In an July 7 editorial, the Wall Street Journal says “No.” But the editorial has a number of errors that make this conclusion suspect. The editorial also compares the incidence of U.S. highway fatalities and U.S. combat deaths in Iraq, a comparison that is both technically incorrect and hard to interpret.
The Journal editorial, titled “Safe at Any Speed” (subscription may be required), claims that there has been a steady decline in the highway fatality rate since 1994, and therefore repeal of the 55 miles per hour nationwide speed limit in 1995 has been vindicated. This appears to be based largely on a 1999 analysis by Stephen Moore, who at the time was director of fiscal policy studies at the Cato Institute and now is a member of the Journal editorial board.
HIGHWAY FATALITY RATES HAVE BEEN DECLINING FOR DECADES
Statistics show steady declines in highway fatality rates, which are usually measured per 100 million vehicle-miles traveled. This decline is real, but it has been going on for decades. The Journal editorial compares 1995 (1.73 fatalities/108 VMT) and 2005 (1.46 fatalities/108 VMT), but not, for example, 1985 (2.1, as reported by Moore). It’s not surprising fatality rates continued to decline after the nationwide 55 mph speed was repealed. To estimate the effects of repeal, the secular decline first must be removed. It’s possible that this rate slowed because of repeal. The Journal didn’t control for this, but Moore didn’t do so in his 1999 paper, either.
Moreover, the 1974 federal speed limit that was repealed in 1995 was not widely enforced, at least in its later years. As the Journal editorial notes, the “federal speed limit was arguably the most disobeyed and despised law since Prohibition.” This comparison is highly inapt: the federal speed limit probably was much more disobeyed than Prohibition. Unlike the illegal possession of alcohol, speed limit violations were (and remain) endemic and carried little or no social opprobrium. There was no organized crime profiteering on speed limit violations, and no Elliot Ness to bring violators to justice. Firms make a decent legal living selling radar guns to the police and radar-gun detectors to drivers who want to evade them.
In any case, the federal speed limit might have been despised but it could only be routinely disobeyed because the States decided to enforce it lazily. And that leads to another analytic error.
IT’S VERY HARD TO ESTIMATE THE EFFECTS OF REPEALING AN UNENFORCED LAW
Sophisticated methods are necessary to measure subtle changes, and they might be insufficient no matter how sophisticated they are. Simply looking at charts of aggregate statistics, which appears to be the dominant form of Moore’s 1999 analysis, is not sufficient. Peer reviewed studies have been described showing significant increases in fatality rates resulting from higher average speeds on rural interstate highways. These results are intuitively plausible because the laws of physics are hard to deny. Still, offsetting reductions in rural non-interstate fatalities might have occurred. For example, federal speed limit repeal could have shifted traffic from slower rural non-interstate roads with higher risk of head-on collusion to faster rural interstates that have effective median dividers.1 Moore says he controlled for this in his 1999 paper (see footnote 14 and accompanying text), but he did not show his work.
A high quality analysis may be complex, in no small part because drivers constantly adapt their behavior to changing conditions. including changes in speed limits and their enforcement. For libertarians like Moore, for whom the federal speed limit “became a symbol of the heavy hand of the federal nanny state,” repeal itself is an important “changed condition” to take into account. Some states that previously refused to enforce the federally-imposed speed limit may have responded by aggressively enforcing their own speed limits — even if numerical speed limits didn’t change. For safety enthusiasts, it is possible that drivers adapt to engineering and policy innovations that are intended to reduce risk by engaging in more risky behavior. (The analysis of offsetting behavior is most clearly associated with the work of Prof. Sam Peltzman of the University of Chicago School of Business. A nontechnical taste of his work in this area can be found in a 2004 lecture.)
COULD SEAT BELT USE BE MORE IMPORTANT THAN SPEED LIMITS?
Speed is not the only risk factor for highway fatality risk. Others include:
- engineering (including both vehicle and highway design)
- driver skill (including the special risks posed by very young and very old drivers)
- road sharing (including risks posed by heavy trucks on others, and by others on drivers of heavy trucks)
- distraction (including fighting children, eating and drinking, and cell phones)
Seat belts are notable because they are one of the oldest engineering remedies, yet a significant fraction of drivers and vehicle occupants remain unbelted. According to the National Highway Transportation Safety Administration, highway fatalities are concentrated among those who don’t wear seat belts.
|Seal Belt Usage||Proportion of Fatalities|
NHTSA believes that universal seat belt usage would “save thousands of lives,” and that conclusion seems plausible. But the number of thousands of lives saved might be disappointingly small. Voluntary seat belt usage could be a proxy for other risk attributes, such as driver skill or risk aversion. Thus, the implied riskiness of seat belt nonuse, as shown in the table above, might be overstated if those who refuse to use seat belts are just higher risks on the highway.
WHAT DO HIGHWAY FATALITIES HAVE TO DO WITH COMBAT DEATHS IN IRAQ?
The Journal also offers comparative risk statements that are both technically incorrect and logically indefensible:
The tragedy is that 43,000 Americans still die on the roads every year, or about 15 times the number of U.S. combat deaths in Iraq.
This is either a technical error or a knowingly misleading statement. OIF casualties have occurred over a span of 1,206 days. Each of the years 2003 through 2006 has 365 days. It is inappropriate to compare any two dynamic phenomena using different time periods. U.S. highway fatalities average about 105 per day. U.S. combat fatalities in Operation Iraqi Freedom (OIF) have averaged 1.67 per day. Thus, the correct ratio of highway to Iraq combat fatalities is 63, not 15. (We show our work below.)
Our best guess is that the Journal divided annual highway fatalities from all causes by total U.S. fatalities from all causes during OIF. We cannot imagine why.
Highway fatalities are not anything like combat deaths except insofar as both are tragedies for the individuals and families involved. Highway fatalities are accidents experienced by civilians engaged in routine economic activity. Combat fatalities are intentionally inflicted by an enemy. This is like comparing fatalities from fire in commercial buildings in 2000 and 2001, and treating the World Trade Center as a routine conflagration.
It’s unclear what purpose the Journal might have thought would be served by making this comparison. We don’t know how to interpret any ratio between highway fatalities and combat deaths in Iraq. Does the Journal mean U.S. highways are a metaphorical “war zone”? Is the public supposed to care more about highway fatalities because they occur at 15 (actually 63) times the rate of combat deaths? Because the ratio was calculated in an obviously incorrect manner, we are convinced that the comparison wasn’t the product of cogent analysis. But as we have described above, other significant elements of the editorial were problematic as well.
1 The Journal of Safety Research is published by the National Safety Council, an organization that opposed repealing the nationwide speed limit. The authors of this article are all past or current employees of the Insurance Institute for Highway Safety (IIHS). IIHS plausibly has an interest in discerning that repeal resulted in more fatalities, but this is analogous to Moore and the Cato Institute’s interest in proving otherwise. Neutral Source does not criticize research because of author bias or source of funding. Results should be judged on the merits of the analysis — which Neutral Source has not done in either case.
Days elapsed from March 19, 2003, through July 7, 2006:1,206
Total combat deaths if 15-fold ratio to highway deaths was true:
Total 2003: 38,477
Average daily 2003: 38,477/365 = 105.4
Total 2004: 38,253
Average daily 2004: 38,253/365 = 104.8
Total 2005 (projected April 20, 2006): 38,963
Average daily2005 (projected): 38,963/365 = 106.7
Average annual highway fatalities: 43,000
Divided by 15: 2,867
Total U.S. fatalities OIF from all causes:2,540