There were considerable differences between my current regression and the previous one. Foremost several multicollernear factors were replaced with stronger independent variables.
The following variables were removed, maintenance, vehicle miles, and secondary. Maintenance was an erroneous variable that was an incorrect measure of fatality rate. Due to population differences it became apparent that every variable, if possible, should be replace with a ratio per capita as oppose to a total. It accounts for population differences between states which could have a serious bias on the data. My data set has been completely normalized. The previous study included a measure maintenance and maintenance per capita. This was multicollernear as well. Vehicle miles were replaced with miles per capita which accounts for population differences as well. The last variable which was removed was the dummy variable secondary which was the primary issue in the previous study. Secondary was given a 1 when primary legislation existed and a 0 which secondary legislation existed. Primary on the other hand was given a 1 when primary laws were enacted and a 0 when secondary laws were present. This resulted in a high multicollernearity between the two variables.
This current study now focuses on the differences between primary legislation and secondary only. Two separate regressions were run. First to determine if primary and secondary have different effects with primary as a dummy variable. The second regression was run to determine the exact differences between the two.
Historically risk compensation seems to be a universal factor in human behavior. My previous study attempted to determine if secondary laws, primary laws, and no legislation resulted in differences between accident fatalities. My model was based off a cross sectional data from the year 2000 and was based on Zapoter’s 2001 study.
Based on previous studies conducted on seatbelt legislation an offsetting behavior occurred which resulted in states with seatbelt legislation having a higher motor vehicle death rate that seemingly nullifies the reasons the legislation was enacted initially. The cause is believed to be a result of risk compensation. People tend to driving more reckless under perceived safer conditions. Safety beats lead to this perceived reduction in risk therefore enforcing seatbelt legislation may increase risky driving and cause more fatalities.
My study currently focuses on the effect of primary seatbelt legislation compared to secondary seatbelt legislation. Primary seatbelt legislation enables an officer to stop a driver if he sees the driver is not wearing a seatbelt, secondary legislation enables an officer to ticket the driver for no seatbelt only if he has been stop for another violation. Based on the intensity of the law I predict that primary seatbelt legislation should increase accident fatalities compared to secondary laws as people may be more likely to wear a seatbelt in states with primary laws.
Abstract: The first test was run with Primary as a dummy variable. The goal is to determine whether or not the 17 states with primary laws had a higher fatality rate that the 32 states with only secondary laws. Unlike previous tests the dependent variable is Death Rate which is given on a per state basis. Death rate included both occupants of the vehicle and non occupants, i.e. pedestrians. Other variables I included was Population Density which is given on a population per square mile, Alcohol, the age of the population, temperature, maintenance per capita, miles per capita, and the variable of safety was reintroduced into the equation. I believe that population density could increase accident rates; however areas with the highest population, such as urban areas, a decline in fatalities could occur because people drive slower in urban areas. Alcohol has been a universal cause of traffic accidents so that was included. For age I looked at the percent of the population who was 16 and older but younger than 24. Young drivers tend to drive more recklessly and less experienced. Temperature is believed to have a direct effect on accident fatalities as temperature rises so do pedestrians and vehicles on the road therefore temperature should increase traffic fatalities. Maintenance per capita is believed to decrease car accidents because cars in better condition are safer to driver. Police per capita determines the safety of the area. Police could cause fewer accidents because people drive slower in heavily enforced areas. On the other hand police might only reduce accidents not accident fatalities people might wear seatbelts due to police enforcement and drive recklessly on the highway where accident fatalities are more frequent. The final variable I included was miles per capita. An increase in miles driven may increase accident fatalities as more driving equal high risk.
FATRATE = -2.54591 +0.613984SPEED* - 0.0001POPDEN + 0.137525ALC + 9.546616YOUNG* – 0.17077PRIMARY + 0.0032485TEMP* + 0.003879MAINPERCAP + 0.00066POLICE
- these variables are removed from the final equation
These results suggest that primary legislation actually decreased accidents by 0.17077. I chose two do a two-sided test at the 95th percentile. Its t-value (|-1.4177|) falls between the significant level (±1.860) which means we fail to reject. Despite the intensity of how the law is enforced it is possible that seatbelt legislation is seen as a single entity which results with the current conclusion. Drivers may view secondary and primary as the same type of law once legislation exist people wear seatbelts.
Population Density seemed to decrease accident death rate by 0.0001 which implies that as population density increases accident fatalities rate decrease. This variable had a t-stat of 1.39 which also in between the significant level. The hypothesis is also significant. This result shows that accident fatalities decrease in heavily populated areas.
Alcohol increase fatalities at a significant level. This theory is one of the most tested theories and my data set confirms that in the year 2000 death rate increase by 0.137525 with a significance of 1.145 which falls between ±1.860. Therefore alcohol did increase accidents.
Surprisingly in this data set younger drivers, given as a percent of drivers between 16 and 24, did not have a significant effect on accident fatalities. The regression suggested that accidents increased by 9.546616. This is extremely high however the t-stat is 3.641424 which is well beyond the limit therefore we can reject this hypothesis.
Temperature has a coefficient of 0.032485 indicating that as temperature increase so does death rate by 0.032485. However this value has a t-stat of 4.28 far beyond the limit. This is rejected.
Maintenance per capita increases by 0.003879 suggest an offsetting behavior that maintenance increases accident fatalities. This was significant with a t-stat of 1.87 which is just above the limit. However I removed several variables in my final equation lower the degrees of freedom and thus increasing the critical t-value to 1.9.
Miles per capita decrease by an ever so small amount of 9.6 * 10¬^-6 with a significance of 0.21525 suggesting that drivers become more skilled at they drive and thus reduces the death rate.
Safety was the final factor that was removed in my previous study however in this study it was significant with a t-stat of 0.453306. This suggests that police increase death rate by 0.00066. The cause is hard to determine.
Speed limit was a non normalized dummy variable. A 1 was given to any state with a max speed limit of 65 mph or higher and a 0 for any limit under 65. The problem was that only one state, Hawaii, had a speed limit lower than 65. Therefore this can be determined simply that the US has a 0.61 higher death rate than Hawaii.
There were considerable differences between my current regression and the previous one. Foremost several multicollernear factors were replaced with stronger independent variables.
The following variables were removed, maintenance, vehicle miles, and secondary. Maintenance was an erroneous variable that was an incorrect measure of fatality rate. Due to population differences it became apparent that every variable, if possible, should be replace with a ratio per capita as oppose to a total. It accounts for population differences between states which could have a serious bias on the data. My data set has been completely normalized. The previous study included a measure maintenance and maintenance per capita. This was multicollernear as well. Vehicle miles were replaced with miles per capita which accounts for population differences as well. The last variable which was removed was the dummy variable secondary which was the primary issue in the previous study. Secondary was given a 1 when primary legislation existed and a 0 which secondary legislation existed. Primary on the other hand was given a 1 when primary laws were enacted and a 0 when secondary laws were present. This resulted in a high multicollernearity between the two variables.
This current study now focuses on the differences between primary legislation and secondary only. Two separate regressions were run. First to determine if primary and secondary have different effects with primary as a dummy variable. The second regression was run to determine the exact differences between the two.
Historically risk compensation seems to be a universal factor in human behavior. My previous study attempted to determine if secondary laws, primary laws, and no legislation resulted in differences between accident fatalities. My model was based off a cross sectional data from the year 2000 and was based on Zapoter’s 2001 study.
Based on previous studies conducted on seatbelt legislation an offsetting behavior occurred which resulted in states with seatbelt legislation having a higher motor vehicle death rate that seemingly nullifies the reasons the legislation was enacted initially. The cause is believed to be a result of risk compensation. People tend to driving more reckless under perceived safer conditions. Safety beats lead to this perceived reduction in risk therefore enforcing seatbelt legislation may increase risky driving and cause more fatalities.
My study currently focuses on the effect of primary seatbelt legislation compared to secondary seatbelt legislation. Primary seatbelt legislation enables an officer to stop a driver if he sees the driver is not wearing a seatbelt, secondary legislation enables an officer to ticket the driver for no seatbelt only if he has been stop for another violation. Based on the intensity of the law I predict that primary seatbelt legislation should increase accident fatalities compared to secondary laws as people may be more likely to wear a seatbelt in states with primary laws.
Abstract: The first test was run with Primary as a dummy variable. The goal is to determine whether or not the 17 states with primary laws had a higher fatality rate that the 32 states with only secondary laws. Unlike previous tests the dependent variable is Death Rate which is given on a per state basis. Death rate included both occupants of the vehicle and non occupants, i.e. pedestrians. Other variables I included was Population Density which is given on a population per square mile, Alcohol, the age of the population, temperature, maintenance per capita, miles per capita, and the variable of safety was reintroduced into the equation. I believe that population density could increase accident rates; however areas with the highest population, such as urban areas, a decline in fatalities could occur because people drive slower in urban areas. Alcohol has been a universal cause of traffic accidents so that was included. For age I looked at the percent of the population who was 16 and older but younger than 24. Young drivers tend to drive more recklessly and less experienced. Temperature is believed to have a direct effect on accident fatalities as temperature rises so do pedestrians and vehicles on the road therefore temperature should increase traffic fatalities. Maintenance per capita is believed to decrease car accidents because cars in better condition are safer to driver. Police per capita determines the safety of the area. Police could cause fewer accidents because people drive slower in heavily enforced areas. On the other hand police might only reduce accidents not accident fatalities people might wear seatbelts due to police enforcement and drive recklessly on the highway where accident fatalities are more frequent. The final variable I included was miles per capita. An increase in miles driven may increase accident fatalities as more driving equal high risk.
FATRATE = -2.54591 +0.613984SPEED* - 0.0001POPDEN + 0.137525ALC + 9.546616YOUNG* – 0.17077PRIMARY + 0.0032485TEMP* + 0.003879MAINPERCAP + 0.00066POLICE
- these variables are removed from the final equation
These results suggest that primary legislation actually decreased accidents by 0.17077. I chose two do a two-sided test at the 95th percentile. Its t-value (|-1.4177|) falls between the significant level (±1.860) which means we fail to reject. Despite the intensity of how the law is enforced it is possible that seatbelt legislation is seen as a single entity which results with the current conclusion. Drivers may view secondary and primary as the same type of law once legislation exist people wear seatbelts.
Population Density seemed to decrease accident death rate by 0.0001 which implies that as population density increases accident fatalities rate decrease. This variable had a t-stat of 1.39 which also in between the significant level. The hypothesis is also significant. This result shows that accident fatalities decrease in heavily populated areas.
Alcohol increase fatalities at a significant level. This theory is one of the most tested theories and my data set confirms that in the year 2000 death rate increase by 0.137525 with a significance of 1.145 which falls between ±1.860. Therefore alcohol did increase accidents.
Surprisingly in this data set younger drivers, given as a percent of drivers between 16 and 24, did not have a significant effect on accident fatalities. The regression suggested that accidents increased by 9.546616. This is extremely high however the t-stat is 3.641424 which is well beyond the limit therefore we can reject this hypothesis.
Temperature has a coefficient of 0.032485 indicating that as temperature increase so does death rate by 0.032485. However this value has a t-stat of 4.28 far beyond the limit. This is rejected.
Maintenance per capita increases by 0.003879 suggest an offsetting behavior that maintenance increases accident fatalities. This was significant with a t-stat of 1.87 which is just above the limit. However I removed several variables in my final equation lower the degrees of freedom and thus increasing the critical t-value to 1.9.
Miles per capita decrease by an ever so small amount of 9.6 * 10¬^-6 with a significance of 0.21525 suggesting that drivers become more skilled at they drive and thus reduces the death rate.
Safety was the final factor that was removed in my previous study however in this study it was significant with a t-stat of 0.453306. This suggests that police increase death rate by 0.00066. The cause is hard to determine.
Speed limit was a non normalized dummy variable. A 1 was given to any state with a max speed limit of 65 mph or higher and a 0 for any limit under 65. The problem was that only one state, Hawaii, had a speed limit lower than 65. Therefore this can be determined simply that the US has a 0.61 higher death rate than Hawaii.