Transcript ECON 240 A GROUP 5

AUTO FATALITY FACTS 2007
ECON 240 A
GROUP 5
Yao Wang
Brooks Allen
Morgan Hansen
Yuli Yan
Ting Zheng
Overview



More men than women die each year in motor vehicle crashes. Men
typically drive more miles than women and more often engage in
risky driving practices including not using seat belts, driving while
impaired by alcohol, and speeding. Crashes involving male drivers
often are more severe than those involving female drivers.
We analyze car crash fatality data for 2007, and run several
regressions to try to determine the likely causes of fatality.
We find that being male, being young, and alcohol all significantly
contribute to the probability of dying during a car crash.
Descriptive Statistics
Percentage of vehicle fatalities by gender, 1975-2007,
taken from “Fatality Facts 2007”
The age distribution in car accident
Table One: Histogram of age distribution in car accident
6000
Series: AGE
Sample 1 65535
Observations 65535
5000
4000
3000
2000
1000
0
0.0
12.5
25.0
37.5
50.0
62.5
75.0
87.5 100.0
Mean
Median
Maximum
Minimum
Std. Dev.
Skewness
Kurtosis
38.13513
34.00000
99.00000
0.000000
21.82044
0.789594
3.170004
Jarque-Bera
Probability
6888.654
0.000000
Analysis Description


We gathered our data from the U.S. Department of Transportation’s
Fatality Analysis Reporting System, “FARS.”
We classify drivers by gender, age group, and alcohol consumption
for our independent variables, then run linear probability regressions
to try to find a relationship with our dependent variable, fatality.
Expectations
Based on historical data, we
assume that males drive more
dangerously
In addition, alcohol should play a
very significant role in vehicle
fatalities
We also expect that the very
young and very old age groups
will have higher fatality rates,
due to less experience and
poor coordination, respectively
Dependent Variable: FATAL
STATISTCAL ANALYSIS
Fatal vs. Male
Method: Least Squares
Date: 12/03/08 Time: 15:17
Sample(adjusted): 1 65535
Included observations: 65535 after adjusting endpoints
Variable
Coefficient
Std. Error
t-Statistic
Prob.
MALE
0.096006
0.004052
23.69246
0.0000
C
0.369374
0.003278
112.6726
0.0000
R-squared
0.008493
Mean dependent var
0.432212
Adjusted R-squared
0.008478
S.D. dependent var
0.495387
S.E. of regression
0.493283
Akaike info criterion
1.424563
Sum squared resid
15946.01
Schwarz criterion
1.424840
F-statistic
561.3325
Prob(F-statistic)
0.000000
Log likelihood
Durbin-Watson stat
-46677.35
2.170404
*Male in a car
accident is a bernoulli
variable with 0
(female) and 1(male).
*As shown in the
table, the t-stat and
F-test are both highly
significant.
* The coefficient
shows a 9% increase
in the probability of
death given that you
are male.
Fatal vs. Age
Dependent Variable: FATAL
Method: Least Squares
Sample(adjusted): 1 65535
Included observations: 65535 after adjusting endpoints
Variable
Coefficient
Std. Error
t-Statistic
Prob.
AGE
0.002345
8.82E-05
26.57916
0.0000
C
0.342801
0.003876
88.44925
0.0000
R-squared
0.010665
Mean dependent var
0.432212
Adjusted R-squared
0.010650
S.D. dependent var
0.495387
S.E. of regression
0.492742
Akaike info criterion
1.422369
Sum squared resid
15911.08
Schwarz criterion
1.422647
F-statistic
706.4517
Prob(F-statistic)
0.000000
Log likelihood
Durbin-Watson stat
-46605.49
2.161306
*Fatal is a Bernoulli
variable set up as: 0
(alive) and 1(death). A
motorist either lives or
was fatally wounded.
The t-stat and F-test
are both highly
significant, with very
low probabilities.
*Durbin-Watson stat
is close to 2, which
indicates there is not
enough evidence of
autocorrelation.
*Coefficient of age
means the probability
of death will increase
0.2345% per age.
Fatal vs. Alcohol
Dependent Variable: FATAL
As expected,
alcohol plays a
part in motor
vehicle fatalities.
Method: Least Squares
Sample(adjusted): 1 65535
Included observations: 65535 after adjusting endpoints
Variable
Coefficient
Std. Error
t-Statistic
Prob.
ALCOHOL
0.010260185
0.000481
21.288930
3.1471935e-100
C
0.39117213
0.002726
143.45659
0
R-squared
0.00686833
Mean dependent var
0.432211724
Adjusted R-squared
0.00685322
S.D. dependent var
0.495387226
0.4936868
Akaike info criterion
1.42619961
15972.139452
Schwarz criterion
1.42647709
F-statistic
453.218549
S.E. of regression
Sum squared resid
Log likelihood
-46730.9955
Although the
coefficient is small,
it is still a positive
factor in fatalities.
Total Regression
Dependent Variable: FATAL
Method: Least Squares
Sample(adjusted): 1 65535
Included observations: 65535 after adjusting endpoints
Variable
Coefficient
Std. Error
t-Statistic
Prob.
MALE
0.110633
0.004028
27.46802
0.0000
AGE
0.002614
8.77E-05
29.81126
0.0000
ALCOHOL
0.011945
0.000478
24.97006
0.0000
C
0.212337
0.005301
40.05712
0.0000
R-squared
0.029674
Mean dependent var
0.432212
Adjusted R-squared
0.029629
S.D. dependent var
0.495387
S.E. of regression
0.487993
Akaike info criterion
1.403030
Sum squared resid
15605.37
Schwarz criterion
1.403585
F-statistic
668.0061
Prob(F-statistic)
0.000000
Log likelihood
Durbin-Watson stat
-45969.78
2.153951
It is apparent that all 3
factors (male, age and
alcohol) all have a
positive effect in
automobile fatalities
Age*Alcohol vs. Fatal
1.0
This graph shows that old
drinkers are more
dangerous drivers than
young drinkers (higher
probability of a fatal crash)
FATAL
0.8
0.6
0.4
0.2
0.0
0
200
400
600
AGE_ALCOHOL
800
1000
Dummy variable regression of age
Dependent Variable: FATAL
Method: Least Squares
Date: 12/03/08 Time: 20:33
Sample(adjusted): 1 65535
Included observations: 65535 after adjusting endpoints
Variable
Coefficient
Std. Error
t-Statistic
Prob.
YOUNG_PEOPLE
0.453953
0.002388
190.0875
0.0000
RETIRED_PEOPLE
0.392729
0.007240
54.24392
0.0000
MIDAGE
0.328425
0.003968
82.76246
0.0000
R-squared
0.015119
Mean dependent var
0.432212
Adjusted R-squared
0.015089
S.D. dependent var
0.495387
S.E. of regression
0.491636
Akaike info criterion
1.417888
Sum squared resid
15839.45
Schwarz criterion
1.418304
F-statistic
502.9973
Prob(F-statistic)
0.000000
Log likelihood
Durbin-Watson stat
-46457.63
2.171381
Young_people=1*(age<20)+0*(age>20)
Retired_people=1*(age>65)+0*(age<65)
Midage=1-Young_people-Retired_people
We generated 3 dummy
variables for young,
middle aged and retired
people
The results indicate that
young people are the
most dangerous, then
retired, and lastly middle
aged drivers.
Fatal vs. Drunk men
Dependent Variable: FATAL
Method: Least Squares
Date: 12/04/08 Time: 03:43
Sample(adjusted): 1 65535
Included observations: 65535 after adjusting endpoints
Variable
Coefficient
Std. Error
t-Statistic
Prob.
MALEALCOHOL
0.016395
0.000520
31.51293
0.0000
C
0.391478
0.002315
169.0980
0.0000
R-squared
0.014927
Mean dependent var
0.432212
Adjusted R-squared
0.014912
S.D. dependent var
0.495387
S.E. of regression
0.491680
Akaike info criterion
1.418052
Sum squared resid
15842.53
Schwarz criterion
1.418329
F-statistic
993.0647
Prob(F-statistic)
0.000000
Log likelihood
Durbin-Watson stat
-46464.01
2.138243
If you are drunk AND
male, you are more
likely to die than if
you were just a
random drunken
motorist.
Fitted vs. Age
0.7
FITTED
0.6
Male
0.5
0.4
Female
0.3
0.2
0
20
40
60
AGE
80
100
This table shows the fitted value
of “fatal” against “age.” Note
that the survival rate is highly
dependent on sex; a female in a
car accident has a greater
chance of surviving.
Middle aged drinkers
FITTEDALCOHOLMID
0.7
This represents a
somewhat more accurate
representation of the
population of drinking
drivers by eliminating
under-aged drivers and the
elderly.
Drunk
0.6
0.5
0.4
It shows that for this age
group, drinking significanlty
increases the chances of
death.
Sober
0.3
0.2
20
30
40
50
MID_AGE
60
70
CONCLUSIONS:
*The results were fairly self-explanatory and
consistent with our expectations, and most
people’s common sense.
*Males are more likely to be involved in a auto
fatality, alcohol increases the chances as well, and
being young makes you a more dangerous driver,
in general.
*Google has some crazy pictures of car accidents!
Sweet pic!