banner of Methodology for Determining Motorcycle Operator Crash Risk and Alcohol Impairment

3. Detailed Report of Discussion

Survey Study

There are many ways in which a survey study could be conducted. Possibilities include telephone surveys, mail surveys, and in-person surveys at motorcycling events. A survey would attempt to understand the effects of alcohol on crash likelihood by asking riders to recall their drinking behavior in the recent past, along with information regarding crashes and close calls. There is evidence that people can remember recent drinking experiences well enough to estimate the number of drinks consumed over a given period. This information, along with rider weight, height, gender, and other factors, can be used to estimate BAC. Typically, surveys will ask a participant to recall a specific incident in the recent past (within a set time window) in which drinking took place. A recent study on drowsy driving used a similar approach (Royal, 2003).

In many cases, the use of surveys requires OMB clearance that can add significantly to the time and complexity involved in performing the study. This report will not assume that OMB approval will always be an issue because of the possibility that the work could be funded by other organizations that are not subject to OMB guidelines.

As with any survey, the method used to identify participants and the response rate of participants has a great effect on the validity and generalizability of results. The American Motorcyclist Association’s experience with surveys in their magazines has been that only one percent will respond. A random survey to identify enough motorcyclists to have a valid survey would require a great deal of effort. Narrowing the search by focusing on people with motorcycle licenses would miss the large proportion of riders who are riding unlicensed, which would lead to skewed results. Focusing on people with registered motorcycles may be a better strategy, though there is no guarantee that such a person rides on a regular basis. A mail-out survey of motorcyclists in Pennsylvania, used for a study by McKnight and McKnight (1989), received a 66 percent response rate using registration data. Collecting information at a motorcycling-oriented event would skew results toward reflecting only the type of rider who would attend that particular event. Some have suggested that motorcyclists may be suspicious of the motives of the safety community such that it may be difficult to get a high degree of cooperation from some segments of the riding community. A balancing belief is that riders enjoy talking about issues related to riding and may be forthcoming with information once any initial hesitance is overcome. A good way to get a “foot in the door” would be to approach riders with someone who is (or is perceived to be) a member of the riding community. In the form of mail or phone surveys, this could be done by having the survey nominally sponsored by a motorcycling organization. In the case of in-person surveys at motorcycling events, surveys would be conducted by people who were visibly identifiable as being riders. In the case of door-to-door surveys, interviewers would actually ride to the house.


It may be possible to administer a survey by including questions in an existing survey, such as those being administered by the Motorcycle Safety Foundation, American Motorcyclist Association, National Household Transportation Survey, Mortality Followback Survey, National Drinking and Driving Survey, National Household Study on Drug Abuse, and Behavioral Risk Factor Surveillance System (more information is provided on these under “General Issues, Data Sources” near the end of this report). In some cases, data from these surveys can be linked to existing data (e.g., FARS).


Although it may be possible to estimate a rider’s BAC based on survey results, the estimate will be very rough compared to an actual BAC measurement. Participants may incorrectly recall, or misreport information. Other self-reported information (e.g., crashes, mileage) may be inaccurate. Although it may be possible to include survey questions in an extant survey, historically many of these surveys require that agencies requesting additions make a contribution of funds or resources, which eliminates some of the advantage of piggybacking on these surveys.


Surveys tend to be costly, mostly because it takes several attempts at contacting the same people to get a sufficiently high response rate. If interviewers attempt to reach motorcyclists by surveying random households, costs will be extremely high due to the relatively few motorcyclists. The Motorcycle Industry Council conducts such a survey. For the 1998 survey they called 270,000 households to get 1,500 motorcycle owner interviews. They also interviewed 1,500 non-motorcycle-owning households. Based on their experience they estimate that the telephone interview portion of such a survey would cost about $350,000. Surveys that attempt to reduce costs by targeting motorcyclists more directly run the risk of having biased results, depending on the avenue used to reach riders, and could still become costly in an attempt to get a sufficient response rate.

It may be possible to get survey questions answered by piggybacking questions on another survey. Although it seems as though this may be a way to get survey questions answered for less cost, it should be noted that some of the surveys discussed as having potential for this purpose have historically asked for large sums of money to include questions on their survey.