DADRL - Comments on Tessmer 2004

Comments on Tessmer DRL Study

In the Fall of 2004 NHTSA released a study suggesting possible benefits of daytime running lights (DRLs). The study was performed by Joseph Tessmer of NHTSA, and the text will be referred to as "Tessmer 2004." This study falls well short of validating any effectiveness of DRLs. In particular the results from Tessmer 2004 are inconclusive and even contradictory, and thus do not support mandating DRLs in the United states at this time. Because DRL effectiveness cannot be positively established some relief from the glare problems is urgently needed.

First off, in Tessmer's powerpoint presentation, which was released prior to his paper, he acknowledges that the findings for two vehicle collisions were not statistically significant by his definition of a desired p-value (p<0.05), yet he still concluded that the Simple Odds analysis adequately demonstrates that DRLs are effective at crash reduction. This claim was also made in the report that was released this past Fall.

As Elvik noted in - his 2003 intermediate report, Simple Odds values respond to other factors besides DRLs, and thus a stronger signal from Simple Odds might be obtained. The signal, however, may not be particularly accurate in determining the effect of DRLs. The discussion that Elvik et.al. conduct under section 5.1, "Choice of estimator of the effects of DRL" (pp75-79) is particularly to the point.

Tessmer wrote that his reviewers asked him to also perform an Odds Ratio analysis to confirm the findings of Simple Odds. The justification for looking at the Odds Ratio results is to better isolate the effects of DRLs. However, Tessmer decided not to give equal coverage to the Odds Ratio analysis, omitted mention of the adverse finding in his conclusions, and relegated the treatment to an appendix.

That the Odds Ratio method did not back up the results of Simple Odds should be of concern to NHTSA. The matter that the p-value was somewhat higher for the Odds Ratio analysis than for Simple Odds should not be the decisive factor in establishing DRL effectiveness. As Phillips and Goodman write in Epidemiol Perspect Innov. 2004; 1: 3., "Researchers still frequently present results as if statistical significance and p-values are useful decision criteria, and decision makers are left with inadequate information."

One very important point regarding the input data is that it does not adequately distinguish between lit and unlit vehicles. While Tessmer separates the vehicle types based on VIN, it is not uncommon to find one out of twenty (5%) or more drivers who have manually activated their headlights on their non-DRL vehicles during clear, bright, conditions. On overcast days this percentage can be much higher. For example, on a recent winter trip to Minnesota, I estimated that roughly 50% of drivers of non-DRL vehicles were using headlamps under overcast conditions, even though the visibility at ground level was in excess of a mile. Because Tessmer's input data also included dusk and dawn conditions when many drivers will have manually activated their headlamps, the inaccuracy in the vehicle count grows considerably. How can a non-DRL vehicle (a Ford in some of Tessmer's data) which is involved in a collision be properly counted if the driver has activated its headlamps? Without a proper accounting for this factor, any analysis becomes an exercise in futility.

Another troubling statistic that has not been addressed by NHTSA is the sharp increase in motorcycle fatalities in the US. Given that there has been a similar increase in DRL usage over the same period, this should have created an alarming wake up call within NHTSA. Motorcyclists have realized the negative effects of automobile DRLs on their conspicuity and have taken action by upping the ante --- many cyclists now ride with their full intensity high beams on all the time. Not only do we have a problem with automobile DRLs, we're facing a growing problem with motorcyclists attempting to compete with the sea of automobile headlamps on our streets and highways.

Japan's decision not to implement automobile DRLs because of the DRLs' masking effects upon motorcycle conspicuity is of major significance. Japan's experiments on the masking effects are particularly telling (TRANS-WP29-GRE-51-10e). Specifically, NHTSA's proposed DRL intensity reductions from 1998 do not even come close to what the report recommended for adequate motorcycle conspicuity. While 1500 cd will likely still allow low beam headlamp implementations, Japan's study seems to indicate that something more along the lines of our parking lamp standards is the appropriate intensity for DRLs.

At this point, NHTSA cannot proceed with a cost effectiveness study regarding mandatory DRLs because the two internationally adopted measures of effectiveness contradict each other by showing respectively an increase and reduction of accidents from DRLs.

Finally, we've noticed a recent trend away from DRLs. While shopping for a new truck the author of this web page found that GM is now equipping some of its vehicles with a DRL interrupt on its light switch. This move by DRLs' biggest proponent should speak volumes regarding public acceptance. Add to this Toyota's dropping of hardwired DRLs as standard equipment on many of its vehicles. It's becoming obvious that Americans are not accepting DRLs and want the return of the control of their vehicle's lighting. A decade after the introduction of DRLs in the US, Ford and Daimler-Chrysler only offer DRLs on fleet type vehicles, and given that these large corporations are concerned with selling vehicles and not alienating customers, it should be another indication that DRLs are not doing their intended job and are not being accepted by American motorists.

What sort of action would we like to see this Spring? For starters, let's get the DRL intensity down where it needs to be: in the range of 250-650 cd. There are still far too many high beam (e.g., GM and BMW) and low beam implementations (several) causing far too much glare and masking vulnerable road users. Even the dedicated DRLs on the new Chevrolet light trucks are far too bright and cause masking problems with its turn signals. Second, any consideration to mandating DRLs should be dropped. Until a better understanding of DRL effectiveness is available, including secondary effects, proclaiming a mandate is premature. Third, in light of Japan's analysis of DRL masking effects, NHTSA might even discourage daytime headlight usage and recommend parking lamp usage for those who feel the need to illuminate their vehicles during the day.