By Daniel Stern, DVN General Editor
LEDs give full intensity immediately they’re powered, while incandescent bulbs take a relatively sluggish 250 milliseconds to reach their full intensity. Therefore, when the equipped car’s driver steps on the brake, LED stop lights are better because they give following drivers an earlier warning and more time to react. So goes the theory, and it is the generally accepted wisdom on the subject.
Notwithstanding any obvious points that might spoil a comparison—something like a minimally-compliant LED stop light versus a maximally-compliant incandescent stop light—it’s difficult to imagine a reason why the theory wouldn’t hold good in practice. That’s one reason why a new study just released by the US NHTSA is very interesting and well worth reading: Effectiveness of LED Stop Lamps for Reducing Rear-End Crashes: Analyses of State Crash Data (pdf) examined crash involvement of various vehicle models that had switched, one year to the next, from incandescent to LED main and/or central stop lights or vice versa. The verdict? No clear benefit to LEDs! Such a result might be hard to accept and easy to dismiss, but let’s look carefully at exactly what it means. The study’s abstract says, inter alia :
 |
Overall, the analysis does not support a firm conclusion about whether LED stop lamps and LED CHMSL are more effective than incandescent lamps. The main analysis shows a significant overall 3.6% reduction in rear-impact crashes with LED. On the other hand, a non-parametric analysis not only fails to show improvement in significantly more than half the models, but actually shows an increase in rear impacts with LED for 9 of the 17 make-models that switched to LED. It was just the favorable results for high-sales vehicles (…) that pulled the overall result into the plus. |  |
So it’s not necessarily that LED stop lights aren’t better, or that they are worse, it’s that the cars scrutinised in this study could not demonstrate a substantial significant safety benefit to LED stop lights. How come? Again from the abstract: