Scientists have long known that such before and after studies may be misleading because the changes found may result from factors other than the law change.

The institute study found that due to the law changes, there were reductions in nighttime fatal crashes among youthful drivers in eight of the nine States studied. There was considerable variation in these reductions, from 6 to 75 percent. Based on all nine States, there was an average reduction of 28 percent. In only one State, Montana, was there an increase in nighttime fatal crashes. The substantial variation in results among the States was to be expected, since they differ in size, population, region, and other respects. In particular, the numbers of drivers in the affected age groups involved in fatal crashes varied tremendously, from a low of 28 drivers in Montana to a high of 538 in Michigan during the 21 months following the law change.

It is well known that such sample size variations lead to considerable variations in results, and it is precisely for this reason that we emphasized overall results rather than those from individual States.

It was estimated from the institute's work that each year about 730 fewer young drivers would be involved in fatal crashes in the United States if the drinking age for alcohol were raised to 21 in every state.

Michael M. Birkley, representing the National Licensed Beverage Association, told the subcommittee on October 4, 1983.

The evidence strongly indicates that raising the legal drinking age is likely to result in more rather than less alcohol abuse among the underaged population, as it did in at least three of the eight States which have recently raised their legal drinking ages, and for which sufficient consistent data have been analyzed.

He claimed that in only one State among eight studied, Michigan, was there a reduction in highway crashes. In addition to the three States in which he claimed there were increases in crash involvement, according to Mr. Birkley, there were four others in which no significant changes occurred. Mr. Birkley concluded that the insurance institute's prediction of a 28-percent reduction was incorrect in seven of the eight States studied. However, Mr. Birkley failed to mention that of the seven States he claimed did not conform to the institute's prediction, six had in fact been included in the Institute's study, and were part of the basis on which the 28percent, average reduction was computed. Five of these six States, including two in which Mr. Birkley reported increases in crash involvement, showed reductions in fatal crashes.

Mr. Birkley referred to the institute's research results only in the case of Montana, the single State in which an increase was shown. Instead of mentioning the reductions in the other States, he claimed to have evidence showing increases or no changes subsequent to raising the drinking age, but most of these data were derived from simple before-after analyses, without appropriate comparison groups. Therefore, they are inappropriate for drawing inferences about the effects of changing the drinking age.

In addition, he ignored two other States, Tennessee and New Hampshire, included in the Institute's study. Both of these States showed reductions in fatal crash involvement. His conclusions,

based as they were on inadequate data and a selective review of the literature, are not valid.

Birkley has also claimed that:

None of the major drinking age impact studies, including those of the Insurance Institute for Highway Safety, found any measurable change in highway crash rates among 16-and 17-year-old drivers attributable to lowering or raising the legal drinking age in any jurisdiction.

That statement is incorrect. In fact, the institute's 1974 study of areas that lowered the legal minimum drinking age indicated "a significant increase in involvement in fatal crashes of drivers under 21 in areas that changed the law." This occurred not only among the 18- to 20-year-olds to whom the law change applied, but also to a somewhat lesser degree among 15- to 17-year-olds.

Finally, Birkley claimed that a study by Cook and Tauchen on lowering the drinking age supports his own position, namely, that an 18-year-old drinking age is preferable. However, Birkley failed to cite the conclusion of Cook and Tauchen's study:

A reduction in the minimum drinking age from 21 to 18 for all alcoholic beverage types will result in an increase in the auto fatality rate for 18- to 20-year-olds of about 7 percent, and a somewhat smaller increase for 16- to 17-year-olds.

We are confident in concluding that the cumulative effect of minimum legal drinking age reductions during the early 1970's was to cause a substantial increase in 18- to 20-year-old auto fatality rates averaging about 150 lives per year during the mid-1970's.

Based on its research, the Insurance Institute for Highway Safety has concluded that raising the minimum legal drinking age to 21 in all States would go far toward reducing the annual toll of motor vehicle deaths in the United States, particularly the deaths of young people and of others with whom they are involved in crashes. That conclusion stands.

Mr. KELLEY. We would be happy to answer any questions, Mr. Chairman.

[Testimony resumes on p. 259.]

[The prepared statement of Mr. Williams and Mr. Kelley follows:]

Statement of Allan F. Williams and Ben Kelley, Insurance Institute for Highway Safety, Before the Subcommittee on Commerce, Transportation, and Tourism, U.S. House of Representatives Committee on

Energy and Commerce, October 19, 1983

The Insurance Institute for Highway Safety is an independent public service organization. It is a nonprofit, tax-exempt scientific and educational organization dedicated to reducing the losses deaths, injuries, and property damage resulting from crashes on the nation's highways. Supported by companies writing most of the motor vehi. cle insurance in the United States, the Institute conducts research involving a wide range of factors that contribute to the huge losses resulting from highway crashes.

As part of this research, the Institute has studied the effects on highway crashes of changes in the legal minimum drinking age. Such research is the focus of this testimony, which you have invited us to present today.

Between 1970 and 1975, some 29 states reduced the legal minimum age for purchasing alcoholic beverages. The minimum age had been 21 in most of these states. It was reduced to 18, 19, or 20 in most cases, 18. Work conducted by the Institute in the mid-1970s found that in states that lowered the drinking age, there were significant increases in fatal crashes of drivers under 21 years old, as compared to states where the drinking.age.laws were not changed. (1) Studies by other researchers in the United States and Canada have also found that lowering the drinking age increases crashes. (2)

Beginning in 1976, there has been a trend toward raising the drinking age to 19, 20, or 21. The majority of states that lowered their drinking ages in the early 1970s have raised them, although usually not back to the previous levels. In 1981, the Institute studied the effects on fatal highway crashes of raising the legal minimum drinking age (3) (Attachment A). All states were studied that had raised their drinking ages and for which sufficient post-law data were available. Nine states were included in the study. (Five were excluded because their law changes were too recent for their effects to be measured at the time of the study.)

Each of the Institute's studies was carefully planned to isolate the effects of drinkingage changes, using established and well-known principles of scientific research design. For example, since numbers of highway crashes and fatalities fluctuate widely over time (for a variety of reasons, known and unknown), simply determining the crash involvement rate of young drivers in a state after the drinking age is raised, and comparing this with the rate before the law was changed, is insufficient. Scientists have long known that such before-after studies may be misleading, because the changes found may result from factors other than the law change.

To rule out the possibility that changes observed in youthful crash involvement in states that raised their drinking ages were merely part of a regional trend, each of the nine states studied was paired with a neighboring comparison state in which the legal minimum drinking age remained unchanged throughout the study period. To rule out the possibility that changes observed in age groups covered by the law were part of a trend occurring at other ages as well, age groups covered by the laws were compared to older drivers, to whom the law changes did not apply.

Comparisons were based on nighttime fatal crashes crashes - which overwhelmingly involve alcohol.

especially single-vehicle

The Institute's study found that, due to the law changes, there were reductions in nighttime fatal crashes among youthful drivers in eight of the nine states studied. There was considerable variation in these reductions from six to 75 percent. Based on all nine states, there was an average reduction of 28 percent. In only one state, Montana, was there an increase in nighttime fatal crashes among young drivers (14 percent).

The substantial variation in results among the states was to be expected, since they differ in size, population, region, and other respects. In particular, the numbers of drivers in the affected age groups involved in fatal crashes varied tremendously, from a low of 28 drivers in Montana to a high of 538 in Michigan, during the 21 months following the law change in those states. It is well known that such sample size variations lead to considerable variations in results, and it is precisely for this reason that we emphasized overall results rather than those from individual states.

On the basis of the Institute's consistent findings of reductions in nighttime fatal crashes, our researchers concluded that just as lowering the drinking age has a negative effect on highway crashes, raising the drinking age has a positive effect. Numerous other researchers have reported similar results (Attachment B). It was estimated from the Institute's work that each year about 730 fewer young drivers would be involved in fatal crashes in the United States if the drinking age for alcohol were raised to 21 in every state.

Michael M. Birkley, representing the National Licensed Beverage Association, told this subcommittee on October 4, 1983: "The evidence strongly indicates that raising the legal drinking age is likely to result in more, rather than less, alcohol abuse among the underage population, as it did in at least three of the eight states which have recently raised their legal drinking ages and for which sufficient consistent data have been analyzed." (4) He claimed that in only one state among eight studied - Michigan was there a reduction in highway crashes. In addition to the three states in which he claimed there were increases in youthful crash involvement, according to Mr. Birkley there were four others in which no significant changes occurred subsequent to raising the drinking age. Mr. Birkley concluded that the Insurance Institute's prediction of a 28 percent reduction was incorrect in seven of the eight states studied.

However, Mr. Birkley failed to mention that of the seven states he claimed did not conform to the Institute's prediction, six had in fact been included in the Institute's study and were part of the basis on which the 28 percent average reduction was computed. Five of these six states - including two in which Mr. Birkley reported increases in crash involvement showed reductions in fatal crashes.

Mr. Birkley referred to the Institute's research results only in the case of Montana, the single state in which an increase was shown. Instead of mentioning the reductions in the other states, he claimed to have evidence showing increases or no changes in youthful involvement in fatal crashes subsequent to raising the drinking age. But most of these data were derived from simple before-after analyses, without appropriate comparison groups. Therefore, they are inappropriate for drawing inferences about the effects of changing the drinking age.

3

In addition, Mr. Birkley ignored two other states Tennessee and New Hampshire included in the Institute's study. Both of these states showed reductions in fatal crash involvement. His conclusions, based as they were on inadequate data and a selective review of the literature, are not valid.

As part of his written testimony, Mr. Birkley also submitted a document entitled "Death and the Legal Drinking Age: A Tri-State Study" (5), which he wrote and in which he claimed that the reductions in crash involvement in Michigan and Illinois after the drinking ages were raised were not as great as reductions in Wisconsin, where the drinking age was not raised. However, this document does not follow accepted procedures of scientific research. Birkley analyzed 1976-1981 Wisconsin data and found a 1981 decline in crash involvement among 18-20 year olds, compared to what he said would have been expected based on 1976-1980 trends. He then compared this 1981 reduction in Wisconsin to a 1980 reduction in Illinois (the first year of the law change in that state) and a 1979 reduction in Michigan (the first year of the law change there). That is, the “comparisons” were based on different years, which invalidates them. (They also involved different age groups in Illinois and Wisconsin, and were based on different measures of crash involvement in Wisconsin, compared to the other two states.)

Birkley claimed that "none of the major drinking age impact studies, including those of the Insurance Institute for Highway Safety ... found any measurable change in highway crash rates among 16- and 17-year-old drivers attributable to lowering or raising the legal drinking age in any jurisdiction." (6) That is incorrect. In fact, the Institute's 1974 study of three areas that lowered the legal minimum drinking age, compared to adjacent areas that did not, indicated “a significant increase in involvement in fatal crashes of drivers under 21 in areas that changed the law .... This occurred not only among the 18-20 year olds to whom the law change applied, but also, though to a somewhat lesser degree, among 15-17 year olds." (1)

Finally, Birkley claimed that a study by Cook and Tauchen on lowering the drinking age supports his own position namely that an 18-year-old legal drinking age is preferred. (6) However, Birkley failed to cite the conclusion of Cook and Tauchen's study: "A reduction in the minimum drinking age from 21 to 18 for all alcoholic beverage types will result in an increase in the auto fatality rate for 18 to 20 year olds of about seven percent, and a somewhat smaller increase for 16 to 17 year olds We are confident in concluding that the cumulative effect of minimum legal drinking age reductions during the early 1970s was to cause a substantial increase in 18-20-year-old auto fatality rates, averaging about 150 lives per year during the mid-1970s.” (7)

Based on its research, the Insurance Institute for Highway Safety has concluded that "raising the legal minimum drinking age to 21 in all states would go far toward reducing the annual toll of motor vehicle deaths in the United States, particularly the deaths of young people and of others with whom they are involved in crashes." That conclusion stands.