Bulletin Vol. 31, No. 2 : April 2014
Honda Accord collision avoidance features: initial results
This analysis examines insurance loss results for two Honda Accord collision avoidance systems: Forward Collision Warning FCW paired with a Lane Departure Warning LDW system and LaneWatch, a passenger side blind spot in formation system. The combined FCWLDW system is associated with reductions in claim frequency for all 5 coverage types examinedcollision 4 percent, property damage liability 14 percent, bodily injury liability 40 percent, medi cal payments 27 percent, and personal injury protection 11 percent. Only the reductions in property damage, bodily injury liability and medical payment frequencies were statistically significant, which is consistent with FCW function to prevent fronttorear collisions. It is also associated with statistically significant reductions in collision claim severity and overall losses, so it does not suffer the same penalty of earlier FCW systems studied by the Highway Loss Data Institute that were enabled by expensive radar units mounted near the front of the car. In contrast, the Accords combined system uses a camera mounted behind the windshield. LaneWatch also shows reductions in physical damage claim frequencies, but the declines are not significant. There currently are not enough data to produce stable estimates for LaneWatchs ef fect on injury claim frequencies, but two of the three injury coverages indicate reductions. This represents the first HLDI evaluation of the effectiveness of crash avoidance systems on highvolume nonluxury vehicles.
Introduction
Collision avoidance technologies are becoming popular among U.S. passenger vehicles, and more and more auto makers are touting their potential safety benefits. However, the actual benefits in terms of crash reductions still are being measured. This Highway Loss Data Institute HLDI bulletin examines the early insurance claims experience for Honda Accord vehicles fitted with two systems.
Forward Collision Warning uses a camera system located behind the windshield to assess the risk of a collision with leading traffic. The warning system has three driverselectable range settings. When a potential crash is detected, lights flash in the headsup display, the FCW indicator blinks, and there is continuous beeping. The system is active only at speeds more than 10 mph and can be deactivated by the driver. At each ignition cycle, the system defaults to the previous onoff setting. Vehicles with Forward Collision Warning also have Lane Departure Warning.
Lane Departure Warning utilizes the same camera as forward collision warning to also identify traffic lane mark ings. Audio and visual warnings will indicate if the vehicle path deviates from the intended lane. The system is func tional at speeds between 40 and 90 mph but does not warn if the turn signal is on or the movement is determined to be sufficiently sudden as to be evasive. The system can be deactivated by the driver. At each ignition cycle, the system defaults to the previous onoff setting.
LaneWatch is Hondas term for a passengersideonly blind spot monitor. A camera mounted behind the external passenger side rearview mirror monitors the passenger side of the vehicle and displays an 80degree field of view on the consolemounted information screen when the turn signal indicator is activated. Reference lines are also pro vided to indicate proximity. Both the turn signal indicator and reference lines are drivercontrollable settings and can be deactivated. An upcoming navigation system maneuver can also be given priority over the LaneWatch display. LaneWatch can be deactivated by the driver. At each ignition cycle, it will default to the previous onoff setting.
All of the vehicles in this study were equipped with rear cameras. As there are no vehicles without this feature their effectiveness cannot be evaluated in this analysis. The vehicles in this analysis may also have been equipped with optional rear parking sensors. This feature was not controlled for in the analysis as the availability of rear parking sensors on a vehicle was not discernible from the VIN. The Touring trim level of the Accord fourdoor was excluded from the analysis because it is equipped with a different forward collision warning system that uses a radar system instead of a camera and includes adaptive cruise control functionality. Consequently, despite similar FCW function, these systems should be evaluated separately. However, there is too little exposure to produce reliable estimates of effectiveness at this time. The total collision coverage exposure was only 3,138 years.
Method Vehicles
Several trim levels are offered on the vehicles included in this study. Trim levels are bundles of vehicle options such as interior materials, engines and comfort, convenience and safety features. For example, the Honda Accord EXL V6 is equipped with a 6 cylinder motor, leather seats and several collision avoidance technologies. The less expensive LX is equipped with cloth seats, a 4 cylinder motor and no collision avoidance technologies. For the Honda vehicles included in this study the trim levels can be determined in the first 10 positions of the VIN. The collision avoidance features in this study are either standard or not available at the trim level. Consequently, by knowing the trim level the presence of the collision avoidance features is known. LaneWatch and the combination of Forward Collision Warning and Lane Departure Warning are offered as standard equipment on several 2013 Honda Accord models trims. Honda Accords without these features served as the control vehicles in the analysis. Table 1 lists total expo sure, measured in insured vehicle years, and the exposure of each feature as a percentage of total exposure.
Table 1: Feature exposure by vehicle series
Make
Series
Model year range
Forward Collision Warning includes Lane Departure Warning
LaneWatch
Total exposure
Honda
Honda
Honda
Honda
Accord 2dr
Accord 4dr
Crosstour 4dr
Crosstour 4dr 4WD
2013 69 83
2013 40 51
2013 72 78
2013 100 100
15,183
157,309
2,408
1,968
Insurance Data
Automobile insurance covers damages to vehicles and property as well as injuries to people involved in crashes. Different insurance coverages pay for vehicle damage versus injuries, and different coverages may apply depending on who is at fault. The current study is based on property damage liability, collision, bodily injury liability, personal injury protection, and medical payment coverages. Exposure is measured in insured vehicle years. An insured vehicle year is one vehicle insured for 1 year, two vehicles for 6 months, etc.
Because different crash avoidance features may affect different types of insurance coverage, it can be important to understand how coverages vary among the states and how this affects inclusion in the analyses. Collision coverage insures against vehicle damage to an atfault drivers vehicle sustained in a crash with an object or other vehicle; this coverage is common to all 50 states. Property damage liability PDL coverage insures against vehicle damage that atfault drivers cause to other peoples vehicle and property in crashes; this coverage exists in all states except Michi gan, where vehicle damage is covered on a nofault basis each insured vehicle pays for its own damage in a crash, regardless of who is at fault.
HLDI Bulletin Vol 31, No. 2 : April 2014 2
Coverage of injuries is more complex. Bodily injury BI liability coverage insures against medical, hospital, and other expenses for injuries that atfault drivers inflict on occupants of other vehicles or others on the road; although motorists in most states may have BI coverage, this information is analyzed only in states where the atfault driver has first obligation to pay for injuries 33 states with traditional tort insurance systems. Medical payment MedPay coverage, also sold in the 33 states with traditional tort insurance systems, covers injuries to insured drivers and the passengers in their vehicles, but not injuries to people in other vehicles involved in the crash. Seventeen other states employ nofault injury systems personal injury protection coverage, or PIP that pay up to a specified amount for injuries to occupants of involvedinsured vehicles, regardless of who is at fault in a collision. The District of Columbia has a hybrid insurance system for injuries and is excluded from the injury analysis.
Statistical methods
Regression analysis was used to quantify the effect of vehicle feature while controlling for other covariates. The co variates included calendar year, model year, garaging state, vehicle density number of registered vehicles per square mile, rated driver age group, rated driver gender, rated driver marital status, deductible range collision coverage only, and risk. For each safety feature studied, a variable was included.
Claim frequency was modeled using a Poisson distribution, whereas claim severity average loss payment per claim was modeled using a Gamma distribution. Both models used a logarithmic link function. Estimates for overall losses were derived from the claim frequency and claim severity models. Estimates for frequency, severity, and overall losses are presented for collision and property damage liability. For PIP, BI, and MedPay, three frequency estimates are presented. The first frequency is the frequency for all claims, including those that already have been paid and those for which money has been set aside for possible payment in the future, known as claims with reserves. The other two frequencies include only paid claims separated into low and high severity ranges. Note that the percentage of all in jury claims that were paid by the date of analysis varies by coverage: 70.6 percent for PIP, 49.3 percent for BI, and 53.7 percent for MedPay. The low severity range was 1,000 for PIP and MedPay, 5,000 for BI; high severity covered all loss payments greater than that.
A separate regression was performed for each insurance loss measure for a total of 15 regressions 5 coverages x 3 loss measures each. For space reasons, only the estimates for the individual crash avoidance features are shown on the following pages. To illustrate the analyses, however, Appendix A contains full model results for collision claim frequencies. To further simplify the presentation here, the exponent of the parameter estimate was calculated, 1 was subtracted, and the resultant multiplied by 100. The resulting number corresponds to the effect of the feature on that loss measure. For example, the estimate of the effect of Forward Collision Warning including Lane Departure Warn ing on PDL claim frequency was 0.15083; thus, vehicles with the feature had 14.0 percent fewer PDL claims than expected exp0.15083110014.0.
HLDI Bulletin Vol 31, No. 2 : April 2014 3
Results
Results for Honda Accords Forward Collision Warning System including Lane Departure Warning are summarized in Table 2. The lower and upper bounds represent the 95 percent confidence limits for the estimates. For vehicle damage losses, frequency and severity of claims as well as overall losses are down. The reductions are significant indicated in blue in the table, with the exception of the decrease in collision claim frequency and property damage liability claim severity.
For the injury related coverage types, all measures of frequency for all coverage types show a reduction. The bodily injury liability and medical payment reductions are significant.
Table 2: Change in insurance losses for Forward Collision Warning and Lane Departure Warning
FREQUENCY
Upper bound
Lower bound
SEVERITY
Upper bound
Lower bound
OVERALL LOSSES
3.8
1.8
668
409
132
71
45
14.0
5.4
418
169
104
26
17
Lower Vehicle damage coverage type bound
Upper bound
Collision 9.1 17
Property damage liability
Injury coverage type
Bodily injury liability
Medical payments
Personal injury protection
21.8
Lower bound
57.7
45.0
27.8
6
Upper bound
13.7 14.4 12.1
FREQUENCY
Upper bound
Lower bound
LOW SEVERITY FREQUENCY
Upper bound
Lower bound
HIGH SEVERITY FREQUENCY
39.5
13.5
66.3
31.8
38.1
75.5
47.3
27.3
4.0
65.0
21.7
74.9
52.2
26.0
10.7
10.3
41.1
5.9
50.4
37.5
16.3
Results for Honda Accords LaneWatch System are summarized in Table 3. Again, the lower and upper bounds repre sent the 95 percent confidence limits for the estimates. Reductions in claim frequency are estimated for both first and thirdparty vehicle damage coverages, yet resulting in somewhat higher claim severity. Loss per insured vehicle year overall losses declined under both property damage liability and collision coverage. However, none of the estimated effects for LaneWatch on collision or PDL losses are statistically significant.
Under injury coverages, the frequency of claims is lower for both MedPay and PIP, but not for BI, and none of the differences is statistically significant. Among paid claims, no clear pattern emerges.
Table 3: Change in insurance losses for LaneWatch
FREQUENCY
Upper bound
Lower bound
SEVERITY
Upper bound
Lower bound
OVERALL LOSSES
2.5
3.1
190
99
409
30
1
7.8
1.1
162
102
392
14
4
Lower Vehicle damage coverage type bound
Upper bound
Collision 7.8 31
Property damage liability
Injury coverage type
Bodily injury liability
Medical payments
Personal injury protection
16.0
Lower bound
22.9
32.0
31.3
8
Upper bound
145.4 29.4 16.7
FREQUENCY
Upper bound
Lower bound
LOW SEVERITY FREQUENCY
Upper bound
Lower bound
HIGH SEVERITY FREQUENCY
7.9
51.0
56.4
15.7
63.1
41.8
19.5
11.1
16.1
58.4
10.0
95.0
44.1
14.9
15.8
3.3
34.2
4.2
65.1
33.5
11.9
HLDI Bulletin Vol 31, No. 2 : April 2014
4
Discussion
This is the first analysis of the effectiveness of crash avoidance features involving a highvolume nonluxury model. For example, in 2013 the Accord outsold the most popular Volvo model Volvo S60 10 to 1. It is also less expensive than midsize cars in the luxury class; 23,270 for the fourdoor Accord compared with a range of 32,64562,495 for midsize luxury cars. As such, results of this analysis may give a better estimate of the potential crash reductions associated with these features than analyses examining only luxury class vehicles.
The results for these two Honda Accord collision avoidance features Forward Collision Warning with Lane De parture Warning and LaneWatch are encouraging. The combined FCWLDW systems show a reduction in claim frequencies across all coverages. The pattern of findings for vehicle damage coverages is consistent with the expected benefits; that is, the reduction in claims is greater for property damage liability coverage than for collision coverage. Forward Collision Warning operates in following traffic and is intended to reduce the occurrence andor severity of fronttorear collisions. These types of crashes are more common among property damage liability claims than among collision claims, as the latter often include singlevehicle collisions. The forward collision warning system does not have autonomous braking, yet the significant 14 percent reduction in property damage liability claim fre quency is at least as large as HLDIs earlier estimates for autobraking systems from Acura and MercedesBenz as well as the lowspeed autobraking system, City Safety, from Volvo HLDI, 2012a, 2012b, 2013. All prior forward collision warning systems evaluated by HLDI have been bundled with adaptive cruise control ACC. The Honda study ve hicles are not equipped with ACC. While the effects of ACC could not be isolated in prior analysis, if used by drivers ACC could reduce the likelihood that drivers get into situations that lead to a crash. Given that the Honda forward collision warning system is not bundled with ACC and the favorable loss results for the Honda system, if ACC did contribute to lower claim rates for previously studied systems where FCW and ACC was bundled then it may indicate that the FCW system from Honda is even more effective than previously studied systems.
It is also interesting to note that the current estimates for the effectiveness of this combined FCWLDW system are greater than the estimates for a Volvo combined FCWLDW system that included autobraking HLDI, 2012c. At the time, it was hypothesized that the presence of LDW was decreasing the effectiveness of the Volvo combined system. This was based on estimates of claim frequency increases associated with LDW presence on Buick and Mercedes Benz vehicles compared with estimates for Acura and MercedesBenz FCW with autobrake systems which did not include LDW. In the context of the current results, LDW may not be as deleterious as previously hypothesized. This would be consistent with the significant reduction in property damage liability claim frequency, another effect not observed for the Volvo combined system. Alternately, the forward collision warning system on the Accord is much more effective than those previously studied. It may be the case that these warning systems are more effective for drivers of mainstream cars than drivers of luxury models.
Hondas forward collision warning system is camerabased and unlike the previously evaluated systems from other manufacturers that use a radarbased system, typically mounted in the vulnerable front grille of the vehicle. Analysis of MercedesBenz and Volvo forward collision systems showed increases in collision claim severity, likely associated with replacement of the radar units in crashes not avoided HLDI, 2012a, 2012b. This analysis of Honda Accords showed a significant decrease in collision claim severity, which may be attributed to the better protected interior location of the camera.
Effects of Hondas LaneWatch, a passengerside blind spot detection system, although not statistically significant, are patterned as expected. Incursion into an occupied adjacent lane would be expected to result in a twovehicle crash that would lead to a property damage liability claim against the encroaching driver. Although neither estimate is statistically significant, the estimated reduction in property damage liability claims is much larger than the reduc tion estimated for collision claims. This is consistent with the fact that the reductions in collision claims from such crashes would be diluted by the many singlevehicle crashes that result in collision claims and are unaffected by the LaneWatch system.
HLDI Bulletin Vol 31, No. 2 : April 2014 5
Limitations
There are limitations to the data used in this analysis. At the time of a crash, the status of a feature is not known. The features in this study can be deactivated by the driver, and there is no way to know how many of the drivers in these vehicles turned off a system prior to the crash. However, surveys conducted by the Insurance Institute for Highway Safety indicate that large majorities of drivers with these types of systems leave them on. If a significant number of drivers do turn these features off, any reported reductions may actually be underestimates of the true effectiveness of these systems.
Additionally, the data supplied to HLDI does not include detailed crash information. Information on point of impact and the vehicles transmission status is not available. The technologies in this report target certain crash types. For example, LaneWatch is designed to prevent sideswipetype collisions. All collisions, regardless of the ability of a fea ture to mitigate or prevent the crash, are included in the analysis.
References
Highway Loss Data Institute. 2012a. MercedesBenz collision avoidance features: initial results. Loss Bulletin Vol. 29, No 7. Arlington, VA.
Highway Loss Data Institute. 2012b. Volvo City Safety loss experience: an update. Loss Bulletin Vol. 29, No. 23. Ar lington, VA.
Highway Loss Data Institute. 2012c. Volvo collision avoidance features: initial results. Loss Bulletin Vol. 29, No. 5. Arlington, VA.
Highway Loss Data Institute. 2013. Acura collision avoidance features: an update. Loss Bulletin Vol. 30, No. 15. Ar lington, VA.
HLDI Bulletin Vol 31, No. 2 : April 2014 6
Appendix A
Appendix A: Illustrative regression results collision frequency
Parameter
Degrees of freedom
Estimate
Effect
Standard error
Wald 95 confidence limits
Chisquare
Pvalue
Intercept
Calendar year
Vehicle model year and series
Rated driver age group
Rated driver gender
Rated driver marital status
Risk
State
2012
2013
2013 Accord 2dr
2013 Accord 4dr
2013 Crosstour 4dr 2WD
2013 Crosstour 4dr 4WD
1420
2124
2539
65
Unknown
4064
Male
Unknown
Female
Single
Unknown
Married
Nonstandard
Standard
Alabama
Arizona
Arkansas
California
Colorado
Connecticut
Delaware
District of Columbia
Florida
Georgia
Hawaii
Idaho
Illinois
Indiana
Iowa
Kansas
Kentucky
Louisiana
Maine
1 8.9726 0.5086 9.9694 7.9758 311.23
1 0.4699 37.5 0.0485 0.5650 0.3748 93.85
0.0001
0.0001
0.0071
0.2375
0.6883
0.0001
0.0001
0.0001
0.0001
0.0010
0.3136
0.0297
0.0001
0.0021
0.0001
0.7368
0.9079
0.9115
0.6135
0.9375
0.7789
0.8346
0.4301
0.6342
0.7805
0.5124
0.3333
0.9514
0.6692
0.7974
0.9755
0.5159
0.6561
0.6907
000000
1 0.2560 29.2 0.0951 0.0695 0.4425
1 0.1087 11.5 0.0920 0.0716 0.2890
1 0.0486 5.0 0.1211 0.1888 0.2860
000000
7.24
1.40
0.16
45.92
68.91
63.12
25.09
10.83
1.02
1 0.3899
1 0.3289
1 0.1733
1 0.1229
1 0.1444
47.7 0.0575
38.9 0.0396
18.9 0.0218
13.1 0.0245
15.5 0.0439
0.2771 0.5026
0.2512 0.4065
0.1306 0.2161
0.0748 0.1710
0.0584 0.2304
000000
1 0.0202 2.0 0.0201 0.0595 0.0191
1 0.1513 14.0 0.0696 0.2877 0.0149 4.72
000000
1 0.2342 26.4 0.0219 0.1913 0.2771 114.50
1 0.2122 23.6 0.0691 0.0768 0.3475 9.44
000000
1 0.1819 19.9 0.0362 0.1111 0.2528 25.32
000000
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
0.1702 15.7
0.0584 5.7
0.0571 5.9
0.2529 28.8
0.0399 3.9
0.1421 13.2
0.1086 10.3
0.4169 51.7
0.2386 21.2
0.1401 13.1
0.3382 40.2
0.5532 42.5
0.0307 3.0
0.2169 19.5
0.1347 12.6
0.0158 1.6
0.3336 28.4
0.2241 25.1
0.5065 1.1630
0.5049 1.0480
0.5135 0.9493
0.5007 0.7285
0.5093 1.0381
0.5061 1.1340
0.5199 1.1277
0.5283 0.6186
0.5015 1.2214
0.5029 1.1258
0.5162 0.6736
0.5718 1.6739
0.5025 1.0156
0.5077 1.2120
0.5245 1.1626
0.5149 0.9933
0.5135 1.3401
0.5033 0.7623
0.5529 1.3036
0.8225 0.11
0.9312 0.01
1.0634 0.01
1.2344 0.26
0.9583 0.01
0.8499 0.08
0.9105 0.04
1.4523 0.62
0.7442 0.23
0.8456 0.08
1.3500 0.43
0.5675 0.94
0.9543 0.00
0.7781 0.18
0.8933 0.07
1.0249 0.00
0.6729 0.42
1.2105 0.20
0.8636 0.16
0.2200 19.7
HLDI Bulletin Vol 31, No. 2 : April 2014
7
Appendix A: Illustrative regression results collision frequency
Parameter
Degrees of freedom
Estimate
Effect
Standard error
Wald 95 confidence limits
Chisquare
Pvalue
Maryland
Massachusetts
Michigan
Minnesota
Mississippi
Missouri
Montana
Nebraska
Nevada
New Hampshire
New Jersey
New Mexico
New York
North Carolina
North Dakota
Ohio
Oklahoma
Oregon
Pennsylvania
Rhode Island
South Carolina
South Dakota
Tennessee
Texas
Utah
Vermont
Virginia
Washington
West Virginia
Wisconsin
Wyoming
Alaska
0250
251500
1001
5011000
099
100499
500
1 0.0668 6.9
1 0.0739 7.7
1 0.2745 31.6
1 0.1709 15.7
1 0.0621 6.0
1 0.1909 17.4
1 0.8978 59.3
1 0.1166 11.0
1 0.2760 24.1
1 0.1079 10.2
1 0.0414 4.1
1 0.0671 6.5
1 0.2583 29.5
1 0.3855 32.0
1 0.3347 28.4
1 0.1924 17.5
1 0.0427 4.4
1 0.1113 11.8
1 0.0099 1.0
1 0.0155 1.6
1 0.2156 19.4
1 0.4461 36.0
1 0.1267 11.9
1 0.0318 3.1
1 0.1737 15.9
1 0.1491 13.9
1 0.0211 2.1
1 0.0784 8.2
1 0.4545 36.5
1 0.0208 2.1
1 0.3746 31.2
0.5025
0.5073
0.5062
0.5097
0.5104
0.5083
0.6712
0.5288
0.5160
0.5177
0.5016
0.5229
0.5009
0.5032
0.6013
0.5026
0.5090
0.5095
0.5021
0.5161
0.5053
0.6124
0.5051
0.5012
0.5206
0.5628
0.5023
0.5051
0.5359
0.5085
0.6712
0.9180 1.0516 0.02
0.9205 1.0682 0.02
0.7176 1.2667 0.29
1.1699 0.8281 0.11
1.0625 0.9383 0.01
1.1873 0.8054 0.14
2.2133 0.4177 1.79
1.1531 0.9199 0.05
1.2874 0.7354 0.29
1.1226 0.9068 0.04
1.0244 0.9417 0.01
1.0918 0.9577 0.02
0.7235 1.2401 0.27
1.3718 0.6008 0.59
1.5133 0.8439 0.31
1.1775 0.7927 0.15
0.9550 1.0404 0.01
0.8873 1.1098 0.05
0.9742 0.9939 0.00
0.9960 1.0270 0.00
1.2060 0.7749 0.18
1.6464 0.7542 0.53
1.1166 0.8632 0.06
1.0142 0.9505 0.00
1.1941 0.8467 0.11
1.2523 0.9540 0.07
1.0056 0.9634 0.00
0.9115 1.0684 0.02
1.5048 0.5958 0.72
1.0175 0.9759 0.00
1.6901 0.9408 0.31
0.8943
0.8842
0.5876
0.7374
0.9032
0.7072
0.1810
0.8255
0.5928
0.8350
0.9343
0.8979
0.6061
0.4437
0.5778
0.7019
0.9331
0.8271
0.9843
0.9761
0.6697
0.4663
0.8020
0.9493
0.7386
0.7910
0.9665
0.8766
0.3963
0.9674
0.5767
0.0001
0.6592
0.0001
0.0001
0.0001
0.1782
0.3720
000000
Deductible range
Registered vehicle density
Forward Collision Warning Lane Departure Warning
LaneWatch
1 0.4765 61.0 0.0315 0.4148
1 0.0776 7.5 0.1759 0.4224
1 0.3064 35.9 0.0274 0.2528
0.5382 229.20
0.2673 0.19
0.3600 125.43
000000
1 0.2548 22.5 0.0339 0.3213 0.1883 56.38
1 0.1506 14.0 0.0219 0.1936 0.1076 47.11
000000
1
1
0.0388
0.0254
3.8
2.5
0.0288
0.0285
0.0954
0.0812
0.0177 1.81
0.0304 0.80
HLDI Bulletin Vol 31, No. 2 : April 2014
8