Survey of Vehicle Owners in the
On-Board Diagnostics Program
- Final Report -
WESTAT
An Employee-Owned Research Corporation

-------
Survey of Vehicle Owners in the
On-Board Diagnostics Program
- Final Report -
Prepared for
Certification Division
Office of Mobile Sources
Office of Air and Radiation
U.S. Environmental Protection Agency
EPA Contract Number. 68-01-7359
Delivery Older 19
Prepared by:
Westat, Inc.
1650 Research Boulevard
Rockville, Maryland 20850
My 18, 1990

-------
TABLE OF CONTENTS
Section	Page
ACKNOWLEDGEMENTS	viii
EXECUTIVE SUMMARY	 ix
1.	INTRODUCTION	 1
1.2 Survey Report	2
2.	SURVEY METHODOLOGY	 3
2.1	The Questionnaire	 3
2.2	The CATI System	 6
2.3	Survey Sample			9
2.4	Data Collection	10
2.5	Quality Control	10
3.	SURVEY RESULTS	13
3.1	Survey Response Rate	13
3.2	Statistical Methods				14
3.3	Survey Results			19
Appendices
A	SURVEY QUESTIONNAIRE 	A-l
B	CATI FLOW CHART	B-l
C	LIST OF VEHICLES IN SAMPLE	C-l
D	CONFIDENTIALITY PLEDGE 	D-l
iii

-------
18
21
21
22
23
24
24
.25
.25
.26
.27
.28
.28
.30
.30
.31
LIST OF TABLES (Continued)
95 percent Confidence interval for percentages by number of respondents and
estimated percentage	
Car make from the sample tape	
Car's model year from the sample tape	
Responses to "When you acquired the car, was it new, that is never previously
titled, or was it a used car?"	
Responses to "Since 1988 , has your car been inspected or tested by the State for
its emissions?"	
Responses to "Approximately how many miles do you have on the car?"	
Coded answers to "Approximately how many miles do you have on the car?" for
statistical analysis	
"Has the engine warning light [malfunction indicator light] ever come on and
stayed on after starting the car?"	
Cross-tabulation of MIL illumination and total mileage, with number of
respondents and column percentages	
Cross-tabulation of MIL illumination and car make, with number of respondents
and column percentages	
Percentage of respondents reporting MIL illumination by car make and total
mileage	
Approximate car mileage at the last malfunction indicator light illumination, derived
from several questions	
Responses to "When the light came on, was the car running satisfactorily or was
there a problem?" (Driveability)	
Owner response to malfunction indicator light	
Owner response to malfunction indicator light illumination used for statistical
analysis					
Cross-tabulation of owner response and driveability with number of respondents
and column percentages	
iv

-------
LIST OF TABLES (Continued)
Table	Page
Table 3-17 Cross-tabulation of owner response and recent emissions inspection with number
of respondents and column percentages	32
Table 3-18 Cross-tabulation of owner response and car mileage with number of respondents
and column percentages			32
Table 3-19 Cross-tabulation of owner response and car mileage for cars with a driveability
problem with number of respondents and column percentages	33
Table 3-20 Cross-tabulation of owner response and car mileage for cars with no driveability
problem with number of respondents and column percentages	34
Table 3-21 Responses to "Is the light still on?"	34
Table 3-22 Responses to "Was the engine light disconnected?" and "What happened to the
light?"	35
Table 3-23 Responses to "Approximately how many miles was the car driven with the light on
before any action was taken?"	35
Table 3-24 Coded answers to "Approximately how many miles was the car driven with the
light on before any action was taken?"	36
Table 3-25 Cross-tabulation of miles driven before repair and driveability with number of
respondents and column percentages	36
Table 3-26 Cross-tabulation of miles driven before repair and recent emissions inspection with
number of respondents and column percentages	37
Table 3-27 Cross-tabulation of miles driven before repair and approximate car mileage with
number of respondents and column percentages	37
Table 3-28 Number and percentage of respondents selecting each malfunctioning component	38
Table 3-29 Number and percentage of respondents selecting each combination of
malfunctioning components	39
Table 3-30 Was the repair successful?	40
Table 3-31 Cross-tabulation of repair success and driveability with number of respondents and
column percentages	40
v

-------
LIST OF TABLES (Continued)
Table	EagS
Table 3-32 Cross-tabulation of repair success and ignition system repair with number of
respondents and column percentages	41
Table 3-33 Responses to "Was the repair or service covered under the manufacturer's
warranty, therefore performed at no cost to you?"	41
Table 3-34 Responses to "How much did the repair or service cost you?"	42
Table 3-35 Coded responses to "How much did the repair or service cost you?"	42
Table 3-36 Cross-tabulation of repair cost and driveability with number of respondents and
column percentages.	43
Table 3-37 Cross-tabulation of repair cost and recent emissions inspection with number of
respondents and column percentages	43
Table 3-38 Cross-tabulation of repair cost and owner response with number of respondents
and column percentages	44
Table 3-39 Significance values for terms in the model of Log(repair cost)		45
Table 3-40 If the respondent reported the car was not ruiming well, response to "After the car
was repaired, did it still have the same problem?"	46
vi

-------
LIST OF FIGURES
Figure	Page
Figure 3-1 Interview and data analysis flow chart with number of respondents	20
Figure 3-2 Percentage of cars with malfunction indicator light illumination by total mileage	47
Figure 3-3 Percentage of cars for which the malfunction indicator light stayed on at the
indicated car mileage	47
Figure 3-4 Owner response versus approximate car mileage for cars which were running
satisfactorily	49
Figure 3-5 Owner response versus approximate car mileage for cars which were not running
satisfactorily	49
vii

-------
ACKNOWLEDGEMENTS
Several members of the EPA Motor Vehicle Emissions Laboratory contributed to this
study. We especially appreciate the support of Christine Mikolajczyk, the EPA project leader, who was
diligent in providing technical direction throughout Our thanks go to John German who supported the
research and Dan Harrison who assisted in defining the database deliverables. We also wish to thank Mel
Kollander of the EPA Statistical Policy Branch for serving as a statistical consultant to the study.
Westat conducted the survey and analyzed the results. The authors of this report are
Danuta B as sett and John Rogers. David Maklan was the Corporate Officer for the study. Danuta Bassett
was the Project Director, John Rogers was the Statistician, David Hill was the Systems Analyst, and Carol
Rader was the Telephone Research Center Project Manager.
Several other individuals at Westat contributed to the study. They are Patricia Barbosa,
Sharon Beausejour, Sandra Gallagher, Linda Gowen, Helen Jewels, Nita Lemanski, Ann Masse, John
Michael, Thomas Milke and Joyce Powell.
viii

-------
EXECUTIVE SUMMARY
The Clean Air Act requires the Environmental Protection Agency (EPA) to promulgate
regulations whereby manufacturers of motor vehicles must certify that their vehicles will comply with
emissions standards throughout their useful lives. To comply with these standards, some manufacturers
have installed dashboard malfunction indicator lights (MIL) that illuminate when emissions components fail.
Currently, these lights are not installed on all vehicles nor are they required or controlled by EPA. Since the
MIL could potentially improve the effectiveness of EPA's certification program as well as overall air
quality, EPA conducted a national survey of motorists to leam how they responded to MIL illumination and
to answer four specific questions:
What is the incidence of MIL as a function of 1) vehicle mileage, and 2)
malfunctioning component?
What is the response to MIL as a function of 1) vehicle mileage, 2) state
inspection, and 3) driveability?
• What is the relationship between distance driven before seeking repair and 1)
vehicle mileage, 2) state inspection, and 3) driveability?
What is the relationship beween average cost and effectiveness (or success) of the
repair and 1) maLfuncticsmng component, and 2) driveability?
The survey was administered by Westat, Inc. between January 11, and March 14, 1990 in
its Gaithersburg, Maryland Telephone Research Center using computer assisted telephone interviews. The
sample for the survey was a self-weighted sample of selected non-commercial vehicles having MIL. At the
time of the survey these were the only vehicles so equiped. The respondent universe consisted of the
owners and principal drivers of these vehicles.
The response rate achieved for the survey was very good reaching 87 percent Interviews
were completed with 2621 owners of General Motors, Chrysler, and Toyota vehicles equiped with MIL.
There were no problems with the questionnaire either with respondent comprehension or cooperation with
refusals accounting for only 5 percent of all eligible cases (142 out of 3017).
The central findings of the survey consistently supported the usefulness of MIL to
motorists. The survey revealed that 30 percent of the vehicles had an illumination and that the incidence of
illumination increased with total mileage on the vehicle. Thirty percent of the respondents experiencing MIL
ix

-------
illumination repotted a simultaneous driveability problem at the time the light appeared. This finding
suggests that in the absence of MIL, the motorists would have no reason to believe that anything was wrong
with the vehicle and might have continued to drive with malfunctioning emissions equipment Only 3
percent of respondents experiencing MIL illumination said that their light was disconnected after the
illumination occurred. This finding suggests that that most motorists considered the signal useful.
In response to the MIL illumination, most motorists (68 percent) sought repair. However,
whether they sought repair depended significantly on both the car mileage and the driveability of the car. If
the car was not running satisfactorily at the time of the illumination, 87 percent of owners sought repair,
independent of the mileage. If the car was running satisfactorily, 60 percent sought repair. This finding
corroborates the usefulness of the MIL to motorists.
Most motorists drove only a few miles after the MIL illumination: the median miles driven
before repair was 10 miles, with 75 percent of vehicles driven less than 50 miles before repair. The number
of miles driven before repair was not significantly related to other variables collected in the survey. Again,
the finding suggests that most motorists paid attention to the MIL. Additional survey results relating to
EPA's specific research questions follow.
What is the incidence of MIL as a function of 1) vehicle mileage, and 2)
malfunctioning component?
The incidence of malfunction indicator light illumination increased as total mileage
increased. The percentages were 18 percent at 0 to 30,000 miles, 30 percent at 30,001 to 50,000 miles, 37
percent at 50,001 to 70,000 miles and 37 percent at greater than 70,000 miles.
The malfunctioning component, if known, most often included the computer control
module and the fuel system. Five percent of respondents reported a problem with the warning light only.
• What is the response to MIL as a function of 1) vehicle mileage, 2) state
inspection, and 3) driveability?
Response to the MIL illumination depended significantly on both mileage and driveability,
but it was not significantly related to recent state emissions inspection. Seventy-seven percent of
respondents with vehicles in the 0 to 30,000 miles category sought repair, 70 percent in the 30,001 to
50.000	miles category, 56 percent in the 50,001 to 70,000 miles category and 63 percent in the greater than
70.001	miles category.
x

-------
Repair response had a significant relationship with driveability, which was defined for the
survey as running satisfactorily at the time of the illumination. If the car was not running satisfactorily, 87
percent of owners sought repair, independent of the mileage. However, even if the car was running
satisfactorily, 60 percent sought repair.
What is the relationship between distance driven before seeking repair and 1)
vehicle mileage, 2) state inspection, and 3) driveability?
Three factors, vehicle mileage, state inspection and driveability, were considered as
possible explanations of the number of miles driven before repair. None of the statistical relationships
between miles driven and any of the three factors were significant
What is the relationship between average cost and success of the repair and 1)
malfunctioning component, and 2) driveability?
Two primary factors were considered as possible explanations of cost of the repair. These
were the malfunctioning component and driveability. The relationship between repair cost and driveability
was not statistically significant.
The cost of repair as a function of malfunctioning component is not conclusive from the
survey. The cost of repair might be expected to depend on the items repaired. Repair of several items
showed a significant relationship with repair cost, however the interpretation of these relationships is
difficult without considering all of the combinations of items which were repaired. Although it can be
concluded that the components repaired affect the repair cost, the survey results are not adequate to
determine the cost as a function of the component repaired.
Two factors were considered as possible explanations of the success of the repair. These
were driveability and the malfunctioning component. The relationship between the success of the repair and
driveability was not statistically significant
The analysis looked at the relationship between repair success and repair of all six
components in the survey. For only one component, the ignition system, was the relationship with repair
success significant When the repair work involved the ignition system 83 percent thought the repair was
xi

-------
successful but when the repair work did not involve the ignition system, 92 percent thought the repair was
successful.
In summary, the survey findings regarding the primary research objective of the study
consistently suggest that motorists are attentive to the illumination of the MIL. Nevertheless, from a
statistical perspective, the results of the survey will apply to future motorists only to the extent that the
factors which affected motorist response are germane in the future.
xii

-------
1. INTRODUCTION
1.1	Survey Background
Section 207(a) of the Clean Air Act, as amended in 1977, requires the Environmental
Protection Agency (EPA) to promulgate regulations establishing a program whereby a manufacturer of
motor vehicles must certify that its vehicles will comply with emissions standards throughout their useful
lives.1 EPA issues certificates of conformity to the manufacturers of vehicles which comply with emissions
standards and satisfy the other regulatory requirements.
In the certification process, manufacturers are required to demonstrate that their vehicles
will comply with emissions standards over their useful life, e.g., 50,000 miles for light-duty vehicles.
However, despite manufacturers' compliance with certification requirements, in-use emissions testing
programs using consumer-owned passenger cars have shown that emissions of vehicles in use can often
exceed the standards. Such exceedances are often due to an emissions control component malfunction, a
need for maintenance, or improper diagnosis or repair of a malfunction. As a result, the public is not
realizing the full benefit of the emissions standards (i.e., cleaner air) due to the excess emissions produced
by malfunctioning vehicles.
To meet existing certification emissions standards, most manufacturers equip their new
vehicles with on-board computers to regulate the fuel metering, ignition, and emissions control systems.
Back-up or default systems in the on-board computers may maintain driveability performance even when
key components affecting emissions control malfunction or fail. Malfunctions or failures of the emissions-
related components on these vehicles often lead to significant increases in emissions without affecting
driveability or fuel economy. Since there is no indication of a problem, the owner is not aware of the
malfunction and is unlikely to seek repair. To remedy this, some manufacturers have provided a
malfunction indicator light on the instrument panel which is triggered by a diagnostic system incorporated
within the vehicle's on-board computer. The malfunction indicator light (MIL) illuminates when an
emissions-related component malfunctions. Most systems also store data about the malfunction in the on-
board computer to assist mechanics in diagnosing and repairing the vehicle. The immediate benefit of an
Much of the teat thai describes the background of the survey is from the Supporting Statement for Surveying and Requcttina
Information from Vehicle Owners in the On-Board Diagnostics Program, prepared by the U.S. Environmental Protection Agency
Motor Vehicle Emissions Laboratory, Ann Arbor, ML, 19S9.
1

-------
on-board diagnostic (OBD) system to the consumer is that it assists the consumer in determining when
service or maintenance is necessary and assists the mechanic in determining what components are in need of
repair.
Currently, OBD systems are not required or controlled by EPA. However, OBD systems
have the potential to improve the effectiveness of the certification program with a net improvement in air
quality. Other potential benefits from OBD include improved repair success and the possibility that OBD
system checks may be able to supplement or replace the current tailpipe test in Inspection and Maintenance
programs.
To determine the usefulness of OBD systems to the consumer and, hence, the
environmental benefit, EPA conducted a survey of vehicle owners in early 1990. The survey asked
questions on vehicle owner response to the illumination of the malfunction indicator light, the distance
owners drive before seeking repair, and the effect of current voluntary OBD systems on repair success.
1.2	Survey Report
This report describes the methodology and results of EPA's Survey of Vehicle
Owners in the On-Board Diagnostics Program. Section 2 covers the survey methodology in
detail. It describes the questionnaire design, the computer assisted telephone interview (CATI) system used
to administer the questionnaire, development of the sample frame and drawing the survey sample, the
survey field period, and steps taken to insure the quality of the data.
Section 3 presents the survey results. The discussion of results is preceded by a report on
survey response, the limitations of the survey data, statistical techniques used in the analysis and the
statistical tables that support the conclusions of the study.
2

-------
2. SURVEY METHODOLOGY
This section describes the design of the Survey of Vehicle Owners in the On-Board
Diagnostics Program and the methods used to collect, edit, and report the data. It covers the
questionnaire, the CATI system used to administer the questionnaire, the sample and quality control
procedures.
2.1	The Questionnaire
EPA and Westat had several discussions to plan the content of the questionnaire.2 The
major content considerations were to limit the number of questions to about 20; to interview the owner of
the vehicle about ownership issues, such as whether the car was new or used, when was acquired; and to
interview the principal driver of the vehicle for all issues related to driveability. and the illumination of the
malfunction light.
Since die primary research objective was la determine response to the iUurainaMn of the
malfunction light, many of the questions were asked only of those motorists who experienced an
illumination of the warning light. Accordingly, the questionnaire divided into two parts. The first part was
fairly short identifying the current registered owner of die sampled vehicle and asking a few ownership
questions leading up to the question on illumination of the light. The second pan of the questionnaire was
administered to only those respondents who had experienced an illumination. It contained questions on
events that took place subsequent to the light's illumination.
The questionnaire was ideal for a CATI application. A hard copy mail instrument would
have required the respondent to follow complicated skip patterns and, in cases where the owner and
principal driver were different individuals, would have required the owner to forward the instrument to
another individual to complete.
Below we discuss the major questions in the in the survey instrument and their respective
limitations. Questions omitted from the discussion can be seen in the full CATI text of the instrument in
Appendix A.
Appendix A J lfc« sure* ^ue: dona tire pruoHK. in. CATI lor.ii
3

-------
Part 1 of the Questionnaire
The interview began with a brief introduction that quickly established credentials (United States
Environmental Protection Agency and Westat, Inc.), and the topic of the survey (dashboard
warning lights). Next, the interviewer verified the telephone number that appeared in the sample.
When the number was not confirmed, the original number was redialed.
Q2 The interviewer asked to speak to the person whose name appeared on the vehicle
registration. If that individual was not at the telephone number dialed, the interviewer
asked for a household telephone number where the owner could be reached and attempted
to contact the owner.
Q5 This question verified that the respondent was the current car owner {defined for the survey
as the person whose name was listed on the vehicle registration). When the respondent
whose name was on the sample list was not the current car owner, the interviewer asked if
the respondent had the name of the current owner. Again, when the current owner's name
was obtained, Westat called the new owner. Westat did not pursue vehicles sold to
commercial establishments such as automobile dealers since the survey was interested in
consumers' responses to the illumination of malfunction lights (Q7).
The CATI system enabled the questionnaire to be tailored to each respondent. For
example, the interviewer's screen displayed not only the name of the registered owner but
also the model and model year of the vehicle sampled (1986 Cavalier Wagon). This CATI
feature eliminated any respondent confusion about which vehicle we were discussing.
Thereafter, every question in the interview that mentioned the vehicle, mentioned it by year
and model.
Q11 The main portion of the interview began with an introduction that established the scope of
the series of questions to follow (/ am going to ask a few questions about your 1986
Cavalier Wagon and its dashboard warning lights).
To put the respondent at ease, the interviewer provided additional information {Before we
begin, you should know that we have no reason to believe that anything is wrong with
your car, we are just gathering information). Then, the interviewer informed the
respondent that participation was voluntary and information provided would be kept
confidential.
The interviewer immediately asked the first question (When you acquired the 86 Cavalier
Wagon was it new, that is never previously titled, or was it a used car?). How the vehicle
was acquired, i.e., purchase, inheritance or gift, was not of interest to EPA. The phrase
"never previously titled" eliminated any confusion such as "It was new to me."
Q12 Question 12 asked if the State tested the respondent's car for its emissions since 1988. The
survey was conducted between mid-January and mid-March 1990. EPA was interested in
recent emissions testing and the scope of the question was designed to include all of 1989
and eliminate all but State testing.
Q13 This question established whether the owner was the principal driver of the car. For the
survey, EPA defined principal driver as the individual who drove the car most of the time.
In cases where the owner was not the principal driver, interviewers were instructed to
conduct the remainder of the interview with the principal driver since this would be the
person most likely to observe the illumination of the warning light
4

-------
Q19 CATI displayed this question only when the owner was not the principal driver. It
introduced, the principal driver to the survey, established, credentials and verified that tie or
she drove the car SO percent of the time. The remainder of the interview was conducted
with, the principal driver.
Q20 This question asked for the approximate mileage on the car. Interviewers were instructed
to record the exact number of miles reported by the respondent These mileage estimates
were later combined into categories established by EPA for analysts.
Q21 Question 21 was the most important question in the survey. H began with a short
introduction (The 1986 Cavalier Wagon has an engine warning light located on the
dashboard which reads....). Once again, the CATI system facilitated the interview. CATI
was programmed to display to the interviewer the appropriate warning message the
manufacturer had installed in the vehicle (check engine, service engine soon, power toss
etc.). This feature reduced potential respondent confusion about which light was of
interest Interviewers were trained to use dashboard diagrams that illustrated the lights in
question to help them answer respondents' questions.
The key question (Has the warning light ever come on and stayed on after starting the car?)
was asked after the interviewer and the respondent understood which light was of interest.
Interviewers were trained to distinguish between the engine warning light and other
dashboard lights such as oil or gasoline signals. Interviewers were also trained to
distinguish between signals that illuminate briefly every time the driver starts the car and
engine warning lights that come on and stay on while the car is being driven. Interviews
with respondents who did not experience an episode of engine wanting light illumination or
who did not know if such an illumination had occurred were discontinued at this point.
Part 2 of the Questionnaire
Q23 This question introduced the second part of the interview which collected information on
what the respondent did in response to the Illumination of the warning light. It asked for
the year in which the light came on most recently. The survey collected information only
on the most recent episode, not on all episodes that may have occurred.
Q24A This question established whether there were any driveability problems at the time of the
episode. Interviewers were trained to distinguish between driveability problems such as
stalling and other problems with the car such as a broken rear view mirror.
Q24 Question 24 asked "Was any action was taken in response to the engine warning light?"
Note that the question did not ask "Did jtou take any action...?" Rather it intentionally
included action taken by anyone not just the driver, Preliminary research, conducted in
preparation for the survey, revealed that illumination of the engine warning light was a
relatively rare event and. we did not want to lose any episodes because someone other than
the principal driver took action in response to the illumination. "Any action" was defined
as any step taken toward finding out why the light came on or any action taken toward
repairing the car because the light came on.
Question 24 was designed to separate respondents who did take action from those who did
noL Respondents who did not take action were asked only if the light was still on and the
interview was terminated. Respondents who took action continued the interview.
5

-------
Q26 This question asked the respondent approximately how many miles the car was driven
with the light on before any action was taken. EPA was interested in consumers' reactions
to the illumination of the light and number of miles driven was thought to be one way to
measure how quickly respondents acted. Interviewers recorded the exact number of miles
reported by the respondents.
Q28 This question asked what the first action was after the illumination of the warning light.
Responses were of two kinds: seeking more information about the light, and, taking the
car for repair. Respondents who sought more information were asked what action they
took after obtaining the information.
Those who immediately sought repair were asked if the repair was made by themselves, by
a friend or relative or by a professional. Those who used a professional were asked to
describe the repair shop as a dealership or and independent shop (Q30).
Q31 This question asked the respondent which of a list of items were repaired when the car was
first serviced. The list, which was developed by EPA for the survey, grouped specific
items. For example, the fuel system included the pump, lines, filter, injector carburetor or
idle speed equipment. These items were read to respondents who were unsure of what was
repaired. In addition, interviewers encouraged respondents to review their repair receipts
before answering.
Q32 This question asked if the repair was successful. Successful was defined as satisfactory to
the client. Respondents who said that the repair was not successful were asked additional
questions: "Since the first repair attempt, was the car ever successfully repaired?" (Q33).
Those that responded in the affirmative were also asked what was repaired (Q34) using the
same list as in Q31.
Q35 This question asked if the repair was covered under warranty and therefore at no cost. If
the respondent paid any part of the cost, interviewers recorded a "No" response.
Q36 This question asked for the cost of the repair. Interviewers recorded the exact amount
provided by the respondents. These amounts were later combined into categories
established by EPA for analysis.
Q37 This question, which asked if the warning light was disconnected, was included as an
indication of the extent to which the light was considered useful or a nuisance.
Q38A This question (After the repair, did the car still have the same problem?) was asked only of
respondents who reported a driveability problem in response to Q24A.
2.2	The CATI System
Westat used computer assisted telephone interviewing (CATI) to administer the survey.
CATI provided several advantages over traditional methods of telephone data collection such as:
6

-------
•	The skip pattern logic of the questionnaire was fully automated so as to eliminate
interviewer choice in question branching which, in this survey, was considerable;
•	Question wording choices, including the responses from previous questions were
performed for the interviewer by the CATI software;
•	Validity edits on response codes were perfonned during the interview so that
invalid codes could not be entered into the data file;
•	The need for routine data retrieval was eliminated because editing was performed
during data collection; and,
•	Post-data collection machine editing was minimized.
Westat's system of CATI software, called the Cheshire System, was developed in-house
especially for use on larger government surveys that demand high standards of quality for deliverable data
sets. Several features of the Cheshire System were important in this survey. These and their importance in
the survey are described below.
The data dictionary is a file that contains in one place all the definitions and attributes of
all the permanent variables used during data entry or CATI interviewing. For each variable in this survey,
the data dictionary contained a variable name compatible with SAS and other software packages used to
process the data; a variable label or description; and allowable codes and other code descriptions.
Information that was entered into the data dictionary also defined the record format for the data to be
collected.
The screen painter is a full screen, interactive editor which allowed staff to write and
revise directly into the system "questions" with interviewer instructions. The screen painter, which has the
capability of producing highlighting, underlining, blinking and box enclosures, was used to generate hard
copy documentation of all screens exactly as they were used in the survey. These hard copy versions were
used during interviewer training where specific items were called to the attention of interviewers improving
their comprehension of the question and the screen layout.
The flowchart for this survey is presented in Appendix B. It documents all branching or skip patterns used in this short
questionnaire.
I
Hard copy screens are displayed in Appendix A of this document.
7

-------
Programs written in the Cheshire System flow language contained the statements that
"drive" the interview. These statements were collected into modules. One main module started the
interview. The main module was accompanied by other modules needed in separate sections of the
interview.
Modularization allowed the programming of the interview to be broken into units so that a
number of people could work on the interview simultaneously and revisions could be introduced without
major impact In this survey, Q38A is an example of a followup question added late in the development of
the flow language.
The Audit Trail File feature of Westat's CATI played an especially important part in this
survey. This file, maintained for each interview that was conducted, recorded every keystroke entered
through the keyboard. If an interview was prematurely terminated and later restarted, the file was used
during restart to simulate all key strokes entered during the previous interview attempt. This process
restored all permanent and temporary variables to their pre-existing values, allowing consistency checks
between the two interviewing sessions to be performed accurately. In addition, the interviewer could
backup into the data collected during the earlier session providing continuity in completing the interview. In
this survey, it was an important feature when the owner and the principal driver were not the same person
and the principal driver had to be reassured that the owner had sanctioned the interview.
Automatic Telephone Dialing was used for this study. This dialing was accomplished
by using a device connected between the CATI terminal and the interviewer's telephone jack which accessed
the source data file. This feature relieved the interviewing staff of the time-consuming burden of pressing a
long sequence of telephone number digits, eliminated re-dialing due to interviewer error and ensured that the
number dialed was the same as the number recorded in the source data file.
Finally, the Cheshire system generated frequencies and cross-tabulations directly
from the on-line database as the survey progressed. These reports enabled staff to monitor several
unknown aspects of the study such as whether the telephone numbers in the sample were "good," whether
the sample was out of date in terms of vehicle ownership, and, most important, whether the incidence of the
illumination of the malfunction light (upon which the sample was based) was as projected.
8

-------
2.3	Survey Sample
The sample design used for this survey was a self-weighted sample of selected non-
commercial vehicles having malfunction indicator lights.5 EPA specified the vehicles which were to be
included in the sample and also specified that commercial, government and institutional vehicles were to be
excluded.
The respondent universe for the survey consisted of the owners or principal drivers of
these vehicles. However, because the survey design called for data to be collected by telephone, actual
interviewing was restricted to that pan of the population residing in housing units with telephones available.
As of 1986, this constituted 94 percent of all households in the U.S. Given the requirement of vehicle
ownership and the model years covered by the study, it is reasonable to assume that more than 94 percent of
owners/drivers of eligible vehicles have a telephone available in their housing units.
Finally, the sample design also took into account the availability of current motor vehicle
registrations throughout the United States. Due to state restrictions on access to vehicle registrations, the
sample for the survey did not include the following 14 states:
Alaska	New Jersey
Arkansas	New Mexico
Connecticut	Oklahoma
Georgia	Pennsylvania
Hawaii	Virginia
Indiana	Washington
Kansas	Wyoming
The sample frame of eligible vehicles was extracted from the R.L. Polk one percent sample
of current motor vehicle registrations. The R.L. Polk one percent sample was first sorted by model year
and within model year by manufacturer and model. Next, a sampling interval that yielded a survey sample
of 4,000 vehicles was computed and systematically applied to the frame with a random start. This sample
size was estimated to be sufficient to yield the target number of 2,500 completed interviews with eligible
vehicle owners or principal drivers assuming vehicle eligibility and response rates of approximately 80
percent.
Appendix C contains the vehicle list EPA selected to be used for the survey
9

-------
The design called for completed interviews to be of two types: eligible respondents who
did not experience an episode of malfunction light illumination and eligible respondents who did experience
an episode of malfunction light illumination. The latter group, estimated to total approximately IS percent
of the sample were also to be interviewed about their actions taken in response to the illumination.
2.4	Data Collection
Westat conducted the survey in its Gaithersburg, Maryland Telephone Research Center
(TRQ. On Wednesday January 10, 1990 Westat trained the 23 interviewers and shift supervisors who
were to administer the questionnaire. The training lasted from 6:00 P.M. until 9:00 P.M. On January 11,
interviewing commenced and the survey was completed on March 14,1990. There were no problems with
the questionnaire, either with respondent comprehension or cooperation. In fact, there were relatively few
refusals (142 cases out of 3,017 cases, or 5%) in the survey.
2.5	Quality Control
Several measures were used to ensure the quality of the data. A TRC project manager was
assigned to direct the day-to-day data collection activities. Assisting the project manager was a staff of shift
supervisors and clerical staff.
All interviewers received intensive training which included Westat's four hour General
Interviewing Training Techniques, four hour Teletrain (CAT!) training and three additional hours of project
specific training. The agenda for the project specific training included one hour devoted to an overview of
the survey, question-by-question specifications and survey procedures. The remaining two hours were
used to for interactive lectures and dyad role play.
For the interactive lecture, Westat used a large screen data-graphics display unit which
worked in coordination with a computer terminal and projected the questionnaire material at the front of the
room. After an opening explanation of the concepts to be covered in the lecture, the trainees were led
through the questionnaire by calling on individuals to adopt the role of the interviewer while the trainer
played the respondent. The interactive lecture consisted of 8 different scenarios progressing from very
simple to more complicated and sometimes hostile situations.
10

-------
After the trainees completed the lecture session, they were separated into pairs. Within
each dyad, one trainee took the role of the interviewer while the other played the respondent using a
prepared script. Trainees reversed roles with each script The role-plays were conducted in the telephone
center so that each interviewer had an opportunity to log-on, enter a complete interview while on the
computer, and log-off. Trainers and supervisors monitored the role-plays using the TRC monitoring
equipment.
During the actual interviewing for the survey, Westat used standard monitoring techniques.
Using extension telephones and displays linked to interviewer CRTs, shift supervisors silently monitored
10 percent of each interviewer's completed work. Monitoring was heaviest during the first few days of the
survey when project staff joined TRC staff to oversee the data collection and resolve any unforeseen
problems.
Westat made nine attempts each to contact and interview the owner and the principal driver.
These attempts were staggered on different days of the week and at different times of the day over the data
collection period. We used a call pattern that spanned several weeks to reach as many respondents as
possible. A call pattern that spans only a few weeks will miss a significant number of persons and will
introduce a potential source of bias. Our experience shows that individual respondent schedules (jobs,
classes, vacations etc.) have a more negative impact on the level of response when call attempts are limited
to a short time spaa
All data collected for the survey were collected with an assurance that the respondents'
answers would remain confidential. This assurance was given early in the interview and was supported in a
number of different ways. All Westat personnel including interviewers, coders, and professional staff, sign
a statement stating that they will maintain the confidentiality of all survey data.''" Access to the study was
limited to Westat employees all of whom have signed such a confidentiality statement
Data security was also protected by Westat's CATI Cheshire System which has built in
procedures for data protection such as the audit trail file, discussed earlier, and by computer rooms with a
non-interruptable power supply and protection against fire and flooding. This survey was also protected by
our regular schedule of disk-to-tape backups. All files that have been created or modified during a given
Appendix O contains Westal's Confidentiality Pledge.

-------
day are copied from disk to tape at the end of the day. All backup tapes are picked up daily by a computer
storage contractor and are stored off-site in an environmentally controlled, secure storage facility.
12

-------
3. SURVEY RESULTS
This section contains the results of the survey. It provides the response rate and how it
was calculated, the statistical methods used in the analysis, limitations of the data, and the results of the
analysis.
3.1	Survey Response Rate
The data collection procedures for the survey directed considerable effort toward achieving
a high rate of response. These efforts were successful resulting in a very good response rate of 87%. The
calculation of the response rate (2,621/3,017) is described below.
Of the total sample of 4,000 vehicles, 983 cases were found to be ineligible for the
purposes of this study. There were 241 non-working numbers in the Polk sample. Commercially owned
or dealer owned vehicles accounted for loss of 226 additional cases; information on current ownership was
obtained in Questions 5,6, and 7. In 131 cases where the telephone number in the Polk sample had been
correctly dialed, the respondent never heard of the owner. In 39 instances the individuals listed as the
registered owner in the Polk sample said that they never owned the vehicle. Finally, in 346 cases, the
individual listed on the Polk sample was not the current owner of the vehicle and the current owner was not
known to the previous owner or could not be located.
There remained 3,017 eligible cases. Of this number there were 2,621 completed
interviews, 142 refusals, 70 ring-no-answer, 41 language problems, 41 cases of non-response due to death
or illness, 37 cases where the respondent was not available during the field period, 32 cases where the
respondent was not reached after the maximum number of calls (9), 16 cases where the current owner's
telephone number could not be obtained, 10 cases where the name of the current owner was not known,
and 7 cases where the principal driver could not be located.
13

-------
3.2
Statistical Methods
3.2.1	Descriptive Statistics
This report includes descriptive statistics of all variables used in the statistical analyses.
Where the variable is derived from several questions in the interview, the responses for the individual
questions are not presented, however the derived variable is described in terms of the original questions.
The variables are either categorical (with responses in categories, such as "Yes", "No", or
"0 to 100 miles") or numerical (such as number of miles). For categorical variables the report provides
summary statistics on the number of respondents answering the question, the response categories, and the
number and percentage of respondents providing answers in each category. The percentages in the tables
may not add to 100 due to rounding. For numerical variables the mean, standard deviation, and selected
percentiles of the data are presented.
For the analysis, EPA identified several variables which were of particular interest. For
variables of interest, a statistical analysis is used to determine which factors might be useful for explaining
the patterns in die data.
3.2.2	Statistical Analysis
The statistical analysis follows the following general steps:
1)	Select those factors which might reasonably provide an explanation for the pattern
of results in the variable of interest, i.e. select a "model" for the data;
2)	Fit the model using a statistical program;
3)	Use a hypothesis test to determine if the model (or factors selected) provide a good
explanation, or "fit," based on statistical criteria; and
4)	Interpret the significant relationships.
For the data from this survey, two basic models were used in the data analysis, linear
models for numerical variables and log-linear models for categorical variables. Each of the procedures is
described briefly below.
14

-------
Linear models
Linear models (including analysis of covariance, analysis of variance and linear regression)
are used to analyze continuous data such as the miles driven before repair and the cost of repair. These
models cannot handle missing or "don't know" responses in the variable of interest, i.e.independent
variable. The explanatory variables can be continuous variables such as miles driven (i.e. when using
regression), discrete variables such as answers to yes/no questions (i.e. when using analysis of variance),
or both continuous and discrete variables (i.e. when using analysis of covariance).
Selecting the model which best describes a dependent variable can be a multiple step
process involving a combination of selecting the terms in the model, fitting the model, and, based on the
residuals, selecting alternate models to fit. These methods make certain assumptions about the data which
must be reasonably satisfied, e.g., that the errors have a normal distribution and with constant variance. At
times the data must be transformed, at least in part, to obtain data which meet these assumptions. For
instance, data which must be non-negative (such as repair cost and miles driven before repair) often require
transformation. If a transformation is used in the analysis, the results of the analysis will be presented in
the original units.
When using linear models, such as analysis of covariance, the results will include a
summary of the significant terms in the model, confidence intervals where appropriate, and the conclusions
which can be made from the analysis. The linear models were fit using either the SAS (Statistical Analysis
System) procedure PROC GLM or using a program written for the Macintosh PC by SAS called JMP
which performs the same analyses. When using linear models, we report on all factors which were
significant or which are of particular interest.
Categorical Data analysis
Categorical data analysis is used for analyses in which both independent and independent
variables are categorical such as the responses "Yes" and "No." Although this analysis can be quite
complicated, it is usually accomplished using an analysis of contingency tables and a chi-square statistic.
Contingency tables are tables where each row corresponds to a different response category on one variable
(in this case the variable of interest or the dependent variable), and each column corresponds to a response
category of a second independent variable. Within the table, each cell has the number of respondents with
the combination of responses to the two questions indicated by the row and column. Contingency table
15

-------
analysis requires that all variables are coded into discrete categories. This analysis can easily handle "don't
know" responses, because "don't know" is just another category of response.
There are some assumptions behind the contingency table analysis which must be met in
order to obtain correct results. In particular, the number of cases which fall into each cell must not be too
small Often a lower limit of S cases is used; i.e., at least 5 cases must fall into each cell. If some cells do
not have enough cases, rows and columns can be combined (i.e., two or more response categories are
combined into one) until all cells have at least 5 cases. Usually categories which have similar meaning or a
similar expected distribution of cases are combined. Alternatively, categories with few cases in some cells
can be excluded from the analysis. In general, collapsing categories is preferred to excluding categories.
For the contingency table analysis in this report, the "don't know" category was excluded
from the analysis if there were less than 5 cases in any cells in the category and if there was no logical
reason to combine these cases into another category. In all other situations, categories were combined if
there were too few cases in any celL Where it was necessary to combine categories, the same categories are
combined for all analyses for consistency. With many possible analyses to consider, the optimal strategy
for combining categories is a matter of judgement Fortunately, the results of the statistical analysis are, in
general, not greatly affected by the exact choice of categories to be combined.
Numerical, or continuous variables can be analyzed using contingency table analysis by
"coding" or dividing the continuous variable into categories, e.g., miles driven before repair might be
divided into two categories of 0 to 100 miles and more than 100 miles. Contingency table analysis as
described above has only one independent variable and one dependent variable in each analysis. Each 2-
dimensional table is evaluated using a chi-square statistic. If there are multiple factors, all of which might
explain the variable of interest, either 1) multiple tables can be evaluated, each with one independent
variable, or 2) a more complicated analysis can be performed called log-linear modelling, which can be
thought of as an analysis of a 3 or higher dimensional table. In general, the multi-dimensional approach is
preferred as important interactions may be missing when using separate analyses of single predictors.
When using log-linear models, we report on all factors which were significant or which are of particular
interest
The analysis of categorical data was performed using either the SAS procedure PROC
FREQ for 2-dimensional contingency analysis or the JMP program for fitting log-linear models. Log-linear
models as fit by JMP are not available in SAS for the IBM-PC. For analyses using contingency tables, the
results include any 2-dimensional tables useful in understanding the data, the chi-square statistic, and the
conclusions which can be made from the analysis.
16

-------
Hypothesis Tests and the Model's "p value"
An hypothesis test is used to evaluate whether a selected model provides a good description
of the data, or fits the data. The "p-value" for the hypothesis test is used to evaluate the statistical
significance of the model. It is the probability that the observed relationship between the independent and
dependent variables could have occurred by chance alone when in fact, there is no relationship between the
variables. For example, if the p-value is .05 (5%) then there is a 1 chance in 20 that the relationships
described by the model might not really exist and that this relationship in the data occurred by chance. In
this case the relationships found in the data are usually said to be "statistically significant" since small p-
values imply that the relationships are unlikely to be due to chance alone.
When there are multiple factors in a model which might explain the variable of interest,
there is a p-value for each factor. The p-value for all factors in the model must be small to conclude that the
model fits the data, i.e., provides a good explanation for the patterns in the data. Following generally
accepted statistical practice, only factors with a p-value of less than 0.0S are considered to be statistically
significant2
Interpretation of the Significant Relationships
The objective of the statistical modeling is to identify a relationship between variables
which describes as much of the variation in the data as possible using only significant factors. There are
usually several models which fit the data reasonably well, in which case there are several interpretations
which might be consistent with the data. For example, because used cars typically have higher mileage than
new cars, if the incidence of malfunction light illumination increases with increasing mileage, the incidence
will also be higher for used cars than new cars. Because the new/used status and the mileage are related,
either one (or both) can be used to explain the incidence of light illumination. If a factor is statistically
significant, any other variable (whether measured by the survey or not) which is related to that factor may
also be related to the variable of interest.
Whether using contingency table analysis or linear models, the results do not imply a causal
relationship. Statistical analyses can only be used to establish a statistically significant relation between
2 Note that there are several types of p-values provided by the software. It is desirable for each factor in the model to have a
significant p-value and that the p-value for "lack of fit" to be non-significant.

-------
factors. Thus, if the data show that more owners who report recent emissions inspections get their vehicles
repaired, this does not imply that the presence of emissions inspections influences whether owners get
repairs, merely that these two items are related. It might be that other factors which cause owners to care
more about their care' performance also affect the installation of emissions inspection programs (e.g. an
environmentally conscious citizenry).
Once the significant factors are identified, their relationships with the variable of interest are
determined by looking further at the data. For example, observation of a cross-tabulation of owner
response versus recent emissions inspection will indicate if the presence of emissions inspection programs
is associated with increased or decreased repair frequency of repair. When reporting the results of the
statistical analysis, the statistically significant factors will be presented along with their respective p-values,
and a brief description of the relationships present in the data.
When summarizing numerical data and percentages, confidence intervals are often used. A
confidence interval is a range of values within which the parameter being estimated (e.g. average repair
costs) is likely to lie. Following standard practice, 95 percent confidence intervals were used in the
analysis. This means that confidence intervals will, on the average, include the true value 95 percent of the
time. Confidence intervals have been presented for the important results. For cases where the confidence
interval on a percentage has not been presented in the report, an approximate range for the 95 percent
confidence interval can be determined from Table 3-1. For example, if 12.5 percent of the 300 respondents
answered "Yes", then the approximate 95 percent confidence interval would be 12.5% ±3.6% percent or
8.9 percent to 16.1 percent.
Table 3-1 95 percent Confidence interval for percentages by number of respondents and estimated
percentage
Number of
respondents

Estimated percentage


10% or
90%
25% or
75%
50%
2621
±1.1%
±1.7%
±1.9%
792
±2.1%
±3.0%
±3.5%
535
±2.5%
±3.7%
±4.2%
270
±3.6%
±5.2%
±6.0%
100
±5.9%
±8.5%
±9.8%
18

-------
3.2.3
Sampling Versus Non-sampling Error
The error, or variation in the data can be attributed to a combination of sampling and non-
sampling error. Sampling error refers to the variation in survey results which are due to the random sample
selection process. The statistical tests used to identify appropriate models for the data only evaluate
sampling error.
Non-sampling error refers to all other sources of error or variation in the data. Some
sources of error could result in bias, i.e. results which are consistently above or below the value which the
researcher is tying to estimate from the data. For example, questions which are poorly worded may not be
understood, resulting in incorrect answers according to the intention of the questioa Respondent error can
also contribute to error, for example the owner may not like to admit that he spent a lot to get his car fixed
and "remembers" a lower cost than was actually incurred.
Non-sampling error can also be introduced if the non-respondents, those who refuse or are
unable to answer the interview questions, would provide different answers than the respondents. In this
case the results would correctly describe only cars represented by the respondents and not all cars in the
survey. Because the response rate for this survey is high (87%), any bias due to non-response is likely to
be small.
3.3	Survey Results
The statistical results are presented below in the order in which they were collected during
the telephone interview. Figure 3-1 shows the information obtained at different points in the interview and
the number of respondents answering selected questions.
Of the 4000 names on the sample tape, 2621 car owners and principle drivers were located
and completed the interview. The following section summarizes the data obtained from these 2621
respondents.
19

-------
Figure 3-1 Interview and data analysis flow chart with number of respondents
20

-------
3.3.1
Data From All Respondents
The sample tape provided information on the make, model year, and state in which the car
was registered at the time the frame was constructed. Table 3-2 and 3-3 respectively summarize the make
and model year for the cars owned by the survey respondents. Note that by design, the sample only
included Chrysler, General Motors, and Toyotas.
Car Make
Table 3-2 Car make from the sample tape
Car Make
Number of
Respondents
Percentage of
Respondents
Chrysler
271
10%
General Motors
2155
82%
Toyota
195
7%
Total
2621
100%
As Table 3-2 shows, the vast majority of respondents drove General Motors cars,
therefore, For subsequent statistical analyses, the two smaller categories, Chrysler and Toyota, were
combined into an "Other" category.
Model Year
Table 3-3 Car's model year from the sample tape
Model Year
Number of
Respondents
Percentage of
Respondents
1981
131
5%
1982
178
7%
1983
283
11%
1984
439
17%
1985
467
18%
1986
441
17%
1987
332
13%
1988
348
13%
1989
2
<1%
Total
2621
100%
21

-------
As can be seem from Table 3-3, the median model year is 1985. While this variable was
not used directly in the subsequent analyses. It was used to calculate the approximate mileage at the time the
light came on in those cases where the malfunction indicator light stayed on.
State of Registration
The sample included cars from 34 states, with the number of cars per state varying from 1
(D.C.) to 256 (Ohio). States with more than 100 cars in the sample were (in order from largest to smallest)
Ohio, New York, California, Michigan, Florida, Illinois, Texas, Massachusetts, Wisconsin, and
Minnesota.
Regulations for the 1988 and 1989 model year for cars sold in California were different
than for the rest of the country. Only 28 (1%) of the respondents owned 1988 or 1989 model cars
registered in California. Of these, the malfunction indicator light came on in only one car. Because of the
small number of vehicles involved, differences associated with the different California regulations were not
considered in the statistical analyses.
In addition to the information on the sample tape, the following information about each car
was obtained from the respondent: whether the car was new or used when acquired, whether the car had
recently been subject to state emissions inspection, the total miles on the car at the time of the interview, and
whether the malfunction indicator light came on and stayed on. These data are summarized in tables 3-4
through 3-8.
Acquired New or Used
As can be seen from Table 3-4, 62 percent of vehicles were new when acquired. For the
statistical analysis, the "Don't know" category and the "Used" category were combined.
Table 3-4 Responses to "When you acquired the car, was it new, that is never previously titled, or was
it a used cat?"
Acquired New or Used
Number of
Respondents
Percentage of
Respondents
New (never previously titled)
1615
62%
Used
1002
38%
Don't know
4
<1%
Total
2621
100%
22

-------
Recent State Emission Inspection
Table 3-5 summarizes the responses to the question about recent state emissions inspection.
The objective of the question about recent emission inspection was to determine if the presence of a state
emissions inspection program might have affected the owner response to malfunction indicator light
illumination. The answer to the question above was a measure of that expectation. If the respondent's
circumstances (location, expectation of inspection) when the light last came on were different than those at
the time of the interview, the answer to the question above may not reflect the circumstances when the light
came on. Other measures, such as zip code at the time of registration or telephone exchange at the time of
the interview have similar measurement problems.
Table 3-5 Responses to "Since 1988 , has your car been inspected or tested by the State for its
emissions?"
Recent Emissions Inspection
Number of
Respondents
Percentage of
Respondents
Yes
1062
41%
No
1460
56%
Don't know
99
4%
Total
2621
100%
For the statistical analysis, the "Don't know" category and the "No" category were
combined, leaving two categories.
Total Mileage
Table 3-6 summarizes the survey results pertaining to the number of miles on the car at the
time of the interview. For subsequent statistical analysis, the total mileage data was coded into the 4
categories shown in Table 3-7.
23

-------
Table 3-6 Responses to "Approximately how many miles do you have on the car?"
Number of "Don't Know's"
111
Number of Data Values
2510

Miles Driven
Mean
52,687
Standard Deviation
32,729
Percentiles

Minimum
0
5%
12,000
25%
30,000
50% (Median)
50,000
75%
70,000
95%
105,000
Maximum
555,000
Table 3-7 Coded answers to "Approximately how many miles do you have on the car?" for statistical
analysis
Total Mileage
Number of
Respondents
Percentage of
Respondents
0 to 30,000 miles
655
25%
30,001 to 50,000 miles
699
27%
50,001 to 70,000 miles
586
22%
Greater than 70,001 miles and
"Don't know"
681
26%
Total
2621
100%
Malfunction Indicator Light Illumination
The survey was primarily interested in respondents who knew that the malfunction
indicator light came on and could therefore answer questions about the event. Consequently the No and
Don't Know responses were treated the same by CATI and are grouped together for analysis purposes. Six
respondents did not know if the light had come on or stayed on. These results are summarized in Table 3-8
#

-------
Table 3-8 "Has the engine warning light [malfunction indicator light] ever come on and stayed on after
starting the cai?n

Number of
Percentage of
MIL Illumination
Respondents
Respondents
Yes
792
30%
No or Don't know
1829
70%
Total
2621
100%
The malfunction indicator light came on and stayed on in 792 cars, or 30.2 percent of the
2621 cars covered by the interviews. Thus we can state with 95 percent confidence that the true proportion
of cats with malfunction light illumination is between 28.S percent to 32.0 percent.
Respondents for whom the light never came on or who did not know were thanked for
their participation in the survey. Respondents for whom the light came on and stayed on were asked
questions about their response to this light illumination. These responses are summarized in section 3.3.2.
Analysis of Incidence of Light Illumination Versus Mileage
An analysis of malfunction indicator light illumination versus total mileage was performed
using contingency table analysis. Table 3-9 cross-tabulates whether the malfunction indicator light came on
versus the total mileage on the car.
Table 3-9 Cross-tabulation of MIL illumination and total mileage, with number of respondents and
column percentages.
"Has the engine
warning light ever come
on and stayed on after
starting the car?"
"Approximately how many miles do you have on the
car?"
Greater than
70,001 or
30,001 to 50,001 to Don't know
0 to 30,000 50,000 70,000
miles miles miles
Total
Yes
No or Don't know
117(18%) 209(30%) 217(37%) 249(37%)
538(82%) 490(70%) 369(63%) 432(63%)
792
1829
Total
655(100%) 699(100%) 586(100%) 681(100%)
2621
Statistical Analysis
Chi-square = 73.4 Df = 3 p < .0001
As can be seen in Table 3-9, the percentage of cars in which the malfunction indicator light
came on increased with increasing total miles on the car. The differences in the incidence of malfunction
25

-------
indicator light illumination among the mileage categories are statistically significant at the .OS level since the
p-value is less than 0.05 and thus these differences are too great to be due to chance (sampling variation)
alone. While the table is not shown, the relationship between light illumination and whether the car is used
or new is also statistically significant (p = 0.0051). When both factors are included in a log-linear model,
only the mileage is significant Note that for used cars, the current owner may not have known if the light
had come on and been repaired by the previous owner.
An analysis of malfunction indicator light illumination versus car make was performed
using contingency table analysis. Table 3-10 cross-tabulates these results.
Table 3-10 Cross-tabulation of MIL illumination and car make, with number of respondents and column
percentages.
"Has the engine warning light
ever come on and stayed on
after starting the car?"
Make of car
Total

General Motois
Chrysler or Toyota

Yes
717 (33%)
75 (16%)
792
No or Don't know
1438 (67%)
391 (84%)
1829
Total
2155 (100%)
466 (100%)
2621
Statistical Analysis
Chi-square = 53.6
II
<4-1
Q
p<.0001
As can be seen in Table 3-10, the percentage of cars in which the malfunction indicator
light came on is higher for General Motors cars than for Chrysler and Toyota cars combined. The
difference in the incidence of malfunction indicator light illumination between car makes is statistically
significant at the .5 percent level. Because the total miles driven on the General Motors cars may be greater
than for the Chrysler and Toyotas in the sample, a further analysis is required to determine how the
incidence of malfunction indicator light illumination depends on make for cars driven the same distance.
Table 3-11 confirms that General motors cars have had a higher reported incidence of malfunction indicator
light illumination for a given total mileage than the cars made by Chrysler and Toyota. This result was also
confirmed using log-linear models.
:: 7

-------
Table 3-11 Percentage of respondents reporting MIL illumination by car make and total mileage.
"Approximately how many miles do
you have on the car?"
Maker of the car
General Motors Chrysler or Toyota
0 to 30,000 miles
30,001 to 50,000 miles
50,001 to 70,000 miles
Greater than 70,001 miles and
"Don't know"
21% 9%
32% 18%
39% 26%
39% 19%
Statistical Analysis (JMP)
Car make (Df = 1) Chi-Sq. = 37.7 p < .0001
Total Mileage (Df = 3) Chi-Sq. = 56.6 p < .0001
Although differences between makes of car may not be relevant for the purposes of
regulation, it is important to remember that the survey conclusions were based on the existing mix of car
makes and mileage and the incidence of illumination for those cars. Future response to malfunction
indicator light illumination may differ, in part, because the mix of car makes will be different
3.3.2	Data from Respondents for Whom the Malfunction Indicator Light Stayed
On
Questions about the last time the malfunction indicator light stayed lit and the response to
the last malfunction indicator light illumination were asked of those 792 respondents for whom the light
came on. Responses to these questions are summarized in this section.
Approximate Car Mileage at Malfunction Indicator Light Illumination
In the interview, the following question was asked: "Thinking back to the most recent time,
in what year did the engine warning light come on?" The responses to this question were not used directly
in the statistical analysis. On the assumption that the mileage to consider in the analysis is the mileage at the
time the light comes on rather than the mileage at the time of the interview, the year in which the light came
on, the total miles on the car, and the model year of the car were used to estimate an approximate mileage
for the last malfunction indicator light illumination event The following formula, based on the assumption
that the miles driven per year is the same each year, was used.
/Year of MIL illumination - Model year\ ^ , ,
Approximate car mileage = ( Year of Interview - Model year ) Total mileage
27

-------
In 20 cases where the year in which the light came on was unknown and in cases where the the total miles is
missing, the approximate car miles will be unknown. The approximate car mileage was coded to a
categorical variable for the analysis because it is based on categorical variables. This approximate car
mileage is used in several analyses because it provides a better statistical fit than using the coded total
mileage. Table 3-12 summarizes the coded values for the approximate car mileage used in the statistical
analyses. Note that the term "approximate car mileage" is used to describe the approximate mileage when
the light came on and the teim "total mileage" is used to describe the repotted mileage at the time of the
interview.
Table 3-12 Approximate car mileage at the last malfunction indicator light illumination, derived from
several questions
Approximate car mileage
Number of
Respondents
Percentage of
Respondents
0 to 30,000 miles
253
33%
30,001 to 50,000 miles
205
27%
50,001 to 70,000 miles
154
20%
Greater than 70,001 miles
156
20%
Total
768
100%
Driveability
Table 3-13 summarizes the responses about whether the car was running satisfactorily at
the time of the malfunction indicator light illumination. In later discussion, this is called "driveability"
Table 3-13 Responses to "When the light came on, was the car running satisfactorily or was there a
problem?" (Driveability)

Number of
Percentage of
Driveability
Respondents
Respondents
Satisfactorily
552
70%
Problem
234
30%
Don't know
6
1%
Total
792
100%
At the time the light came on, 29.5 percent of the respondents reported that the car was not
running satisfactorily (i.e. had a driveability problem). The 95 percent confidence interval for the
percentage of cars not running satisfactorily is 26.4 percent to 32.7 percent. For the statistical analysis the
"Don't know" and "Satisfactorily" categories were combined.
28

-------
Owner Response
Several interview questions were used to determine the owners response to the malfunction
indicator ligfit illumination. The definition used in this analysis assumed ibax getting repair after getting
information or after using a dashboard diagnostic computer was the same owner response as getting repair
without first getting other informalioa This definition of owner repair combines the responses from several
questions. The owner response categories aie:
Dealer repair.
Independent repair
Self ir
Info/diagnostics only:
No Action:
Disconnected:
Unknown:
Professional repair by a dealer as the first action or the action after getting
information or using a dashboard diagnostic computer.
Professional repair by other than a dealer as the first action or the action after
getting information or using a dashboard diagnostic computer. Note that this
definition combines those cases where the professional repair was from an
independent shop (165 cases) or OTHER (2 cases, one specified as vocational
school, one unspecified).
Respondent Cried to repair or service the car as die first action or the action after
getting information or using a dashboard diagnostic computer.
No action was taken after getting more information or after using the dashboard
diagnostic computer
No action was reported in response to the engine light warning, even to explain the
light going off (if it did). If the respondent took no action and does not know if -Jhe
light is on, the light is assumed to be off, using the same logic as used in the data
collection.
No action was reported other then disconnecting the light to turn it off. (Those
whose lights are disconnected after getting repairs are not included here).
One case.
A breakdown of these responses is shown in Table 3-14. For the statistical analysis, the
eight categories above were collapsed into three: "Dealer Repair", "Non-Dealer Repair" (Independent repair,
Se2/repaf'r, and Fneitd's repair) and "No repair" (TjiformaDarVDiagnostics onlv, No Action, Disconnecting
the MIL arid Don't Know). Although the several responses which are combined into "No Repair may be
of interest, Ltiere are not enough responses in each category to perform separate analyses. These collapsed,
categories are summarized in Table 3-15.
29

-------
Table 3-14 Owner response to malfunction indicator light
Owner Response
Number of
Respondents
Percentage of
Respondents
Dealer Repair
298
38%
Independent Repair
167
21%
Self Repair
41
5%
Friend's Repair
29
4%
Information/diagnostics only
19
2%
No Action
236
30%
Disconnecting the MIL
1
<1%
Don't Know
1
<1%
Total
792
100%
Table 3-15 Owner response to malfunction indicator light illumination used for statistical analysis
Owner Response
Number of
Respondents
Percentage of
Respondents
Dealer Repair
298
38%
No Repair
257
32%
Non-Dealer Repair
237
30%
Total
792
100%
In response to the malfunction indicator light illumination, most owners (535 of 792)
sought repair, with 37.6 percent seeking dealer repair and 29.9 percent seeking other repair. Approximately
one third (32.4%) did not attempt to repair the car in response to the light illumination, with a 95 percent
confidence interval from 29.2 percent to 35.7 percent The 95 percent confidence interval for the percentage
of respondents seeking dealer repair is 34.3 percent to 41.0 percent, and for non-dealer repair, 26.7 percent
to 33.1 percent.
In one case the respondent reported not knowing what response was made to the light
illumination, however later the respondent specified the components which were repaired. These responses
are inconsistent. This inconsistency was resolved by using only those cases in the dealer or non-dealer
repair categories above when summarizing the survey results for the repair. If effect, the responses for this
one case about the repair were ignored in the analysis.
30

-------
Respondents who reported no repair were asked if the malfunction indicator light was still
on and then thanked for their participation in the survey. Respondents who attempted some repair were
asked questions about that repair. These responses are summarized in section 3.3.3.
Analysis of Owner Response Versus Driveability, Recent Emission
Inspection, and Mileage
Three primary factors were considered as possible explanations of owner response,
whether the car was running well at the time the malfunction indicator light came on, whether the car had a
recent emissions inspection, and the mileage on the car at the time of the illumination incident. Cross-
tabulations and statistical analyses for these three factors are displayed in Tables 3-16 through 3-18.
Table 3-16 Cross-tabulation of owner response and driveability with number of respondents and column
percentages.
Owner Response to malfunction
indicator light illumination
"When the light came on, was the car
running satisfactorily or was there a
problem?"
Total

Problem
Satisfactorily or
"Don't know"

Dealer Repair
116(50%)
182 (33%)
298
No Repair
32 (14%)
225 (40%)
257
Non-Dealer Repair
86 (37%)
151 (27%)
237
Total
234 (100%)
558 (100%)
792
Statistical Analysis
Chi-square = 53.8
Df= 2
p< .0001
As can be seen from Table 3-16, the statistical relationship between owner response and
driveability is highly significant Owners are more likely to seek repair if the car has a problem when the
light comes on than if the car has no problem.
31

-------
Table 3-17 Cross-tabulation of owner response and recent emissions inspection with number of
respondents and column percentages.
Owner Response to malfunction
indicator light illumination
"Since 1988 , has your car been inspected
or tested by the State for its emissions?"
Total

Yes
No or "Don't know"

Dealer Repair
132(39%)
166 (36%)
298
No Repair
97 (29%)
160(35%)
257
Non-Dealer Repair
106 (32%)
131 (29%)
237
Total
335 (100%)
457 (100%)
792
Statistical Analysis
Chi-square = 3.24
C4
II
<**
Q
p = .198
The owner response to the malfunction indicator light illumination is not significantly
related to whether the respondent reported a recent state emissions inspection. Although it might be
suspected that the existence of an inspection program might result in increased repair rates, the differences
are not statistically significant and may be due to random variation.
Table 3-18 Cross-tabulation of owner response and car mileage with number of respondents and column
percentages.
Owner Response to
malfunction indicator
light illumination
Approximate car mileage at time of malfunction
indicator light illumination


0 to 30,000
miles
30,001 to
50,000
miles
50,001 to
70.000
miles
Greater than
70,001
Total
Dealer Repair
141 (56%)
83 (40%)
32(21%)
37 (24%)
293
No Repair
58 (23%)
62 (30%)
67 (44%)
58 (37%)
245
Non-Dealer Repair
54 (21%)
60 (29%)
55 (36%)
61 (39%)
230
Total
253 (100%)
205 (100%)
154 (100%)
156 (100%)
768
Statistical Analysis
Chi-square = 68.3
Df=
6
pc.0001
Owner response is significantly related to the approximate car mileage at the time of the
malfunction indicator light illumination. In this analysis, the approximate car mileage at the time of
illumination provided a more significant statistical result than using the total mileage at the time of the
interview. As a general rule, the incidence of repair decreased with increasing car mileage. When the car
was repaired in response to the malfunction indicator light, use of dealer repair decreased and use of non-
dealer repair increased with increasing car mileage.
32

-------
Since driveability may have been a function of mileage, an analysis using log-linear models
was performed and confirmed that the two factors which were important for understanding owner response
are the car mileage and driveability (p-values based on JMP are: Approximate car mileage chi-
square = 66.2, Df = 6, and p < .0001, Driveability chi-square = 50.7, Df = 2, and p < .0001). Although
the new/used status and the make of the car were also significantly related to the owner response when
considered alone, these factors were not statistically significant after adjusting for the mileage and
driveability.
Table 3-19 shows the relationship between owner response and car mileage for cars with a
driveability problem and Table 3-20 shows similar information for cars with no driveability problem. As
can be seen from these tables, the relationship between owner response and driveability depends on whether
the car is running satisfactorily.
Whether owners sought repair depended significantly on both the car mileage and the
driveability of the car. If the car was not running satisfactorily, 87 percent of owners sought repair,
independent of the mileage. However, owners were less likely to go to a dealer for service as the car
mileage increased. If the car was running satisfactorily, 60 percent sought repair with decreasing incidence
of both repair in general and dealer repair, with increasing car mileage.
Table 3-19 Cross-tabulation of owner response and car mileage for cars with a driveability problem with
number of respondents and column percentages.
Owner Response to
malfunction indicator
light illumination
Approximate mileage at time of malfunction indicator
light illumination


0 to 30,000
miles
30,001 to
50,000
miles
50,001 to
70,000
miles
Greater than
70,001
Total
Dealer Repair
49 (67%)
37 (61%)
9 (22%)
19(37%)
114
No Repair
9(12%)
7(11%)
7 (17%)
6(12%)
29
Non-Dealer Repair
15(21%)
17 (28%)
25 (61%)
27 (52%)
84
Total
73 (100%)
61 (100%)
41 (100%)
52 (100%)
227
Statistical Analysis
Chi-square = 30.9
Df =
6
p< .0001
33

-------
Table 3-20 Cross-tabulation of owner response and car mileage for cars with no driveability problem
with number of respondents and column percentages.
Owner Response to
malfunction indicator
light illumination
Approximate mileage at time of malfunction indicator
light illumination


0 to 30,000
mites
30,001 to
50,000
mites
50,001 to Greater than
70,000 70,001
mites
Total
Dealer Repair
92 (51%)
46 (32%)
23 (20%) 18 (17%)
179
No Repair
49 (27%)
55 (38%)
60 (53%) 52 (50%)
216
Non-Dealer Repair
39 (22%)
43 (30%)
30 (27%) 34 (33%)
146
Total
180 (100%)
144 (100%)
113(100%) 104(100%)
541
Statistical Analysis
Chi-square-= 49.6
Df = 6
p< .0001
Is the Light Off or On
Before completing the interview, respondents were asked if the light was on or off at the
present time. In 84.7 percent of the cars in which the malfunction indicator light came on and stayed on, the
light was off at the time of the interview, often but not always as a result of the actions of the respondent.
The 95 percent confidence interval for the percentage of cars with lights off at the time of the survey is from
82.2 percent to 87.2 percent These results are summarized in Table 3-21.
Table 3-21 Responses to "Is the light still on?"

Number of
Percentage of
Light Still On
Respondents
Respondents
Qi
92
12%
Off
671
85%
Don't Know
29
4%
Total
792
100%
Is the Light Disconnected
Based on 559 respondents who were asked the question at the end of the interview, in only
2.3 percent of these cars the light was disconnected. One light was disconnected without any repair being
attempted. In the remaining 12 cases, the light was disconnected as a result of the repair effort. These
results are summarized in Table 3-22.
34

-------
Table 3-22 Responses to "Was the engine light disconnected?" and "What happened to the light?"
Light Disconnected
Number of
Respondents
Percentage of
Respondents
Yes
13
2%
No
546
69%
Not asked
233
29%
Total
792
100%
3.3.3	Data from Respondents Who Attempted Repair
For those 535 respondents attempting repair, additional questions to obtain information
about the distance driven before repair, the items repaired, the success of the repair and the cost of the repair
were asked. These items are discussed in this section.
Miles driven Before Repair
Table 3-23 Responses to "Approximately how many miles was the car driven with the light on before
any action was taken?"
Number of "Don't Know's"
45
Number of Values
490

Miles Driven
Mean
521
Standard Deviation
3760
Percentiles

Minimum
0
5%
0
25%
3
50% (Median)
10
75%
50
95%
1,000
Maximum
50,000
Of the 535 respondents who sought repair, the median miles driven before repair is 10
miles, with 95 percent of cars driven 1000 miles or less before repair. For the statistical analysis, the total
35

-------
mileage data was coded into the two categories shown in Table 3-24. The "don't know" responses were
not coded for analysis. Most (78%) of the respondents drove SO miles or less before getting repair.
Table 3-24 Coded answers to "Approximately how many miles was the car driven with the light on
before any action was taken?"

Number of
Percentage of
Miles Driven Before Repair
Respondents
Respondents
0 to 50 miles
380
78%
Over 50 miles
110
22%
Total
490
100%
Analysis of Miles driven Before Repair Versus Driveability, Recent
Emissions Inspection, and Mileage
Three primary factors were considered as possible explanations of number of miles driven
before repair was obtained, whether the car was running well at the time the malfunction indicator light
came on, whether the car was subject to emissions inspection (or in the case of the survey, had been
recently inspected), and the mileage on the car at the time of the illumination incident Cross-tabulations and
statistical analyses for these three factors are displayed in Tables 3-2S through 3-27.
Table 3-25 Cross-tabulation of miles driven before repair and driveability with number of respondents
and column percentages.
Miles driven before repair
"When the light came on, was the car
running satisfactorily or was there a
problem?"
Total

Problem
Satisfactorily or
"Don't know"

0 to 50 miles
154 (80%)
226 (76%)
380
Over 50 miles
38 (20%)
72 (24%)
110
Total
192 (100%)
298 (100%)
490
Statistical Analysis
Chi-square = 1.28
Df = 1
p = 0.258
Based on the results in Table 3-26, the statistical relationship between miles driven before
repair and driveability is not statistically significant
36

-------
Table 3-26 Cross-tabulation of miles driven before repair and recent emissions inspection with number
of respondents and column percentages.
Miles driven before repair
"Since 1988 , has your car been inspected
or tested by the State for its emissions?"
Total

Yes
No or "Don't know"

0 to SO miles
158 (74%)
222 (80%)
380
Over 50 miles
55 (26%)
55 (20%)
110
Total
213 (100%)
277 (100%)
490
Statistical Analysis
Chi-square = 2.46
Df= 1
p = 0.116
The relationship between miles driven before repair and recent emission inspection is not
significantly significant
Table 3-27 Cross-tabulation of miles driven before repair and approximate car mileage with number of
respondents and column percentages.
Miles driven before
repair
Approximate mileage at time of malfunction indicator
light illumination


0 to 30,000
miles
30,001 to
50,000
miles
50,001 to
70,000
miles
Greater than
70,001 or
"Don't
know"
Total
0 to 50 miles
142 (79%)
112(82%)
58 (72%)
61 (71%)
373
Over 50 miles
38 (21%)
24(18%)
23 (28%)
25 (29%)
110
Total
180 (100%)
136(100%)
81 (100%)
86 (100%)
463
Statistical Analysis
Chi-square = 5.71
Df=
3
p = 0.127
Miles driven before repair is not significantly related to the approximate car mileage. In this
analysis, the results were similar whether the approximate car mileage at the time of illumination or the total
mileage at the time of the interview was used. Analysis of variance was used to model the numerical miles
driven before repair. None of the factors considered had a significant relationship with the miles driven
before repair. Although these factors might explain some of the patterns in the miles driven before repair,
the data set is not large enough to identify any significant relationships.
37

-------
Malfunctioning Component
Information on the malfunctioning component was only collected if the malfunction
indicator light came on. For this analysis, the definition of the "malfunctioning component" is specified in
terms of the results from the survey as follows: The malfunctioning component is the last component or set
of components which, when repaired, resulted in a successful repair or, if a successful repair was not
achieved, those components reported as repaired. Questions were asked about the repair of six items (fuel
system, ignition system, oxygen sensr- computer control module, exhaust gas recalculation system
(EGR), and the engine warning lights . Any combination of these six items may have been selected.
In most cases only one or two items w> :lected. The malfunctioning component may be one item (e.g.
"fuel system") or several items (e.g. "fu_. and ignition systems").
In Tables 3-28 and 3-29 the information on the malfunctioning component is summarized
in two ways, 1) the number of respondents specifying each component, and 2) the combination of
components which were repaired.For the data in table 3-28, a respondent may have selected multiple items.
Table 3-29 shows the most commonly specified combination of items which were repaired.
Table 3-28 Number and percentage of respondents selecting each malfunctioning component
Malfunctioning component
Repaired
Not
repaired
Don't
Know
Total
Fuel System
135 (25%)
206 (39%)
194 (36%)
535
(100%)
Ignition System
118(22%)
223 (42%)
194 (36%)
535
(100%)
Oxygen Sensor
72 (13%)
269 (50%)
194 (36%)
535
(100%)
Control Module
138 (26%)
203 (38%)
194 (36%)
535
(100%)
EGR System
53 (10%)
288 (54%)
194 (36%)
535
(100%)
Engine Warning Light
68 (13%)
273(51%)
194 (36%)
535
(100%)
38

-------
Table 3-29 Number and percentage of respondents selecting each combination of malfunctioning
components
Malfunctioning Components
Number of
Respondents
Percentage of
Respondents
Unknown
194
36%
Control Module only
50
9%
Fuel System Only
45
8%
Ignition System Only
36
7%
Warning Light Only
26
5%
Oxygen Sensor Only
21
4%
Fuel System and Control
Module
16
3%
Control Module and Warning
Light
13
2%
EGR System Only
12
2%
Fuel and Ignition Systems
12
2%
Ignition System and Control
Module
10
2%
33 other combinations
100
19%
Total
535
100%
Statistical analysis using the malfunctioning component are based on responses to the
questions about each item repaired. Although prediction of the malfunctioning component was not of direct
concern in the survey, as a result of the analysis, the following significant relationships were found: the
ignition system was repaired more often in cars with higher mileage, the oxygen sensor was repaired more
often in used cars than new cars (even after adjusting for mileage), the fuel system was repaired more often
in cars with a driveability problem, and the indicator light was repaired more often in cars which were
running satisfactorily.
Repair Success
The success of the repair is judged from the answers to the two questions: "Was the repair
successful?" and if not, "Since the first repair attempt was the car ever successfully repaired?". Note that
the questions which ask if the repair was successful defined success as satisfactory to the respondent.
Successful repair was indicated by an affirmative response to the questions about repair success, i.e.
successful repair the first time or successful repair ever will be considered successful repair. Because the
flow through the questionnaire considered "don't know" to be equivalent to indicating a successful repair,
39

-------
the same criteria will be used to define success. Note that the survey results do not provide data to compare
the repair success in cars with and without the malfunction indicator light
The repair was judged successful 88.8 percent of the time. The 95 percent confidence
interval for the percentage of cars receiving successful repair in response to malfunction indicator light
illumination is from 86.1 percent to 91.5 percent Table 3-30 summarizes the success of the repair work.
Table 3-30 Was the repair successful?

Number of
Percentage of
Repair Success
Respondents
Respondents
Yes
475
89%
No
60
11%
Total
535
100%
Analysis of Repair Success Versus the Malfunctioning Component and
Driveability
Two primary factors were considered as possible explanations of repair success; whether
the car was running well at the time the malfunction indicator light came on, and the malfunctioning
component Cross-tabulations and statistical analyses for these two factors are displayed in Tables 3-31 and
3-32.
Table 3-31 Cross-tabulation of repair success and driveability with number of respondents and column
percentages.
Repair success
"When the light came on, was the car
running satisfactorily or was there a
problem?"
Total

Problem
Satisfactorily or
"Don't know"

No
20(10%)
40 (12%)
60
Yes
182 (90%)
293 (88%)
475
Total
202 (100%)
333 (100%)
535
Statistical Analysis
Chi-square = 0.56
ii
<4-1
Q
p = 0.453
Based on the results in Table 3-31, the relationship between successful repair and
driveability is not statistically significant
40

-------
The analysis looked at the relationship between repair success and repair of all 6
components. For only one component was the relationship with repair success significant, shown in Table
3-32. The repair was reported to be less successful when the ignition system was repaired.
Table 3-32 Cross-tabulation of repair success and ignition system repair with number of respondents and
column percentages.
Repair Success
"Did the Repair work involve the ignition
system including spark plugs, distributor, or
timing"
Yes No Don't know
Total
No
20(17%) 18(8%) 22(11%)
60
Yes
98 (83%) 205 (92%) 172 (89%)
475
Total
118(100%) 223(100%) 118(100%)
535
Statistical Analysis
Chi-square = 6.18 Df= 2
p = 0.045
Warranty
Table 3-33 Responses to "Was the repair or service covered under the manufacturer's warranty,
therefore performed at no cost to you?"
Warranty Repair
Number of
Respondents
Percentage of
Respondents
Yes
194
36%
No
331
63%
Don't know
10
2%
Total
535
100%
In 36.3 percent of the cases, repairs were covered by warranty, with a 95 percent
confidence interval for the percentage of cars covered by warranty of 32.2 percent to 40.3 percent.
Cost of Repair If Not Covered by Warranty
As shown in Table 3-34, for those repairs not covered by warranty, the median repair cost
is $100 with 95 percent of the repairs costing $907 or less.
41

-------
Table 3-34 Responses to "How much did the repair or service cost you?"
Number of "Don't Know's"
71
Number of Data Values
270

ReoairCost
Mean
$246
Standard Deviation
$446
Percentiles

Minimum
$1
5%
$1
25%
$35
50% (Median)
$100
75%
$230
95%
$907
Maximum
$4,000
For the statistical analysis, the numerical cost was coded into the two categories shown in
Table 3-35.
Table 3-35 Coded responses to "How much did the repair or service cost you?"

Number of
Percentage of
Repair Cost
Respondents
Respondents
0 to $100
141
52%
Greater than $100
129
48%
Total
270
100%
Analysis of Repair Cost Versus Malfunctioning Components and
Driveability
Two primary factors were considered as possible explanations of cost of the repair,
whether the car had a driveability problem and the malfunctioning component. Two other factors which
were important were the owner response (dealer versus non-dealer repair), and whether the car had a recent
state emissions test Cross-tabulations and statistical analyses for these factors are displayed in Tables 3-36
through 3-39.
42

-------
Table 3-36 Cross-tabulation of repair cost and driveability with number of respondents and column
percentages.
"How much did the repair or
service cost you?"
"When the light came on, was the car
running satisfactorily or was there a
problem?"
Total

Problem
Satisfactorily or
"Don't know"

$0to$100
53 (47%)
88 (56%)
141
Greater than $100
60(53%)
69 (44%)
129
Total
113(100%)
157 (100%)
270
Statistical Analysis
Chi-square = 2.20
ii
Q
p = 0.138
Based on the results in Table 3-36, the statistical relationship between repair cost and
driveability is not statistically significant
Table 3-37 Cross-tabulation of repair cost and recent emissions inspection with number of respondents
and column percentages.
"How much did the repair or
service cost you?"
"Since 1988 , has your car been inspected
or tested by the State for its emissions?"
Total

Yes
No or "Don't know"

$0 to $100
50 (43%)
91 (59%)
141
Greater than $100
67 (57%)
62(41%)
129
Total
117(100%)
153 (100%)
270
Statistical Analysis
Chi-square = 7.48
Df = 1
p = 0.0063
Based on the results in Table 3-37, the statistical relationship between repair cost and recent
emissions inspection is statistically significant. Repair costs were greater when a recent state emissions test
was reported.
Based on the results in Table 3-38, the statistical relationship between repair cost and
owner response is statistically significant Repair costs were higher using dealer rather than non-dealer
repair. Note that non-dealer repair includes self repair and repair by a friend, which could be expected to be
less expensive.
43

-------
Table 3-38 Cross-tabulation of repair cost and owner response with number of respondents and column
percentages.
"How much did the repair or
service cost you?"
Owner response
Total

Dealer Repair
Non-dealer Repair

$0to$100
43 (39%)
98 (62%)
141
Greater than SI 00
68 (61%)
61 (38%)
129
Total
111 (100%)
159 (100%)
270
Statistical Analysis
Chi-square = 13.7
Df= 1
p = 0.0002
The cost of repair might be expected to depend on the items repaired. Repair of several
items showed a significant relationship with repair cost, however the interpretation of these relationships is
difficult without considering all of the combinations of items which were repaired. When using log-linear
models where several factors were considered at the same time, no one combination of components
appeared to best explain the repair cost.3 Although it can be concluded that the components repaired
affected the repair cost, the results from the survey were not adequate to determine the costs as a function of
the items repaired.
Since the repair cost is a numerical variable, analysis of covariance was also used to model
the repair costs. In general, using analysis of covariance is more sensitive to relationships in the data than
using log-linear models with the coded repair cost Because the cost data were highly skewed (with many
low values and few very high values) the log transformed data were used in the analysis. In addition, the
log transformed total miles was used in the analysis. Because the results can be more difficult to interpret
and depend on the transformation used, the general results are provided without a detailed explanation of the
model.
As with the log-linear analysis of the coded repair costs, the owner response, recent
emissions inspection, and the components being repaired were related to the repair cost. In addition, the
repair cost increased with increased total mileage. The repair cost also depended on the driveability of the
car. Cars with driveability problems in general had a higher repair cost The p-values from the analysis of
covariance are shown in Table 3-39.
The final model included terms for fuel and ignition system repairs.
44

-------
Table 3-39 Significance values for terms in the model of Log(repair cost)
Factor
P-value
Driveability
Owner response
Recent emissions inspection
Log transformed total mileage
Factors associated with the components
repaired
p = 0.0066
p < 0.0001
p = 0.0035
p = 0.0047
p = 0.013 and 0.020
Assuming the model is correct, the following cost ratios can be determined. Repair costs if
the car was not running well were greater than costs if the car was running well by a factor of 1.29 (with a
95 percent confidence interval of 1.07 to 1.54). Repair costs using dealer repair were greater than non-
dealer repair by a factor of 1.52 (with a 95 percent confidence interval of 1.27 to 1.83). Repair costs when
the car had a recent emissions inspection were greater than with no recent inspection by a factor of 1.31
(with a 95 percent confidence interval of 1.09 to 1.58).
Because different models can be fit to the data, the models may be sensitive to the
transformation used, and the model may not apply beyond a narrow range of costs represented by the data,
the factors above should be used with care. The following general statements might be made. As a rule of
thumb, cars with driveability problems cost more, by a factor of 1.3, to fix than those with no driveability
problem. Dealer repair cost more than non-dealer repair by a factor of about 1.5. Repair costs for cars with
recent emissions inspections were greater than for other cars by a factor of about 1.3. These factors do not
apply in extraordinary circumstances. Repair costs also tend to increase with car mileage. Finally, repair
costs depended on the components repaired. However, because of the many combinations of items being
repaired, a breakdown of the costs could not be determined from the survey data.
Was Driveability Problem Corrected
Those respondents who reported a driveability problem were asked if the same problem
still existed. The answers to this question are summarized in Table 3-40
45

-------
Table 3-40 If the respondent reported the car was not running well, response to "After the car was
repaired, did it still have the same problem?"
Driveability Problem Remains
Number of
Respondents
Percentage of
Respondents
Yes
37
19%
No
153
77%
Not Asked
9
5%
Total
199
100%
3.3.4	Summary of Results
The survey interviewed 2621 owners of General Motors, Chrysler, and Toyota cars cars
from 34 states. The cars included the 1981 to 1989 model years. Eighty-two percent of the cars were made
by General Motors and 88 percent came from the 1983 through 1988 model years. The owners reported
that 62 percent of the cars were acquired new and 38 percent used. Forty-one percent of the respondents
reported that their cars had state emissions inspections since 1988. The median mileage on the cars at the
time of the interview was 50,000 miles.
The malfunction indicator light came on and stayed on in 792 cars, or 30.2 percent of the
2621 cars covered by the interviews. Thus we can state with 95% confidence that the true proportion of
cars with malfunction light illuminations is between 28.5 percent to 32.0 percent. The incidence of
malfunction indicator light illumination increases as total mileage increases as shown in Figure 3-2. The
incidence also depends on the make of the car.
Of the 792 cars in which the malfunction lights came on and stayed on, the approximate
mileage at which the light last came on is shown in Figure 3-3. At the time the light came on, 29.5 percent
of the respondents reported that their car were not running satisfactorily (i.e. had a problem). The 95
percent confidence interval for the percentage of cars not running satisfactorily is 26.4 percent to 32.7
percent.
46

-------
Figure 3-2 Percentage of cars with malfunction indicator light illumination by total mileage
Percent of cars
with malfunction 20% +
light illumination
0 to 30,000
miles
30,001 to
50,000 miles
50,001 to
70,000 miles
Greater than
70,000 miles
Total Mileage
Figure 3-3 Percentage of cars for which the malfunction indicator light stayed on at the indicated car
mileage
Percent of cars	.
with malfunction
light coming on
Oast occurence)
0 to 30,000
miles
30,001 to
50,000 miles
Total Mileage
50,001 to
70,000 miles
Greater than
70,000 miles
47

-------
In response to the malfunction indicator light illumination, most owners (S35 of 792)
sought repair, with 37.6 percent seeking dealer repair and 29.9 percent seeking other repair. Approximately
one third (32.4%) of the respondents did not attempt to repair the car in response to the light illumination
(with a 95 percent confidence interval from 29.2 percent to 35.7 percent). The 95 percent confidence
interval for the percentage of respondents seeking dealer repair is 34.3 percent to 41.0 percent, and for non-
dealer repair, 26.7 percent to 33.1 percent Whether owners sought repair depended significantly on both
the car mileage and the driveability of the car. As shown in figures 3-4 and 3-5, if the car was not running
satisfactorily, 87 percent of owners sought repair, independent of the mileage. However, they were less
likely to go to a dealer (more likely to go to a non-dealer) for service as the car mileage increased. If the car
was running satisfactorily, 60 percent sought repair with decreasing incidence of both repair in general and
dealer repair, with increasing car mileage.
In 84.7 percent of the cars in which the malfunction indicator lights came on and stayed on,
the lights were off at the time of the interview, often but not always as a result of the actions of the
respondent Of the 559 respondents who were asked the question, 2.3 percent of them said that their light
was disconnected. For most of these cases, the light was disconnected even through repair was soughL
Of the 535 respondents who sought repair, the median miles driven before repair is 10
miles, with 95 percent of these cars driven less than 1000 miles before repair. The miles driven before
repair was not significantly related to other variables collected in the survey. The malfunctioning
component, if known, most often included the computer control module and the fuel system. In 13 percent
of the cars, there was a problem with the warning light itself. Five percent of respondents reported a
problem with the warning light only.
The repair was judged successful 88.8 percent of the time. The 95 percent confidence
interval for the percentage of cars receiving successful repair in response to malfunction indicator light
illumination is from 86.1 percent to 91.5 percent The repair success was greater when the ignition system
was not one of the components being fixed.
48

-------
Figure 3-4 Owner response versus approximate car mileage for cars which were running satisfactorily
Percent of
Respondents
~ No Repair
E3 Non-dealer Repair
¦ Dealer Repair
0 to 30,000 30,001 to 50,001 to Greater
miles 50,000 70,000 than
miles miles 70,000
miles
Approximate Car Mileage
Figure 3r5 Owner response versus approximate car mileage for cars which were not running
satisfactorily
Percent of
Respondents
100% T
90% ¦¦
60%
50%
40%
30%
20%
10%
0%
0 to 30,000 30,001 to
miles 50,000
miles
50,001 to Greater
70,000 than
miles 70,000
miles
~ No Repair
E3 Non-dealer Repair
¦ Dealer Repair
Approximate Car Mileage
49

-------
In 36.3 percent of the repairs, the costs were covered by warranty, with a 95 percent
confidence interval for the percentage of cars covered by warranty under the survey conditions of 32.2
percent to 40.3 percent. For those repairs not covered by warranty, the median repair cost was $100 with
95 percent of the repairs costing $907 or less. The repair cost depended on several factors, including
driveability, the items repaired, whether the car had a recent emission inspection, type of repair and the
mileage on the car. As a rule of thumb, repair of cars with driveability problems cost more, by a factor of
1.3, than chose with no problems. Dealer repair cost more than non-dealer repair by a factor of about 1.5.
Repair costs for cars with recent emissions inspections were greater than for other cars by a factor of about
1.3. Repair costs tend to increase with car mileage. Finally, repair costs depend on the components
repaired. However, because of the many combinations of items being repaired, a breakdown of the costs
could not be determined from the survey data.
50

-------
Appendix A
SURVEY QUESTIONNAIRE
IN CATI FORMAT

-------
3o
< El
Ot.Ol El
H*llot ny nut t» {	>,	I
I'n calling fro* M*»tat« a rtstarch firn, for th» Unttad
Stat* Environmental Protect ion Agancy. Ut ar* conducting a
itudy about daahboard warning light* in can.
I'H trying to raach < >-< >-< }.
Did I dial tht correct ttltphon* nurtbtr?
( )
1.	YES
2.	NO
3.	ENTER NEW PHONE NUMBER AND RESIAL
4.	GO TO RESULT CODE SCREEN
5.	REDIAL
A-l

-------
Do you hav» • t»l«phon» nunbor it which
 can b* rtachod?
( )
1.	YES
2.	NO

-------
Scr««fi nui* E2
01.02 E2
Nay I pltat* tp»ak to <	>?
( )
1.	OWNER SPEAKING
2.	AVAILABLE
3.	NOT AVAILABLE
4.	NO LONGER AT THIS PHONE NUHBER
A-3

-------
Scr««n n«w< E4
01.03 E4
Uhat it that t»l»phon» nunbtr including art* cod*?
A-A

-------
Scr««n run*' INTR02
01.043 INTR02
Htlloi By nan* It <	>.
I'm calling fro* W»»t«t> • rvtaarch fir*. for th» United
States Environmental Protection Agency. We art conducting a
study about dashboard warning light* In cart.
CPRESS RETURN]
A-5

-------
Scrwn nmi E3
01.03 E3
According to our r«eordii you art the owner of
{RARE* HOTEL AND YEAR	>.
It that correct?
< »
1.	YE9
2.	NO

-------
Do you hava tha nan* of tha currant ownar?
( )
1.	YES
2.	NO

-------
ScrtM niM* E7
01.07 E7
Wfcat it th* naM of lft« owwr?
( )
1.	DEALER
2.	INDIVIDUAL
NAME l		
FIRST " " "" LAST"

-------
Dom <
li*» in thli houttholil?
( )
1.	YES
2,	NO

-------
Do |fou ftavt a t*l«pfton» nunbvr «t which
t	>
cm b* r»tch»d?
( )
t. VES
2. NO
A-ie

-------
Hig 1 pltat* tp«ak to <
( )
1.	AVAILABLE
2.	NOT AVAILABLE

-------
ScrMn rtiM( CL0SE.3
01.1099 CtOSE.3
Thank you for your tin* «nd cooperation.
CPRESS RETURN!
A-12

-------
Sctmb niM< Ell
01.11 Ell
1 m going to ask you a feu questions about the dashboard warning
lighti on your .	I
Your participation in this study it voluntary and th» info mat ion you
provide will be kapt confidential. Before we begin> you should know
w» have no r«a*on to believe that anything it wrong with your cart we
are just gathering information.
When you acquired th» { nodal	> was it new*
that is never previously titled, or we* it a used car?
( )
1.	NEW I NEVER PREVIOUSL TITLED)
2.	USES
A.-13

-------
Scr«»n nan** E12
101.12 E12
b*»n inspected or	by th« Stal* for Missions?
( )
1.	YES
2.	NO
A-14

-------
Scr«m na*** E13
01.13 E13
Art you th» principal drlv»r of th» (nod*I	). i
that 1»> it th* car drivvn by y°u	of th» tin*?
( )
1.	YES
2.	NO
A-15

-------
Scr»*n njM« £14
01.14 £14
Ufca it vtw £rincip»l driw?
FIRST NME	LAST KANE
A-16

-------
Scr««f* niM< E15
01.13 £13
Do** <	> liv» in thit hout*hold?
I )
1.	YES
2.	MO
k-11

-------
Scr»»n naf»* £14
01.14 £14
Do you hiv» a t*l*phon» nimtwr *t which
{	>	1
can b» r**ch«d?
< )
u res
a. *Q
A.-18

-------
Screen nane^ E17
01.17 E17	I
i
i
l
What is that telephone nunber including area code?
A-19

-------
Scr««n nan*< E18
01.18 E18
nay I pleat* ipiak to {PRINCIPAL DRIVER	>?
( )
1.	AVAILABLE
2.	NOT AVAILABLE
A-20

-------
Screen nane« E19
01.19 E19
Hello? Hy nan* is {	}.
1> calling fron Ueitat> a retearch fir*. for the Unitad
Slat** Environmental Protection Agency. Me ere conducting a
ttudy about dathboard warning light» in can.
< OUNER »1	> told u* that you art the
principal driver of the < YEAR HAKE MODEL	>.
that i»> the < MODEL	> it driven by you no»t
of tht tint. It that correct?
( )
1.	YES
2.	NO
A-21

-------
Screen nan*» STATNENT
01.193 STATItWT
I an going to a*k a f«w quettiorii about tha dashboard warning Ugftti
on your (HODEl	>. Your participation in thii
study it voluntary and tha int*ornation you provid* yil! b* *apt
confidant!al. Bafora wa oagln. you thould know w* hava no raaten to
b*U»v» that anything it wrong nith you carl wa ara just gathering
information.
tPRESS RETURN TO CONTINUE!
A-22

-------
Scrttn ni««< £20
01.20 E20
Approximately hou n
-------
Scr»«n na«*« E21
01.21 E21
Th»  hit an
•ngin* warning light locatad on th« dashboard.
< In Chryiltr* th* engine warning light nay read on* of the	>
( following! power loss. power United or check engine.	>
Hat the warning light ever con* on and stayed on after starting the car?
f )
1.	YES
2.	NO
A-24

-------
Screen nan»< E22
01.22 E22
CPR06E: IN THE {YEAR. MODEL	>
THE < WARNING SICN> LIGHT 15 LOCATED ON THE {ALLOCATION	>
PART OF THE DASHBOARD. I
Sine* you acquired the < MODEL	>. ha* the
{ WARNING SIGNAL	> con* on whilt the car uai being driven?
( )
1.	YES
2.	NO

-------
Scr**n nan*< CL0SE_4
101.2299 CL0SE_4
I
I
I
I
I
I
I	Tho»* ar* all th* qu*»tion» I hav* at thlt tin*.
I	Thank you v*ry nuch for your tin* and cooperation.
I
I
I	[PRESS RETURN]
A-26

-------
Scr»*n nan*< E23
01.23 E23
C<	>3
U« art only int»r»it»d in th# 58ii_CiSJDi tin* th» »ngin» warning light
can* on and stayed on aftar starting th» car. Thinking back to th# *o»t
r*c«nt	in uhat ytic did th» *ngin« warning light com* on?
19( )
A-27

-------
5cr««n n*A*< E24A
01.233 E2«
11	>11
Wh*n tht light cia« cm w*< you ear running latliftctorily
or uat th»r» • probltn?
( )
1.	SATISFACTORILY
2.	PROBLEM
A-28

-------
Scr»#n riant: E2*
101.24 £24
I
I
I
in	jj	i
I
I
I	U«t any action takan in reipontt to tha angint warning light?
I
I	< )
I
I	1. YES
I	2. NO
A-29

-------
Scr»«n nant< E23
101.23 E23
I
I
I
I	t<	>]	I
I
I
I	It th» light ttlU on?
I
I
I	( )
l
I	1. YES
I	2. NO
A-30

-------
Scrt«n niMi E26
01.26 E26

>1
Uh«t h*pp*n»d to th» Ught?
( 1
1.	LIGHT TURNED OFF BY ITSELF
2.	LIGHT WAS DISCONNECTED
3.	RESPONDENT TOOK ACTION
A-31

-------
Scr»«n na*«« E27
101.27 E27
I
I
I	C<	>3	I
I
I
I	Approximately how nany nilct wat tht < MODEL	>
I	driven with th» light on btfor* any action ua« takan?
I
I
I	NILES (	)
A-

-------
Screen niMi E2S
01 .2& saa
Ei	)1
I am going to read a thort lilt and I'd like you to tell me
which of the fallowing best deicribet the ficii action taken
to repair tht car after the engine warning light cane on?
< >
1.	Tht car wat taken to a profej»ional For service or repair.
2.	>»u attempted U ripilr or- imlcf the car yourtelF.
3.	A friend or relative attempted to repair or service the car
4.	You tried to get more information about the problem
5.	You uteo the ciagnoitic computer on your dathboard to
learn more about the problem.
A-33

-------
Scrtcn ntM< £28 A
1.281 E2BA
S£iIC you tri»d to g»t nor* information
«baut th» probltn did you ...
( >
t. h*v» th» cir ttk«n to » prof»inon»l for rtptir or itrvic*
Z. eid ifou iLtMPt to nptir cr s»rvi(» the e»r youriitr
3.	did ¦ Fri»rtd or rtlativ# att«npt to rtptir or tcrvic* th« c«ri or
4.	did you t»k« no further action?
A-34

-------
Scraan na*a« E29
101.29 E29
SfilC you uiad tha diagnottic eonput»r. did you..
( )
1.	havt th# car takan to a professional Tor repair or sarvica>
2.	did you attanpt to rapair or sarvica tha car youri*lP> or
3.	did a friend or ralativa attanpt to rapair or sarvica tha car?
4.	did you taka no further action?
A-35

-------
ScrMfi ni«*i E30
01.30 E30
Which of th» following but d»»crlb»» th* repair shop
whtr* tht car was first tak»n? Was it a...
< >
1.	d»al*rship> or an
2.	ind*p*nd*nt shop
91.OTHER
SPECIFY:	
A-36

-------
Screen nane< E3I
01.31 E31
C<	>3	I
I an going to read a short lilt of itent that nay have been repaired. At
1 rtad each iten. pleat* tell «e whether that repair was attempted when the
<	nODEL	> uat ficii serviced after
the engine warning light cane on. At thit tine. I an not interested in
whether the repair was successful. You can answer thete yet or no.
Did the repair work involve...	C1*VES. a«N0J
a.	the fuel sytten> that it> injectort> carburetor,
or idle tpeed controller?
b.	the ignition tytten including spark plugi> distributor, or tilling?
c.	the oxygen sensor?
d.	the computer control nodule?
e.	the EGR> that is> the exhaust gat recirculation systeM?
f.	a problen with the engine warning light. itself?	( ) I
A-37

-------
Scr»#r» n«n*< E32
01.32 E32
ful?
t )
1.	YES
2.	NO
A-38

-------
Scrttn nan*: E33
01.33 E33
Sine* th« first repair att«npt> wai th» car »v»r »ucc»t»fuUy rtpairtd?
( )
1.	YES
2.	NO
A-3S

-------
Screen nane» E34
01.34 E34
I an going to read the lift of itens that nay have been repaired again.
As I read each iten. please tell ne whether that iten was involved
when the car was iySUilfyllu repaired. You can answer these yet or no.
Did the repair work involve...	tt«YES> 2*N01
a.	the fuel systen. that is. injectors. carburetor.
or idle speed controller?	( )
b.	the ignition lysten including spark plugs, distributor! or tining?	( )
c.	the oxygen sensor?	( )
d.	the conputer control nodule?	( )
e.	the EGR. that is. the exhaust gas recirculation systea?	( )
F. a problen with the {WARNING SIGNAL) light, itself? ( ) I
A-40

-------
•*n niMi E33
01.33 E3S
Uai tha rapair or tarvica covarad undar tha nanufacturtr's
warranty, tharafora parPornad at no co«t to you?
( )
1.	YES
2.	NO
A-41

-------
Scr»tn nan*: E36
101.36 E36
I
I
I
I
I
I
I	Hoy «uch did th» repair or itrvlct cott you?
I
I
I	«( ).00
A-A 2

-------
Ua« th» »ngin» warning light dlsconn»ct*d?
( )
1.	YES
2.	NO
A-43

-------
Scr««n nan*: E3SA
01.373 E38A
You n*ntlon*d »arli»r that at th» tin* the *ngin» warning
light can* on. your car wat not running satitfactorily. Aft»r
th» car va* r*palr»d> did it itiU have th» *an* probl»n?
( )
1.	YES
2.	NO
A-44

-------
Scrftn n*n*( E39
01.379 E39A
You nvntionvd •¦rlivr thit jhggiuy
A.-45

-------
Scr»*n nan*< E38
01.38 E38
U	>J
Ii th» light itiU on?
( )
I. YES
3. NO
A-46

-------
Scr«*n nan*« CL0SE1
101.9998 CL0SE1
I
I
I
I
I
I	Thot* ar» all th» qu»ttiont I hav* at thit tint.
I	Thank you v*ry nuch for your tin* and cooperation.
I
I
I	[PRESS RETURN!
A.-47

-------
Scr»»n nanei CLOSE.5
01.9999 CLOSE 3
Thank you very nuch for your tin* and cooperation.
Your participation has h*lp»d to nak* th» study
a tuccfii.
(PRESS RETURN]
A-A 8

-------
Scr»»n nwti KEWHOH
I 9.99 NEWPHON
I
t
I	CEnter new phone nunbtr or :«ro to l»»v» tcr«*nl
I
I
I
t	I >
A-49

-------
Scr»*n ntft*< RSLTSCRN
99.0 RSLTSCRN
ENTER APPROPRIATE RESULT CODE.
( )
2. REFUSAL
9. INTERIM
NW. NON-WORK!NC
NA. NO ANSWER
NR. NON-RESIDENTIAL
LP. LANGUAGE PROBLEM
RB. FINAL REFUSAL
nC. MAXIMUM CALLS
0. OTHER
Nl. NEVER HEARD OF OWNER
N2. NEVER OWNED CAR
N3. NOT AVAILABLE DURING FIELD PERIOD
A-50

-------
Scrttn nan*( ElSB
199.099 EISA
Ii {	> »v»r available at
thli phon» nunb*r?
( )
1.	YES
2.	NO
A-51

-------
Screen nane« RESTART
99.2 RESTART
[THIS IS A RESTART CASE) SOME OF THE INTERVIEW UAS ADMMISTERED
IN A PREVIOUS SCREEN. J
nay I speak with <	>?	1
Htllot fty nan* ii <	>. I an calling back
fron Weitat« Inc.. for th» United States Environmental Protection
Agency. The other day we ipoke to you about a atudy of dashboard
warning light! in cart. We would like to conplete the interview
at this tine.
( )
1.	CONTINUE
2.	GO TO RESULT
A-52

-------
THIS CASE HAS BEEN CODED < >
[PRESS RETURN TO CONTINUE]
A-

-------
Scrtfft run*: CALLBACK
999.91 CALLBACK
I would lik# to «aka an appointnant to tpaak to
< >.
Mould tha bait tin* to call back 6a...
( >
1.	Utar today»
2.	iom othar day. or
3.	no (pacific Una?
A-54

-------
Scrttn na**i TODAYAPP
999.92 TODAYAPP
RESPONDENT TIHE	)s( ) AN > 1 ( >
Ptl ¦ 2
A-55

-------
Scr»»fi n»n»» SCHEDAPP
999.93 TODAYAPP
MONTH ( > DAY ( t
RESPONDENT TIKE < ):< > An > 1 ( )
PH » 2
A-56

-------
cr#»n nan®i NQSPECAP
999.94 NOSPECAP
Should m* call back ...
( )
1.	d«y>
2.	•v*ning>
3.	or M**ktnd?
:-7

-------
Appendix B
CATI FLOW CHART

-------
-CZD
irrnQWCTisv
AJO
«i»im
twin
OTTO IB:
E- P. A. - WARKIGTG LIGHTS STUOY
S
iTt OLLLBUX
^ ao ts n ^
tiC.JO»
uci.il «ra
CZD
-Q « » i )
J urture J
Bit

ffU#
QVWTlO* 11



3 <*61
It MiiC-tfia MtYOI
ui i rakii uw

tvi* «t Tilts fROU (¦

ro

*U*»DL?


cevvuti 1
ITlfMfOri

1
.

1-'
1
-------
E.P.A. - WARNING LIGHTS STUDY
B-2

-------
( " )
B-3

-------
B-4

-------
Appendix C
LIST OF VEHICLES

-------
CONDENSED MAKE/MODEL LIST
MANUFACTURER
MODEL YEARS(S)

MODEL NAME
Chrysler
85,
86,
87,
88

Aries
Chrysler
86




Aries Miser
Chrysler
87,
88



Aries Wagon
Chrysler
85,
86,
87,
88

Caravelle
Chrysler
86




Caravelle Miser
Chrysler
87,
88



Charger
Chrysler
84,
85,
86,
87,
88
Daytona
Chrysler
88




Dynasty
Chrysler
84




E Class/New Yorker
Chrysler
87,
88,
89


Horizon
Chrysler
85,
86,
87,
88

Lancer
Chrysler
87




Lancer Convertibl -
Chrysler
86




Lancer Miser
Chrysler
85




Laser
Chrysler
84,
86,
87


Laser/Daytona
Chrysler
84,
85 ,
86,
87,
88
Lebaron
Chrysler
84,
85,
86,
87,
88
Lebaron Convertible
Chrysler
85,
86,
87,
88

Lebaron GTS
Chrysler
86.




Limousine
Chrysler
85,
86,
87


New Yorker
Chrysler
88




New Yorker Turbo
Chrysler
88




New Yorker
Chrysler
87,
88



Omni
Chrysler
85,
86,
87,
88

Reliant
Chrysler
86




Reliant Miser
Chrysler
85,
86,
87,
88

Reliant Wagon
Chrysler
87,
88



Shadow
Chrysler
87,
88



Sundance
Chrysler
85,
86,
87,
88

Town & Country Wagon
Chrysler
87,
88



Turismo
Chrysler
84,
85,
86,
87,
88
600
Chrysler
84,
85,
86


600 Convertible
Chrysler
86




600 Miser
General Motors
87,
38



Beretta
General Motors
83,
84,
85,
86

Bonneville
General Motors
82




Bonneville Model G
General Motors
82




Bonneville Model G Wagon
General Motors
83




Bonneville Wagon
General Motors
84




Brougham/Devi lie (RWD)
General Motors
83,
84,
85,
86,
87, 88
Caballero Pickup 2WD
General Motors
85,
86,
87


Calais
General Motors
82,
83,
84,
85,
00
o\
00
CO
Camaro
General Motors
86,
87,
88


Caprice
General Motors
83,
84,
85


Caprice Wagon
C-l

-------
General
Motors
81






Catalina Bonneville Wago
General
Motors
81






Catalina Bonneville
General
Motors
83,
84,
85,
86,
87,
88

Cavalier
General
Motors
83 /
84,
85,
86,
87


Cavalier Convertible
General
Motors
83 ,
84,
85,
86,
87,
88

Cavalier Wagon
General
Motors
82,
83 ,
84,
85,
86,
87,
88
Celebrity-
General
Motors
84,
85,
86,
87,
88


Celebrity Wagon
General
Motors
81,
82,
83,
84,
85,
86,

Century


87,
88





General
Motors
84






Century Estate Wagon
General
Motors
81,
85,
86,
87,
88


Century Wagon
General
Motors
81,
82,
83
84,
85


Chevette
General
Motors
86






Chevette CS
General
Motors
83,
84,
85,
86,
87


Cimarron
General
Motors
82,
83





Citation
General
Motors
84,
85





Citation II
General
Motors
87,
88





Corsica
General
Motors
82,
83,
84




Corvette
General
Motors
81.
82,
83
84,
85


Custom Cruiser Wagon
General
Motors
81






Cutlass
General
Motors
88






Cutlass Calais
General
Motors
82,
83 ,
84 ,
85,
86,
87,
88
Cutlass Ciera
General
Motors
82 ,
83,
84,
85,
86,
87,
88
Cutlass Cruiser Wagon
General
Motors
81






Cutlass Supreme Calais
General
Motors
82,
83,
84,
85



Cutlass Supreme
General
Motors
81






Cutlass Wagon
General
Motors
81,
82,
83,
84



Delta 88
General
Motors
85






Delta 88 Royale
General
Motors
86,
88





Deville
General
Motors
81,
82,
83




Devilie Brougham
General
Motors
83,
84,
85,
86,
87,
88

El Camino Pickup 2WD
General
Motors
81,
a ft
82,
83,
84,
85,
86,
87,
Eldorado
General
Motors
0 0
84,
85





Eldorado Convertible
General
Motors
81,
82,
83,
85



Electra -
General
Motors
84






Electra (FWD)
General
Motors
84






Electra (RWD)
General
Motors
84






Electra Estate Wagon
General
Motors
84,
85,
86,
87,
88


Fiero
General
Motors
82,
83,
84,
85,
86


Firebird
General
Motors
88






Firebird Trans Am
General
Motors
82,
83,
84,
85,
86,
87,
88
Firenza
General
Motors
86 .
87,
88




Firenza Cruiser
General
Motors
si i
w v ,
84,
85




Firenza Cruiser Wagon
General
Motors
5 7






Fleetwood Deville
General
Motors
88






Fleetwood
General
Motors
85






Fleetwood Brougham (RWD)
General
Motors
84






Fleetwood Deville (FWD)
General
Motors
85,
86 ,
87,
88



Grand Am
General
Motors
81,
82,
83,
84,
85,
86,
87
Grand Prix
General
Motors
81,
82,
83,
84 ,
85


Impala Caprice
General
Motors
81,
82





Impala Caprice Wagon

-------
Genera
Motors
82





J2000
Genera
Motors
82





J20Q0 Wagon
Genera
Motors
81,
88




Lemans
Genera
Motors
81





Lemans Safari Wagon
Genera
Motors
81,
82,
83, 84,
85


LeSabre
Genera
Motors
81,
82 ,
83, 85



LeSabre Electra Wagon
Genera
Motors
81,
82,
83, 84,
85,
86 ,
87
Limousine
Genera
Motors
82,
83




Malibu
Genera!
Motors
82,
83




Malibu Wagon
Genera
Motors
82,
83 ,
84, 85,
86,
87,
88
Monte Carlo
Genera;
Motors
81,
82,
83, 84



Ninet-Eight
Genera!
Motors
34





Ninety-Eight II
Genera!
Motors
85





Ninety-Eight Regular
Genera!
Motors
82,
83 ,
84



Omega
Genera!
Motors
83,
84 .
35, 86



Parisienne
Genera!
Motors
83,
84 ,
85



Parisienne Wagon
Genera!
Motors
82,
83 ,
84



Phoenix
Genera
Motors
81,
82,
83, 84,
85


Regal
Genera!
Motors
82,
83




Regal Estate Wagon
Genera!
Motors
81,
82,
83, 84,
85


Riviera
Genera!
Motors
85





Riviera Convertible
Genera
Motors
81,
82,
83, 84,
85,
86,

Seville


87,
88





Genera
Motors
82,
83 ,
84, 85,
86,
87,

Skyhawk
Genera
Motors
09
83,
84,
85 ,86,
87,
88

Skyhawk Wagon
Genera
Motors
82,
83 ,
84, 85,
88


Skylark
Genera
Motors
85





Somerset Segal
Genera
Motors
S6,
87




Somerset Lark
Genera
Motors
85,
86,
87, 88



Sunbird
Genera
Motors
85,
86,
87, 88



Sunbird Convertible
Genera
Motors
85,
86,
87, 88



Sunbird Wagon
Genera
Motors
81,
82,
83, 84,
85


Toronado
Genera
Motors
83 ,
84,
85, 86



1000
Genera
Motors
83





2 000
Genera
Motors
83





2000 Convertible
Genera
Motors
84





2000 Sunbird
Genera
Motors
84





2000 Sunbird Convertiole
Genera
Motors
84





2000 Sunbird Wagon
Genera
Motors
83





2000 Wagon
Genera
Motors
82,
83,
84, 85,
86,
87,
88
6000
Genera
Motors
84,
85,
86, 87,
88


6000 Wagon
Toyota

83,
84,
85, 86,
87,
88

Camry
Toyota

87,
88




Camry Wagon
Toyota

93,
84,
85, 86,
37,
88

Celica
Toyota

87,
88




Celica Convertible
Toyota

83 ,
84,
85, 86



Celica Supra
Toyota

84 j
85,
88



Corolla
Toyota-

88





Corolla All-?rac Wagon
Toyota

85,
86,
87



Corolla Sport
Toyota

83,
8 4 ,
85, 86,
87,
88

Cressida
Toyota

83,
84 ,
85, 86,
87


Cressida Wagon
Toyota

85,
86 ,
87, 88



MR2
Toyota

86 ,
87,
88



Supra
C-3

-------
Appendix D
CONFIDENTIALITY PLEDGE

-------
WESTAT, INC.
EMPLOYEE OR CONTRACTOR'S ASSURANCE OF CONFIDENTIALITY OF SURVEY DATA
Statement of Policy
Westat is firmly committed to the principle that the confidentiality of individual data obtained through
Westat surveys must be protected. This principle holds whether or not any specific guarantee of confidentiality
was given at time of interview for self-response), or whether or not there are specific contractual obligations to
the client. When guarantees have been given or contractual obligations regarding confidentiality have been
entered into, they may impose additional requirements which are to oe adherea to strictly.
Procedures for Maintaining Confidentiality
1 All Westat employees and field workers shall sign this assurance of confidentiality. This
assurance may be superseded by another assurance ror a particular project.
2.	Field workers shall keep completely confidential the names of respondents, all information or
opinions collected in the course of interviews, and any information about respondents learned
incidentally during field work. Field workers shall exercise reasonable caution to prevent access
by others to survey data in their possession.
3.	Unless specifically instructed otherwise for a particular project, an employee or field worker,
upon encountering a respondent or information pertaining to a respondent that s/he knows
personally, shall immediately terminate the activity and contact her/his supervisor for
instructions.
4.	Survey data containing personal identifiers in Westat offices shall be kept in a locked container or
a locked room when not being used each working day in routine survey activities. Reasonable
caution shall be exercised in limiting access to survey aata to only those persons who are working
on the specific project and who have been instructed in the applicable confidentiality
requirements for that project.
Where survey data have been determined to be particularly sensitive by the Corporate Officer in
charge of the project or the President of Westat, such survey data shall be kept in locked
containers or in a locked room except when actually being used and attended by a staff member
who has signed this pledge.
5.	Ordinarily, serial numbers shall be assigned to respondents prior to creating a machine-
processible record and identifiers such as name, address, and Social Security number shall not,
ordinarily, be a part of the machine record. When identifiers are part of the machine data
record, Westat's Manager of Data Processing shall be responsible tor determining adequate
confidentiality me assures in consultation with the project director: When a separate file is set up
containing identifiers or linkage information which could be used to identify data records, this
separate file shall be kept locked up when not actually being used each day in routine survey
activities.
6.	When records with identifiers are to be transmitted to another party, such as for keypunching or
key taping, the other party shall be informed of these procedures and shall sign an Assurance of
Confidentiality form.
7.	Each project director shall be responsible for ensuring that ail personnel and contractors involved
in handling survey data on a project are instructed in these procedures throughout the period of
survev performance. When there are specific contractual obligations to the client regarding
confidentiality, the project director shalf develop additional procedures to comply with these
obligations and shall instruct field staff, clerical staff, consultants, and any other persons who
work on the project in these additional procedures. At the end of tne period of survey
performance, the project director shall arrange for proper storage or disposition of sucvey data
including any particular contractual requirements for storage or disposition. When required to
turn over survey data to our clients, we must provide proper safeguards to ensure confiaentialitv
up to the time of delivery.
8.	Project directors shall ensure that survey practices adhere to the provisions of the U.S. Privacy
Act of 1974 with regard to surveys of individuals for the Federal Government. Project directors
must ensure that procedures are established in each survey to inform each respondent of the
authority for the survey, the purpose and use of the survey, the voluntary nature of the survey
(where applicable) and the effects on the respondents, if any, of not responding.
PLEDGE
I hereby certify that I have carefully read and will cooperate fully with the above procedures. I will keep
completely confidential all information arising from surveys concerning individual respondents to which [ sain
access. I will not discuss, disclose, disseminate, or provide access to survey data and identifiers except as
authorized by Westat. In addition, I will comply with any additional procedures established by Westat For a
particular contract. I will devote my best efforts to ensure that tnere is compliance with the required
procedures by personnel whom I supervise. I understand that violation of this pledge is sufficient grounds for
disciplinary action, including dismissal. I also understand that violation of the privacy rights oiindividuals
through such unauthorized discussion, disclosure, dissemination, or access may make me subject to criminal or
civil penalties. I give my personal pledge that I shall abide by this assurance of confidentiality.
Signature	
Date 	

-------