2. Frequent use of statistics without reference to the definitions accompanying them, resulting in some confusion (hearings, pt. I, p. 282).

3. Frequent use of NSC statistics, in two patterns:

(a) Frequently used without attribution when they were sufficient for the user's purpose (especially our cost estimates); and

(b) Usually used with attribution if they were felt to be inadequate.

4. Almost complete lack of recognition of State and local uses of accident statistics and reports, and coupled therewith, nonrecognition of the fact that such mass statistics (right or wrong) are a byproduct of their major purpose-use at the State or local level to resolve litigation and to identify problems: locations, drivers, and vehicles, individually or in groups.

5. Almost complete unawareness of the fact that research is needed to find "causes," and that mass statistics can only report certain objective events and give indicators or clues as to causative factors.

Further discussion of accident records and accident statistics, for simplicity, ought to be in the context of NSC's statement and appendixes submitted to the subcommittee, much of which pertains to this subject. For example, appendix 2, "The Traffic Accident Problem". Here two questions could be asked (1) Is this analysis meaningful and useful? and (2) If so, are the data sufficiently accurate to justify the conclusions? The answers to both questions are: Yes, judging from the deliberations of a 20 agency task force (hearings, p. 113 ff.)

In addition, perhaps the papers summarizing NSC's traffic accident data project, plus this discussion, should be viewed as another appendix to NSC's statement.

History of work to improve accident records

The council is probably the agency which has most frequently published material on records and statistics, their inadequacies, and ways to improve them. The council has certainly done the most work to improve them.

Prior to World War II, when the council had grants for and used some general funds on accident records, the field saw great improvements. The war interrupted the work.

Since the war there was not a single response from Government or private sources to many NSC requests for grants for records improvement, until 1965. Many NSC requests for funds were turned down. Therefore, until recently, records improvement stagnated.

In 1963 the Bureau of Public Roads received a report on the records situation, which BPR commissioned from Northwestern University (subcommittee hearings, pt. I, p. 467). Based on this and other analyses, the Council's Traffic Conference in July 1964, held a 2-day conference of 36 officials, and university and other professionals at the University of Illinois. The conferees developed many promising lines for traffic accident records improvement. Subsequently, an ad hoc committee of six conferees developed plans for a traffic accident data project under the NSC Traffic Conference. The council was able to budget some new grant funds and some of its general funds for 1965 for administration of the project. The project is now underway. A 5-year plan for research and demonstration, which should culminate in new manuals, procedures, etc., is being executed. It has been emphasized that other agencies, especially the Federal Government, should fund the necessary research and demonstration projects. Already, this is happening with $150,000 of cooperative funds of the Highway Research Board and other BPR funds to a dozen or more States for developing new procedures. Obviously more funds could be used.

The execution of the work outlined by this project will be the primary factor in whether or not accident records are improved at the source and are increasingly useful at State and local levels. A byproduct will be improved accuracy of national estimates.

Accident costs

The NSC estimate of accident costs was used a number of times in the hearings. (In pt. I, President's Committee for Traffic Safety, p. 109; Secretary Connor, pp. 410 and 435; O'Mara, p. 483. Only the latter credited NSC.) No criticisms were voiced.

The NSC estimate is based on a large number of special studies (including the one cited by Moynihan, p. 258) and other continuing data series. The estimates have been reviewed, and been improved by, economists from a variety of agencies over a period of many years.

The costs tabulated are defined to include: wage losses of injured or killed, administration and claims settlement costs of insurance, and property damage.

(See app. 8a, pp. 4 and 5 for fuller details.) The need for more depth studies of costs is described by NSC (app. 3, item 2). Such studies could measure the indirect costs to employers from time lost or time worked at reduced efficiency, the overhead costs of hospitals, wage loss attending funerals, etc., etc. It could be that the real total cost of traffic accidents is several times the present estimate of specific costs. Should we stop using the present estimate until we get broader data? Not likely, the present estimate shows clearly the economic justification for increased accident prevention. Further, it is not clear expensive data collection for larger scope of definitions is warranted on a continuous basis, nor is it clear that larger numbers in the million and billion categories produce more action.

Moynihan testimony

Daniel P. Moynihan, then Assistant Secretary of Labor, presented a detailed analysis of statistics deficiencies. The general points he makes, and more, were made in the Northwestern University report (p. 467) almost 2 years earlier, the Urbana conference 9 months earlier, and in other published articles even earlier. These deliberations and reports also included thoughtful analysis of ways to improve statistics. It is unfortunate that this field has not had the continuing attention of enough interested people during the last decade. Men of Moynihan's stature could make a real contribution if their interest was sustained.

Mr. Moynihan does tend to be rather freewheeling in his rhetoric, his testimony is internally inconsistent, it includes some serious errors, and it is at variance with other testimony at the hearings. Further, it shows some unawareness of literature of his own department. Since he went into considerable detail in the hearings, some of his comments should be analyzed, especially as they conflict with other testimony.

Moynihan says, "The reason we have done almost nothing about the problem of traffic safety is that we know almost nothing about it." This statement seems absurd, but a few examples should suffice.

The 646 pages of the hearing record are filled with reports of activities and accomplishments by the Post Office, the President's Committee, the General Services Administration, the U.S. Public Health Service, the Interstate Commerce Commission, the Department of Commerce, and the Department of Defense. All of the representatives of these agencies felt they could do more, especially with more funds, but none said they were doing "almost nothing" and none said the funds they were now spending on programs were wasted.

The annual traffic inventory reported that 1,063 cities had accident spot maps the first tool in identifying high accident locations.

Mr. Whitton reported "conclusive evidence of measurable, and often drastic reductions in accidents" from improvements of high accident locations and went on to describe the growing size of the BPR spot improvement program (p. 465). The NSC Traffic Inventory shows that State and local traffic engineers put 300,000 man-days of study into design of spot improve ments in 1 year.

The same inventory also reported 10 million convictions for hazardous traffic violations. Unless we assume no deterrent effect, and no educational accomplishments by courts, plenty was done.

The above convictions, plus accident reports, were the basis on which driver license authorities counseled, warned, instructed, and suspended to the total of 3,500,000 actions. Research has shown the efficacy of a "point system" in such programs.

Moynihan says, "I will risk the generalization that the most important step forward towards the solution of any social problem occurs when you learn to measure it." He then comments on traffic accident statistics, some of them in regard to the accuracy of national totals. But he cites no social problems for which there are national statistics as good or detailed as those on traffic accidents.

Moynihan analyzed categories of statistics:

1. Number of accidents.-Moynihan incorrectly projected a sample of a 4-percent sample of the Nation to get his projection of 40 million accidents. The sample was not accidents, it was vehicles. Correcting for this error, and for the urban character of the sample, could cut his guess in half.

The NSC estimate of 12 million accidents is based on city and State summary reports. The NSC has tried to make allowances for underreporting in some

places. Further, the total of city and State reports is more complete than insurance data.

It is most certainly true that many minor accidents do not come to the attention of the authorities. Many of the "fender dents" occur in parking lots, driveways and garages. It is not clear that the expense (and trouble to drivers) to get such reports on a continuing basis would be warranted. Certainly we need special studies of samples, like the one Moynihan cities, but we need to interpret them carefully.

Under this caption, Moynihan praises the directional analysis table developed by NSC and used in many States and cities. The data have been very useful, for example in showing the importance of defenses against particular kinds of accidents in the driver improvement program (app. 10).

2. Injuries in traffic accidents.-There is unnecessary, but understandable, confusion all through the hearing record on the number of injuries. Very few used definitions. Some seemed to pick whatever figures was higher, without consideration as to whether definitions affected relevance to what they proposed or thought.

Moynihan did say there were diffeernces in definitions, but then proceeded to ignore the definitions by saying the difference in estimates was "a probable error equal to the size of the numbers." He's plain wrong.

The smallest sample used in collecting any of the data he cites is that of the National Health Survey, and in this case, the professional statisticians handling the survey compute the sampling error at less than 4 percent.

The National Safety Council tallies cover 100 percent samples from nearly all States, many of which divide their injuries into three categories of severity as provided for by a set of definitions developed by the best experts in the field, and which have been accepted as a standard by the American Standards Association.

Later in his statement, Moynihan suggests use of a national statistical sample, which is what the National Health Survey is.

As few users of injury estimates take the trouble to point out:

The NSC estimate of 1,700,000 injuries is based on an ASA standard which defines an injury as one which prevents a person from performing any of his usual activities for a full day beyond the day of the accident.

The National Health Survey estimate of 3,308,000 motor vehicle injuries includes categories as follows:

Bed disabling_.

Other, with activity restriction_.
Other, medically attended----

1,291, 000 950,000 1,067, 000

There are many other special studies tending to confirm the accuracy of both sets of estimates. The reconciliation is discussed on page 2, Accident Facts (app. 8a). It was also discussed more fully in an August 1958, article in Traffic Safety, "How Many Are Hurt?"

It is another type of mistake to say or imply that these estimates are equivalent to persons "seriously, or permanently injured." The NSC estimate of permanent disabilities from traffic injuries was 140,000 for 1964 still a shocking number, but only 4 percent of the PHS overall estimate.

3. Fatalities in traffic accidents.-Moynihan concedes little difficulty with this statistic, but then implies that the number of deaths is not a full measure of the total accident problem. On this point he is right but hardly original. Many years ago, the National Safety Council presented this point, for a recent example see an article in Traffic Safety, April 1962, "The Fallacy of Fatals." Moynihan says that the traffic death rate is a good statistic, but a "useless one." He points out, correctly, some nonsafety factors which can affect the rate. But is it "useless"? Death rates were cited in the hearings by Senator Ribicoff (p. 2), and by representatives of the Post Office (p. 17), President's Committee (pp. 85-112), Interstate Commerce Commission (pp. 396, 400, 401), Department of Commerce (pp. 459, 480, 483), and Department of Defense (p. 595).

Interestingly, the exhibits Moynihan introduced quote him using death rates, as follows:

"The death rate on the New York State Thruway *** dropped last year to 0.873 * * *" (p. 320).

Table showing comparative death rates for different kinds of transportation (apparently the NSC table) (p. 298).

(Exhibit 93 further discusses death rates.)

4. Rates for deaths, injuries, and accidents.-Moynihan makes the observation that death and injury rates probably have an inverse ratio and he "expects this is primarily a function of vehicle density." This is hardly news, nor does it vitiate their use. Several articles on this point have appeared in Traffic Safety over the last 4 years, and in several issues of Accident Facts. (Traffic Safety, April 1962, p. 8; April 1963, p. 8; July 1964, p. 12. Accident Facts, 1961, p. 46; 1962, p. 45.)

As to the mileages used to compute mileage death rates, Moynihan says, "at the best, a very questionable number." But, the mileage estimates are very carefully worked out by the Bureau of Public Roads and are accepted by the Congress in its consideration of highway planning.

5. Geographical and temporal distribution.-Moynihan here criticizes the holiday death counts tallied by the newspaper wire services. He apparently accepts the validity of the number of deaths, but questions whether the rate of deaths is really worse on the holidays. Since he earlier called death rates "fraud," his interest in having a death rate to evaluate the holiday situation is inconsistent. [Incidentally, rate estimates were first calculated by the U.S. Bureau of Public Roads (pp. 87-89, "The Federal Role in Highway Safety, 1959") and have been extended since by the NSC. They are higher on holidays.]

Also under this heading, he complains of the lack of ongoing data regarding time of day of accidents, and praises two special studies.

Tallying accidents by time of day on each day of the week is very time consuming and costly, and the figures have been found to change very little from one year to another. Consequently, although States have the data, many tally them only from time to time to check patterns. This does not minimize their usefulness, but it does free limited resources for other, more valuable statistical work. 6. Types of vehicles involved.-Type of vehicle is provided for on accident reports as Moynihan correctly points out, and this information is available in State files.

The States have not (to our knowledge) summarized the data, presumably because a tally of accidents by make and model of car would show primarily that there are more of certain cars on the street. Obviously, there will be fewer Ramblers in accidents than Chevrolets because there are fewer Ramblers on the street.

Exposure data would also have to be collected by make and model of car, and for precision this should be in terms of numbers of vehicles, miles driven, where and when driven, and by whom.

Formerly, even the data on numbers were in another file from accidents, and expensive to tally. Now, with the aid of modern computers, some matching of registrations with accident information could be done to delineate the problem. For example, an analysis of variance by make or model for day and night driving, or urban and rural driving, or for young or older owners, or for directional analysis might give some additional clues.

7. Types of driver failure.-Moynihan correctly points out that the investigating officer, who usually is not investigating at all but merely reporting an accident, is not able to identify the true driver failure which "caused" the accident. Moynihan states that it requires research to identify basic driver errors and that this cannot be accomplished from traffic safety statistics, but then he incorrectly lists this as a deficiency of such statistics.

8. Types of vehicle failures.-Again, Moynihan mixes up statistics and research. He properly points out that traffic officers who report information on vehicle failure "have neither the training nor the time to determine, in detail, what may have gone wrong with the vehicle, or what may have been originally wrong with the design of the vehicle."

To point out that data is meager on this point is true, but Moynihan should list this under the heading of research voids and not imply that traffic safety statistics is the villain.

Information on reported vehicle failures was tabulated from accident reports many years ago, but the tabulations were discontinued for the very reasons which Moynihan points out.

The current data on tires as factors in turnpike accidents (see main NSC statement) may turn out to be more useful, partly because turnpike conditions of use are more homogeneous.

9. Types of drivers involved.-Under this heading, Moynihan indicates that the occupation of the drivers would be "most helpful." Such tabulations were made at one time, but most States discontinued them. Now that mass communications are being more carefully designed, the data might be used as to targets.

Perhaps they should be renewed. The occupation groups which comprise the most workers have the most accidents, so no rathes would be implied; that is an area for research, not traffic statistics.

How are mass statistics used at State and local levels? One simple example illustrates this. It was formerly common to blame accidents on nonresident or out-of-State drivers. Mass statistics showed that more than 90 percent of the drivers had their accidents within 25 miles of their place of residence. Preventive efforts were then redirected to local people.

10. Types of roadway and environmental failures.-Moynihan concedes that on this subject no national statistics are needed. It is primarily a local problem, and local highway departments have the records they need. The book "Traffic Control and Roadway Elements" (Automotive Safety Foundation) contains much valuable data.

In his oral testimony, Mr. Moynihan said, "Today there are probably two and a quarter million truck and delivery drivers alone. The fact is that probably the single largest discrete cause of occupational accidents today is automobile accidents. I think that is the case-we don't have any statistics on the subject so we really can't say with any certainty."

For years, the National Safety Council has reported just such figures in Accident Facts, based on reports from State labor departments (app. 8a, p. 31). The information not only shows the relative frequency of injuries in total, but also shows this data for fatalities, permanent injuries, and temporary injuries separately.

As then Assistant Secretary of Labor, it is surprising that he was not familiar with a publication of his own Department which shows injury rates for such categories of truck and delivery drivers as "Motor freight transportation," and "Trucking, local and long distance."

Some State labor department figures show voluminous details relating to vehicle collisions, such as rear end, head on, etc., and injuries in such noncollision accidents as sudden stops or starts, jackknifing, load shifting, etc. Many more details are summarized in Accident Facts (p. 32).

After all this discussion, it still appears that the work of the traffic accident data project is the most significant plan to improve records and statistics. The urgency of records improvement is apparent as the accident problem more and more gets the attention it deserves.

EXHIBIT 93

STATISTICS COLLECTION

In response to the question regarding the Census Bureau's ability to collect information and compile death rates, this probably could be done by the Census Bureau, although this Bureau is not collecting any motor vehicle statistics, or other mortality or morbidity.

The National Vital Statistics Division of the National Center for Health Statistics tabulates certain information on motor vehicle deaths from death certificates, but even this agency has no facilities uniquely suitable for compiling and publishing death rates, especially on a current basis. To have maximum usefulness, such rates must be determined and made available promptly, as is now done by the National Safety Council. NVSD could do this, but it would not be a byproduct of their present activities. Death certificates which NVSD collects could not be used for this purpose because there is no urgency in reporting or compiling such data. (Death totals from death certificates are not available until nearly a year after the end of the year in which the deaths occurred.)

To report monthly traffic deaths on a current basis as is now being done by the National Safety Council, NVSD would have to set up a new and special service, presumably relying on the same State authorities who now report to NSC.

NVSD is uniquely qualified to compile certain additional information regarding motor vehicle deaths from data on the death certificates, but such special tabulations were discontinued in 1949 despite pleas by the council and motor vehicle officials to have the tabulations continued. These special tabulations covered such categories as day of month of accident, hour of accident, duration of