Their average improvement will be compared with that of a similar control group that will not take the course.

(b) Attitudes.-A test of attitudes relevant to the course is being developed and will be used in subsequent research at a New England branch of a major industrial firm.

(c) Driving records.-The council is presently trying to interest an appropriate university research department in studying the effect of the course on accidents and violations.

6. Multiple regression study of the effects of safety activities on the traffic accident problem

Research was undertaken to answer the question: What are the reasons for the differences in fatal and nonfatal motor vehicle accident frequencies in the various States?

The hypothesis studied was: Differences in motor vehicle accident frequencies can be accounted for by

(a) Differences in environmental and physical conditions (vehicles and population density, weather, terrain, etc., called nonprogram factors); and (b) Differences in safety activities (called program factors).

In testing the hypothesis, 199 program items from the 1960 motor vehicle inventory (an annual questionnaire of State safety-related activities) and 19 nonprogram items comprised the independent variables. Various measures of accident experience including the milage death rate, bodily injury and property damage accident claims frequency rates, and the death ratio (ratio of actual to expected deaths when the States were equalized for size) were used as dependent variables. Multiple regression analysis in conjunction with contingency table analysis (for discrete items) was used to investigate the degree of association of the independent variables with the various dependent variables.

It was found that the major portion of the differences in States' motor vehicle death rates was associated with such nonprogram items as density (cars per mile of road), percent of population which was 65 years or over, weather (minimum temperatures and precipitation) and increase in vehicle registrations. But after adjusting for nonprogram items, many program items were significantly associated with the rate differences. Program areas in which activities were found to be significant were traffic engineering, enforcement, safety education, safety organization activities, vehicle inspection, and use of records.

Neither of the analytic techniques used prove cause-and-effect relationships but they do provide measurements of the degree of variables. Based on the degree of variables. Based on the degree of association indicated by these analyses, the evidence supports the hypothesis that both nonprogram and selected program items are factors in the differences in States' motor vehicle accident frequencies. 7. Research on the costs of accidents

In 1964, the council contracted with the Stanford Research Institute for the preliminary phase of a study of accident costs, entitled "A General Framework for Analysis of the Costs of Traffic, Home, and Public Accidents." The overall objectives of the study are to

(a) Develop a methodology for assessing the economic costs of traffic, home, and public accidents and to identify the specific segments of the community that suffer various adverse effects.

(b) Develop a methodology for analyzing and comparing the effectiveness and cost of alternative means of alleviating such accidents.

(c) Demonstrate the use of the methodology by assessing the cost and effectiveness of alternative remedial measures relating to these three types of accidents in a sample area.

1. Traffic accident data project

III. ACCIDENT RECORDS

The traffic accident data project was initiated in recognition of the need for increased development and implementation of traffic records and accident information. This need has been expressed by researchers, police, engineers, motor vehicle administrators, and other users of such information. Further, there is need for continuous attention to policies, standards, definitions, procedures, and uses pertaining to traffic accident records.

The specific function of the project is to stimulate improvements in traffic accident systems by providing means of reaching agreement on specific objectives 49-959-66-pt. 3-7

and standards. Through its staff and a steering committee, needed research and experimentation will be encouraged. All technical and professional groups having responsibility in this area had been invited to participate in the project and to provide technical assistance in field service.

In order to define the project's objectives in greater detail, a 5-year program has been developed. The first year is largely devoted to research into accident data user needs, data collection, accident investigation and processing of data. At the conclusion of the research phase, standards will be developed for investigating accidents, collecting and processing of data and uses of data. During the third and fourth year of the project, demonstrations of pilot systems will be undertaken in selected cities and States. The fifth year will be devoted to review, publication, and promotion of reporting systems developed by the project.

In addition to the development of reporting systems, there will be corollary work in development of new definitions and measurements and rates. There will also be developed curriculum and course materials for training of personnel in operations of model accident reporting systems. This development will be followed by Statewide and regional workshops designed to promote training. (A statement of the 5-year program and chart are to be found in app. L.) 2. Data analysis

The statistics division of the National Safety Council has reported to it on a regular and continuing basis, the summarized accident experience of a majority of the States and hundreds of cities. These data are further summarized by the council and projected to national totals, relating to such items as regions, urbanrural, day-night, time of day, day of week, month, type of accident, circumstances of accidents, direction of vehicle at time of accident and place of accident, drinking, speeding, age and sex of driver, and many, many more. Thirty pages in the council's annual publication Accident Facts are devoted to such motor vehicle statistics which show trends as well as current figures.

In addition to the above routine tabulations, special statistical analyses are done, of which the following are examples:

(a) Program correlation study. In this study, correlations were run between States' traffic death rates and their safety program activities, as well as nonprogram items such as vehicle density, age of population, weather, terrain, per capita income, etc. In general, it was found that most of the differences in rates were associated with the nonprogram items, but after adjusting for this, significant relationships were found for many program items in such areas as traffic engineering, enforcement, safety education, and others.

(b) Effect of increase in young drivers.-Young drivers have sharply higher accident rates than drivers of other ages, a fact which has been determined from the reports of State motor vehicle departments. These high rates not only contribute to the high overall death total, but with the numbers of young drivers increasing faster than drivers of other ages, the impact of such changes on the death count is important. This study traces the birth rate since the 1930's and projects this in terms of new drivers and deaths in the years ahead.

(c) Speed and motor vehicle deaths.-Accidents at higher speeds are more serious than those which occur at lower speeds. The extent of such increase in severity is revealed by State motor vehicle death summaries. Other reports from the U.S. Bureau of Public Roads shows how average speeds have been increasing through the years. This study combines these two elements and shows the impact of increasing speeds on the death count.

(d) Motor vehicle deaths on holidays.-Special reports have been collected from State motor vehicle departments covering most of the major holidays. These show deaths by hour of day and day of the holiday period, and type of accident on the same basis. Also included are drinking and speeding as factors in the accident, age of drivers, and age of victims, seat belt usage and multiple death collisions. Purpose of trip, type of vehicle involved, and distance from driver's home, are items that have been reviewed on a selective basis.

(e) Small car studies.-A half dozen reports of accidents involving small cars have been completed by State motor vehicle departments. These were reviewed and summarized in one report which revealed the general conclusion that small cars apparently are not involved in accidents any more frequently than large cars, but when accidents do happen, the fatality rate among small car occupants is about twice as high as it is among occupants of larger cars. Another report translated the higher small car casualty rates and the increase in the numbers of such vehicles into national totals.

The above are examples only of the special study work being done by the council's statistics division. Other studies, complete or in progress, cover

motor vehicle accidents with bicycles, motor scooters, motorcycles, tires, seat belts, motor vehicle deaths related to gross national product, and others.

IV. ENCOURAGEMENT OF PUBLIC SUPPORT OF RESEARCH

1956.-School and college conference research needs study publicized. September 1957.-Traffic safety services chart showed great deficiences in research-publicized continuously ever since.

April 23, 1958.-General statement before the Roberts' committee.

June 1958.-Two NSC representatives on panel which produced "Analysis of Responsibility and Capability of Public Health Service in Accident Prevention." 1959.-Inventory of 1958 State expenditures for research, returned to the States in analyses recommending expansion.

April 3, 1962.-Board of directors approves policy of support and encouragement for U.S. Public Health Service research programs.

April 10, 1963.-Statement to House Subcommittee on Public Health and Safety in support of Public Health Service Intramural Laboratory.

March 11, 1964. Statement on behalf of USPHS research appropriations before House Subcommittee on Labor, Health, Education, and Welfare. July 31, 1964.-Same to Senate subcommittee.

March 29, 1965.-Letter to House subcommittee reiterating and enlarging March 11, 1864, statement.

Continuous.-Publicize action program, including research section.

Continuous.-Assistance to State and local councils in their support efforts on behalf of State university safety centers.

Mr. JOHNSON. Senator Nelson is speaking tonight to the Winter Hazards Committee which is researching the studded tires in his home State of Wisconsin.

Senator RIBICOFF. Now, what has the council done in the area of determining engineering requirements for the safe design of automobiles?

Mr. PYLE. Not as much as we should, sir.

Senator RIBICOFF. How do you think you could do more?

Mr. PYLE. We sincerely hope that in the course of the immediate future we can establish an appropriate section devoted to this subject. We intend to explore this possibility.

We have talked about it a great deal in recent months and I think it has real possibilities. This would be a platform, so to speak, or an area in which persons with interests of this type could come and present their cases and be assured helpful attention.

Senator RIBICOFF. Can you tell us the nature of some of the technical information that you have been able to furnish governmental agencies responsible for drafting modern legislation in relation to motor vehicles? Have you played any role there?

Mr. PYLE. May Mr. Johnson respond to that?

Mr. JOHNSON. You are talking in terms of the uniform vehicle code, the provisions in the code for a wide variety of regulations, including motor vehicle equipment. We endorse the uniform vehicle code.

Senator RIBICOFF. You have endorsed it, but I mean you haven't initiated anything in that regard?

Mr. JOHNSON. We don't introduce legislation in the State legislature. However, many State and local safety councils are active in this field all the way from asking somebody to introduce the legislation to helping them draft it, and so forth.

TRAFFIC SAFETY STATISTICS

Senator RIBICOFF. Now, questions have been raised over the adequacy of the council's use of the death rate per vehicle miles traveled

as a measurement of traffic safety progress. The fatality rate for 100 million-vehicle miles traveled declined steadily from 11.4 in 1940 to 5.2 in 1962, and then began rising in 1963, and then registered 5.7 in 1964.

However, when the total U.S. population is used as a base, the fatality rate seems to remain unchanged. Traffic deaths per 100,000 population total 26.1 in 1940; 23 in 1950, and 24.9 in 1964, which sug gests that a motorist is just as likely to be killed now within a given year as in previous years.

Are you satisfied with the National Safety Council's current statistical methods, that they are as useful a reflection of traffic safety as you can devise, or do you think there is a better way?

Mr. PYLE. Senator, we have with us Mr. Gene Miller, who is the director of our statistics division and one of the best known names in America in this regard.

May we have him comment?

Senator RIBICOFF. Certainly. Anyone you have can come up and answer that question.

Mr. MILLER. The mileage death rate is one way of looking at the frequency of deaths in relation to the exposure to death, and for that purpose it says, but often people trying to look into this rate think it is things you don't really measure.

Senator RIBICOFF. Can you come forward? The reporter can't hear you. Will you take one of these seats?

Mr. MILLER. I was saying that the mileage death rate which relates numbers of deaths to miles driven by vehicles is one way of measuring the death problem. This is exactly what it does measure, the number of deaths in relation to exposure to death. But this is all that it does

measure.

And many people find fault with this measure because they feel it is not measuring the total accident problem, and it does not measure the total accident problem, because the accident problem includes many types of accidents besides fatalities.

But experts through the years have searched for a better way of measuring the accident problem, and all of the experts, including people from the Federal Establishment, have not been able to find a better method.

As one activity of this new accident data project which the council is setting up, there will be further investigation and further committee activity in the area of searching for a better way of measuring the accident problem.

Senator RIBICOFF. I am just curious.

Basically you perform the function that ought to be done by the Census Bureau. Why doesn't the Census Bureau collect data on traffic accidents? Have you any idea?

Mr. MILLER. I don't know why they do not. In fact, in terms of the total problem, I am not sure what could be done more than we are currently doing.

EXHIBIT 91

DEATH RATE STATISTICS

The mileage death rate is useful for what it is intended to measure, viz: the number of deaths per 100 million vehicle miles of travel. Since the rate relates deaths to a measure of exposure to death, it is the best indicator of the

death picture. Through the reporting procedure which the council has set up with State traffic authorities (for deaths) and with the U.S. Bureau of Public Roads (for mileage) monthly death rates for the Nation are known by the 25th of each month following. Along with being the best indicator of the death picture presently available, the promptness with which it is determined gives it added value for evaluating both the current situation and trends.

The council also uses two other death rates to measure the death problemthe registration death rate (deaths per 10,000 vehicles), and the population death rate (deaths per 100,000 population). The registration death rate relates deaths to the number of vehicles, but does not give consideration to how much the vehicles are used. For this reason, it is not as useful as the mileage death rate, and it is employed principally where mileage figures cannot be estimated. The principal advantage of the population death rate is that it permits expressing motor vehicle deaths in the same terms as deaths from accidents in the home, in public places, at work, on farms, etc. It is a general measure of the problem, but since it completely disregards the sharp rise which has taken place through the years in the number of motor vehicles, which is the agent of death, it has limited use as a measure of the motor vehicle problem.

Actually, no one rate can provide a complete measure of the motor vehicle death problem, because there are so many elements involved-vehicles, people, roads, and many variations of each from place to place. The Council has experimented with various ways of measuring the death problem, and in fact, one suggestion was included in "The Federal Role in Highway Safety," page 154 (H. Doc. No. 93, published Mar. 3, 1959).

In July 1961, the chairman of the Traffic Conference of the National Safety Council appointed a special committee on "Revision of Motor Vehicle Traffic Death Rate Basis. The committee consisted of the best statistical minds in the traffic field and was chaired by J. Stannard Baker, Traffic Institute, Northwestern University. Other members of the committee and the organizations they represented were: A. D. Battey, National Safety Council; David M. Baldwin, Institute of Traffic Engineers; J. W. Bethea, President's Committee for Traffic Safety; Dr. Alexander V. Monto, U.S. Public Health Service; Charles W. Prisk, U.S. Bureau of Public Roads; John E. Smiley, Insurance Institute for Highway Safety; William James, U.S. Public Health Service; Roy Haeusler, Chrysler Corp.; Burton W. Marsh, American Automobile Association; and Fletcher N. Platt, Ford Motor Co.

The committee met for a full day on September 11, 1961, after which it concluded: (1) that motor vehicle officials were not ready for the council's proposed method, (2) that there was no method that was good and yet simple for measuring the problem, and (3) that the separate rates-the mileage rate, backed up by the registration and population death rates-should continue to be used, even though there were deficiencies in each.

Under the council's new accident records data project (see exhibit 9), attention again will be directed toward a study of these rates. What the outcome will be cannot be guessed at this time, but the problem will be examined thoroughly.

It is also important to understand that death rates do not and cannot measure the total motor vehicle accident problem. Statistical studies which the council has summarized have shown that nonfatal accidents-which represent 90 to 95 percent of the accident problem in terms of costs-occur with different frequencies than fatal accidents.

In general, in those areas where death rates are high, nonfatal accident rates are low, and where death rates are low, nonfatal rates are high. Several articles on this subject have been run in the Council's Traffic Safety magazine. The first of these, titled "The Fallacy of Fatals" appeared in April 1962.

EXHIBIT 92

ACCIDENT STATISTICS AND RECORDS

The record of the hearings (pts. I and II) of the Subcommittee on Executive Reorganization are remarkable in several ways:

1. Frequent use of statistics-intermixed with frequent criticism of statistics, ranging on to "we know almost nothing about it" (the problem).

Γ