Page 1

HHTSA TECHNICAL REPORT

DOT KS-805 684

THE EFFECTIVENESS OF THE 55 MPH

NATIONAL MAXIMUM SPEED LIMIT

AS A LIFE SAVING BENEFIT

1 '

OCTOBER 1880

Prepared by:

U.S. DEPARTMENT OF TRANSPORTATION

Nalonal Highway Traffic Safety Mnktration

Office of Driver and Pedestrian Programs

1. Report No.

2. Government Access~on No.

DOT-HS-805 694

I

4. f ttle and Subtatle

( The Effectiveness of the 55 MPH National Maximum

' Speed Limit As a Life Saving Benefit

7

i

!

7. A~?ho*'s

I

\Johnson, P., Klein, T.M., Levy, P., Maxwell, D.

II 9 Performong Orgontrot~on Name ond Address

Technical Report Documentotion Page

3. Rec~plenc s Catalog NO.

I

I

1

----

1

5. Report Date

October 1980

b Poformlng Orgon~zataon Code

NTS-41

8. Performtng Orgon~zateon Report No.

I

I L

( 10

1

Work Unt? No. (TRAIS)

,

Contract 01 Grant No.

I

I

Type 01 Report and Perood Covered

i

J

i 400 7th St. S.W.

Washington, D.C. 20590

i

1 12. Soonsov~ng Agency Name and Address

I

I

I 15. Supplementary Notes

I

16. Abstract

U.S. DOT, National Highway Traffic Safety Administration.

11.

13.

4

I

The following report contains an analysis of the life saving benefits resulting

from the 55 mph N?lSL from 1974-1979. Monthly fatality data from 1970-1979 was used

in a time series model to arrive at the estimated safety benefits (lives saved).

The time series model relates changes in monthly fatalities to changes in monthly

vehicle miles traveled, introduction of safety improvements and the implementation

of the 55 mph NMSL law. Increases in highway fatalities in 1976-1979 compared to the

1974-1975 level led to a detailed examination and analysis of the composition of

these fatalities in order to determine possible causes for the increases. Based upon

the available data, it was concluded that 55 mph compliance had eroded somewhat in

1977 and 1978 thus resulting in some fatality increases. The statistical time series

model estimated annual life saving benefits as follows:

LIVES SAVED

I

i

1974

7,532

1975

7,532

1976

7,216

1977

6,794 e

1978

6,423

1979

6,454

TOTAL

41,951

I

17. Key Words

18. Distri.bution Statement

55 XPH Speed Limit, Box-Jenkins Time Seri

Analysis, Intervention analysis, compliance,

speeding involvement, FARS.

19. Securtty Class~l. (01 thts report)

20. Security Clorsil. (of thl s page)

21. No. al Pog.s

22. P19ce

I

I

Form DOT F 1700.7 (8-72)

Reproduction of completed poge authorized

i

rru

Irk'

rmk

L HI

hmhl

A great many reasons have been gut forth in attempting to explain the dramatic

reduction in highway fatalities ohserved to have begun in January 1974, the

offici a1 date of the imposition of the 55 mph National Raximum Speed Limit

(NMSL). The traffic safety literature contains a wide range of conflicting

views as to the causal factors and magnitudes of these reductions in highway

fatalities. These views range from the entire reduction to half of the reduction

in highway fatalities being attributable to the lower speeds. A recent National

Safety Council report 11 claims that "reduced speeds still are a major factor

in keeping down the dezth total, accounting for 44 percent of the difference

between actual and expected totals for 1977. In numerical terms, there would

have been 5,550 additional deaths in 1977, if there had been no 55 mph laws,

and consequently no reduced speeds .I8

The following report contains an analysis of the life saving benefits resulting

from the 55 mph NMSL from 1974-1979. Monthly fatality data from 1970-1979

was used in a time series model to arrive at the estimated safety benefits

( 1 ives saved). The time series model relates chanqes in monthly fatalities

to changes in monthly vehicle miles traveled, introduction of safety improvements

and the implementation of the 55 mph NMSL law.

Increases in highway fatalities in 1976-1979 compared to the 1974-1975 level

led to a detailed examination and analysis of the composition of these fatalities

in order to determine possible causes for the increases. Based upon the

available data, it was concluded that 55 mph compliance had eroded somewhat

in 1977 and 1978 thus resulting in some fatality increases.

The statistical time series model estimated annual life saving benefits as

follows:

LIVES SAVED

THE EFFECTIVEMESS OF THE

55 HPH NATIONAL WIM SPEED LIMIT

AS A LIFE SAVIS BENEFIT

Introduction

Six years have elapsed since the passage of the 55 mph NMSL. A large number

of studies regarding both the fuel saving and safety benefits of this law

have been published. A review of the traffic safety literature reveals (1) a

great variation in the estimated safety benefits of the 55 mph NMSL, and

(2) a lack of national level evaluations covering the total period since

implementation.

Tlis National Highway Traffic Safety Administration (NHTSA) strongly supports

the 55 rnph NMSL and believes that it is one of the most effective countermeasures

to have been used in reducing highway fatalities. In a memorandum by Joan Claybrook,

NHTSA Administrator, dated July 6, 1977, life savings resulting from the

55 mph NMSL were estimated to be between 4,000 and 5,000 per year based upon

analyses up to that time. In October 1979, an NHTSA/FHWA Task Force was

established to review all previous studies of the safety benefits of the

55 mph NMSL and develop a technical report indicating a consensus of the

safety benefits. Due to the lack of valid updated studies, and differences

of opinion among task force members, only a survey report containing a range

of life saving estimates was released. It is strongly felt by the authors

of this paper, that the safety benefits of the 55 mph NMSL have been underesti-

rnated and it is the purpose of this paper to present the authors' views and

conclusions.

The Nature of Speed

Any analysis of the effectiveness of a national speed reduction program such

as the 55 mph NMSL should be accompanied by a discussion of the nature of

speed and why one should expect reductions in fatalities and serious injuries

res u! ting from a reduction i n speed.

The traffi c safety literature is replete with statistical analyses of the

effect of speed and speed changes on accidents and injuries. A most widely

cited study conducted by Solomon -2/ indicates that:

o

Accident severity increased as speed increased, especially at speeds

exceeding 60 mph.

o The fatality rate was highest at very high speeds and lowest at

about the average speed.

o The accident involvement rate and the injury and property damage

rates were highest at very low speeds, lowest at about the average

speed of all traffic and increased at the very high speeds, particularly

at night. Thus, the greater the variation in speed of any vehicle

from the average speed of all traffic, the greater its chances

of being involved in an accident.

3ne can concl~jde that as speed increases from the average speed, accident

involvement rates and severity of injuries increase. Why does this occur?

What factors in the speedhafety relationship cause these events to take

place? The simplest element of the speed/safety relationship which can be

isolated is the concept of the dissipation of kinetic energy resulting from

an accident. Energy is calculated using the vehicle mass, a constant and

the soeed of the vehicle, a variable under the control of the driver.

Wadsworth 3/ defines a motor vehicle accident as an unwanted rate of exchange

of kinetic-energy qiven by the following formula:

K i n e t i c Energy = 4Mass x

Any increase in speed (velocity) increases the kinetic eneray to be dissipated

by the squaw of that velocity rather than the velocity itself. For example,

a 20 percent increase in speed from 50 to 60 mph w i l l result in a 44 percent

increase in kinetic ener9.y to be dissipated, thereby increasing the severity

of accidents and associated injuries.

For any given instant of time, traffic flow.on a segment of road can be describ-

ed by at least two important statistical measures: the mean or averaqe speed

and variation in speed, usually expressed by the standard deviation of the

speed distribution. Solomon showed empirically that the qreater the variation

in speed from the average speed of all traffic, the greater the chance of

being involved in an accident, hence a higher accident involvement rate.

R. Michaels 4/ showed that both accident involvement rates as well as injury

and fatality-rates vary directly with changes in the standard deviation of

travel speeds. As one approaches the average speed, the accident involvement

rates are minimized.

E. Hauer 51 concluded that there is a very stronq correlation between Solomon's

involveme?it rate curves and the number of passive and active overtakings

by passing and passed vehicles. Based upon a mathematical formulation, he

concludes that the number of overtakings is minimal when vehicles travel

at the median speed*. As a vehicle departs from the median speed i n either

direction, it is either overtaken by or it overtakes other vehicles. The

greater the departure, the greater the number of overtakings. The most impor-

tant consideration however, is that overtakings vary di rectly vi th the mafrni tude

of the deviations from the median speed, which in turn leads to vehicle conflicts

resulting in increased accident involvement rates.

Empirically, there is very strong evidence that as the average speed decreases,

a corresponding decrease in the standard deviation is noted. In an analysis

conducted by Cerelli 6/, a reduction in the standard deviation of vehicle

speeds was noted on a11 major rural highway systems after the 55 mph NMSL

was imposed. This was accomplished by greatly reducing the number of vehicles

traveling at higher speeds thereby compressing speeds toward a lower average

speed. Although there is m analytic or theoretical basis for predicting

the reduction in speed variation resulting from a reduction in average speed,

* Ihe median speed is defined as that speed at which half of the drivers

are traveling above and half are traveling below.

it can be shown empirically that the resulting compression of higher speeds

results in a smaller standard deviation and hence less dispersion among the

individual vehicle speeds. Less speed variation in turn results in a smaller

number of overtakings which decreases the probabil ities of crash involvement.

In sumnary, based upon the evidence, it can be argued that lower, uniform

speeds, acceptable to the pub1 ic and properly enforced, produce lower accident

involvement rates. In addition, lower speeds also result in less severe

injuries in the event of an accident.

A National Experiment

Since the imposition of the 55 mph NMSL, the motor vehicle drivinq public

has had six years to modify its driving behavior in keeping with the letter

or spirit of the 55 mph NMSL. If any modification of this behavior has taken

place, it should bc evident in the accident statistics collected during this

time frame. Based upon the previous speed discussion, one should expect

to observe national reductions in accidents, fatal ities and injuries on those

roads affected by the 55 mph NMSL, along with reductions in speed.

One can view this as a simple pre/post experiment comparing accident statistics

before and after the imposition of the law for significant changes in accident

jevels. However, many events took place during the six .year period which

precludes one from performing a simplistic analysis of speed reduction effective-

ness. Specifically, highway, vehicle and driver improvements are continually

being introduced in small increments and must be accounted for. Sufficiently

large changes in the amount of travel as measured by vehicle miles traveled

(VMT) have taken place over the period to be considered an important factor.

From 1970-1979, VMT increased over 33.5% (from 11,181 to 14,931 hundred mill ion

miles). From 1976 on, over half of the States repealed their mandatory motor-

cycle helmet laws resulting in a decrease in helmet use and an increase in

the frequency of fatal head injuries 7/. The mix of vehicle size has been

steadily changing from the larger full size cars down to the compact and

sub-compact sizes thereby increasing the potential for increased severity

of injury. Smaller cars offer less protection in that the kinetic energy

dissipation takes place over a smal ler surface. Theref ore, greater coll apse

and higher injury severity can be expected. Also, some analysts claim that

the recession of 1975 has had significant impact in reducing fatalities during

this period 8/. In addition to these confounding factors, accident data is

not avail able for 55 mph versus non-55 mph roads during the pre and post

eval uati on periods.

Evaluation Approach

A statistical approach using time series analysis 9/ was selected to derive

an impact estimate of the 55 nph NMSL as a life saying benefit. The period

c~vered is the 1970 through 1979 time frame using national monthly fatalities

as the impact measure. The approach taken is to evaluate the baseline fatality

level (non-55 rnph period) for 1970 through 1973 and compare it to the 1974

through 1979 fatality level (55 mph period) considering the confounding factors

previously mentioned. The ultimate result w i l l be the development of a statisti-

cal evaluation model which contains all of the factors considered to signifi-

cantly affect fatalities for the ten year period 1970 through 1979 to determine

the steady stste or averase chanae attributable to the 55 mph NMSL. Further

detailed analysis w i l l be performed on the 1975 through 1979 data to adjust

the steady state change due to the change in the degree of 55 mph noncompliance

which may have taken place during this time frame.

The statistical model consists of a mathematical relationship which excresses

the degree to which the safety index, vehicle miles traveled (a measure of

the volume of travel activity) and the presence of the reduced speed limit,

affects monthly fatalities during the nine year period from January 1970

-December 1979. A least squares statistical technique known as Box-Tiao

Intervention Analysis 10/ was utilized for the multivariate analysis to minimize

the variation betweenfFie model and actual data. The mathematical expression

known as the transfer function relates the time series VMT, speed, and safety

index to fatalities.

One of the problems associated with the analysis of time series accident

data (data collected over equal time intervals) is that the data tends to

be dependent. This means that each paint could be correlated with previous

data points. For example, seasonality and trends in the data represent depen-

dence or autocorrelation. In the case of 12-month seasonality, each data

point is related to a data point occurring 12 months previously. High and

low accident volumes occur during various months of the year, which cause

a seasonal pattern. Trends a1 so represent dependence and autocorrel at ion

in that the data points in an upward trend for each month are generally numeri-

cal ly larger than the previous month. Therefore, each month's accidents

can be expressed as a function of the previous month's accidents. Data depen-

dence or autocorrelation must be accounted for before any meaningful analysis

can be conducted. The Box-Jenkins Time Series Analyses 9/ approach has been

used to determine the time series parameters and the tra?isfer function estimates.

This technique is a generalization of the linear regression model:

where the basic assumption that the covariance (e ,e ) = 0 for ifj represents

a severe constraint for application to traffic aclid4nt data due to factors

such as seasonality. The Sox-Jenki ns technique re1 ies heavily on the autocorre-

lation function (ACF) for the identification of the correlation (normalized

covariance) structure; and permits the parsimonious use of time series parameters

to account for this dependence. Parsimony is the practice of using the least

possible number of parameters for adequate representation.

Design Approach

An ideal design approach for the evaluation model would produce a measure

of the change in the fatal and injury accident levels before vs. after the

imposition of the 55 mph NMSL on two sets of roads: those roads whose speed

limits were reduced to 55 mph versus those roads whose speed 1 imit remained

unchanged, for the period 1970 through 1979. A comparison of the changes

in level would then lead one to conclude whether the 55 mph NMSL was an effec-

tive life saving countermeasure. Since this design must be treated as a

----

q~asi-experimental design, i.e., one in which the sampling units beinq measured

are lli~tselected at random, a mu1 tiple time series desian is most suited

fcr controll'na against a11 threats to internal validity 111. In a multiple

t imp ser ies design, two or more time ser ics are examined =fore and after

some intervention point (imposition of the 55 mph NMSL) to determine if a

chznqe ir trend or a change in the level of fatalities has occurred in the

ex?e.!rnenta:

time series without a corresponding change in a comparison series.

I f 50, cne car then conclude with some statistical confidence that the change

was due to tne intervention effect, provided the intervention was the only

effect beainning in 1974 and continrrinq through the entire time period.

This approach has been successfully applied to the analysis of State fat a1it.v

data, where the data could be segregated by posted speed limit 12,13/.

However, i a this case, since fatalities could not he separated b y s t e e

speed l'vit roa+, total fatalities were used as the impact measure. The

assumption is that if any impact resulted from the imposition of the 55 mph

NVSi, ;t wt~uld oe reflected as a change in the level of fatalities nn 55

mph posted roads which are embedded in total fatalities. In addition, due

to the lack of monthly injury accident data, it was not possihle to measure

t.lw effect of the 55 mph NMSL on the injury accident level.

Evo?utim of the 55 WH Evaluation Uodel

!n the development of a statistical model, an impact measure is selected

rfh'cb reflects the ultimate measure of the process or svstem beinq evaluated.

Ifipgt measures or explanatory variables are selected which loqically affect

the impact measure. The statistical analysis generates the relationships

between the explanatory variables and the impact measure. The goal is to

explain as mi zh of the variation in the impact measure using the least nunber

of logical 1.y related explanatory variables.

The purpose of this part of the analysis was to derive an evaluation model

which yields an estimate of the life saving benefits attributable to the

55 mph NMSL. The final model should relate fatalities to factors affecting

fatallties over time. The first step in the process was to develop a univari-

ate model for fatalities which is based solely on its awn past history.

The univariate model can only be used to forecast future activity based entirely

on the past; it cannot relate the effect of one variable to another. However,

the analyst gains a better understanding of the time series characteristics

of fztalities from the rleconposition of the raw data. Figure 1 is a araph

of monthly fatalities with its 12 month moving average.

FIGURE 1

U. S. Total Fatalfiles

30-08y Definition

With 12 Month Moving Avc~sge

Two observations are imnediately evident:

o

The series is dominated by a distinct seasonal pattern /annual

cycle), and

o

There was an unusually large drop in fatalities between 1973 and

1974.

In addition, the years 1974-1976 were at approximately the same level, with

upward growth during 1977-1978, and a leveling off of fatalities in 1979,

Figure 2 is a graph of the autocorrelation function (ACF) of raw fatalities.

FIGURE 2

Graph Of Autocorrebtbn Function Of Fataliths

.lo

2

4

6

8 10 12 14 16 18 20 22 24

Time Lags

The nondecaying sinusoidal wave pattern, with a period of 12 time lags (months)

is characteristic of the seasonal pattern evident in Figure 1. The series

is nonstationarp in that there appears to be a sustained high correlation

with previous observations at, for example, 12 month intervals ( i .e., observa-

tions at tinbe period t are correlated with time periods t-12, t-24, t-36,

etc. ) . This nonstationarity was eliminated by performi ng a 12 month seasonal

difference of the fatality data:

producing the series depicted in Figure 3 in deviations from the mean.

" Where the p r o c e s m o t remain in equilibrium about a constant mean

-3

lcvcl i.e., having no natural mean. -9/

FIGURE 3

I

U. S. Total Fatakities

( 7 2 Month Differenced)

The seasonal pattern has disappeared from the series, as expected fa1 though

high correlation still exists at lag 12, but not lags 24, 36, etc.). The

large drop in fatalities that occurred between 1973-1974 is manifested as

large negative deviations for the differe~ced data '749'73.

The leveling off of fatalities during 1975-1976 is evidenced by the next

two years of data, and the growth during 197791978 is represented by the

high concentration of positive deviations during '77-'76 and '78-'77. The

autocorrclation function of the seasonally differenced series appears in

Figure 4.

FIGURE 4

Graph Of Autocorrehtbn Function W (7-8 12) Y.

This slowly decaying ACF, which indicates a high degree of serial correlation,

result c mostly from the large negative deviations during '74-'73, and suggests

a change i r ~level in this series. This can be accounted for in either one

of two ways:

(1j a regular difference of the already seasonally differenced series; or

(2) through the introduction of exogenous (external ) variables in the

form of a mu1 tivariate analysis.

For the univar iate analysis, a regular difference

was taken producing the series in Figure 5

U. S. Totat FataMks

(12 Month And Regular Differencing)

zt - zt-i - zt-r2 + zt-13

The series no longer exhibits concentrations of large negative and positive

deviations. Stati onarity has been achieved by approprf ate differencing,

this being supported by the ACF in Figure 6.

FIGURE 6

Graph Of Autocorrebtion Fmctkn (7-B)(7-B 12)Y,

Significant correlations (spikes) appear at lags 1 and 12, but in general,

the ACF is zero elsewhere, with no discernible pattern. At this point, a

tentative model was postulated which accounted for the spikes at time periods

1 and 12.

where Qi = moving average parameter of order i

= Norm8lly and independelrtly distributed ervors NID (o,#),

at

referred to as white noise.

Usinq a least squares procedure, the parameters of the model were estimated

as follows:

The numbers in parentheses below (or above) the equation represent standard

errors of the estimate. This convention will be used throughout the paper.

Diagnostic checks appl ied to the residual series revealed no model inadequacies,

thus this was the final univariate model for monthly fatalities.

In the next phase of model development, an explanatory variable for the effect

of the 55 mph NMSL was introduced as an intervention variable in the form of

zeros and ones to represent the absence or presence of the 55 mph NMSL respectively.

The purpose of the intervention variahle was to detect either a change in

trend or a change in level of fatalities occurring from November 1973 through

December 1979. The effect of this intervention variable (SPEED) contains

the sun total of all of the effects associated with the implementation of

the 55 rnph NVSL, m.,the effect due to enforcement, public information

and education,. etc., and not solely the effect of posted speed limit signs.

The months i n which the 55 mph speed l i m i t became effective i n each State

can be seen in Table 1-14/.

TABLE 1

Months in llhich 55-H Speed Limit

Became Effective in the 50 States

November 1973

Connecticut

December 1973

Alaska

January 1974

Ar 3zona

Delaware

Florida

Arkansas

Hawaii

Mary1 and

California

Massachusetts

New Hampshire

Idaho

New Jersey

North Caro 1 ina

Ma1 ne

New York

North Dakota

Nevada

Oregon

Pennsylvania

Texas

Rhode Island

Virginia

Utah

Vermont

West Virginia

Yisconsi n

Washington

February 1974

March 1974

~ ~

Colorado

Alabama

Geor g i a

I l l i n o i s

Kentucky

Indiana

Louisiana

Iw a

South Carol ina

Kansas

Michigan

Minnesota

Mississippi

Mf ssouri

Montana

Nebraska

New Mexico

Ohio

Oklahoma

South Dakota

Tennessee

Wyomi ng

As can been seen, not a1 1 States reduced their maximum speed limits to 55

mph in the same month. Therefore, November 1973 was selected as the first

date of intervention nationally. The dummy variable SPEED (Xt) is a step

function characterized by:

Xt = 0, t < November 1973

1, t 2 November 1973

Since the ultimate goal of the analysis was to relate the intervention variable

X

to fatalities and since the fatality series must be seasonally differenced

tb'induce stationarity, the following model was hypothesized:

where:

Yt = fatalities in month t

(1 - 812) = seasonal differencing

Wo

= impact of 55 mph NtlSL

Xt = 55 mph intervention variable

b = delay time before impact is felt

nt = nofse series, of the form used in the univariate model

for time series characteristics (ARIMA) .

Equation (5) can be rewritten as:

Y, = roXt-b + Nt;

where (1-B

12

) Nt = nt.

Figure 7 is a graph of (1-8

12

) Xt, expressed in deviations from the mean.

FIGURE 7

lnterventrbn Varhrblb Sped

( f 2 Month Differenced Series)

A comparison of Figure 7 with Fiaure- 3, (I-B 12) Y , supports the appropriate-

ness of Xt in explaining at least some of the chakges that occurred between

1973-1974. The first step in the madel building process is the identification

of the form of the transfer function, which encompasses functions with step

changes (v(0) = w(B)) as well as gradually increasing effects (dB) = w(B)/

&(B)), accounting for any possible delay time (t-b) between activity and

effect. The crosscorrelation function fCCF) is used for transfer function

model identification. The CCF exhibited a statistically significant negative

spike at lag 0 (November la731 and nothina elsewhere. The model forv hypothesized

contained just an w term to account for a step change in fatalities. After

inspection of the A ~ Fof the noise series (n = Yt - wo Ift), a tentative

model was proposed with the fol lowing estimates:

Diagnostic checks of the residuals indicated no model inadequacies. The

55 mph NMSL law is estimated to have forestalled an average of 687.41 fatalities

per month between November 1973 - December 1979.

There can be no doubt that equation (6) consisting of one intervention and

the noise series is a simp1 istic representation of fatalities from 1970-1979.

However is the model a valid representation of what actually occurred? To

answer this question, consider the "process" which generates fatalities.

Durin~the period 1970-1973, fatalities were in a state of equilibrium (growing

at an average rate of 700 fatalities per year, or 1.33%). In November 1973,

the process was loshocked" by some intervention resulting i n a change to a

new but lower level. During 1974-1976, fatalities remained at this lower

level in a state of equilibrium 145,196 -- 44,525 -- 45,523). Therefore,

regardless of the factors influencing fatalities, such as increased exposure

due to increased vehicle miles traveled and the increase in relative safety,

the process was transformed from one state of equilibrium to' another, but

at a much lower level.

The model represented by equation (6) does not reflect changes in exposure

levels as measured by vehicle miles traveled (VMT) nor does it directly account

for the effect of the fuel shortage of 1973-1974. Since different regions

of the country experienced fuel shortages at different time periods with

varying degrees, explicit mdel representation was considered unlikely.

However, it was felt that the introduction of the VMT variable in the model

would account for changes i n expnsure nation all.^, includina the changes result-

ing from the fuel shortage. VMT data is generated by the States using motor

vehicle gasoline revenue and consum~tion data and then supplied to the Federal

Highway Administrati on (FHWA). Although not a1 1 States fcT1-w a uniform

procedure in deriving these estimates, the data i s assumeu LO 3e consistert

between years.

Therefore, in order to account for the fuel shortage and exposure (travel )

a new model was hypothesized of the form:

yt = w1 vt + wZ xt + nt

(7)

where Vt = vehicles miles traveled (x10

8

)

Monthly vehicle miles traveled (VMT) is plotted in Figure 8.

FIGURE 8

U. S. Vehicle Mi&$ Travelbd

With 12 Month Moving Average

(FHWA-Traffic Vo/ume Trends-Table SA)

l&r

With the exception of 1974 and 1979, the series is increasing over time.

Inspection of the autocorrelation function (ACF) sugqested the need for a

seasonal (12 month) difference as depicted i n Figure 9.

FIGURE 9

U. s.Total VMT

(12 Month Differenced Senhs)

The large negative deviations occurring during the differenced periods of

1974 and 1979 indicate the reductions in travel during these years. An August

1975 NHTSA Technical Report -6/ makes the following observations on the change

in VMT durinq 1973-1974.

TABLE 2

PERCENT CHANGE IN INDIVIDUAL MONTHLY TRAVEL 1974 vs. 1973

Jan Feb Mar Apr May Jun Jul Auq

Sept Oct Nov

Dec

kin Rural

-6.6 -9.6 -9.0 -5.7 -3.6 -3.2 -1.5 -1.4 -2.7 -1.4 -2.9 6:9

jcal Rural

-3.8 -5.1 -4.0 -2.4 -1.3 -1.4 -0.1 -1.0 -0.8 0.6 -1.0 4.2

.ban

-2.9 -8.4 -7.7 -4.2 -2.9 -2.4 -1.0 0.8 -1.3 -1.1 -0.6 4.4

I1 Systems -4.2 -8.5 -7.7 ,-4.5 -3.0 -2.6 -1.1 -1.1 -1.8 -1.0 -1.4

5.2 1

i

PERCENT CHANGE I N CWLATIVE MONTHLY TRAVEL 1974 vs. 1973

r

Jan Feb Mar Apr Way

Jun Jul Aug

Sept Oct Nov

Dec

Main Rural

-6.6 -8.1 -8.4 -7.7 -6.7 -6.0 -5.3 -4.7 -4.5 -4.2 -4.1 -3.3

Local Rural

-3.8 -4.4 -4.3 -3.7 -3.2 -2.8 -2.4 -2.2 -2.0 -1.8 -1.7 -1.3

Urban

-2.9 -5.6 -6.3 -5.8 -5.1 -4.7 -4.1 -3.7 -3.4 -3.2 -3.0 -2.4

All Systems

-4.2 -6.3 -6.8 -6.2 -5.5

I -4.9

I -4.3 -3.9 -5.6 -3.4 -3.2 -2.6

"The basic facts depicted in these tables are:

(1) Reduction in travel, which started in July 1973, continued through November

1974, and averaged 2.6 percent for the year.

(2) The reduction was oreater in main rural than in urhan and local rural

areas.

(3) The reduction was more severe during the months of January through March,

with a peak in February of 8.5 percent."

I t is also worth noting the similarity between the seasonally differenced

VMT series and the seasonally differenced fatality series during '740' 73.

(Figures 3 and 9)

The final univariate model dorsved from the VMT series was:

The use of VMT as an independent variable appears justified both from an

exposure point of view and ability to account for fuel shortages. This was

further supported when the cmsscorrelation with fatalities which yielded

a statistically significant spike at lag zero of .46 (approximate standard

error of .11). A transfer function/interventi on model usinq fatalities as

a function of VMT and SPEED, is depicted in equation (9):

The estimate of 1 ives saved due to the 55 mph NMSL rose from 687 (equation

('6)) to 866. A close investigation of the implications of this model will

explain how this change came about. Equation (9) explains fatalities as a

function of vehicle miles traveled, an intervention variable for speed limit

and time series characteristics . As mentioned earl ier, VMT is a pred~minan t 1y

increasing series. Thus the implication is that fatalities should also be

increasing due to increased exposure. This is graphically portrayed in Figure

10, a forecast of fatalities as a function of VMT only.

FIGURE 10

Forecast Of Fataliths As A Firnctbn W VMT

The coefficient (2.41) for VMT i n equation (9) can be considered as a fatality

rate for the 1970-79 time period. For every change of 100 million vehicle

miles traveled on the average a change of 2.41 fatalities can be expected.

However, because this fatality rate is calculated as a constant, it does

not consider the declining fatality rate that actually occurred and therefoiae,

overstated the number of fatalities due to exposure or VMT. The speed l i m i t

dumny variable operates in a decrimenting manner such that the greater the

number of fatalities generated by the constant fatality rate, the greater

the negative effect would be evidenced by the speed limit dumny variable.

As a result of the VMT constant, a higher negative coefficient was calculated

to counteract the overstated expected fatal ities due to increasing VMT which

led to an overstatement of 55 mph NMSL impact.

This observation led to the conclusion that the model with both VMT and SPEED

was not adequate in explaining the behavior of fatalities.

As a result of the establishment of the U.S. Department of Transportation,

automotive safety standards, driver education and training programs, highway

design and construction improvements have had a very positive impact on the

quality of travel in terms of safety. When one observes a steadily declining

fatality rate (fatalities/lOO million VMT) since 1968, it is reasonable to

conclude that the introduction of these factors has contributed to this decline.

A third variable was introduced to account for the differential changes between

VMT and fatalities. This variable represents a measure of safety. An increase

i n VMT, acconpanied by no change in fatalities w i l l cause a reduction in .

the fatality rate per hundred million VMT, i.e., an increase in safety.

Figure 11 is a graph of VMT and fatalities, indexed with 1967 as the base.

Figure 11

Fa talities and Vehicle Miles Truvehd

(1967 = 1.00)

Years

The shaded -area from 1967 through 1973 represents the qrowth in VMT over

and above the growth in fatalities, i.e., fatalities forestalled due to an

increase f~ relative safety on the nation's roads (lower probability of a

fatality per given VMT). In 1974, the large reduction in fatalities occurred

independent from the increased safety on the roads. To illustrate this point,

a fatality index projection is shown in Figure 11. It should be noted that

the area between the VMT index and the fatality projection index (safet.~)

was still increasing through the late 1970's. Critics of the 55 mph NMSL

have proposed that up to 50% of the fatality reduction in 1974 should be

attributed to increased safety in the areas of motor vehicle and highway

improvements. This does not follow from Fiqure 11. Figure 11seems to suggest

that most ifnot all of the reduction in fatalities beginning in 1974 should

be attributed to reduced exposure (VMT) i n 1974 and the 55 mph NMSL. I n

addition the safety effect was still increasing. In order to account for

this safety effect, a third independent variable in the form of the fatality

rate was introduced. A major problem with using the fatality rate as a SAFETY

INDEX was that it also included a change to a lower level, presumably caused

by the 55 mph NMSL.

FIGURE 12

FataRy Rate

Per Hundred Million VMT

(With 12 Month Moving A m p )

5.40

This change in level and an interruption of the smooth downward trend of

the fatality rate can be seen in Figure 12. The 12 month differenced series

is depicted in Figure 13, with the now familiar pattern of negative deviations

durinq '74-'73.

This was presumed to result from to the 55 mph NMSL, since the series has

already been normal ized for exposure (mi leage) . An intervention model was

developed for the fatality rate/speed limit relationship, to determine the

necessary adjustment to be made for the change in level due to SPEED.

The reduction in the fatality rate in 1974, beyond the expected trend, was

estimated to be .575. This estimate (.575) was added to each observation

in the fatality rate series beginning in November 1973, to arrive at .an adjusted

fatality rate called SAFETY for the total period 1970-1979.

FIGURE 14

Comparison Of Fatality Rate

Per 100 Million VMT And Safety Index

With 12 Month Moving A vwage

Adjusted Safety Index

Figure 14 portrays the adjusted safety index projecting what would have occurred

had there been no 55 mph NMSL. This series was used in the final evaluation

model, to offset the growth in VMT. A schematic of the traffic safety system

is presented in Figure 15.

FIGURE 15

Fatality System

VMT and SAFETY operate in opposite directions in order to determine the fatality

level .

The final evaluation model, combinin fatalities as a function of VMT, SAFETY

?

and SPEED was estimated with the fol owing results:

Thus, the steady state estimate of life savings attributed to the 55 mph

NMSL is 627.65 per month (7,532 per year) for 1974-1979. The estimate for

55 mph impact has been reduced to the magnitude of the estimate found in

equation (6), the SPEED only model. The necessity of adjusting the estimate

of 55 mph impact comes from the fact that this estimate (-627.65) represents

a steady state average over the total period rather than a specific estimate

.

for each year. In order to reflect the dynamics of traffic safety, the steady

state estimate was adjusted in order to account for variability in noncompliance

between years. The analysis which follows uses fatalities by speeding involve-

ment for the years 1975-1979 to adjust this steady state estimate. Note

also that the coefficient of VMT has increased to the more conventional magnitude

of 3.52.

Indications of Noncomliance w i t h the 55 WH Speed L i m i t

A review of the fatalities in Figure 6, shows that there has been an increase

i n 1977-1979 compared to the 1974 through 1976 level. Many changes in the

traffic safety level occurred during the 1977-1979 time frame including increasing

noncompliance with the 55 mph NMSL. This section w i l l address these issues

in order to obtain the true effect of the 55 mph NMSL during 1976-1979.

An analysis of fatalities was conducted using data from the Fatal Accident

Reporting System (FARS) in an effort to adjust the steady state estimates

for the effect of the 55 mph NMSL derived in the previous section evaluation

model. The reporting of the variable posted speed limit was not 100% complete

as can be seen in Table 3.

TABLE 3

It is evident that the percent of known posted speed limit fatalities has

been increasing over the years. For the purpose of this analysis, the unknown

fatalities were redistributed into posted speed 1imit categories based on

the distribution of the 80-88 percent known fatalities resulting in a two-

by-two table.

The determi nation of speeding involvement in a fatal ity presented somewhat

of a different problem. The selection process was based upon three data

elements reported to FARS:

(1) Posted speed limit,

(2) Pre-crash traveling speed, and

(3) Too fast for conditions.*

If the pre-crash traveling speed of any vehicle in a fatal accident exceeded

the posted speed limit by at least 5 mph, those fatalities were defined to

be speeding involved. In the absence of either posted or traveling speed,

too fast for conditions was the determining factor. Tables 4A through 4E

depict the output from the algorithm.

TABLE 4A

FATALITIES VS. POSTED SPEED LIMIT

AND SPEEDING INVOLVEKNT 1975

SPDIIYVOL

POSTED

f REOUEMC I

!

PERCENT

!

ROU P C 1

8

COL PCT

' L E S S l M l ! 5 5 * P M RO!UNKMOYN !

'\ 5 s

!ADS

!

! TOTAL

-----------------*--------*-.-..--.4.--.-.-.+

MO S P f t D l W 1WVO ' 114ld ! 1

! 6

! 30294

! 25.64 ! 28.39 ! 14.01 ! 68.04

! 37.69 ! 41.72 ! 20.58 !

! 2

3 ! 1

! 71.75 !

-----------------+--------*--------*-.---.--+

SPtEOlhG IMVOLUE ' 5317 ! 8459 ! 245 ! 14231

! 11.94 ! 16.51 ! 1

! 31.96

! 37.30 ! 5

3 ! 17.25 !

' $1.77 ! 33.82 ! 28.25 !

* Too fast for conditions is driver tactor 44 in FARS.

TABLE 48

FATALITIES VS. POSTED SPEED LIMIT

AND SPEEDIWG INVOCVEIENT 1976

~ ~ E O U E W C V

!

PtnCEWT

!

now PCT

!

COL PC1

'LESS TMA !s5lPH IO!UNKWOYI !

'h 55

!ADS

!

! TOTAL

"---------------*--.....-*--.-..-.*-.--.---+

MO SPEEDING JNvO ! 11714 ! 12751 ! 6326 ! SO817

! 2 5 7 ! 26.02 ! 13.90 ! 67.70

' 0

! 41.39 ! 20.13 !

' 3

! 6558 ! 71.01 !

----0------------*--.--...*..-.-.--+..-..---+

SPLLDlk6 INVOLVF ' 5627 ! 6

! 2164 ! 16?06

! 1

1 ! 14.71 ! 5

6 ! S2.30

! Jb.90 ! 45.53 ! 17.17 !

! 31.62 ! 34.62 ! 21.99 !

-----------------.----.---+.-....-.4-.--.-

I01AL

17161

19450

8912

45523

37.76

3

1

100.00

TABLE 4C

FATALITIES VS. POSTED SPEED LIMIT

AND SPEEDIWG INVOLVEMENT 1977

f IEQUEhC v

#

PCR C E ~ T

n

ROY CCT

!

COL PCT

!LEss I ~ A!ssn~n10 !UWIIOYI

!

!N 55

!ADS

!

! TOTAL

-..--.--------..-+---...--*...-----

*.-+

NO SPttol~blnvo ! 12530 ! 14284 ! 1220 ! 32034

! 26.17 ! 29.63 ! 10.90 ! 66.91

! 1

! 44.59 ! 16.30 !

! b6.65 ! 66.07 ! 69.99 !

-.+

TABLE 4D

FATALITIES VS. POSTED SPEED LIMIT

AND SPEEDIS INWKVEKNT 1978

SPOIMVOL

POSTED

FREQUEkC v

'

PERCEl T

!

ROY PC1

!

COL PCT

!LESS

~o!urr~our

T N A ! ~ ~ ~ P N

!

! 5

!ADS

!

! TOTAL

----------r-----r+r-------+--------+o---o---*

NO sPEEoxn6 1wvO ! 13192 ! 159a7 ! 4258 !

! 27.40 ! 31.7b ! 8064 !

! 40.52 ! 46.97 ! 12.91 !

! 3

! 67.02 ! 68.97 !

,---------.------*,-------+--------*------.-*

SPEEDING JkVOLVE ' 6510 ! 1847 ! 1914 !

! 12.93 ! 15043 ! 3.81 !

! 39.96 ! 48.28 ! 11.76 !

! 32.07 ! 52.98 ! 31.03 !

-----------------*--------+---.----*-------o*

TOTAL

20302

23854

6974

40.36

47.40

12.2r

TABLE 4E

FATALITIES VS. POSTED SPEED LIMIT

AND SPEEDING INVOLVEMENT 1979

F REOUE NC V

'

PECCFNT

!

ROY PC1

'

COL PIT

'LESS tnr!ssnrw ~O!UMKMOUN !

! k 5 5

!ADS

!

! TOTAL

-----------------*--------.---r--------+--------+

NO SPCtDlNG INVO ' 14174 ! lSb72 ! 4414 ! 34460

! 2s-17 ! 50.72 ! 5

! b7.54

4i.ri ! 4 5 4 ! 12.81 !

' 67-29 ! 65.97 ! 74.76 !

-----------------*--------,-----*

SPEEOIhG INVOLVE ' bVC8 ! 8085 ! 1490 ! 16563

! 13.70!

1 5 6 5 2.92!

32.46

! 42-19! 8

1

9.00!

' 32m?l! fb.OS!

21.21 !

-----------------*--------*.-------*-------.+

TOTAL

21362

23757

5904

51023

61-87

6

11.57

100m00

Several observations are worth noting:

(1) The percent of speeding involved fatalities for all posted speed

limits has remained relative'y constant across the years (31.96 --

32.30 -- 33.09 -- 37.37 -- 32.46);

(2) The same is true for percent speeding involved within 55 and less

than 55 posted roads; and

(3) The pel-cent unknown posted speed limit has been gradually declining.

The redistribution of unknown posted speed limit fatalities yield the following

results in Table 5.

TABLE 5

Fatalities the to speeding involvenent by posted speed limit

Nm-55

Change

-55

Change

Using the estimate for lives raved due to the 55 rnph NMSL i n formula (lo),

627.65, yields an annual reduction of approximately 7,532 lives saved per

year. This estimate can be adjusted by using the fatality changes on 55 rnph roads

from Table 5 fop 1976-1979 to arrive at the benefits in Table 6 below.

TABLE 6

Lives saved due to 55 lnph mSL

It should be noted that no reductions to the 55 rnph impact should be made to

account for fatality increases either on roads posted less than 55 rnph or

on 55 rnph roads where speeding was not a contributing factor. Therefore,

for the six year period 1974-1979, almost 42,000 lives have been saved due

to the slower, more uniform speeds resulting from the 55 mph NMSL.

Analvsis of S ~ e dand S

~

D Chanaes

Adjustment to model estimates for changes in noncompliance were derived by

analyzing changes in speeding involvement in fatal accidents. In ordcr to

corroborate gbse estimates, speed data was analyzed. When the 55 rnph NMSL

was imposed,d$ls effect should have been reflected in the speed statistics

derived fm&;fbc FHMA speed monitoring system. This system's data was used

to analyze fese speed changes over time. Annual speed certification data

for FY 1968-1979 was obtained. The average speed and 85th percentile speed

for the U.S. were used to determine:

1. If a considerable reduction in speed occurred timely to the imposition

of the 55 mph NMSL; and

2. If any changes in speeds can be shown for 1976-1979, coinciding

with changes in fatalitfes/fatality rate.

These two determinations, tied to comparable trends in fatal ities, injuries,

fatality rates and injury rates serve as a basis for validation of the estimate

of the number of lives saved due to the 55 mph NMSL.

Trends in speed monitoring data for 1968-1979 were used to determine if the

behavi or of average speed and 85th percentile speed coincided with the imposi -

tion of the 55 mph NVSL. Speed measureme~t data is obtained by FHWA from

the States. For the years 1968-1975 speed data was reported on a FY basis

to FHWA and represented measured speeds for all free flowinq vehicles on

main rural roads only. For FY 1976-1979 speed data was reported quarterly

to FHWA as part of the 55 ~nph speed limit monitoring program. Data for these

years represent measures of speeds of free-flowing vehicles on "straight,

level segments of a State's highway system having a posted or a1 lowable speed

of 55 mph," 15/ i.e., not restricted to main rural roads. Therefore, the

annual figurz were obtained for 1976-1979 by averaging the quarterly data

and are somewhat different from the pre-1976 data. Snme caution should be

taken since pre-1976 data represents speeds on "main rural roadsu and post-

1976 data represents speeds on "55 mph highways."

Annual data for 1968-1978 was obtained for:

average speed

85th percentile speed

total fatalities

total injuries

injury rate

Interstate fatalities

Interstate injuries

Interstate fatality rates

Interstate injury rate

The rates in Table 7 are fatalities and injuries per 100 million VMT. Data

for fatalities, injuries and rates were obtained from the FHWA publication

"Fatal and Injury Accident Rates" for 1975 16/ and 1976 17/. Data

for 1977-1978 was estimated by FHWA based ortrends in malities, injuries

and VMT for earlier years.

In order to make the speed data trends comparable to the trends for fatalities

and injuries, indices were calculated using 1973 as the base year (100) to

reflect changes of the pre- and post-period imposition of the 55 mph NMSL.

Table 7 shows the data adjusted to reflect 1973 as the base year along with

the original data.

Based on Table 7 data, the following observations can be made:

c

Pvorage cpeeds, in;, - & " -

speeds, t?r I.,. r

C a t a l i t y rate

and the Interstate fatality rate declined in 1974-19/e, began to

increase i n 1977.

o Total fatalities and Interstate fatalities declined in 1974-1975,

began to increase i n 1976.

o Total fatalities, fatality rate, Interstate fatalities, Interstate

fatality rate, and Interstate injury rate exhibited a substantial

drop (greater than ten percent) in 1974 from 1973 levels. The

drop in average speeds was approximately five percent, for 85th

percentile speeds, it was approximately 2 percent.

Table 8 shows the distribution of fatalities by highway systems for 1973-

1977. In 1974, the Interstate and Federal Aid Primary systems showed the

largest percent decreases in fatalities from the previous year. These systems

would be where the 55 mph NMSL would have the most impact. For 1975, all

systems except Federal Aid Secondary and Urban (FAS/U) cantinued decreases

over the previous year. The increase in fatalities for FAS/U could be attribut-

able in part to the anticipated FHWA highway system realignment 171 which

reclassified highways using functional criteria (usage characteristics) instead

of administrative criteria (funding characteristics) . The Interstate and

Federal Aid Primary systems showed percent increases in fatalities for bath

1976 and 1977. The largest percent increase in fatalities in 1977 was on

the highway system almost exclusively 55 mph, i.e., the Interstate system

with a 17 percent increase.

TAME. 7

OUIGINCU, MTA

Interstate

Interstate 05th

Year

Amage

Speed

U.S.

Fstalitles

Fatality

Rate

U.S.

Injurles

Injury

Rate

Interstate

Fatal itles

FatalIty

Rate

Interstate

InJur ies

Injury

Rate

Percent 11

Speed

TABLE 8

Distributions of Fatali ties by Highway System, 1973-1977

YEAR

Interstate

Other FA

Primary

FA Secondary

and Urban

Non

FA

Year

Totals

SYSTEM TOTALS 19,173

84,801

75,354

62,700

242,028

(5-yr.

rach cell gives number of fatalities and percent change from previous year.

The distributions of fatalities by highway system were analyzed using the

chi square test statistic 18/. The calculated chi square value with 12 degrees

of freedom was 527.7. Thexagnitude of the calculated value, which is statisti-

cally significant, is due in part to the magnitude of the values in the table.

However, by observing the largest contributions to the chi square value,

it can be seen where the greatest differences in the fatality distributions

1ie.

For the highway systems, the Interstate and FAS/U fatalities made the largest

contributions to the chi square statistic. This was due to 1973 Interstate

fatalities being much qreater than following years. The contribution from

FAS/U is due to the increase in fatalities for 1975 over 1974, again, probably

attributable Ib the system realignment. For the year totals, the.largest

contributionb'b the chi square value were for 1975, which had the lowest

fatality tots# of the five years, and for 1973, the highest fatality total

for the five years.

This comparison of the fatality distributions by highway system demonstrates

that beginning with 1974 there was a considerable drop in the fatality level.

In addition, the greatest reductions in fatalities for 1974 occurred on those

highway systems which would be most inpacted by the 55 mph NMSL -- the Interstate

and other Federal Aid Primary Systems. Also, the increase in fatalities

in 1977 is most evident in the change in fatalities on the Interstate --again

the highway system most impacted by the 55 mph NMSL and consequently most

aff ec ten by increasi ng noncompl iance.

These characteristics of the fatality distributions by highway system can

be matched to similar behavior in the various speed measures.

Table 9 contains the distribution of average speed for the 50 States for

FY 1976-1979. In FY 1976, approximately two years after imposition of the

55 mph NMSL, 59% of the States had an average speed of less than 56.0 mph.

In FY 1977 40% of the States had an average speed of less than 56.0 mph.

For FY 1978, this percentage decreased to 30% of the States. For FY 3979,

the percentage increased to 54% of the States.

TABLE 9

Average Speed

Distribution of States

Percent of States

less than 56.0 mph

59%

These proportions were tested for year to year differences using the parametric

test (t test) for differences in proportions 18/ and compared to the normal

distribution critical value for 5% level of smificance. There was a signifi-

cant difference in the proportion of States for FY 1976 vs. FY 1977. No

other differences were significant.

Table 10 shows the distribution of 85th percentile speed for the States for

FY 1976-1979. In FY 1976, 44% of the States had 85th percentile speeds of

less than 61.0 mph. For FY 1977 this proportion of States was 34%, decreased

to 30% for FY 1978, and increased to 46% for FY 1979. These proportions

were also tested for year to year differences at 5% level of significance.

There was a sign ificant difference between the proportion of States with

85th percentile speeds of less than 61.0 mph for FY 1978 vs. FY 1979. No

other differences were significant.

TABLE 10

85th Percentile Speed

Distribution of States

RangeFlph

Percent of States

less than 61.0 rnph

44%

Table 11 shows the d istrihution of States for percent exceeding 55 mph for

FY 1976-1979. The proportion of States with 50% of the vehicles and above

exceeding 55 rnph was 64% i n FY 1976, increased to 78% in FY 1977, t o 84%

i n FY 1978, and decreased to 70% in FY 1979. Therefore i n FY 1977 and FY

1978, a larger number of States had compliance levels of 50% o r lower. These

proport ions were tested for year to year differences. The difference in

the proportions for FY 1978 vs. FY 1979 was significant at the 5% level of

significance.

TABLE 11

Percent Exceeding 55 MPH*

Distributim of States

Below 30%

30-39%

40-49%

50-59%

60-693

70-79%

80-89%

*Conpi 1ed from FHWA Speed Sumnaries

for the 50 State for FY 1976-79.

During 1976-1978, a large proportioq of States experienced increases in their

dverage speed, 85th percentile speed, fatalities, and fatality rates. Table

12 shows those States with increases for 1976-1978 (fatality rate data not

avail able for 1979).

TABLE 12

1976 vs. 1978

No. of States with

No

Change Increase Decrease

Average Speed

4

32

14

85th Percentile Speed

9

27

14

F a t a l i t i e s

12

32

6

Fatality Rates

8

24

18

From 1976-1978, 32 States' average speed increased, 27 States' 85th percentile

speed increased, 32 States' fatalities increased, and 24 States' fatality

rate increas.:?.

Tables 13A-13D show those States that experienced increases in average speed

and 85th percentile speed with increases in fatalities and fatality rates

for 1976-1978. In Table 13A, 21 States experienced both an increase in their

average speed and an increase in fatal ities. In Table 138, 17 States experienced

and increase inaverage speed and the fatality rate. In Table 13C, 17 States

had increases in 85th percentile speed and fatalities. In Table 130, 15

States had increases in both 85th percentile speed and the fatality rate.

TABLE 13A

AVERAGE SPEED

F

No

A

Change

Increase Oecrea se

Total

T

A

No Change

0

8

4

12

L

I Increase

3

21

8

32

T

I Decrease

1

3

2

6

E

S Total

4

32

14

50

TABLE 138

AVERAGE SPEED

No

Change

Increase Decrease

Tot a1

No Change

1

5

2

8

Increase

1

!7

6

34

Dec yea se

2

!0

6

18

Tot a1

4

32

!4

50

TABLE 13C

85TR PERCENTILE SPEED

No

Change

Increase Decrease

Tota1

Flc Change

2

7

3

12

Increase

6

17

9

32

Decrease

1

3

2

6

Tota1

TABLE 13D

85TH PERCENTILE SPEED

Change

Increase Decrease

Tota1

A

T R

No Chanae

1

4

A A

L T

Increase

3

15

I E

T

Decrease

5

8

Y

Tota1

9

27

3

8

6

24

5

18

14

50

In addition to a large proportion of States (32/50) experiencing an increase

in the averaqe soeed, a large proportion of States has also experienced an

increase in the percentage of vehicles exceeding 55 mph, i.e., a decrease

in the percent compliance. The average percent compliance was obtained hy

'

averaging the quarterly values for each State.

The ranae and median of the averane speed distributions and the 85th percentile

speed distributions for FY 1976 - FY 1979 are shown below. The median of

average speed values increased in FY 1977 and FY 1978, decreased for FY 1079.

A similar result occurred with the 85th percentile speed.

TABLE 14

Ranges of Speed Measures (WH)

Average Speed

Minimum

Median

Max imuw

85th Percentile Speed

Minimum

Median

Maximum

From these observations of the distributions of the speed measures for the

States it can be seen that the increases in average speed, percent exceeding

55 mph, and 85th percentile speeds were concurrent with increases in fatalities.

Table 15 shows the distributions of average percent compliance for FY 1974-

1979. The median of the distribution of the States' percent compliance de-

creased i n FY 1975, increased in FY 1976, and decreased in FY 1977 and 1978.

However, it appears that more States are in the ranges of 30% t o 49% compl iance,

beginning with FY 1977. There appears to be a shift downward in compliance

levels as evidenced by the larger compliance intervals (50% compliance and

above) having fewer States.

TABLE 15

AVERAGE PERCENT C W L IANCE

FY 1974-1979

Interval

Be1 ow 30%

30-39%

40-49%

50-59%

60-69%

70- 79%

^'q ?7d Above

Minimum

Median

Maximum

* Data from 40 States.

+*Data from 29 States.

Tables 16A and 168 show those States that experienced decreases in percent

compli ance and increases in fatal itieslfatal ity rates for 1976-1978. Twenty-

two States that experienced a decrease in the compliance level (i.e., an

increase in the percent of vehicles exceeding 55 mph) -

also experienced an

increase in fatalities.

Seventeen States experienced a decrease in the compliance level and an increase

in the fatality rate 1976-1978. Therefore, the shift downward i n h e States'

compliance levels appears to be associated with increases in fatalities and

fatality rates.

TABLE 16A

X C o w liance

N0

Change

Increase Decrease

Total

T

Increase

I

-

E Decrease

0

3

3

6

S

Total

0

17

33

50

TABLE 168

S Collpli ance

No

Change

Increase Decrease

Total

F

A No

T R Change

0

3

5

8

A A

L T Increase

0

7

17

24

I E

T Decrease

0

7

11

18

Y

Total '

0

17

33

50

The analysis of the impact of the 55 mph NMSL has been conducted in three

phases. Initially, the global model was developed as a starting point to

assess the magnitude of the effect of the reduced speed limit during 1974-

1979. This pointed to the following conclusions:

o The introduction of safety improvements for the vehicle, driver

and the environment has been generally effective in maintaining

the level of fatali ties, offsetting the expected increases due

to increased exposure (VMT).

o The initial impact of the 55 mph NMSL was fairly constant for the

years 1974-1 975.

o

Duringtheperiod1976-1979therewassianificant upwardgrowth

i n fatalities, as compared to 1974-1975.

The second-phase of the analysis was an investigation of what factors contribut-

ed to the increase in fatalities in 1977-1978. This led to an analysis of

fat a1 ities due to speeding involvement by posted speed 1 imit roads. From

this analysis, adjustments were made to the evaluation model estimates, to

account for the effect of noncompl iance. The reconci 1 i ati on of these factors

appears in Table 17.

TABLE 17

MODEL RECOWCILIATION (1ives saved)

55 mph NMSL

7,532

7,532

7,532

7,532

7,532

7,532

Noncompliance with 55

-316

-738

-1,109

-1,078

Total lives saved due

t o 55 nph NMSL

-

rn m-6,-

Based on the analytic formulation on paqe 23, 8 QW rr-ci!ec:e

ivtev1:al f3r

55 mph impact on monthly fatalities in 1974-1975 would yield (597,658).

The value 627.65 per month or an annual 7,532 in Table 17 represents the

best estimate of the 55 mph NVSL impact on the reduction in fatalities.

I t should be remembered that the adjustments made in phase I1 and reflected

in Table 17 are the result of the analysis of fatalities by speedina involvement

and posted speed l i m i t from FARS.

The third phase consisted of analyzing the speed monitoring data to investigate

55 mph compliance over time. This analysis showed that:

o Retween 1977-1978 there was a shift downward in compliance levels

for the individual States.

o The averagespeedand the85thpercentilespeed, after reaching

a low i n 1976 began to rise i n 1977.

o From 1976-1978, 32 States' average speed increased.

o

From 1976-1978, 27 States' 85th percentile speed increased.

o From1976-1978, 22 States experienced a decrease in compliance

level and

-an increase in fatalities.

The results of the phase three malysis appear to validate the model adjustments

derived in phase two.

In conclusion, the 55 mph NMSL is one of the most effective countermeasures

to have been used in reducing fatalities. The effect of the 55 mph NMSL

on fatalities depends heavily upon the compliance level present on the Nation's

roads.

B i b l iography

,

Factors Contributing to the Decrease in Motor Vehicle Fatalities from

0

0

,

Solomon, DO.Accidents on Pain Rural Highways Related to Speed, Driver and

Vehicle. Bureau ot Public Roads, Washington, 0. C., July 1964.

,

Wadswort h, J. The Effect of Conventionally Enforced Maximum Speed Limits

on Motor Vehicle Accidents. National Research Council of Canada, Motor

vehicle Accident btudy Group, Technical Note No. 9, OIHawa. November

1966.

. Michaels, I?. and Schneider, C. The Energy Crisis, Characteristics of Traffic

Flows and Highway Safety. Accident Analysis and Prevention, Volume 8,

pp. 263-270. Pergamon Press, 1976.

, Hauer, E. Accidents, Overtaking and Speed Control. Accident Analysis

and Prevention, Volume 3, pp. 1-13. Pergamon Press, 1971.

, Cerrel 1 i, E. The Effect of the Fuel Shortage on Travel and Highway Safety.

NHTSA T e c h n i c A . 1J.S. Department ot Transportation,

August 1975.

. The Effect of Motorcycle Helmet Usage on Head Injuries, the Effect of

Dsage Lam on Helmet Wearing Rates - A Preliminary Report, January 1979

Yela, D. F. Review of Inforrnatim on the Safety Effects of the 55 MPH Speed Limit

i n the United States, DOT HS-802 383 , May 1917, PP. 3.

Box, G.E.P and Jenkins, G.M. Time Series Analysis Forecasting and Control.

Holden-Day, 1970.

Box, 6.E.P and Tiao, G.C. Intervention Analysis with Applications to

Economic and Environmental Problems. Journal of the American >tatistical

Association, March 1975.

Campbell, D.J. and Stanley, J. C. Experimental and Quasi-Experimental Design

for Research, Chicago Rand-McNally Co. 19b6.

Johnson, P., Klein, T.M., Levy, P., Effect of the 55 nrph Speed Limit Law on Fatal

Crashes i n Texas, DOT-HS-802-172, July 1918.

Klein, T.M., Effect of the 55 mph Speed Limit on Traffic Accidents i n I11inois,

DOT-HS-805-400, A p r i l 1980.

Kahane, C. J., Lower Speed Limits, Reduced Speeds, Fewer Deaths, DOT- HS-801-607,

August 1975.

FHWA Quarterly Speed Sumnary (July-September 1979)

16. Fatal and Injury Accident Rates on Federal-Aid and Other Highway Systems,

T815 u.5. Department of Transportation, Federal Highway Aamlnlstration,

m.

17. Fatal and Injury Accident Rates on Federal-Aid and Other Highway System,

Tm6 U.S. Department of Transportation, tederal Highway Admlnlstration,

September 1978.

18. Hogg, R .V., Craig, A.T. ? Introductim to Mathematical Statistics, 4th

edition, Macmillan Publishing Co., Inc., 1918.