Index | Table of contents | Technical Report | Executive Summary | Background | Rear Window Defrogger and Wiper Information | State Crash Data | Analysis Databases | Analysis Method | Hypotheses | Conclusions | Appendix A
Analysis Databases
The analysis databases are combinations of the rear-window defogger file and the State crash files. Linking the rear-window defogger file to the State crash files created the analysis files. The files were linked by matching the vehicle make, model, and model year for cars as defined in Ward’s. In Florida, Maryland, and Michigan State data files, vehicle make and model was decoded from the Vehicle Identification Number (VIN); model year was not decoded from the VIN, but taken directly from the variable available on the State data files. In Pennsylvania for calendar years 1988-2000, make and model was decoded from the VIN; model year was not decoded from the VIN, but taken directly from the variable available on the State data file. For calendar years 1980-1987, very few VINs were included on the Pennsylvania data files, so make, model and model year were taken directly from the variables available on the State data files. The resulting analysis databases are vehicle-oriented files, with one record for each car that was involved in a crash. Since the rear-window defogger file only contains cars with 1973-2001 model years, the analysis files also only contain cars with 1973-2001 model years. The Michigan analysis file only has 1973-1992 model years and Florida only has 1973-2000 model years.
We believe rear window defoggers may have the potential to reduce crashes in which a driver was backing up or changing lanes before the crash. These tasks can be facilitated by vision through the rear window. Only cars with rear damage were included in the analysis. Consequently, rear impacts in which the vehicle maneuver was backing up or changing lanes were considered the only crashes that should be reduced as a result of rear window defoggers (i.e., treatment group). Vehicles with rear damage that were stopped prior to the crash were considered the control group. Since these vehicles were not moving at all and did nothing to cause the crashes, we can be relatively certain that rear-window defoggers would not have prevented them. Therefore, the critical parameters besides the presence or absence of rear window defoggers that must be defined in each State file are rear damage and vehicle maneuver (backing up and changing lanes vs. stopped).
Every State has its own unique ways of coding a vehicle’s impact location. “Rear damage” cannot be defined exactly the same in each State, but at least the definitions can be made as similar as possible. Rear damage was defined to include any impact into the rear portion of the car where the driver might at least have had a chance to see the other vehicle behind it, either indirectly (via rear view mirror) or directly through the rear window. This includes rear-corner impacts and in some cases even impacts resulting in damage to the back portion of the side of the car, because sometimes when a car changes lanes into the path of another vehicle traveling in the same direction damage to the side of the car is possible. Given this extended definition of “rear damage,” the following cars were included in the analysis:
State |
Definition |
Percent of Cars* that have Rear Damage |
|---|---|---|
|
|
|
* 1973 – 2001 model year domestic cars with rear defogger information.
Every State also has its own unique ways of coding “backing up,” “changing lanes,” and “stopped” maneuvers. Florida, Maryland, and Pennsylvania have a variable called Veh_man1 that indicates the vehicle’s maneuver just prior to impact. Michigan has a “driver intent” variable that indicates the driver’s intended maneuver just prior to impact. The following shows the definition for “backing up,” “changing lanes” and “stopped” by State:
State |
Maneuver |
Definition |
|---|---|---|
|
Changing Lanes Changing lanes |
|
*Includes some vehicles that were slowing rather than stopped. Both “slowing” and “stopped” vehicles may be considered part of the control group, since vision through the rear window is not an issue.
If rear window defoggers were effective, they would reduce crashes in which a driver is backing up and changing lanes – relative to the number that would have been expected if the cars had not been equipped with rear window defoggers. T here is no direct way to count the crashes that were prevented, nor is there any way to determine if the rear window defogger was activated during the pre-crash maneuver. The basic approach was to study the change in proportion of crashes that were relevant, assuming that the presence or absence of defogger does not affect the occurrences of non-relevant crashes. The analytical methodology chosen for this study controls for some demographic characteristics of the drivers along with environmental and vehicle factors.
Ideally, only make-models that did not offer rear window defoggers as optional equipment in an earlier model year but offered them as standard equipment on the next model year should be analyzed. But there were no make-models where this happened. Another ideal group to analyze would have been make-models that had a big increase in the percentage of defoggers in consecutive model years. But there were very few make-models that had an increase of 50 percentage points or more and only about 30 make-models had an increase of 20 percentage points or more in consecutive model years.
In order to have a large sample of cars, the sample included make-models with a large volume of crashes: make-models that had at least 500 cases in Pennsylvania where the cars were changing lanes, backing up or stopped just prior to impact and had rear damage. Pennsylvania was used because there were 20 calendar years of data. This ensured the sample of vehicles would include a wide variety of model years that have various percentages of defoggers. If Florida or Maryland had been used then the sample would have been mostly more recent model years that have a high percentage of cars with defoggers. If Michigan had been used then the sample would have been mostly earlier model years that had very few make-models with defoggers.
But since only 1981-1991 CY of data were available in Michigan, the mix of cars with a large volume of crashes was different in Michigan than in Pennsylvania, so Michigan’s database included 34 additional make-models not included in the other States’ databases. These additional make-models had at least 750 crashes in Michigan. Appendix A contains a list of the make-models included in the analysis. The list includes for each make-model the percentage of rear window defoggers by model year and the number of crashes in each State. Appendix A also contains a list of additional make-models included in Michigan’s database.
Table 8 shows the percentage of defoggers by model year for six make-models. Some make-models that met the above criteria were excluded from the analysis because the percentage of defoggers jumped all over the place over the span of several model years. The fluctuating percentages are unusual, since for most cars the percentage gradually increases by model year. Intuitively, it appears that some of the reported percentages could be erroneous, but without another source of percentages it is hard to know. So these make-models were excluded from the analysis. For example, the Chevrolet Camaro was excluded from the analysis for this reason (see Table 8). From model year 1982 to 1987, about 60 percent of Camaros had rear defoggers. Then the percentage fluctuated between model year 1988 and 1997. During that time, it dropped into the 50’s; plummeted to a low of 23.4 percent in model year 1990; rose to the 56.7, 62.3, and 78.4; and dropped again into the 60’s before it was consistently in the 70’s.
Table 8
Percentage Of Rear Window Defoggers By Model Year For Some Make-Models
Model Year |
Ford Escort (1981-90 design) |
Chevrolet Caprice (1977-96 design |
Ford Crown Victoria (1979 - design) |
Ford Taurus (1986-95 design) |
Chevrolet Citation |
Chevrolet Camaro (1982 - design) |
|---|---|---|---|---|---|---|
1977 |
. |
32.0% |
. |
. |
. |
. |
1978 |
. |
34.5% |
. |
. |
. |
. |
1979 |
. |
40.6% |
38.5% |
. |
43.7% |
. |
1980 |
. |
45.8% |
39.1% |
. |
45.8% |
. |
1981 |
36.5% |
45.9% |
61.2% |
. |
45.0% |
. |
1982 |
38.3% |
50.1% |
64.5% |
. |
47.2% |
61.8% |
1983 |
35.9% |
57.0% |
75.3% |
. |
53.4% |
63.6% |
1984 |
35.8% |
60.7% |
63.3% |
. |
63.6% |
65.8% |
1985 |
42.2% |
63.5% |
80.1% |
. |
61.1% |
66.3% |
1986 |
60.0% |
70.5% |
80.9% |
91.1% |
. |
68.9% |
1987 |
67.5% |
76.7% |
86.7% |
89.2% |
. |
62.5% |
1988 |
80.6% |
72.6% |
89.3% |
93.2% |
. |
55.0% |
1989 |
86.5% |
77.1% |
87.5% |
89.7% |
. |
53.8% |
1990 |
88.0% |
39.4% |
89.4% |
91.5% |
. |
23.4% |
1991 |
. |
81.9% |
94.0% |
100% |
. |
56.7% |
1992 |
. |
87.6% |
95.3% |
99.6% |
. |
62.3% |
1993 |
. |
88.8% |
97.1% |
100% |
. |
78.4% |
1994 |
. |
91.7% |
97.7% |
100% |
. |
69.7% |
1995 |
. |
88.3% |
100% |
100% |
. |
63.6% |
1996 |
. |
98.2% |
100% |
. |
. |
65.4% |
1997 |
. |
. |
100% |
. |
. |
64.0% |
1998 |
. |
. |
100% |
. |
. |
78.8% |
1999 |
. |
. |
100% |
. |
. |
73.9% |
2000 |
. |
. |
100% |
. |
. |
73.2% |
2001 |
. |
. |
100% |
. |
. |
89.1% |
Increase |
51.5 |
66.2 |
61.5 |
8.9 |
17.4 |
27.3 |
Model years where the percentages were much higher or lower than expected relative to previous and subsequent percentages were excluded from the analysis. These data may be correct or erroneous, but since there is no way to know, they were excluded from the analysis. Model year 1990 for the Chevrolet Caprice was excluded from the analysis for this reason (see Table 8). The percentage of defoggers steadily increased from 32.0 percent in 1978 to 98.2 percent in 1996 except for a severe drop to 39.4 percent in 1990.
Table 9 shows the number of cases in the analysis databases by crash type and State. At this point, the analyses were limited to Florida and Michigan because they have larger samples of total cases and far more changing lane and backing cases. Maryland has fewer than 75,000 total cases. Although Pennsylvania has almost 175,000 total cases, only 4 percent are changing lane and backing cases.
Table 9
Number Of Cases in Analysis Databases By Crash Type And State
State |
Changing Lane & Backing Crashes |
Stopped Crashes |
Total |
|---|---|---|---|
Michigan |
71,668 |
250,158 |
321,826 |
Florida |
44,761 |
276,357 |
321,118 |
Pennsylvania |
7,791 |
174,669 |
182,460 |
Maryland |
13,121 |
61,604 |
74,725 |