                variables that                                 completeness   
                should be                                      on FARS data,  
                reported on.a                                  though some    
                                                               differences    
                                                               exist in state 
                                                               definitions of 
                                                               a CMV fatal    
                                                               crash and may  
                                                               result in an   
                                                               inflated or    
                                                               deflated       
                                                               rating.        
                                                                              
                                                               Does not       
                                                               represent      
                                                               current status 
                                                               of             
                                                               completeness.  
                                                                              
                                                               Does not       
                                                               measure the    
                                                               completeness   
                                                               of CMV         
                                                               information    
                                                               within a crash 
                                                               report         
                                                               (missing       
                                                               variables).    
Timeliness   All reportable  Percentage of                  Only reflects  
                CMV crash       Crash Records                  records that   
                records are     Reported within 90             are uploaded   
                available for   Days                           into MCMIS and 
                state and                                      have not been  
                federal         The percentage of              changed or     
                analytical      State-reported                 edited.d       
                purposes in a   fatal and                                     
                useful          non-fatal crash                Backlogs of    
                timeframe.      records reported               crash reports, 
                                within 90 days in              once entered,  
                FMCSA           the MCMIS database             can negatively 
                recommends that for carriers over              affect         
                CMV crash data  a 12-month time                timeliness     
                be reported     period.c                       rating.        
                within 90 days                                 
                of the crash                                   
                occurrence.                                    
Accuracy     All data within Percentage of                  Only measures  
                reportable CMV  Matched Crash                  accuracy of    
                crash records   Records                        one variable   
                are accurate                                   (identity of   
                and reliable.   The percentage of              the motor      
                                State-reported                 carrier-the    
                                fatal and                      U.S. DOT       
                                non-fatal crash                number);       
                                records in the                 neglects to    
                                MCMIS database for             measure the    
                                interstate                     accuracy for   
                                carriers and                   other          
                                intrastate                     recommended    
                                hazardous material             variables.     
                                carriers over a                
                                12-month time                  
                                period that were               
                                matched to a motor             
                                carrier in MCMIS.              
Consistency  Crash data      None                           N/A            
                should be                                      
                collected                                      
                uniformly.                                     
                Officials                                      
                should use the                                 
                same standards                                 
                and apply the                                  
                same criteria                                  
                uniformly                                      
                across                                         
                jurisdictions.                                 
                                                               
                FMCSA provides                                 
                guidelines and                                 
                criteria for                                   
                reporting CMV                                  
                crashes.                                       

Source: FMCSA and Volpe National Transportation Systems Center. Data
Quality Standards are based on GAO's review of data quality guidelines
from a variety of sources including FMCSA, NHTSA, and Data Nexus Inc., and
includes references to the Model Minimum Uniform Crash Criteria, National
Governors' Association Elements and Criteria, and the American National
Standard Manual for Classification of Motor Vehicle Traffic Records.

aIn 1992, FMCSA adopted the National Governors' Association (NGA)
recommended data elements, requiring these data to be collected and
reported on motor carrier crashes.

bFMCSA created an "Crash Consistency Overriding Indicator" to indicate
consistency in reporting of non-fatal CMV crashes. The indicator flags
states that may be experiencing major problems in reporting crash data.
The Crash Consistency Overriding Indicator is the percentage of
state-reported non-fatal crashes as compared to a three-year average of
reported non-fatal crashes. States that have reported fewer than 50
percent of non-fatal crash records for the current year based on the
previous three-year average of non-fatal crash records are flagged and
receive a rating of "poor" regardless of their ratings in any of the other
data quality indicators. This indicator is also limited because it only
identifies extreme cases of under-reporting and does not assess if there
is a substantial increase in reported non-fatal crash records.

c Includes both interstate carriers (those carriers that operate between
states) and intrastate carriers (carriers that operate only within one
state) though FMCSA is only responsible for regulating interstate
carriers.

dFMCSA's MCMIS currently does not have the ability to track the original
upload date when a crash record is edited or changed. Therefore, records
that are changed after the original upload are not used in the calculation
for timeliness. For additional information see appendix II.

o The data quality map is limited in its ability to meaningfully monitor
and track CMV crash data quality over time. Since the State Safety Data
Quality map was first issued, a large majority of states have been rated
as "good" or "fair" for the completeness, timeliness, and accuracy
measures. Since the first issuance of the map, over 90 percent of the
states have been rated "good" or "fair" for completeness, and about three
quarters of the states have been rated "good" or "fair" for crash
timeliness and crash accuracy. However, our review shows that states need
to continue to make improvements. In addition, since the majority of
states are rated as "good" or "fair" in measures, it is also difficult to
measure any progress made because many states have already reached the
highest rating. For example, as of the June 2005 data quality map, 41
states were rated "good" in crash completeness. Because these states have
already reached the highest rating for this measure, it may prevent
measuring any subsequent progress in this data quality standard.

There are other problems with using the data quality map to track trends.
Based on our review, a state's rating can temporarily decline for a
variety of reasons-even for implementing improvements. Hence, for that
period of time, the data quality map is not accurately reflecting a
state's true data quality status. Officials in one state told us that the
state was implementing a new electronic system which in the long term
would improve its data quality greatly. However, current CMV crash
reporting was slowed during implementation, and the state's rating went
down from "fair" to "poor."

o The data quality map's ratings for overall data quality combine data
from crashes with data from FMCSA's inspections, making it difficult for
map users to obtain an overall picture based solely on crash data. For the
overall data quality rating, individual crash measures for completeness,
timeliness, and accuracy are combined with inspection measures for
timeliness and accuracy. For each of the five individual crash and
inspection measures a state will receive a rating of "good," "fair," or
"poor."36 A state's overall rating depends on how well the state did
across all five measures (see table 4).

Table 4: Overall Data Quality Rating

                                        

Rating             Thresholds for overall data quality rating              
Good   No "poor" and a minimum of one "good" in a crash or inspection      
          measure                                                             
Fair   Maximum of one "poor" in a crash or inspection measure              
Poor   Two or more "poor" ratings in crash or inspection measures or a red 
          flag in the Crash Consistency Overriding Indicator                  

Source: FMCSA.

Currently, users can only view CMV crash data quality by individual
measures (completeness, timeliness, and accuracy). Separating the
inspection data and presenting a specific overall CMV crash data quality
rating (a combined rating composed of completeness, timeliness, and
accuracy for crash data only) would enhance a state's ability to
understand its crash data status and to monitor progress in improving the
information.

o While the State Safety Data Quality map provides a description of the
methodology used, it does not identify limitations to the methodology.
While many of these limitations to the map are acknowledged by FMCSA, they
are not publicly displayed on the State Safety Data Quality map Web site.
The absence of this information limits users' understanding of the map's
data and increases the potential for incorrect deductions and improper
map-based decisions.

FMCSA officials are aware of many of the limitations that we have
identified and we recognize their efforts to improve the State Safety Data
Quality map to date. However, they do not have a formal plan in place to
implement improvements. Further, it is important to acknowledge these
issues so that users understand the limitations of the data quality map as
a tool. During our review we learned that not only was the data quality
map consulted in awarding SaDIP funds, it was also consulted when awarding
Motor Carrier Safety Assistance Program (MCSAP) High Priority funds.37
Further, according to officials at FMCSA and Volpe National Transportation
Systems Center, there has been some discussion of expanding the usage of
the map in the future for CMV crash data efforts. As the data quality map
gains wider use, it will become even more important that these limitations
are addressed.

Conclusions

The grant program and FMCSA's collaborative efforts with states have had a
positive impact on improving the quality of states' crash data, therefore
ultimately enhancing the ability of both federal and state governments to
make highway planning and safety enforcement decisions. While states have
made progress in improving the quality of this data-in terms of
timeliness, completeness, accuracy, and consistency-much remains to be
done. With additional funding through the reauthorization of the Safe,
Accountable, Flexible, Efficient Transportation Equity Act: A Legacy for
Users and as states refine and target areas needing further improvement
with respect to their crash data, it is expected that additional states
will participate in the program. FMCSA will need a more formal framework
to better ensure that the decision-making process for awarding funds to
SaDIP program applicants is conducted uniformly.

FMCSA's efforts to improve the quality of commercial motor vehicle crash
data have brought considerable attention to the issues associated with
poor commercial motor vehicle crash data. Providing states with funding to
improve their CMV data quality and publicizing a rating through a data
quality map are incentives that work in tandem to maximize states'
efforts. It is clear that states pay attention to their ratings of "good,"
"fair," or "poor" on the data quality map. However, the limitations we
identified highlight some important concerns with the data quality map's
ability to measure progress and accurately portray states' commercial
motor vehicle crash data quality. As FMCSA continues to make improvements,
it will be important for these ratings to become more precise, so that
FMCSA and the states can obtain the clearest picture possible of the
progress being made.

Recommendations for Executive Action

We recommend that the Secretary of Transportation direct the Administrator
of FMCSA do the following:

o Establish specific guidelines for assessing state proposals for SaDIP
grants in order to better assess and prioritize states' funding requests
and provide uniformity in awarding funds.

o Increase the State Safety Data Quality map's usefulness as a tool for
monitoring and measuring commercial motor vehicle crash data by ensuring
that it adequately reflects the condition of the states' commercial motor
vehicle crash data and continues to motivate states in their improvement
efforts. Specifically, FMCSA should develop a plan for assessing and
improving the data quality map's methodology. In addition, FMCSA should
display an overall crash data rating separately from the inspection
rating, and provide information on the limitations of the State Safety
Data Quality map and the underlying ratings on FMCSA's Analysis and
Information (A&I) Online Web site.

Agency Comments and Our Evaluation

We provided a draft of this report to the Department of Transportation for
its review and comment. The department agreed with our findings and
recommendations in this report. Department officials provided some
technical comments and some minor additions to provide more detail on
FMCSA's training efforts.

We will send copies of this report to the interested congressional
committees, the Secretary of Transportation, and other interested parties.
We will make copies available to others upon request. In addition, the
report will be available at no charge on GAO's Web site at h 
ttp://www.gao.gov.

If you or your staff has any questions about this report, please call me
at (202) 512-6570. Contact points for our Offices of Congressional
Relations and Public Affairs may be found on the last page of this report.
Key contributors to this report are listed in appendix VI.

Katherine Siggerud Director, Physical Infrastructure

Objectives, Scope and MethodologyAppendix I

Congress asked us to review the Federal Motor Carrier Safety
Administration's (FMCSA) program for helping states improve their
commercial motor vehicle crash data. As part of this review, we were asked
to describe the benefits obtained through the program, identify what can
be done to improve the effectiveness of the program, and address concerns
regarding crash data raised in a February 2004 Department of
Transportation (DOT) Inspector General's report. The specific objectives
of this report were to explain (1) what is known about the quality of
commercial motor vehicle crash data and what states are doing to improve
it, and (2) the results of FMCSA's efforts to facilitate the improvement
of the quality of commercial motor vehicle crash data sent to the federal
government.

To provide information on the quality of states' commercial motor vehicle
(CMV) crash data and efforts to improve it, we reviewed grant applications
and project information from 39 states. Of these 39 states, 34
participated in the Safety Data Improvement Program (SaDIP) and 5 states
either chose not to participate in the program or their proposals were not
accepted. We also conducted site visits to six of these states--Georgia,
Minnesota, North Carolina, Ohio, Oklahoma, and Washington. We chose our
case study states based on a variety of criteria, including participation
in the SaDIP grant program, the type of agencies with which the state
works under SaDIP, the number of CMV crashes in the state, the number of
reported CMV crash fatalities, and data quality map ratings. To help
ensure that our states reflected a variety of experiences, we chose states
that had different combinations of these criteria. To understand the
insights and experiences of states that no longer participated-or had
never participated-in SaDIP, we also interviewed officials in Michigan,
Missouri, New Hampshire, and New Jersey.1  While the results from the case
studies and interviews cannot be projected to the universe of states, they
are nonetheless useful in illustrating the uniqueness and variation of CMV
crash data systems and the challenges states face in improving them.
During our case study visits we met and discussed the status of state
crash data systems with a variety of traffic safety data officials.2  The
discussions included gathering information on FMCSA's data quality
criteria3  but also included, for those participating in the program,
state objectives and progress made with SaDIP funds. For additional
perspective, we also interviewed key experts from organizations
responsible for the development of crash data systems and models used by
FMCSA, carrier industry officials, and public interest organizations.
Finally, we conducted a literature review of studies published by the
University of Michigan Transportation Research Institute (UMTRI). The
Institute plans to conduct studies in all states to determine where
problems are occurring in the collection of CMV crash data and in the
reporting of this data to FMCSA. As of September 2005, studies had been
conducted in eight states. We reviewed the studies and determined the
methodologies used in determining whether Motor Carrier Management
Information System (MCMIS) and state CMV crash data were sound. We
reviewed the studies conducted by UMTRI and determined that the
methodologies used in assessing the quality of the data states submit to
FMCSA were sound and that the studies provided sufficiently reliable
results for our purposes. Through site visits, a review of grant
applications, interviews with relevant stakeholders/experts, and these
studies, we were able to determine shortcomings in the reliability of
FMCSA's CMV crash data. However, we determined that the data was
sufficiently reliable for our purpose of case study selection.

To provide results of FMCSA's efforts to facilitate the improvement of the
state CMV crash data quality, we conducted interviews with officials from
participating states and from FMCSA concerning the administration and
management of SaDIP. We also analyzed the guidance and support provided by
FMCSA to states for CMV data improvement efforts and assessed FMCSA's role
in coordinating CMV data quality initiatives. In addition, we reviewed
FMCSA's State Safety Data Quality map and assessed the methodology used by
FMCSA in evaluating states' crash data quality. We interviewed officials
and key experts at FMCSA and the Volpe National Transportation Systems
Center responsible for developing and managing the data quality map. We
also interviewed state officials from states participating-and not
participating-in the SaDIP program to find out their views on the data
quality map and its use as a monitoring tool for CMV crash data
improvements.

We conducted our review from February 2005 through November 2005 in
accordance with generally accepted government auditing standards.

State Safety Data Quality Map LimitationsAppendix II

This appendix explains, in greater detail than the body of our report, the
concerns we (and others) have raised about the limitations in the
methodology FMCSA uses to develop ratings for the State Safety Data
Quality map. The measures FMCSA employs to measure the completeness,
timeliness, and accuracy of CMV crash data quality are limited, and do not
provide comprehensive measurements of these attributes as established by
the general data standards discussed in the body of this report. As a
result, the ability to draw conclusions about the actual quality of a
state's data is limited.

The definitions FMCSA uses for each of the crash measures are shown in
table 5, together with the criteria that constitute a rating of "good,"
fair," or "poor." In the sections that follow we explain the limitations
associated with each measure, followed by other limitations that stem from
the current methodology.

Table 5: State Safety Data Quality Map Measures for CMV Crashes

                                        

Data quality                Measure                       Criteria        
     standard                                                                
Completeness The measure compares CMV fatality data Good = match to FARS  
                in the MCMIS database against those in is greater than or    
                the Fatality Analysis and Reporting    equal to 90 percent   
                System (FARS). The FARS data is                              
                available through 2003.                Fair = match to FARS  
                                                       is between 80 and 89  
                                                       percent               
                                                                             
                                                       Poor = match to FARS  
                                                       is below 80 percent   
Timeliness   This measure reflects the percentage   Good = the percentage 
                of state-reported crash records        of reported records   
                uploaded to the MCMIS database within  within 90 days is     
                the 90-day standard.                   greater than or equal 
                                                       to 85 percent         
                                                                             
                                                       Fair = the percentage 
                                                       of reported records   
                                                       within 90 days is     
                                                       between 60-84 percent 
                                                                             
                                                       Poor = the percentage 
                                                       of reported records   
                                                       is less than 60       
                                                       percent               
Accuracy     This measure reflects the percentage   Good = the percentage 
                of state-reported crash records (fatal of matched records is 
                and non-fatal) that were matched to a  greater than or equal 
                motor carrier in MCMIS over a 12-month to 95 percent         
                period.                                                      
                                                       Fair = the percentage 
                                                       of matched records is 
                                                       between 85 and 94     
                                                       percent               
                                                                             
                                                       Poor = the percentage 
                                                       of matched records is 
                                                       below 85 percent      

Source: FMCSA.

Completeness: Overall Completeness is Based on Fatal Crashes Only

A key limitation in FMCSA's measure of crash data completeness is the
inability to evaluate completeness against nonfatal CMV crashes.
Currently, the total number of nonfatal crashes occurring on a
state-by-state basis is

unknown; no baseline exists against which to measure these records.1 
Consequently, FMCSA is limited to measuring crash data completeness with
fatal CMV crashes, which is approximately 3 percent of all CMV crashes.2

Even within this narrowed dataset, the use of FARS3 as the basis of
comparison poses other limitations:

o There are notable differences in definitions used in the two databases.
Most importantly, MCMIS can be subject to individual state definitions
that may differ from FARS. According to officials at the Volpe National
Transportation Systems Center, the range of these state-driven
definitional differences is unknown. However, based on anecdotal evidence
received from states and reported by Volpe, these differences can vary by
state and can account for about 4 percent of crash records.4 Some examples
are below:

o A crash fatality resulting from the private use of a large truck may not
meet the criteria as a reportable crash for MCMIS, but it is considered a
large truck crash fatality by FARS.

o MCMIS defines a large truck as any truck greater than a 10,000 gross
vehicle weight rating. Many states do not collect the gross vehicle weight
rating of vehicles and instead define a large truck based on an older
FMCSA reporting criterion of greater than six tires or another definition
of their choosing. As such, counts of large trucks derived from state
crash databases (and reported to FARS) may be inconsistent with counts of
vehicles that were reported to MCMIS.5

The effect of these definitional differences is evident when comparing
MCMIS fatal CMV crashes to FARS data-the state's index on completeness for
crash data rises above 100 percent. The extent to which this occurs is
substantial: in our analysis of the June 2005 issuance of the map, 24 of
51 states (47 percent) had a completeness measure of greater than 100
percent.  6

o FARS data is not current. FARS is released once per year, detailing the
prior calendar year's crash statistics. The lag between the most recently
available FARS data and the most recent map issuance can be considerable.
For example, the 2004 FARS annual report is not scheduled to be released
until fall 2005 and as a result, the baseline for completeness data in the
June 2005 map is based upon 2003 crash data.

Timeliness: Timeliness is Not Based on All Reported Crashes

The timeliness measure currently relies on a subset of records states
submit to MCMIS. Any record that has been edited or changed since it was
originally entered into MCMIS is not included in the calculations.7 If a
record is edited, the initial upload date is replaced with the date it was
updated. Because FMCSA's timeliness rating is based on the percentage of
crash reports uploaded within 90 days of the crash, and edited records no
longer reflect their initial upload date, they cannot be used in the
timeliness calculation without distorting a state's rating. The
consequence, however, is that timeliness is not measured against the
entire universe of crashes in MCMIS-the more records a state edits, the
fewer records its timeliness rating represents. FMCSA has acknowledged
this problem with the edited records and is taking steps to resolve it.

Another limitation with this measure is that efforts to reduce backlogs of
crash records-a positive effort-can have a negative effect on the
timeliness rating. If a state submits a backlog of reports from CMV
crashes that occurred more than 90 days previously, and the crash took
place during the period of time which the FMCSA rating covers, then the
state's timeliness rating will be negatively affected. Conversely, if a
state has a large backlog, its current timeliness rating may not be a
meaningful representation of timeliness because the methodology has no way
to capture those records that are accruing in the backlog.

Accuracy: Accuracy is Based on Only One Variable

The measure for accuracy on the data quality map is based only on a match
of a CMV crash report in MCMIS, against a registered carrier in MCMIS. The
U.S. Department of Transportation (DOT) assigns each interstate motor
carrier a unique identifier number-the DOT number. For this measure,
accuracy is evaluated primarily based on the degree to which a carrier's
DOT number is the same as identified for a crash in MCMIS.8 Therefore, a
crash report with a missing or invalid DOT number may be considered
inaccurate, even if the rest of the information on the report provides
accurate information on the crash. Currently, the data quality map does
not assess the accuracy of any other data elements; however, FMCSA has
been working with states to improve their collection of CMV crash
information. For example, recently FMCSA distributed visor cards to state
officials that explain how to determine who the carrier is, and where the
correct carrier's DOT number can be found. See appendix III for copies of
these cards.

FMCSA Reportable Crash, CMV, and Carrier Identification Visor
CardsAppendix III

The following are copies of visor identification cards that FMCSA created
as educational tools for law enforcement officers. The cards can be placed
in officer's sun visor and referenced to determine whether a CMV crash
should be reported, for identifying a vehicle as a CMV, and for
identifying the correct carrier involved in the crash. FMCSA provided
these to states to distribute to enforcement officers with the intention
they will increase the officers' ability to properly identify a CMV and a
reportable CMV crash.

Figure 3: FMCSA Reportable Crashes

Figure 4: Reportable Commercial Motor Vehicle Configurations and Cargo
Body Type

Figure 5: Responsible Carrier and Correct DOT Number Identification

SaDIP Grant and Cooperative Agreement Distribution by StateAppendix IV

Table 6: Distribution of SaDIP Grants by State

                                        

        State           FY2002    FY2003      FY2004      FY2005        Total 
Alaska             $198,000         0           0    $218,626     $416,626 
California                0              $100,000           0      100,000 
Connecticut         234,056         0           0           0      234,056 
Georgia             269,820         0     409,000     553,733    1,232,553 
Indiana             179,321   242,423           0           0      421,744 
Kansas                    0         0           0     373,436      373,436 
Kentucky            199,000         0           0           0      199,000 
Maryland            145,600         0           0           0      145,600 
Massachusetts       100,000   249,972           0           0      349,972 
Michigan                  0   115,845           0           0      115,845 
Minnesota           363,000         0     620,000           0      983,000 
Montana                   0         0           0     328,564      328,564 
Nebraska                  0         0       2,000       3,342       $5,342 
Nevada              427,443         0           0     350,000      777,443 
New Hampshire             0    98,068           0           0       98,068 
New Mexico                0         0           0     250,000      250,000 
New York                  0         0     300,000           0      300,000 
North Carolina            0         0     193,350           0      193,350 
Oklahoma                  0         0     131,111           0      131,111 
Pennsylvania              0         0           0   2,000,000    2,000,000 
Rhode Island         25,000         0      26,000           0       51,000 
South Carolina      350,000         0           0           0      350,000 
South Dakota        270,000         0           0           0      270,000 
Tennessee                 0         0     436,027     149,334      585,361 
Texas                     0   148,424     533,611           0      682,035 
Utah                      0         0     433,500     128,853      562,353 
Vermont                   0         0           0     134,000      134,000 
Washington                0         0     803,935           0      803,935 
West Virginia       157,500         0           0     603,943      761,443 
Total            $2,918,740  $854,732  $3,988,534  $5,093,831  $12,855,837 

Source: FMCSA.

Note: This data is current as of September 2005.

Table 7: Distribution of SaDIP Cooperative Agreements by State

                                        

      State            FY2002   FY2003       FY2004       FY2005        Total 
Colorado                 0        0            0   $1,460,000   $1,460,000 
Georgia                  0        0            0      612,051      612,051 
Iowa              $728,065        0            0      226,400      954,465 
Louisiana          829,625        0            0      569,000    1,398,625 
Ohio               406,330        0      $91,000      224,400      721,730 
Oklahoma                 0        0    1,150,390            0    1,150,390 
Tennessee          281,954        0      219,543      426,382      927,879 
Vermont                  0        0            0      136,500      136,500 
Virginia            42,300        0            0      431,617      473,917 
Total           $2,288,274       $0   $1,460,933   $4,086,350   $7,835,557 

Source: FMCSA.

Note: This data is as of September 2005.

SaDIP Case Study StatesAppendix V

As part of our work, we conducted six case studies to examine how states
are working to improve commercial motor vehicle crash data. Our visits to
the six states yielded additional information about crash data quality
improvement activities, the nature of their efforts, and the extent of
progress made. States were chosen on a wide variety of factors, including
crash data quality and participation in the SaDIP program.

Georgia

Georgia received a total of $1,844,604 between 2002 and 2005 (see table 8)
to conduct crash data improvement projects. During this period, Georgia
made significant improvement in its crash data quality, despite undergoing
a major state government reorganization process.

Table 8: Georgia SaDIP Funding History

                                        

      Award year      Funds awarded      Agency            Award type         
2005             $612,051           FMCSA       Cooperative agreement      
2005             $553,733           FMCSA       Grant                      
2004             $409,000           FMCSA       Grant                      
2002             $269,820           NHTSA       Grant                      

Source: FMCSA.

SaDIP Projects

Georgia received its first SaDIP grant from the National Highway Traffic
Safety Administration (NHTSA), and this grant was used to accomplish
several projects, including: (1) hiring temporary employees to extract CMV
crash reports from backlogged paper copies and microfilm them; (2)
precoding the paper crash reports in preparation for their entry into the
state repository system; (3) renting a mobile trailer to house temp
employees; (4) developing a system to electronically transfer crash data
from the state repository to the national database; (5) adding edit checks
to the state crash database; (6) purchasing a new microfilm
scanner/reader; and, (7) hiring a contractor to update the state's crash
report manual. Progress on these projects was reported quarterly to NHTSA.

Georgia received two additional grants after FMCSA became the lead agency
for the SaDIP program. The first grant was used to maintain the temporary
employees. Additionally, these funds were used to hire a statistician as a
SaDIP advocate. This grant allowed Georgia to eliminate its reporting
backlog from 2003 to the present. The second grant that Georgia received
was used to fund its electronic crash records system. This system will
electronically transfer crash records from local jurisdictions to the
state repository system. This project has been put on hold following a
state government reorganization.

Crash Data Collection/Reporting Issues

Georgia historically has had the worst CMV crash data quality rating in
the nation (see crash data quality statistics, table 9). This is due to
the fact that until recently, Georgia was not submitting crash data to
FMCSA. Georgia's current overall rating is "good," and it is rated "good"
for completeness and "fair" for timeliness and accuracy. For all other
rating periods prior to March 2004, Georgia had an overall rating of
"poor." This improvement in its data quality is due in large part to the
SaDIP program. Georgia's non-reporting was due to a technical problem, but
the state was able to correct it in a timely manner because of the SaDIP
funding it received.

Table 9: Georgia Crash Data Quality Statistics (Percentages)

                                        

                    March     June     September    December    March    June 
                     2004     2004          2004        2004     2005    2005 
Completeness       117      115            95          95       95      95 
Timeliness           6        8            27          48       65      80 
Accuracy            84       82            82          83       94      93 

Source: FMCSA.

SaDIP funds are the only federal funds currently being used to address
data quality improvement in Georgia. Because the state's data quality was
so poor, both FMCSA and the Federal Highway Administration indicated they
would withhold federal funding unless Georgia improved its reporting. It
was at this point that Georgia applied for SaDIP, and began focusing on
its data quality. As of July 1, 2005, Georgia began a major state
government reorganization process, and officials were unsure of how this
would affect the quality of its crash data because all collection and
reporting functions are moving to new departments. Officials expect to see
a decrease in the state's rating during the transition period, but do not
expect long-term problems.

State/Federal Coordination Issues

State and federal coordination on issues related to traffic safety data
has not always been effective in Georgia. Between 1998 and 2002, Georgia
had only submitted 250 fatal CMV crash records to FMCSA, yet between 2002
through 2003 Georgia submitted 410 CMV fatalities to FARS. Georgia's
Department of Public Safety, the agency responsible for crash reporting,
did not have a functional computer system to upload the state's crash file
to FMCSA. The problem was discovered in 2002 and FMCSA offered assistance
to fix the problem, but the Department of Public Safety did not accept the
offer. FMCSA then threatened to withhold Georgia's highway safety funds
until the data issues were resolved; at this point the Department of
Public Safety applied for-and received-SaDIP funding, and began work to
resolve the crash data problems.

Data Quality Map

Georgia officials stated that the data quality map provided an incentive
for the state to make data quality improvements in order to raise its
rating from "poor" to "good." State and federal officials acknowledge that
the map is one indication of the progress the state has made in improving
its CMV data quality.

Georgia officials never disagreed with their state's rating on the data
quality map. However, officials felt that the map did not recognize
improvements that were taking place within the state to improve crash
data. The state plans to continue to make improvements to the crash
records system even after the state reaches the "good" rating.

Minnesota

Minnesota received a total of $983,000 in 2002 and 2004 (see table 10) to
conduct crash data improvement projects. During this period, Minnesota
received safety data improvement grants from NHTSA and FMCSA.

Table 10: Minnesota SaDIP Funding History

                                        

      Award year           Funds awarded           Agency       Award type    
2004                         $370,000        FMCSA        Grant            
2004                         $250,000        FMCSA        Grant            
2002                         $363,000        NHTSA        Grant            

Source: FMCSA.

SaDIP Projects

Minnesota received its first SaDIP grant while NHTSA was administering the
program. This grant went to the Department of Public Safety, Office of
Traffic Safety-the agency that receives NHTSA highway safety funds. The
grant was used to fund two projects, including an electronic post-crash
inspection form (Minnesota conducts in-depth post-crash inspections for
crashes involving CMVs), and to purchase software that will allow for
electronic transfer of CMV crash records between Driver and Vehicle
Services and the state patrol (both agencies within the Department of
Public Safety). In Minnesota, Driver and Vehicle Services collects crash
reports from law enforcement agencies, and the state patrol is the office
in charge of reporting crashes to FMCSA.

The Minnesota state patrol received subsequent funding after FMCSA became
the lead agency on the SaDIP program. These funds have been used to
provide computer hardware so that state patrol officers can access the
Internet and submit their crash reports electronically through an online
form. The second component of the grant is to provide ongoing CMV crash
training to law enforcement officers throughout the state. This training
will focus on correctly identifying CMVs and teaching officers why CMV
crash data is important, and what it is used for at the state and national
levels.

Crash Data Collection/Reporting Issues

Before SaDIP funds were received, the Minnesota state patrol had limited
access to Driver and Vehicle Services records. As a result, the state
patrol was only able to report crashes to FMCSA that were investigated by
their officers or were sent directly to the state patrol by local
jurisdictions. Now all CMV records that are available at the Driver and
Vehicle Services are reported to FMCSA (see crash data quality statistics,
table 11).

Table 11: Minnesota Crash Data Quality Statistics (Percentages)

                                        

                    March     June     September    December    March    June 
                     2004     2004          2004        2004     2005    2005 
Completeness       117      117           116         123      121     121 
Timeliness          72       59            58          50       67      81 
Accuracy            78       74            82          86       91      93 

Source: FMCSA.

The State of Minnesota encountered difficulty receiving crash reports from
local law enforcement agencies. In addition, law enforcement officers
incorrectly identified CMVs on crash reports. If law enforcement officers
do not correctly identify the vehicle involved in the crash as a CMV, then
the crash may not be extracted from the state crash file for submission to
FMCSA.

State/Federal Coordination Issues

Minnesota officials said that the state now has a crash data users group
as a subsection of its Traffic Records Coordinating Committee. This group
allows all users of crash data to discuss their data needs and
limitations, and provides a forum for making collaborative recommendations
regarding crash data improvements.

Minnesota officials and the FMCSA Division Administrator work very closely
with each other to monitor and improve CMV crash data quality. Neither of
these groups feels that the FMCSA data quality rating map is an accurate
portrayal of a state's data quality. For example, Minnesota's rating
decreased due to the fact that it was implementing its electronic data
transfer software, and during that period crash records were not reported
to FMCSA in a timely fashion. The map showed the state's rating as "poor"
during this period, which did not reflect the fact that the state was
making strides in improving its data quality.

Minnesota uses several sources of funding to improve its overall crash
data and CMV crash data. In addition to FMCSA funding, NHTSA and state
funds have been used to improve the state's data quality.

Data Quality Map

Minnesota crash data has gone from an overall "fair" rating to "poor" to
its current rating as "good." The drop in the state's rating was actually
a result of the state not being able to report crashes during the
implementation of SaDIP-funded improvement projects. After the electronic
systems were fully implemented, the state's rating improved to "good" for
completeness, and "fair" for timeliness and accuracy.

Officials in Minnesota recognize that the map is an effective tool for
getting the attention of decision-makers. While the map may not be an
accurate assessment of the state's current data quality, both state and
federal officials recognize that it does help to spotlight Minnesota's
data quality status. Federal and state officials use the data quality map
as a measure of the state's progress in improving data quality. State
officials also use the map as an indicator of the success of the state's
SaDIP project. Finally, officials in Minnesota said that the data quality
map has been used to identify data quality problems in the state,
including timeliness and accuracy, and to create projects to improve those
measures.

North Carolina

North Carolina received a total of $193,350 in 2004 (see table 12) to
conduct crash data improvement projects. While North Carolina has
excellent crash data reporting at the state level, North Carolina
continues to have complications reporting that data to FMCSA. North
Carolina is a priority state for FMCSA given the fact that it is one of
the top five states for commercial motor vehicle accidents in the nation.

Table 12: North Carolina SaDIP Funding History

                                        

      Award year           Funds awarded           Agency       Award type    
2004                         $193,350        FMCSA        Grant            

Source: FMCSA.

SaDIP Projects

North Carolina is using its SaDIP grant to conduct two projects. First,
the state conducted an analysis to determine what differences existed
between the state data file and the federal crash file, and to try to
determine why these differences were occurring. The completed analysis did
not appear to fully address North Carolina's biggest problem, which is the
data transfer between the two files.

North Carolina is also using SaDIP to update a backlog it has in entering
crash reports into the state data file by providing overtime pay to
employees.

Crash Data Collection/Reporting Issues

Historically, North Carolina has had poor CMV crash data (see table 13).
North Carolina is also one of FMCSA's priority states because of the large
amount of crashes that take place in the state. It was rated as having
"poor" data for each period since FMCSA began rating states.

Table 13: North Carolina Crash Data Quality Statistics (Percentages)

                                        

                   March    June     September    December   March      June 
                    2004    2004          2004        2004    2005      2005 
Completeness       82      82            77          77      77 77 
Timeliness          0     100             0           0       1  2 
Accuracy           81     100            81          80      80 81 

Source: FMCSA.

North Carolina is unique in that it does not have a problem receiving
crash reports from localities across the state. In other states, this is
the biggest problem contributing to the states' data quality issues.1 In
North Carolina the problem is getting records from the state crash file
into the FMCSA data systems. FMCSA officials worked extensively with North
Carolina to fix the file compatibility issues that are creating the
problem exporting the data to FMCSA, but North Carolina has failed to
correct its information technology problems. According to state officials,
the biggest issue in North Carolina is backlog. Even if the state corrects
its backlog, the state's data rating will not improve because the records
are not transferring correctly into the federal database. Additionally,
North Carolina does not require certain elements to be captured on its
crash report or in its state database that are required in the federal
database. Specifically, North Carolina does not require that motor carrier
DOT numbers are collected, which may contribute to the state's "poor"
accuracy rating.

While North Carolina has very good data at the state level, the state's
largest problem is reporting that data to the federal database. It appears
that North Carolina's SaDIP projects are not focusing on this issue.

State/Federal Coordination Issues

North Carolina coordination between state agencies involved in CMV crash
data has not always been effective. A major reorganization of the state's
highway safety agency contributed to this lack of coordination, but the
state is working to improve cooperation among these agencies with regards
to traffic records. The largest factor contributing to North Carolina's
poor crash data quality appears to be a lack of understanding by some
state officials regarding how to convert a crash data file to the correct
format in order to submit it to FMCSA. The SaDIP grant has provided the
state the opportunity to review its crash data problems, a project that
the state would not have conducted otherwise. The state is also using its
own funds to implement an electronic crash reporting system in order to
get crash data into the state data file more quickly.

Data Quality Map

State and federal officials in North Carolina use the data quality map to
measure the progress the state is making with its SaDIP grant and with
data improvements overall.

North Carolina state officials use the map as an incentive for
implementing data improvements and they appear motivated to change North
Carolina's rating for the better. The map has also been used in North
Carolina to identify data quality problems and target improvement efforts,
although more still needs to be done.

Both federal and state officials expect to continue to make improvements
to the state's traffic records infrastructure after the state achieves the
highest data rating.

Ohio

Ohio received a total of $721,730 between 2002 and 2005 (see table 14) to
conduct crash data improvement projects. Ohio has been very proactive in
addressing traffic safety data concerns and the state has paid particular
attention to using crash data for planning.

Table 14: Ohio SaDIP Funding History

                                        

    Award year      Funds awarded         Agency            Award type        
2005                  $224,400     FMCSA          Cooperative agreement    
2004                   $91,000     FMCSA          Cooperative agreement    
2002                  $406,330     NHTSA/FMCSA    Cooperative agreement    

Source: FMCSA.

SaDIP Projects

Ohio was one of the first 5 pilot states to participate in the SaDIP
program. It received funds through a cooperative agreement with the
General Services Administration (GSA) and NHTSA. Through this agreement,
the state created electronic crash reporting capabilities, purchased
handheld devices, modified crash reporting software, and is providing
training to law enforcement officers to help them properly identify CMV
crashes.

Until recently, when FMCSA gained authority for the SaDIP cooperative
agreements, Ohio had not submitted any progress reports to NHTSA on the
status of these projects.

Crash Data Collection/Reporting Issues

Ohio has a "fair" overall data rating, with "good" ratings for
completeness and accuracy, but a "poor" rating for timeliness (see crash
data quality statistics, table 15). Ohio has been very aware of the
importance of CMV crash data and state officials have been working to
improve it for a long time, even before the SaDIP program.

Table 15: Ohio Crash Data Quality Statistics (Percentages)

                                        

                    March    June September 2004    December    March   June 
                     2004    2004                       2004     2005   2005 
Completeness        95      95             95          95       95     95 
Timeliness          15      17             11          23       40     46 
Accuracy            84      97             97          98       98     98 

Source: FMCSA.

The state of Ohio has over 1,000 local law enforcement jurisdictions that
are responsible for reporting CMV accidents. Both large and small
jurisdictions are likely responsible for Ohio's "poor" crash data
timeliness rating. The state has reduced the time lag in receiving crash
reports from local jurisdictions from 62 days in 2004 to 30 days in 2005.

State/Federal Coordination Issues

Ohio state agencies involved in the collection and reporting of CMV crash
data appear coordinated with each other, and have an active Traffic
Records Coordinating Committee, which includes participation by FMCSA
division officials. Ohio uses multiple sources of funding to address data
quality issues, including state funds, and the state is proactive in data
improvement projects.

Data Quality Map

State and federal officials said that the map provided one gauge of Ohio's
data quality, but they felt the map would be more useful if it were
updated more often. The map has brought more attention to data quality
issues in Ohio and is included in reports to state leaders.

State and federal officials in Ohio said that data quality improvements
would still be a top priority regardless of the data quality map. Ohio
officials also stated that the inaccuracies in the map had a big effect on
staff morale. Local officials said they are working hard to improve their
state's data quality, but the map does not accurately capture that
improvement. State officials recognize the map provides a major incentive
for implementing data quality improvements and maintaining the state's
standing as a leader in traffic data quality.

Oklahoma

Oklahoma received a total of $1,281,501 in 2004 (see table 16) to conduct
crash data improvement projects. Oklahoma's traffic records coordinating
committee has taken the lead in the coordination of crash data projects in
the state.

Table 16: Oklahoma SaDIP Funding History

                                        

     Award year       Funds awarded        Agency          Award type         
2004                  $1,150,390      FMCSA      Cooperative agreement     
2004                    $131,111      FMCSA      Grant                     

Source: FMCSA.

SaDIP Projects

Oklahoma began participation in the SaDIP program in July 2004 when it
received a grant to purchase computer equipment, and then entered into a
four year cooperative agreement with FMCSA in December 2004 to support its
long term data quality improvement plans. This includes conducting a
traffic records assessment, hiring a SaDIP coordinator, funding data entry
for the record backlog, revising the state's crash report, and initiating
a mobile data collection pilot project.

Oklahoma officials said that the projects that are taking place using
SaDIP funds would not have been funded otherwise. The SaDIP grant and
long-term cooperative agreement have allowed the state to focus on CMV
data and to begin to make data-driven decisions in its highway safety
planning.

Crash Collection/Reporting Issues

Until recently Oklahoma had a nine month crash report backlog to be
entered into the state system (see crash data quality statistics, table
17). The backlog was primarily due to insufficient state resources.
Overtime hours funded by SaDIP has helped to alleviate this backlog.

Table 17: Oklahoma Crash Data Quality Statistics (Percentages)

                                        

                    March     June     September    December    March    June 
                     2004     2004          2004        2004     2005    2005 
Completeness       104      104           102         102      101     101 
Timeliness          93       92            93          92       90      83 
Accuracy            84       86            87          86       86      84 

Source: FMCSA.

State/Federal Coordination Issues

Oklahoma has been working with the GSA to receive payments on its
cooperative agreement, and officials said that this has caused the state
some confusion.

Oklahoma's crash data divisions are housed in the state's Department of
Public Safety and Department of Transportation. While crash records staff
in both agencies work closely together to make sure that the state's crash
file is complete, Oklahoma officials were unaware of the value of CMV
crash data at the national level, or of the criteria that FMCSA uses to
rate CMV crash data.

Oklahoma is using funds from several sources to improve its entire crash
data system, including funds provided by NHTSA and the Federal Highway
Administration. SaDIP funds are the only funds that the state is using
specifically for CMV data improvements.

Data Quality Map

While high-ranking officials in Oklahoma's Department of Public Safety and
the Highway Patrol are familiar with the data quality map, state officials
involved with the SaDIP program were less familiar. We found that state
officials involved with the SaDIP program had a detailed understanding of
their state's data quality, but had not used the data quality map as an
indicator of progress for the SaDIP cooperative agreement.

Oklahoma officials stated that they would undertake their data improvement
program regardless of whether or not the state was ranked "good." State
officials, for example, recognized that more improvements could be made in
the state's timeliness measure even though the state had a "good" rating
in this category.

Washington

Washington received a total of $803,935 in 2004 (see table 18) to conduct
crash data improvement projects. The state's traffic records coordinating
committee is leading an electronic information processing initiative
designed to reduce crash reporting inefficiencies and help the state meet
national traffic safety goals.

Table 18: Washington SaDIP Funding History

                                        

      Award year           Funds awarded           Agency       Award type    
2004                         $188,460        FMCSA        Grant            
2004                         $615,475        FMCSA        Grant            

Source: FMCSA.

SaDIP Projects

SaDIP is primarily being used to implement an electronic data feed between
the Washington Department of Transportation and the state patrol. This
will allow records to be submitted instantly to the state patrol when they
are entered into the state department of transportation database. It will
also make records searchable so that eligible CMV crashes that are
misidentified and not sent to the state patrol can be identified as CMV
crashes, improving the completeness of the state's CMV crash data.

SaDIP is also being used to eliminate a six-month backlog of crash reports
that need to be entered into the state's database housed at the Washington
Department of Transportation. Until reports are entered into the state
Department of Transportation system, they cannot be transferred to the
state patrol electronically, nor can those reports incorrectly identified
as non-CMV crashes be easily identified. Until this takes place,
Department of Transportation employees identify CMV reports among all
crash reports and provide the state patrol with paper copies of the
reports.

Crash Collection/Reporting Issues

The largest problem that Washington State has with its crash data is
receiving the entire crash report, including supplemental form, from the
law enforcement offices that generate them (see crash data quality
statistics, table 19).

Table 19: Washington Crash Data Quality Statistics (Percentages)

                                        

                    March     June     September    December    March    June 
                     2004     2004          2004        2004     2005    2005 
Completeness        96       96           103         103      108     108 
Timeliness          99       98            98          99       99      99 
Accuracy            95       96            96          97       97      97 

Source: FMCSA.

Washington State also has a problem with incorrect identification of CMVs
on police accident reports. If police do not correctly identify the
vehicle involved in the crash as a CMV, then it may not get extracted from
the state crash file for submission to FMCSA.

State/Federal Coordination Issues

Washington State has very good cooperation among state agencies involved
in crash data collection and reporting. The state also has a good working
relationship with its FMCSA division office. It also uses FMCSA and state
funds to address data quality issues, making decisions on how to use these
funds effectively through an active Traffic Records Coordinating
Committee.

While SaDIP initiatives in Washington State are a topic of discussion at
the Traffic Records Coordinating Committee meetings, the committee
coordinator was not specifically aware of FMCSA's grant-making process and
how those grants are accessed and then allocated by the state. Whereas
NHTSA funds are accessed via the Traffic Records Coordinating Committee
forum, FMCSA funds are processed within the Washington State Patrol's
Commercial Vehicle Division. To date there has been no problem targeting
safety data funding to immediate priorities as identified by the Traffic
Records Coordinating Committee, but this is an area that may need to be
addressed in the future.

Data Quality Map

Washington State officials recognize that the data quality map is an
important indicator of the state's progress in improving its crash records
system. According to state officials, the data quality map has been used
to measure the progress of the state's data quality improvements in
general.

State officials also indicated that they would continue to make
improvements to their traffic records systems regardless of the state's
data quality rating. Specifically, the Traffic Records Coordinating
Committee is helping to coordinate the state's electronic reporting
system.

Even though the state has consistently ranked "good" on the map since its
inception, state officials report that the map has been helpful for
identifying areas where the state can make data quality improvements.

GAO Contact and Staff AcknowledgmentsAppendix VI

Katherine Siggerud (202) 512-6570 or [email protected]

In addition to the contact named above, Randall Williamson, Assistant
Director; Tida Barakat; Jonathan Carver; Bert Japikse; Sara Ann
Moessbauer; Fred Swenson; Sharon Silas; and Stan Stenersen made key
contributions to this report.

(542054)

www.gao.gov/cgi-bin/getrpt?GAO-06-102.

To view the full product, including the scope

and methodology, click on the link above.

For more information, contact Katherine Siggerud, (202) 512-6570,
[email protected].

Highlights of GAO-06-102, a report to congressional committees

November 2005

HIGHWAY SAFETY

Further Opportunities Exist to Improve Data on Crashes Involving
Commercial Motor Vehicles

Large trucks make up 3 percent of the nation's registered vehicles, but
they were involved in 11 percent of all fatal crashes in 2003. To reduce
the fatality rate, the Federal Motor Carrier Safety Administration (FMCSA)
sets national safety goals and works in partnership with states to reach
them. Crash data collected by states and submitted to FMCSA is key to
these efforts, and to be fully useful, this data must be complete, timely,
accurate, and collected in a consistent manner. GAO addressed (1) what is
known about the quality of commercial motor vehicle crash data, and what
states are doing to improve it, and (2) the results of FMCSA's efforts to
help states make improvements.

What GAO Recommends

To ensure uniformity in awarding data improvement funds to states, FMCSA
should establish specific guidelines for assessing and awarding state
funding requests.

Also, in order to address limitations in its data quality map, FMCSA
should develop a plan for assessing and improving the map's methodology,
and it should provide a crash specific data rating and limitations of the
map on its Web site.

The Department of Transportation agreed with our findings and
recommendations in this report.

Overall, commercial motor vehicle crash data does not yet meet general
data quality standards of completeness, timeliness, accuracy, and
consistency. For example, FMCSA estimates that nearly one-third of
commercial motor vehicle crashes that states are required to report to the
federal government were not reported, and those that were reported were
not always accurate, timely, or consistent. States are undertaking four
types of activities to improve data quality, including analyzing existing
data to identify problems and develop plans for addressing them, reducing
backlogs of data that have not been entered into state-level databases,
developing and implementing electronic data systems, and providing
training. As a result of these efforts, states have recently improved both
the timeliness and the number of reportable crashes submitted to FMCSA.

FMCSA has two main efforts to support states in improving their reporting
of commercial motor vehicle crash information-a commercial vehicle crash
data improvement program and a data quality rating system-and both appear
to be beneficial. Through the data improvement program, FMCSA has provided
nearly $21 million in discretionary grants to 34 states from 2002 through
2005. These grants have ranged from $2,000 to $2 million and have helped
states conduct a variety of data improvement activities. GAO did not find
problems with FMCSA's oversight of the program, but we did note that FMCSA
does not have formal guidelines for awarding grants to states. As state
participation in the program increases, formal guidelines and systems
would likely assist FMCSA in prioritizing states' requests and ensuring
consistency in grant awards.

FMCSA's second major effort, a tool for rating states' data quality, has
proven to be an important tool for states to use in improving their crash
data as well. These results are presented in a map that rates each state's
data quality as "good," "fair," or "poor." According to both FMCSA and
state officials, the map and the underlying rating system serve as an
incentive for states to improve their crash data. While the map is useful,
GAO identified problems in the methodology used for developing ratings.
These problems may potentially lead to drawing erroneous conclusions about
the extent of improvements that have been made, and discourage states from
continuing to devote attention and resources to areas needing improvement.

FMCSA's June 2005 data quality map showing each state's overall data
quality rating for crash and inspection data.
*** End of document. ***