Is Cycling Safety Correlated with State Transportation Policies? RESULTS Elizabeth Luther | Michal Pinto | UP 504 Prof. Scott D. Campbell | Project Results | April 2, 2008
Methodology Summary	We were limited by the amount of data available, and ended up focusing our analysis on whether cycling safety is correlated with a combination of helmet laws and traffic statistics.
Unit of Analysis	State
Dependent Variable	Average of 2000-2005 cycling fatalities per 100,000 persons, by state
Independent Variables	We created the Luther-Pinto Helmet Safety Index (LPHSI) as a ranking for states helmet policies. Scores ranged from 0 (no helmet laws for cyclists) to 107 (extremely strict helmet laws for cyclists) and were based on four criteria: 1. li = The existence of a statewide helmet law 2. mi = The existence of municipal helmet laws 3. ni = Whether the state or its municipalities had age caps on helmet laws (we separated these into three categories: 15 and under, 20 and under, all ages) 4. pi = The year helmet laws were implemented The LPHSI for state i = 50(li) + 4(mi) + [2(ni=15), 4(ni=20), 10(ni=all ages)] + 1(2007- pi) Based on the Luther-Pinto Helmet Safety Index, New York has the most stringent helmet policies, with a score of 107, while fourteen states have no helmet policies whatsoever, with scores of zero. We also looked at a number of traffic statistics by state, including: 1. Total traffic fatalities 2. Percent of traffic fatalities that are cyclist fatalities 3. Percent of traffic fatalities caused by speeding 4. Percent of motor-vehicle fatalities caused by alcohol
Results	Our results were not what we expected. After running multiple correlations and a regression, we determined that, for the most part, cycling safety by state is correlated much more general traffic statistics than it is with helmet policies.

Correlation = 0.064

Shortcomings	Some shortcomings of our analysis: (1) We were unable to find the data necessary to address our initial question (2) Changing our unit of analysis to CITY level data may have yielded stronger results, but we were able to find necessary data (fatalities) only at the state level. (3) Underspecified Model: we were able to find quite a few independent variables we thought might be correlated with cycling fatalities, but due to the aggregation associated with statewide data these independent variables fell short of creating a model to explain cycling safety.
Implications	While the LPHSI is not significantly correlated to cycling safety per our study, we still feel there is an interesting relationship between helmet laws and cycling safety. Looking at the spatial distribution of the LPHSI, states near the coasts and containing large MSAs appear to have more stringent (and more) helmet laws, while states in middle America have less strict or no helmet laws. The safest states, however, are located throughout the country, but primarily in middle America, on the east coast, and in the pacific northwest. This relationship indicates that there are a number of other factors that influence cycling safety throughout the country, for instance: number of cyclists, bike paths, weather, terrain, cycling policies beyond helmet laws, etc.
Recommendations for Future Research	A great follow-up to this process would be to analyze a more robust set of cycling policies at the city level, in addition to addressing the factors discussed above (paths, terrain, weather, speed limits, driver mindfulness, etc).