Costs and benefits of a bicycle helmet law for Germany

van Wee and Rietveld (2013) discuss ethical aspects of using the value of statistical life (VSL) for the ex ante evaluation of transport policy options.

The DDD (difference in difference in differences) estimate is −0.031 with a standard error of 0.015 (Carpenter and Stehr 2011, Table 5), the baseline rate of cycling is estimated to 71 % (Carpenter and Stehr 2011, Table 3). The point estimate therefore is 0.031/0.71 = 4.4 % with a 95 % confidence interval of 0.2–8.5 %.

See Mokhtarian and Chen (2004) for a discussion of travel time budgets.

Furthermore, in the HEAT algorithm, output depends also but only slightly on the number of cyclists.

Hagel and Yanchar (2013) calculate a value of 20–40 % of head injuries in bicycle injuries for Canada and Dinh et al. (2010) report that 25 % of trauma admission registered in the Royal Prince Alfred Hospital, Sydney, Australia, had head injuries.

The odds ratio is different from the relative risk and the odds ratio will always exaggerate the size of the effect, compared to a relative risk. Using the estimate of an odds ratio as risk reduction, as in this study, slightly overestimates the actual risk reduction of bicycle helmets.

If there are only data about injuries, a weighted average of \(S_m=0.027{\cdot}VSL\) per injury is used.

Input data for the algorithm are that 63,510,000 million cyclists, that is, the unhelmeted 87 % of B _f 73,000,000 cyclists in Germany, reduce cycling by 0.18840 km at 124 days and additional walk a distance of 8.0534 km annually, resulting in an annual reduction of cycling W _s additional walking of \(W^s_p\).

Risk factors of cars may be biased by a large share of safe motorways.

Börjesson and Eliasson (2012) find that cyclists generally take the health effects into account when making their choices.

The occasional helmet user has utility losses that are smaller than \(c_iV_{km}^h\) when wearing a helmet and greater than \(c_iV_{km}^h\) when not wearing the helmet. Therefore, losses are greater than or equal to zero when wearing a helmet and greater than or equal to \(c_iV_{km}^h\) when not wearing a helmet. For cyclists not owning a helmet utility losses exceed the gains from helmet protection minus purchasing costs. To cover this range of lower bounds of rational utility losses, the interval \([ 0, c_i \cdot V_{km}^h ]\) is used in a sensitivity analysis (see “Sensitivity analysis” section) and ul is defined as midpoint of the interval.

Broadstock and Collins (2010) show that prices influence the demand for cycling to a greater extent than the income effects.

Normal values are explained in Chapter 2 and 3. Low and high values for rr are the boundaries of a 95 % confidence interval (Elvik 2013). q _head low and high values are from Hagel and Yanchar (2013). Following Bickel et al. (2005, p. 87) , the low value of VSL is VSL/3 and the high value \(VSL\cdot 3\). The low value of C _H is the price net of sales taxes for the cheapest good adult helmet according to Stiftung Warentest (2012). The high value of C _H is the average recommended retail price net of sales taxes for the twelve best-selling helmets sold by amazon.de December 12th, 2013. Low and high values for r are the boundaries of a 95 % confidence interval (see Footnote 2). The value \(ul=0\)  € indicates that there are no losses due to Comfort or Style, the value of \(ul=0.0125\)  € indicates that helmet owners who never wear the helmet in the status quo are rational in the sense that utility losses due to wear are at least as high as the expected (internal) benefits due to protection.

Furthermore, the pure health effect (not including costs of helmets and environmental effects) gets negative if there is more than a 7.5 % reduction of cycling.

Curnow (2003) points out that the design of helmets is based on the theory that linear acceleration is the main cause of brain injury. Rotation of the head, which can even be increased by a helmet (Corner et al. 1987), is ignored.