Econometrics project : what factos lead to deadly car accident throughout the world ?

...

The variables we choose are:

- Number of car per country[1]

- Population in a country

- GDP per country

- Legal age to conduct a car per country

- Alcohol consumption per inhabitant per year

- Km of road network

- CO2 emission

- Diesel price in $/litter

- Maximal speed on road

- Revenue level

- Continent

The number of death on the road per year per country is really simple to find, as many websites gather this data. Other data like number of cars per country, population, GDP, CO2 emission, diesel price per litter, alcohol consumption or the length of the road network were easily found on the Wikipedia page or others website (see sources). Knowing that the Wikipedia page gathers studies coming from other organizations, we had to check whether they were trustworthy or not. For the revenue level come from a document uploaded on the World Health Organization website. It considers that: low revenue level is equal or inferior to $935, intermediary revenue is between $936 and $11455 and high revenue level is equal or superior to $11456. Finally, for the continents we completed with our own geographical knowledge. According to us, this data is relevant when studying road accidents. Indeed, with it, we would like to validate the different assumptions of correlation we stated above.

-

The model and our results

-

Database changes

-

Multiple linear regression model

-

Recommendations and conclusion

-

Recommendations

-

Conclusion

ANNEXES

Fisher test

We use the F-test to check if the model is explained by at least 1 variable (significance of the model). We want to see if here is a linear relationship between at least 1 of the explanatory variables and the number of death.

H0: the model is not significant; all the slopes are equal to 0 (no linear relationship). β1 =0

H1: the model is globally significant. At least one of the slopes is different from 0. β1 ≠0

We have: a=0.05 df1=1 df2 = 6

We found that Fa (the critical value of the F distribution) is close to . Because FSTAT is 107.7, which is higher than Fa, the decision is to reject H0 at a margin of error of 0.05%. Then, significant linear relationships between independent variables and the dependent variable are existing.

VIF

[pic 2]

BIBLIOGRAHY

Link to OMS study about car accident:

http://www.who.int/mediacentre/factsheets/fs358/fr/

Other sources:

http://who.int/violence_injury_prevention/road_safety_status/data/table_a2_fr.pdf

https://en.wikipedia.org/wiki/Road_traffic_safety#Cars

http://www.who.int/en/

---------------------------------------------------------------

...