7.1 Selection of Baseline Method
The first consideration when setting a target is what will be used for a base year value against which the target will be compared. Rolling multiyear averages show long-term trends more clearly than annual counts. The longer the time period for which the average is used, however, the longer it will take for trends to show up in the data. If a multiyear average is used, a State or region will also likely track annual numbers.
7.2 Projection Methods
To compare fatality targets with fatality trends, a linear regression methodology (also known as a "line of best fit") is often used to project future fatality numbers and rates. Most spreadsheet software offers a "linear trend" function, which projects what the fatalities would be in the future if the trend were to continue. It is a good idea to review the "fit" of the linear regression. As each fatality becomes more challenging to reduce in the future, it will be more likely that the trend line flattens out as fewer fatalities are reduced per year. Therefore, States may want to segment the trend line to estimate separate linear trends for different time periods or use nonlinear functions (e.g., exponential, logarithmic, polynomial, or power25) to improve the fit of the trend line with the data. By selecting the R-squared value (in the spreadsheet) to evaluate the fit, different approaches can be compared. The closer the R-squared is to 1, the better the fit of the trend line to the historical data.
To adjust a projection further, other statistical methods such as exponential smoothing or averaging can be used. These techniques are helpful because data collected over time will inherently have some form of random variation. "Smoothing" techniques help reveal more clearly the underlying trend, as well as seasonal and cyclic components. More information about smoothing techniques is available in the Engineering Statistics Handbook by the National Institute of Standards and Technology.
7.3 Feasibility Assessment
Calculation of Annual Rate of Fatality Reduction
While some targets are set for only one year, others are set for as many as 20 years into the future. A check on the level of ambition for a target is to calculate the annual rate of fatality reduction for both the historical trend and the future years and to see how they compare. In the future, as the number of fatalities gets closer to zero, every fatality reduction will become more challenging. Therefore, a significantly larger annual reduction than has been achieved in the past may not be realistic. A Compendium of State and Regional Target Setting Practices developed calculations of annual fatality reductions for each State where a target was available, and provides a reference for benchmarking.
Once a potential target has been defined, agencies should overlay it on the forecasted trend based on past performance. This gives a sense of whether the target seems achievable.