The BAMBi Myth

By Chuck Rhode <CRhode@LacusVeris.com>
Sheboygan, WI
Created: 4 Mar 2005, Updated: 5 Apr 2005

Abstract: Born-again, middle-aged bikers are more common than they used to be. Conventional wisdom holds that, as younger motorcyclists, they gave up the expense and risk of riding to join the workforce and raise families. They are supposed now to be returning to motorcycling in droves — older, yet more carefree, with waxing wealth and waning responsibilities.

The acronym BAMBi is amusing and derogatory because it connotes an ingénue, which is at odds with the image bikers are supposed to cultivate. It is said that these people have not regained their lost street smarts and reinvigorated their rusty riding skills. Rather, opinion suggests, they were minimally competent to begin with. Now that they can afford more powerful machinery, their declining mental and physical capacity to handle it makes them yet more of a menace to themselves than before. Instead of resurgent denizens of the mean streets, they are born-again babes in the woods.

This is the myth. Activists in and out of government underwrite it, and Mainstream Media subscribes to it. Perhaps it is true. This writeup confirms parts of the myth:

From government statistics reaching back 20 years, it is plain that older riders are contributing an increasing share of Motorcycle Rider Fatalities.
There is a relationship between turnover in the United States motorcycle fleet involved in fatal traffic crashes and Motorcycle Rider Fatalities. This writeup implies that driver inexperience is, thus, a candidate for the causative factor.

Implicit in the myth is the notion that the upward trend in Motorcycle Rider Fatalities since 1997 is due to factors that have recently come into play. This writeup contradicts that part of the myth:

Any proposed relationship of Motorcycle Rider Fatalities to driver age or to motorcycle engine displacement is seen to be much poorer for explaining variation in Motorcycle Rider Fatalities over the long haul than a simple relationship to Average Motorcycle Age.
Further, Motorcycle Rider Fatalities bears the same relationship to Average Motorcycle Age now as always, and, moreover …
This relationship accounts for nearly all the variation in Motorcycle Rider Fatalities.

Keywords: born-again, middle-aged biker or born-again biker; BAMBi; BAB; Motorcycle Rider Fatalities; Average Motorcycle Age; National Highway Traffic Safety Administration; NHTSA; Fatality Analysis and Reporting System; FARS; pylab; matplotlib; All Traffic Fatalities; Motorcycle Vehicle Miles Traveled; VMT; Motorcycle Registrations; Bureau of Transportation Statistics; BTS; Percent of Motorcycle Drivers Under 25 Years Old; Percent of Motorcycle Drivers 40 Years Old and Over; Percent of Motorcycles 250 cc or Less; Percent of Motorcycles 1000 cc or More

This writeup may be viewed online at: http://www.LacusVeris.com/BABs/

Background

An Associated Press story circulated widely in January 2005 under this lead, which serves as a comprehensive summary:

"Older motorcycle riders with a lot of disposable income are buying more machines than their aging, out-of-practice bodies can handle ('Motorcycle Boomer Deaths Raise Concerns' par. 1)."

The trend in Motorcycle Rider Fatalities is described in government pamphlets:

Motorcycle rider fatalities, following a longer-term trend, declined each year from 1993 to 1997, reaching a historic low of 2,116 in 1997. Motorcycle rider fatalities increased each year between 1997 and 2003 with a total increase of 1,476 or 70 percent. According to early estimates, 43,220 people died in traffic crashes in 2003, a net change of 405 from 2002. In the same period, motorcycle crash fatalities increased by 348. While other types of crashes have also contributed to the net increase in fatalities, motorcycles, which made up about 2 percent of all registered vehicles and 0.3 percent of all vehicle miles traveled (VMT), accounted for over 8 percent of the fatalities in 2003 compared to 5 percent in 1997, a significant increase as a component of the annual loss of life in traffic crashes (Motorcycle Riders in Fatal Crashes par. 2).

The largest ten-year age group of rider fatalities is the 20 to 29 year olds. However, 46 percent are age 40 or over, a group that has grown steadily from 21 percent in 1993 (Motorcycle Riders in Fatal Crashes par. 5).

Larger motorcycles are figuring more prominently in fatal crashes. The percentage of fatally injured riders of motorcycles with engine displacements of 1,001 to 1,500 cc has risen from about 28 percent in 1993 to about 38 percent in 2002. Mean engine size involved in a fatal crash has steadily increased from 820 cc in 1993 to 999 cc in 2002. Two-thirds of the riders killed on 1,001—1,500 cc engine size were 40 and over years old (Motorcycle Riders in Fatal Crashes par. 6).

Recipients of such news are left to draw their own conclusions.

The statistics merit consideration. Because interpretation of the statistics is interesting and important to broad sectors of the public, the conclusions drawn ought to be stated explicitly, and the basis for them described exactly. The cost of misunderstanding the phenomenon of Motorcycle Rider Fatalities is great. Ineffective strategies based on erroneous conclusions will frustrate rather than facilitate reducing the toll.

Objective

This writeup's goal is to present a numerical relationship between the annual series of Motorcycle Rider Fatalities and some other yet-to-be-determined series. Ideally, the relationship turns out to be linear. In other words, I want to explain the variation in the dependent statistic, Motorcycle Rider Fatalities, by a linear mathematical transformation of some other independent statistic or a linear combination of several. Also, I want to show what factors are not explanatory.

I wrote a library of computer programs in the Python language to generate a family of time-series from the National Highway Traffic Safety Administration's Fatality Analysis Reporting System (FARS) database. I graphed them with the pylab module of matplotlib.

I chose FARS for these reasons:

It is widely quoted. Verifying my results is easy.
It is conveniently available online.
It goes back a long way, having a current range of 28 years from 1975 to 2003.
Its data is painstakingly collected and fairly complete and consistent. Many kinds of year-to-year comparisons are possible despite code revisions through the years.

The FARS database is downloadable by year. The data files consist of a mixture of variable-length text records and, uncompressed, range from 20 to 35 MB per year or, compressed, from 5.5 to 7.0 MB. You may download my extract (3 KB) in comma-separated-value format (.csv) from this site.

Here is a graph of Motorcycle Rider Fatalities from FARS by year:

This is a census, not an estimate, of motorcycle drivers and passengers killed in traffic crashes. Rather arbitrarily, I chose the 21 periods beginning in 1983 to yield, if necessary, the 20 most recent period-to-period comparisons up to 2003.

Motorcycle Rider Fatalities is not linear with time. Because I want to demonstrate that it has a linear relationship to yet another time-series, I find myself seeking a series that must also be non-linear with time.

Motorcycle Rider Fatalities is recovering from the depth of a decline, which it reached about a decade ago. The actual peak was 4,961 in 1980 (Evaluation of the Repeal of Motorcycle Helmet Laws in Kentucky and Louisiana, Table 1).

Normalization

Normalizing Motorcycle Rider Fatalities to some broader statistic would have explanatory value in case the variation perceived in the foreground is really due to the background. I want to try to see the forest rather than the trees. In other words, I want to find another series to use as a deflator for Motorcycle Rider Fatalities.

Here is a graph of All Traffic Fatalities from FARS by year:

Because its inflection points do not coincide with those of Motorcycle Rider Fatalities, All Traffic Fatalities is not a good candidate for a deflator. Here is a diagram of the relationship between the two:
All Fatalities vs Rider Fatalities

A linear relationship would show as a straight line from corner to corner on the scatter plot. This one is not a strong relationship, which indicates that Motorcycle Rider Fatalities derive from causative factors distinct from those of other kinds of traffic fatalities, or at least All Traffic Fatalities taken together.

The National Highway Traffic Safety Administration (NHTSA) acknowledges this (Motorcycle Riders in Fatal Crashes par. 2) but makes no useful attempt to distinguish the causes of Motorcycle Rider Fatalities from All Traffic Fatalities.

Exposure

Comparing the dependent statistic, Motorcycle Rider Fatalities, to independent statistics representing conditions favorable (or unfavorable) to its occurrence might explain its variation. For exposure statistics, I foraged beyond the FARS database. Fortunately, other published series were readily available from the Bureau of Transportation Statistics (BTS).

I took the early portion of Motorcycle Vehicle Miles Traveled (VMT) from Evaluation of the Repeal of Motorcycle Helmet Laws in Kentucky and Louisiana (Table 1). I read the later portion from National Transportation Statistics (Table 1-32). There was overlap. I resolved slight discrepancies by preferring figures from the latter series.
Motorcycle M VMT

This is curious: There is a slight negative relationship. In other words, higher Motorcycle Vehicle Miles Traveled are related to lower Motorcycle Rider Fatalities. I had supposed the opposite. Anyway, the relationship is much weaker than I had expected.

I read the early portion of Motorcycle Registrations from Evaluation of the Repeal of Motorcycle Helmet Laws in Kentucky and Louisiana (Table 1), and I took the later portion from National Transportation Statistics (Table 1-11). Again, there was overlap, but this time, for some reason (It escapes me now what it may have been.), I resolved slight discrepancies by preferring figures from the former series.
MC K Registrations

Motorcycle Rider Fatalities is strongly positively related to Motorcycle Registrations. The relationship is much stronger than that for Motorcycle Vehicle Miles Traveled. Why? Here is my speculation:

Many motorcycles that are registered are rarely ridden.
Drivers who ride rarely are at greater risk due to lack of practice.
Motorcycle Registrations represents nearly all motorcycles and (by implication) drivers, whereas Motorcycle Vehicle Miles Traveled does not.
Thus, Motorcycle Registrations more nearly represents inexperienced drivers than Motorcycle Vehicle Miles Traveled does.
Motorcycle Rider Fatalities is positively related to inexperience and consequently more strongly to Motorcycle Registrations than to Motorcycle Vehicle Miles Traveled.

At the end of this writeup, I present yet another time-series, which I believe more nearly measures motorcycle driver inexperience than Motorcycle Registrations does and which has a stronger relationship with Motorcycle Rider Fatalities, but first I should dispense with the causative factors considered significant by BAMBi Myth adherents.

Driver Age

Motorcyclists, this day and age, are mostly old men, but this was not always so. Twenty years ago, they were mostly young men.

From here on, I subdivide the population of motorcycle drivers who are involved in fatal traffic crashes.

Consider the proportion of motorcycle drivers (who may or may not have been killed) involved in fatal traffic crashes who are under 25 years old. Here is a graph of Percent of Motorcycle Drivers Under 25 Years Old from FARS by year:
Drivers < 25

Obviously, because Percent of Motorcycle Drivers Under 25 Years Old is a steadily decreasing series, it looks as though it might be related to any other time-series. However, I want to explain a non-linear series, so a linear one will not do. The fishhook in the scatter plot, while visually striking, indicates an unexplanatory relationship. The relationship looks somewhat negative in the lower realm since 1997. Perhaps this is what the Media sees. Choosing an arbitrarily long interval to examine and taking the time and trouble to do so saved me from this error.

Here is a graph of Percent of Motorcycle Drivers 40 Years Old and Over from FARS by year:
Riders >= 40

Once again there is a hook in the scatter plot, which, I am sure, catches many people finding a positive relationship in the upper realm since 1997. It is certainly true that older riders are contributing an increasing share of Motorcycle Rider Fatalities. However, over 21 years from 1983 to 2003, there is no straightforward relationship between the number of Motorcycle Rider Fatalities and Percent of Motorcycle Drivers 40 Years Old and Over. To conclude that there is, as BAMBi Myth adherents have done, is to misunderstand the phenomenon.

Strategies based on this erroneous conclusion are bound to frustrate the goal of reducing Motorcycle Rider Fatalities. For instance, targeting a specific age group for public-service advertising will be perceived as unfairly discriminatory. Tightening regulations on older motorcycle riders will be, too. Yielding to such impulses may squander what respect exists for regulations and the regulatory process. Indulging in such folly will waste resources that would be more effectively applied in a less discriminatory manner.

Engine Displacement

Erroneously concluding that there is a relationship between motorcycle engine displacement and Motorcycle Rider Fatalities is discredited by arguments identical to those directed against driver age.

Consider the proportion of motorcycles involved in fatal traffic crashes with engine sizes under 250 cc. Here is a graph of Percent of Motorcycles 250 cc or Less from FARS by year:
<= 250 cc

Here is a graph of Percent of Motorcycles 1000 cc or More from FARS by year:
>= 1000 cc

Over 21 years from 1983 to 2003, there is no straightforward relationship between the number of Motorcycle Rider Fatalities and either of the preceding series based on motorcycle engine displacement.

Vehicle Age

Now, consider the age of motorcycles involved in fatal traffic crashes. Here is a graph of Average Motorcycle Age from FARS by year:
Average MC Age

There is a negative relationship between Motorcycle Rider Fatalities and Average Motorcycle Age, which is stronger than the relationships of Motorcycle Rider Fatalities to other series I have examined. This makes the relationship with Average Motorcycle Age seem a more likely explanation for the variability in Motorcycle Rider Fatalities than the other series.
Further, the relationship of Motorcycle Rider Fatalities to Average Motorcycle Age is about as good in the early time periods as in the later. This means its explanatory power, unlike that of the factors favored by BAMBi Myth adherents, is good across the whole range of time periods examined, and, moreover …
By itself, Average Motorcycle Age explains nearly all the variation in Motorcycle Rider Fatalities in the years from 1983 to 2003. This makes all other explanations (including combinations with other series) unattractive.

The range of variation in the age of the portion of the United States motorcycle fleet, which is involved in fatal traffic crashes, waxes and wanes dramatically. It relaxes and tightens up. It tightens up as more new vehicles are added to the fleet and more old ones are scrapped out of it. Low variation is a measure of high turnover, and low Average Motorcycle Age captures it. The range of variation has been particularly tight recently, indicating high turnover.

Conclusion

Motorcycle fleet turnover as measured by Average Motorcycle Age implies driver inexperience, and this is reflected in Motorcycle Rider Fatalities.

Here is the regression formula:

Motorcycle Rider Fatalities = -692.12 * Average Motorcycle Age + 8093.44

Here are graphs of the predicted values beside the actual values and the differences (residuals) against the actuals:

Actual and Predicted Rider Fatalities

The outlying residuals are from 1983 and 1986. These discrepancies may reflect a deficiency in FARS data encoding in the earlier periods or a deficiency in my interpretation of it. The relationship is better over more recent intervals.

I nominate driver inexperience as the mediator between Motorcycle Rider Fatalities and a portion of motorcycle fleet turnover as measured by Average Motorcycle Age between 1983 and 2003. This is a conservative inference, given the other factors I have examined and rejected; nevertheless, I have to allow other interpretations of the relationship even though I cannot imagine what they might be.

The notion that Motorcycle Rider Fatalities is caused by driver inexperience over the 21 years from 1983 to 2003 leads to strategies for reducing the toll that do not unfairly discriminate against newer bikes or older bikers. What needs to be done is this: Emphasize skills demonstration and promote skills acquisition. In other words, donate time and treasure to the many organizations set up to train motorcyclists of all ages, riding styles, and skill levels.

Surely such strategies are more effective than those based on the erroneous conclusions of the BAMBi Myth.

Works Cited

"Motorcycle Boomer Deaths Raise Concerns." WJLA ABC Channel 7 N. seq. (21 Jan. 2005): 17 pars. Online. Internet. 14 June 2006. Available http://www.usatoday.com/money/autos/2005-01-22-motorcycle-deaths_x.htm. United States. Department of Transportation. Bureau of Transportation Statistics. National Transportation Statistics. Washington: BTS, 2004. Online. BTS. Internet. 3 Mar. 2005. Available http://www.bts.gov/publications/national_transportation_statistics/2004/index.html. ---. ---. National Highway Traffic Safety Administration. Fatality Analysis and Reporting System. Online. 1975-2003. Available ftp://ftp.nhtsa.dot.gov/FARS/. ---. ---. ---. National Center for Statistics and Analysis. Evaluation of the Repeal of Motorcycle Helmet Laws in Kentucky and Louisiana. By R.G. Ulmer and D.F. Preusser. Springfield, VA: National Technical Information Service, 2003. Online. NHTSA. Internet. 3 Mar. 2005. Available http://www.nhtsa.dot.gov/people/injury/pedbimot/motorcycle/kentuky-la03/index.html. ---. ---. ---. ---. Motorcycle Riders in Fatal Crashes. By Umesh Shankar. N.p.: n.p., 2004. Online. NHTSA. Internet. 8 Feb. 2005. Available http://www-nrd.nhtsa.dot.gov/pdf/nrd-30/NCSA/RNotes/2004/809-734/index.htm.

Rhode, Chuck. "The BAMBi Myth." 4 Mar 2005. Lacus Veris. 5 Jan. 2009 <http://www.lacusveris.com/BABs/index.shtml>. Last modified 14 Jun. 2006. Served 701 times between 10 Jul. 2008 and 5 Jan. 2009. Contact mailto:root@lacusveris.com?subject=LacusVeris.