Home - About Us - Riding Safer - Contact Us - Blog - Disclaimer - Links - Sitemap
Risk Hierarchy: Information - Rider Ed - Driver Ed - Conspicuity - Bike Defect - Ultra-Defensive Riding - Crash Avoidance - Injury Mitigation - Crash Scene

book   Finally: a new Motorcycle Crash Causation Study in the USA.

We have no solid information on the causes of motorcycle crashes in the USA. The current study on motorcycle crashes, the Hurt Report, is based on 30 year old data. We do have the numbers, though. A lot of death, and too much pain.

Old HarleyA new study is in the works for the US. It was authorized in section 5511 of the Safe, Accountable, Flexible, Efficient Transportation Equity Act: A Legacy for Users (SAFETEA-LU), in October 2009, following up on legislation from 2005. It is being performed by the Oklahoma State University Transportation Center in the School of Engineering at Stillwater. It is mandated to use the same OECD methodology as the European Maids study. Here's the October 5th Federal Highway Administration press release

The new study is already in trouble. October 13th the LA Times revealed that the study had been scaled back severely. The original 900 crashes to be studied have been pared down to 300. Please see our blog postings on the issue, including a funding analysis. At the time of writing the situation is critical, the stakeholders are in disagreement and we plan to follow the issue closely in the blog, which you should check for the current details.

Bikesafer.com's OSU Crash Causation Study Page

We've created a special section of the site to follow the OSU Crash Causation Study. This badly-needed study is in crisis. We have full information here. We also cover the recently-announced MSF/Virginia Tech 'Naturalistic' study. We had a petition campaign, which we withdrew after the MSF announcement that they have withdrawn their support of the OSU study and are now concentrating on the Virginia study.

For now, what we have to work with is a very old (1981) US-based study, the Hurt Report, and some information from abroad. We dug around, talked to some experts and have some additional commentary. We'll have to wait four years for data from the new study. We try to pull it together, but we'd much rather base important riding and policy decisions on better data.

Why we need this big and expensive new study

You'd think it was self-evident that good decisions are based on good facts, and good facts come from good science, and that's not cheap, but you often hear sentiments to the contrary.

There are various ways of doing scientific research. The classic method is to think up a theory and design a set of experiments to confirm or disprove various aspects of this theory. If a theory holds up to experimentation, and the best available minds review the results, then it's probably a good theory. Every day, in millions of ways, we all live by this process. Many of the products we use daily are tested destructively, which is a common feature of scientific testing.

When talking about motorcycle crash causation, we have a problem. We can't deliberately cause an accident, by definition. But we can do experiments around limited features of bike riding, like this one about bike brakes . This is a very good small piece of scientific research, and the findings are interesting as it contradicts what we thought we knew about motorcycle deceleration forces under braking. It's a useful tool for thinking about riding but doesn't directly address the causes of crashes (see this blog post).

Much motorcycle accident causation research is based on statistics gathered from official sources - accident reports, court appearance records, death certificates and the like. This research is often unreliable, as concluded by this paper from Monash University. A big part of the reason for this is that the reports are done on forms which might be different in each jurisdiction, by police personnel who are trained differently in each city. There is very little quality or consistency in the data collection, and often very few data points are collected on motorcycle crashes in accident reports. You can't take valid conclusions from bad data. This is so far from the scientific method that it is just not science.

There is a middle way, taken from drug research and epidemiology, which is the study of how diseases spread. These researchers can't ethically do standard scientific experiments or destructive testing on people. They've evolved another approach which uses carefully collected data from studies on animals and people, and has control groups so they can spot variations from the norm in the experimental group.

Applied to motorcycle crash causation studies, this method means that you need to carefully examine the scene of accidents and record thousands of items of closely defined data about each. You also gather control information about comparable bikers selected from the general rider population. You must employ trained researchers and use careful quality control procedures to make sure that the data collected is complete, accurate and consistent.

It is necessary to design the data collection forms carefully - text is little use for analysis, so you try to collect yes-no items, numbers, and selections from restricted sets of responses (like red, yellow, blue). You also have to decide which accidents to respond to, i.e. place and time range, and perhaps a method of statistically sampling which accidents to look at.

In statistics, a good rule of thumb is that you can often get good statistical results from maybe one hundred random examples. In a study like we are talking about, you need to study a lot more crashes. An example of this might be rear-end crashes, which are about one tenth of all bike crashes. If we take 925 crashes, like Maids did, then we might have enough rear end crashes on which to base valid statistical conclusions.

Even with the best methodology - like the OECD methodology Maids used, you can still get bad data without some additional data collection. Maids, for example, flubbed the control group selection for part of their study, so was unable to report on the effect of ABS. And both Maids and Hurt, as this blog post discussed, screwed up royally by not correcting for the fact that bikers avoid riding in the rain. They concluded, erroneously, that rain makes no difference to frequency of crashes, which every biker knows is untrue. But these are minor and correctable defects.

This is why studies like Hurt, Maids and the new study proposed for the US are different from all the small studies we have. To get reliable data, a study needs to record hundreds of crashes, and to make it worth while we need to collect thousands of items of data about each. This is why Hurt was so effective and why Maids produced such a large quantity of interesting data for Europe. It also explains why the studies are so expensive. Only a new study along those lines will get us out of the death trap created by our information vacuum.

The small, poor-quality studies, done from public records, that we are getting in the meantime are mostly awful and serve to obscure and mislead rather than clarify. Many of them are also flawed by inbuilt bias, and many bikers have no trouble rejecting their conclusions. This creates a breeding ground for dangerous myths, which we'll look at later. The professional researchers who make a living producing this material say they are fine, but they are not good enough where the rubber meets the road.

We think that the new Naturalistic study also has the potential to provide a controlled, epidemiological style study with good science. This is a technique which is untried in motorcycles, but it appears to have the potential of filling the information vacuum.

Other Cycle Crash Studies  Wikipedia MC Safety article