Decision Rules and Routines

Within the constraint of regulatory philosophy there is considerable room for discretionary judgment. Given the inherent uncertainties in setting safety standards, however, it is unclear how standards-setters structure and exercise this discretion. In other words, what are the rules of evidence, formal and informal? Some version of cost-benefit analysis is the formal answer in the public sector, although agencies are often accused of circumventing the requirement. The mystery widens in the private sector, where cost-benefit analysis is almost never done. Private safety standards often precede any information on accidents or injuries. UL listed woodstoves for decades without the benefit of any injury estimates. So what does the private sector do instead? How does it actually make decisions? The cases suggest that heuristics and other standard operating procedures have evolved to simplify the resolution of these complex safety questions in the private sector.

One technique is to defer to "professional judgment" within certain constraints. In other words, so long as particular provisions do not cost too much, engineers are given free rein in making the decisions. This strategy requires only rough estimates of costs and not necessarily any formal assessment of benefits. It also results in a substantial amount of guesswork. UL prefers to call this "engineering judgment." Others use the phrase "educated guess," with the stress on educated . In either case, the implication is that guesses are guided by a combination of education, experience, and values. This explains both the importance of engineering ethics and the potential for unreasonably strict (but economically inexpensive) provisions in private standards. More costly and salient issues are not resolved so informally; they are subject to the political process described in the next chapter.

A related approach that also does not require extensive information about costs and benefits is to concentrate on ominous hazards such as

electrocution and amputation. As a staff engineer at NFPA put it: "I can tell you how to protect against a hazard. I can't tell you how likely it is to happen." With ominous hazards, the probabilities are unimportant. Participants in both public and private standards-setting express the sentiment that certain hazards are obviously worth regulating. "One case can give you the answer," explains a UL employee. "Somebody loses a finger, let's fix it. I don't need information on a thousand cases." A former CPSC employee agrees: "If the problem is really significant, like the amputation of fingers or hands in the snowblower, you don't have to do a cost-benefit analysis to say that a twenty dollar control is going to pay for itself."

"Through almost intuition you can come up with cost-benefit analysis," explains an NFPA employee, who boasts that this intuition resulted in the requirement for ground-fault circuit interrupters. These devices, added to the National Electric Code in 1975, automatically trip the circuitbreaker when there might otherwise be serious or fatal shock. The irony is that a subsequent cost-benefit analysis conducted by the National Bureau of Standards casts doubt on the wisdom of the standard, concluding that the cost per life saved could exceed $7 million.^[1]

Considering these tendencies in two dimensions—that is, separating estimates of cost from evaluations of benefits—suggests that there are different patterns of preference in the public and private sectors concerning whether and when to err on the side of safety. Four combinations of estimation errors are possible. Two of these patterns are suggestive of regulatory outcomes; the other two are ambiguous. A standards-setting system might tend to under estimate costs and over estimate benefits. Because both estimation errors favor regulation, the resulting standards, relative to other combinations of estimation errors, would tend to be the strictest, bordering on overly protective (see table 6). Public standards-writing is generally thought to have these characteristics. Conversely, a standards-setting system in which costs are over estimated and benefits under estimated would tend to be most lenient, with a danger of being too lax. That is the conventional wisdom about private standards-setting. The other combinations of estimation errors are ambiguous because the errors are in different directions.

Two caveats are necessary before carrying this analysis further. First, these descriptions are not meant to be pejorative. Overestimation, in this context, is not deliberate puffery, nor is underestimation intentionally deceptive. Rather, these terms describe tendencies that are akin to

legal rules of evidence—they encompass various presumptions and informal rules concerning the burden of proof and the resolution of scientific disputes under uncertainty. Second, the typology illuminates differences in the direction, not the magnitude, of these tendencies. Standards classified in any of the four quadrants might be socially desirable, depending on the magnitude of the errors. Even with this ambiguity, the observed tendencies in the two sectors are quite revealing.

Moving first to how costs are estimated in both sectors, it is widely assumed that private standards-setters err on the high side, while government generally errs on the low. Private organizations have first-hand knowledge of costs and are likely to be sensitive to them, erring in the direction of the industry's concerns about profitability. In contrast, government lacks this first-hand knowledge and is likely to discount industry estimates of costs for fear they have been exaggerated for strategic purposes. Government also has no direct interest in, and often little actual concern about, how regulations affect profitability.

This much of the conventional wisdom is largely supported by the case studies. Government probably underestimated costs in all four cases. OSHA's estimates of the costs of the grain elevator rule were widely criticized as being too low. The cost estimates in the other cases, although much closer to the industry consensus, involved various minor errors in the "low" direction. The CPSC used a low figure for testing costs in the woodstove proceeding. The FAA assumed that airlines would buy the most inexpensive smoke detectors and, more significant, did not even calculate the additional training costs of Halon.

As expected, a tendency to underestimate costs does not characterize the four private cases. What occurred instead, however, is not readily apparent from the record. Documentation is scant, and many participants obscure the issues because arguments "against safety" are considered either unsavory or impolitic.^[2] The one apparent exception is NFPA 408, where costs were never explicitly discussed. The standard was set practically without regard to cost. Nevertheless, it appears that the NFPA Agricultural Dusts committee tended to overestimate, not underestimate, the cost of safety measures. Members of the National Grain and Food Association surely did so in lobbying OSHA.^[3] Manufacturers generally agree that UL and AGA "understand" the cost of various proposals and do not tend to underestimate them.

The surprise is on the benefit side, where it is generally assumed that private standards-setters underestimate benefits while government overestimates them. This characterization appears to fit government. Estimates of benefits in the woodstove labeling case, for example, were wildly exaggerated. W. Kip Viscusi argues that the estimated benefits of the gas space heater standard were also exaggerated, although the mistake was probably not very large.^[4] There is good reason to believe that the FAA and OSHA overestimated benefits as well. The private sector, by contrast, did not behave as expected. In only one of the four cases did the private sector play down the benefits of taking additional safety precautions. Members of the NFPA committee for grain elevator safety seem resigned to the notion that few, if any, safety measures can affect the number of explosions. But in the three remaining cases, there was a marked tendency for the private sector to err on the "safe" side. This tendency varied by hazard but was unmistakable in overall terms. There is no substantive basis for believing that most aspects of UL 1482 have any real effect on woodstove safety. Similarly, the AGA/ANSI standard for gas space heaters is filled with requirements that have no obvious effect on safety. The same is true of the NFPA standard for aviation fire extinguishers. All available evidence suggests that the benefits of adding more Halon extinguishers and increasing the training procedures for personnel are likely to be very small. That was not the view of most NFPA committee members, however, who tacitly assumed that the benefits were worth the cost.

In short, government standards-setting was true to form, tending in all cases to the combination of errors that promotes overprotective regulation. By contrast, the private standards-setters were all over the map (see table 7). Only one of the four cases, grain elevators, falls into

the category where underprotective regulation is most likely. In two cases, the results were ambiguous. Although it appears that UL and AGA tend to err in the favor of industry when estimating the cost of safety measures, they also erred on the side of safety in estimating the benefits of various provisions for woodstoves and gas space heaters. Finally, one of the private standards, NFPA 408, falls into the quadrant dominated by government. Based on the available evidence, this standard probably errs farthest in the direction of overprotectiveness, mandating even more than the FAA in an area where the likely marginal benefits are, by any reasonable measure, minuscule. The unexpected diversity of private standards carries over to the normative evaluation of outcomes.