PART TWO
CASE STUDIES

Three
Safety Standards for Grain Elevators

Private regulation is often characterized as weak or diluted, a kind of lowest common denominator of prevailing business practices. Many private standards are considered largely defensive, motivated by the desire to forestall government regulation. Public standards, by contrast, are thought to be more stringent than their private counterparts, sometimes unreasonably so. Public regulation is also considered inordinately cumbersome, laden with procedural requirements and judicial appeals.

These conceptions of public and private regulation are borne out substantially in the case of grain elevator safety. The private standard (NFPA 61B) is generally lax, reflecting industry's desire to avoid meaningful regulation of dust-control practices. The public sector, on the other hand, developed a more demanding standard but spent almost ten years in the process. The Office of Management and Budget shares the conclusion of the National Grain and Feed Association (NGFA) that the OSHA regulation is burdensome and unreasonable. As with so many regulatory issues, however, the relevant data are weak and inconclusive. Proponents and critics advance plausible, but divergent, arguments, each as good as its unsure assumptions. OSHA's grain elevator standard, for reasons argued below, actually seems worthwhile. It fills the gap in existing practices and is likely to produce benefits in excess of costs.

The importance of this case is twofold. First, it contributes to the conventional wisdom by providing specific details about actual behav-

ior and suggesting ways in which this behavior is affected by institutional design and external influences. Two popular explanations of public regulation—"capture" theory and Bernstein's "life-cycle" theory—have been discredited for their incompleteness. The "life cycle" is a vivid metaphor, but it lacks causal content about the force and nature of the regulatory "aging" process.^[1] Similarly, capture theory has been challenged by the rise in social regulation, where government actions are opposed, not favored, by the regulated.^[2] This case study illuminates several aspects of the conventional wisdom about public and private standards. For example, the enforcement ethic in the public sector tempered OSHA's willingness to adjust the regulatory burden on certain low-risk facilities.

Second, this case highlights an aspect of NFPA's behavior that does not fit the conventional image of private regulation. Several provisions in NFPA 61B are surprisingly strict. When judged by the criteria used to evaluate public regulation, a few NFPA provisions appear to be unreasonable. This suggests a previously unrecognized aspect of private regulatory behavior. Private standards-setting is not all "lowest common denominator" politics. There is a mixture of managerial behavior (marked by considerations of politics and economics) and technical behavior (permeated by the ethos of engineers). Managerial considerations explain the biggest weakness in NFPA 61B: favoring one regulatory approach to dust hazards (ignition control) to the complete exclusion of another technique (dust control). But some realms of private decisionmaking are dominated by technical, more than managerial, influences. The "technical" provisions in NFPA 61B are the most stringent. Technical influences help explain why NFPA 61B was developed more than fifty years before OSHA considered writing a standard. This case indicates how professional engineers alter the dynamics of private regulatory decisionmaking. These influences are sketched in greater detail in later chapters; but first the story of public and private efforts to regulate grain elevator safety.

Grain Elevators and the Explosion Problem

Grain elevators come in all sizes and shapes. Some are connected to processing facilities, others serve only as bulk storage. Bulk storage facilities are often grouped into three functional categories that roughly correlate with their size: country elevators (the smallest), inland terminals, and export terminals (the largest).^[3] Differences in function, prod-

uct, and capacity have all been urged as reasons for avoiding safety standards of general application. Feed mills, for example, have a much better safety record than bulk grain storage facilities.^[4] Some "country elevators" handle grain only a few days a year, obviously minimizing the opportunities for accidents. The significance of these distinctions for safety regulation is unclear because throughput—the amount of grain moved through a facility—has a more direct bearing on safety than function or capacity.^[5]

The Explosion Problem

Of all the hazards associated with grain handling, the most serious one is also the most difficult for those outside the grain business to understand. It is grain dust. This dust, generated whenever grain is handled, is easier to ignite and results in a more severe explosion than equal quantities of TNT. One insurance company distributes placards with the warning "Grain dust is like high explosives." Dust explosions account for the vast majority of personal injuries and property losses in grain-handling facilities. There are thousands of small fires in these facilities every year; estimates range from 2,970 to 11,000. But the twenty to thirty explosions a year account for almost 80 percent of the property damage and 95 percent of the fatalities^[6] (see table 4). Making sense of these trends over time is complicated by the changing fortunes of the grain trade. The major explosions in 1977–78 coincided with tremendous increases in wheat exports (primarily to the Soviet Union). The comparatively low number of explosions in recent years, on the other hand, is attributable in part to decreases in grain sales.^[7] Fluctuations in trade patterns notwithstanding, the explosion problem remains the most serious hazard in grain elevators.

Contrary to popular belief, explosions in grain elevators are not caused by spontaneous combustion. To cause an explosion, the dust must be mixed with air to form a dense cloud. (Layered dust only smolders upon ignition.) The airborne dust concentrations must also be above the lower explosive limit (LEL), a condition so opaque that visibility is minimal and breathing is difficult. Concentrations above the LEL occur regularly inside most bucket elevators (described below) but seldom elsewhere in a facility. Also contrary to popular belief, grain elevator explosions are not one big blast. They usually involve a primary explosion and a series of secondary ones.^[8] The primary explosion generates shock waves throughout the elevator, often raising into sus-

Figure 1.  Terminal-Type Grain Elevator

pension layered dust on walls, rafters, equipment, and the floor. Accumulations of as little as one-hundredth of an inch will propagate the flame from an initial explosion. In other words, layered dust provides the fuel to turn a primary explosion—often itself quite minor—into a major one.

There are two basic schools of thought about how to approach the explosion hazard: one, eliminate ignition sources; two, control airborne and layered dust. Although it seems obvious that neither strategy should be pursued exclusively, the debate over safety standards is often between groups that have a strong preference for one approach.

The Ignition-Control Strategy

Industry's inclination, according to a former USDA investigator, is to concentrate on ignition sources. Ignition occurs most frequently in the bucket elevator—a continuous conveyor belt with equally spaced buckets (often metal) that elevates the grain and discharges it into a spout (see figure 1). The top section of a bucket elevator, where the drive is

located, is referred to as the "head." The bottom section, where grain enters the elevator, is known as the "boot." The "leg" connects the head and the boot.

Ignition sources are varied and often notoriously difficult to pinpoint. In a study of fourteen explosions, the USDA identified ten different "probable sources," the largest group (four) being "unknown."^[9] Most studies agree that welding or cutting (also known as "hot work") is the largest known ignition source, accounting for perhaps 10 to 20 percent of all explosions. Other common ignition sources include electrical failure, overheated bearings, foreign metal objects sparking inside the leg, and friction in choked legs. "Jogging the leg"—trying to free a jammed bucket conveyor by repeatedly stopping and starting the driving motor—is a primary cause of friction-induced explosions.

Ignition-source control can take several forms. One is mechanical. Mechanisms including electromagnets and special grates can minimize the problem of metal objects entering the grain stream. Belt speed, alignment, and heat monitors can be used to detect hazardous conditions and shut down the equipment before suspended dust is ignited. The effectiveness of these devices varies, but quality has improved since their introduction to the grain-handling industry ten or fifteen years ago. Another approach to ignition control is behavioral. Employees are instructed not to jog the legs. Rules against smoking are strictly enforced. Permit procedures are instituted to ensure that hot work is done safely, and preventative maintenance schedules are instituted and implemented.

Whatever the combination of mechanical and behavioral requirements, ignition-source control has two limitations. First, there are countless potential ignition sources. The National Academy of Sciences (NAS) has reported the results of two surveys in which the ignition source remained unknown in over half the cases. Virtually every piece of equipment, as well as every grain transfer point, is a potential ignition source. Eliminating or controlling them all seems impossible. Second, as an NAS panelist put it, "ignition control is fine if you have perfect people; but people will make mistakes and then the equipment fails." A common example is jogging jammed conveyors instead of inspecting and digging out the elevator boot. This problem has been recognized for years, yet it persists largely unabated. According to an insurance representative, operators of some country elevators encourage this time-saving but risky practice. In short, everyone agrees that good operating procedures are a good idea. They are also inherently

difficult to enforce. Since operational breakdowns are inevitable, critics consider this loss control method inadequate by itself.

The Housekeeping Strategy

The other school of thought about grain elevator safety is enamored of—some say preoccupied with—"housekeeping" (or dust control). The label can be misleading to the extent that it conjures up only images of brooms and vacuum cleaners. Dust control is aimed at both airborne and layered dust. Airborne dust can be removed by various systems of aspiration, also called pneumatic ("moved or worked by air pressure)" dust control. Pneumatic dust-control systems have four major components: hoods or other enclosures; ductwork; a filter, or dust collector; and an exhaust fan. Layered dust can also be removed automatically, but in all but the largest facilities it is removed manually: with vacuum cleaners, brooms, compressed air, and, in some cases, water.

Both types of housekeeping (pneumatic dust control and layered dust removal) are controversial, but in the first case the disagreement is mostly technical, while in the second it is largely a matter of economics. The technical potential of pneumatic dust control is highly disputed. One grain elevator insurance company touts a system it claims can reduce dust concentrations in the bucket elevator below the lower explosive limit. The technology was not proven to the satisfaction of a 1982 NAS panel, but panel members placed a high priority on continued research in this area.^[10] Many in the industry still consider such technology unavailable. Others claim to have used it successfully.^[11] Both are probably right, as the design of pneumatic dust-control systems is, according to a Cargill engineer, "more of an art than a science." Engineers cannot simply take specifications for the desired concentration of airborne dust (or rate of accumulation for layered dust) and design a system that will perform accordingly. The technology for removing airborne dust is too uncertain. Much depends on how the system is installed and maintained.

Layered dust poses much less of a technical problem. How to remove it effectively is not very controversial (although the virtues of vacuuming versus sweeping are a minor topic of disagreement). For large facilities, dust removal is usually part of normal operations. Sweeping is done at least once per shift at all export facilities. But for many smaller facilities, removing layered dust means hiring additional labor and slowing down operations when they are most profitable. Few quarrel

with the conclusion of the NAS that housekeeping and maintenance are often given low priority and are usually the first tasks postponed when there is a rush of business.^[12] A grain company representative on the NFPA Technical Committee on Agricultural Dusts concurs that "there are many, many filthy ones that get by." A less tactful USDA investigator describes the prevailing levels of housekeeping in grain elevators as "fair to abysmal." The main objections to stricter housekeeping rules is the cost. The controversy, in short, is managerial, not technical.

Trying to estimate the number of "bad apples" is difficult, however. "It depends on which day of the week you are counting," as one industry representative put it. There has been very little counting to date. No one even knows with precision how many elevators there are, let alone when they were all built and what kind of equipment they currently use. Harder still is determining the relative cleanliness or safety levels of grain-handling facilities. No industry group investigates or collects information on explosions. NFPA collects statistics through newspaper clippings and voluntary reporting from the fire service, but this information is hardly comprehensive. A USDA task force concluded that the NFPA estimate of annual fires in grain elevators may understate the actual situation by a factor of five.^[13]

The NFPA Standard for Grain Elevators

Whatever the numbers, insurance companies have worried about grain elevators since a series of big explosions in 1919 and 1921. Responding to calls for an industrywide standard on grain elevator safety, NFPA appointed a Committee on Dust Control in Grain Elevators. Insurance interests have been major participants in NFPA's subsequent efforts. The committee, lacking sufficient information on certain aspects of the explosion problem, hired Underwriters Labs (another organization created by the insurance industry) to investigate methods of controlling floating dust in terminal grain elevators.

The results of the UL study formed the basis for the dust-control provisions in the early versions of the standard, which also contains general operation and design provisions. It has been revised at least five times since then, although many of the provisions in the most recent version can be traced back to 1953. The most significant changes were in 1970, when NFPA added country elevators (formerly governed by a separate standard) to the scope of the standard, and in 1980, when NFPA responded to the threat of imminent government regulation by strengthening the ignition-control requirements for bucket elevators.

The revision process began again with a meeting in July 1985 of the Technical Committee on Agricultural Dusts, and a revised version of NFPA 61B (hereinafter "61B") was adopted the following year.

The Agricultural Dusts committee, chaired for the past fifteen years by a representative of Continental Grain, has approximately two dozen members. Twelve to fifteen attend formal committee meetings, but all actions are subject to letter ballot by the full committee. The largest-segment of committee membership comes from the insurance industry—Industrial Risk Insurers, the Mill Mutuals, the Insurance Services Office of Nebraska, Kemper Insurance Co., and Factory Mutual Research Corp. are each represented. Two other organizations related to the insurance industry, UL and Johnson & Higgins (a brokerage firm), are represented on the committee along with Cargill, the country's largest grain company, and two major grain processors (Kellogg Co. and General Foods). Other representatives include those of a fumigant company, a fire equipment manufacturer, a manufacturer of grain-handling equipment, and several academics. Conspicuously absent from the committee roster is a representative from OSHA. "It wasn't for lack of trying," notes an NFPA staff member. Neither was it always this way. OSHA's policy on participating in private standards-setting activities has varied by administration and by issue.

Certain provisions in 61B have been a source of repeated debate over the years. Such issues, according to the current committee chairman, are brought up practically every time 61B is revised. Some of these largely technical issues are discussed below. But 61B is probably better known for what it does not cover than for what it does. Some provisions, including those on housekeeping, are so general that the standard quite literally requires nothing in particular. Additionally, owing to the retroactivity clause, most provisions apply only to new facilities, even though older facilities actually pose the most serious hazards.^[14]

In its present form, 61B is a remarkably compact regulation, devoting no more than a page or two apiece to chapters on construction requirements, equipment, and dust control. Most provisions are as general as they are brief. "Extraneous material that would contribute to a fire hazard shall be removed from the commodity before it enters the [grain] dryer."^[15] "Boot sections [of the elevator leg] shall be provided with adequate doors for cleanout of the entire boot and for inspection of the boot pulley and leg belt."^[16]

NFPA 61B avoids almost all design details. The complex topic of

explosion venting, for example, is addressed by reference to a separate NFPA venting guide that committee members agree is not particularly applicable to grain-handling facilities.^[17] Similarly, dust-control systems are mentioned, but there are no performance requirements or construction specifications.^[18] In contrast, the National Academy of Sciences has published a 116-page guide to designing pneumatic dust-control systems.

Housekeeping: A Gentlemen's Agreement

The most significant provision in 61B is the basic housekeeping requirement. Unlike the standard later adopted by OSHA, which contains specific action levels and alternatives for dust control, 61B dispenses with the topic by stating simply that "dust shall be removed concurrently with operations."^[19] This language has been known to provoke a good laugh from Agricultural Dusts committee members pressed to explain what it really means.

Whatever it means, the provision is not enforceable. OSHA learned that when it tried to use 61B in support of citations issued after the 1977 explosions. The commission that reviews OSHA citations was unwilling to rely on such an ambiguous provisions.^[20] In actuality, the housekeeping provision is not a "requirement" at all. It is more of a gentlemen's agreement to recognize the problem but leave its solution entirely to the individual operator. "I think that you have to consider that people are going to be reasonable and rational about applying this," explained a Continental Grain representative at an OSHA hearing.^[21]

Whether a more specific requirement would be desirable is the subject of considerable controversy. Leaving that question aside for the moment, the lack of specificity of the housekeeping requirement in 61B is not adequately justified by the reasons most commonly offered in its defense. "If someone can tell us how to be more specific and be scientifically logical" in establishing dust-control requirements, explains an Agricultural Dusts committee member from a major grain company, "we would buy it." Indeed, there is no way to do so—short of banning grain dust—if "scientific logic" demands total safety. Research sponsored by the NGFA indicates that a dust layer as thin as one-hundredth of an inch can propagate an explosion. The NGFA has been accused by its detractors of conducting this research precisely to bolster the "scientific" argument against any standard. But there rarely is a strong

scientific basis for resolving complex problems involving the trade-off between cost and safety. If this call for greater scientific certainty applied equally to all standards-setting, it would largely paralyze the effort. There are too many variables interacting and changing over time to expect anything resembling scientific certainty for each one.

Guesses, commonsense judgments, and just plain arbitrary numbers adorn public and private standards alike. They have to. NFPA 61B is no exception. The requirement that grain driers be cleaned every 168 hours, for example, is not scientifically logical. As it turns out, 168 was chosen because that is how many hours there are in a week—a measure no more scientific than Continental Grain's policy of cleaning them every 48 hours.^[22] Limiting the temperature of hot pipes to 160° F, prohibiting more than 25 percent of a roof from being plastic panel, and suggesting that motor-driven equipment be cleaned at one-hour intervals during operations are further examples of provisions in 61B that are equally susceptible to the charge of scientific infirmity.^[23] In each case, the number is an admittedly arbitrary one, based on the consensus of committee members as to what constitutes a reasonable requirement.

In fact, scientific uncertainty did not prevent Continental Grain or the Factory Mutual Corporation from incorporating an "action level" of one-eight of an inch of layered dust into their own in-house standards for housekeeping.^[24] So why does the NFPA Agricultural Dusts committee demand more of science? One possible reason is that the dust-control problem is more complicated than most issues. Hot pipes present similar hazards under a variety of circumstances; grain dust does not. Dust hazards depend, to some degree, on virtually every aspect of a grain-handling facility, including the product it handles, its sales and operation patterns, the general layout and date of design, and the effectiveness of existing dust-control equipment. In short, some facilities have much less need or ability than others to conduct housekeeping. The image of the "small country elevator"—a mom-and-pop operation with no hired hands, but a line of anxious farmers waiting to unload their grain before it starts raining—is often evoked in this line of argument. Some country elevators have little need for dust control because they have low throughput and no enclosed bucket elevators. Others, it is argued, lack the resources to purchase dust collection systems or hire additional labor. The appropriate action level, if there is such a thing, varies significantly by facility, and the small ones should, the argument continues, be spared the regulatory rod entirely.

This position on dust control contradicts the position taken else-

where in 61B. The Agricultural Dusts committee has gone on record several times against the notion that differences in facilities render a general standard inappropriate. Separate NFPA standards for country elevators were combined with those for other grain-handling facilities in 1973, when the committee decided that "a distinction between types of grain elevators on the basis of capacity or shipping or receiving media is no longer practical."^[25] Similarly, a committee member argued in July 1985 that motion switches should not be mandated on all bucket elevator legs because some country elevators "do not realistically need them." The committee rejected the argument on the grounds that motion switches were generally a good idea and it would be impossible to identify in a standard those situations in which they are not necessary.^[26] The same could be said about dust-control requirements. "Facilities vary," notes a former USDA investigator, "but the hazard scenarios are the same."

The Unspoken Arguments: Liability and Retroactivity

So why is the housekeeping provision in 61B so vague? Two factors other than the limitations of science and the diversity of facilities are at play. One is specific to this issue, the other indicative of a larger force affecting the development of private safety standards. There is an unspoken belief that good housekeeping simply is not the answer. To some, it is a matter of practicality. It would require excessive effort, the argument goes, to keep an elevator clean enough to prevent explosions. "There is no such thing as a clean elevator," quipped one trade association representative. Perhaps the strongest explanation is that a specific housekeeping provision could be legal dynamite. Industry has learned that voluntary standards can and will be used against you in a court of law. Most explosions lead to litigation. According to a retired Cargill executive who testifies in such lawsuits, 61B and a host of related standards are raised in almost every case. The stakes can be very high. The two largest explosions in 1977 resulted in settlements of approximately $25 million each, and a jury recently applied the bane of the tort law—punitive damages—for the first time in a grain elevator case.^[27] The vaguer 61B is on housekeeping, the less powerful a weapon it would be after an explosion.^[28] Some of the vagueness, then, is simply an attempt to make the standard liability-proof. The ill-fated OSHA citations that relied on 61B are testimony to the effectiveness of this strategy.

Success, in this context, breeds more generalities. Substituting what Agricultural Dusts committee members refer to as "motherhood statements" for specific provisions is an increasing trend with 61B. Space heaters shall be located in "suitable places." Fire extinguishers shall be located in "strategic" places. A proposal that all bearings be properly maintained in accordance with the manufacturer's specifications was rejected by the committee in 1985 in favor of a requirement that "all bearings shall be properly maintained."^[29] ("Let's see what a lawyer can do with that one!" mused a committee member at the July 1985 meeting.) Of course, the ability of lawyers to "do something" with the standard is probably equal to (or greater than) the standard's usefulness to its intended audience, grain elevator operators. Without guidance concerning the specifics of "proper maintenance," the lawyers will be as lost as anyone trying to glean any substance from the standard.

The extent to which 61B should contain such generalities appears to be the most significant issue facing the Agricultural Dusts committee. Some think that the committee has gone too far already. The president of one of the only two insurance companies presently underwriting grain elevators—major companies such as Cargill are self-insured—abhors the generality. He considers 61B so vague as to provide no real assistance in evaluating elevator safety. He does not even keep a current copy of the standard in his office. The technical director of the National Grain and Feed Association, on the other hand, considers such generalities the silver lining to an otherwise objectionable standard.

The generality problem indicates a more significant dilemma. The committee is trying to achieve two largely incompatible objectives: having a meaningful standard that can affect safety and provide helpful guidance to elevator operators without providing ammunition that will be used against you whenever an explosion occurs. Nowhere are the dimensions of this problem more accentuated than in respect to retroactivity.

By its own terms, 61B applies only to new facilities. This makes the most sense for design requirements. Compliance costs for many requirements are low in the design stage but very high after construction. Elevators used to be built with combustible materials, for example, and little consideration was given to minimizing ledges and horizontal surfaces or designing conveyors in a manner that would facilitate cleaning with portable vacuum cleaners. These things can easily be remedied in the design process. Correcting them retroactively would practically involve building a new facility.

In reality, 61B can, has, and will be applied to situations other than brand-new elevators. When issuing citations, OSHA often refers to 61B without regard for the date the facility was constructed. Plaintiff's attorneys are similarly inclined and have met with greater success. The retroactivity clause invites this attitude. NFPA 61B attempts to be unusually lenient in this regard. Unlike most building codes, it does not even require compliance in the case of major expansion or renovation—the vast majority of "new" construction. Moreover, the 1980 version also exempted from retroactive coverage a host of operational activities—such as hot-work procedures and housekeeping—that can be carried out without regard to the age or design of the facility.

The opposing viewpoints about retroactivity reflect deep-seated differences concerning the role and nature of private standards. Some of those favoring the broadest possible exemption from retroactive application "really do not want a standard at all," according to an Agricultural Dusts committee member at the July 1985 meeting.^[30] This group is not insignificant. The National Grain and Feed Association, the largest trade association in the field, is unwilling to refer to 61B as a national consensus standard and refuses to participate on any of NFPA's agricultural dust committees.

The implications of this reticence are not lost on the Agricultural Dusts committee. Organized opposition can prevent or at least postpone proposals from being adopted in the NFPA system. Should this private standard become unpalatable enough to the NFPA, or if others in the industry actively opposed it, 61B might fall of its own weight. As one committee member said at the July 1985 meeting, "It behooves us to try and give some relief to those being sued." Nevertheless, most committee members realize that the only real "protection" against retroactive application is to abolish the standard. They understand that others will do as they please in interpreting the retroactive effect of 61B, so it is futile to attempt to prevent this from happening by tinkering with the retroactivity clause. These members also see the danger in exempting too many requirements from retroactive effect. "If this [standard] is too unattractive to the states," warned another committee member at the same meeting, "they might adopt their own requirements." Several proposals to strengthen the tone of the retroactivity clause were rejected in July 1985 primarily because they seemed pointless to the committee.

To the chagrin of those proposing the stronger exemption, the committee adopted a proposal by the representative from Continental Grain

to narrow the scope of the retroactivity clause—that is, to make more provisions apply retroactively. The only provisions in the 1980 version with retroactive effect concerned fumigant usage. In seeking to identify appropriate candidates for retroactive application, the intent, according to the committee chairman, was to include "operational" requirements but not those requiring "even a dime of investment."

The results of the 1985 revisions are revealing because the committee chose to include the housekeeping provisions in the group to be given retroactive effect. This change appears to make 61B more credible. After all, shouldn't old and new facilities alike be expected to do proper maintenance or take precautions to ensure that welding is done safely? In light of the "not even a dime of investment" principle, however, the committee's action seems more significant for what it admits about the housekeeping provisions—that they do not require anything more than what an operator is already doing.

The Uneasy Solution: Change the Packaging

In an effort to placate those who would rather not have a standard, the Agricultural Dusts committee has done everything possible to make 61B a more agreeable document. These changes have largely been a matter of form, not substance. The language in 61B has been toned down considerably over time. References to such unpleasantnesses as "injuries to personnel" have been replaced with less specific references to hazards in general.^[31] The introduction to the 1959 version of 61B warned that: "GOOD HOUSEKEEPING AND CLEAN PREMISES ARE THE FIRST ESSENTIALS IN THE ELIMINATION OF DUST EXPLOSION HAZARDS, CONSEQUENTLY THIS CODE IS NOT INTENDED TO LESSEN IN ANY WAY THE RESPONSIBILITY OF THE OWNER AND OPERATOR IN THIS RESPECT ." This language was moved to the appendix in 1973, and the capital letters were replaced with regular capitalization. The statement was omitted entirely in the 1980 version.^[32] Only the gentlemen's agreement remains.

Placing things in the appendix is a popular compromise. Fully one-third of the 1980 version of 61B is appendix material. The appeal of this compromise is clear: the individual in favor of the material gets it "into the standard" in some form, while those opposed to the specifics take considerable solace in the fact that it is "not a requirement." Employee health and safety, a subject avoided in many NFPA standards but favored by the labor representative on the 61B committee, is addressed in an appendix added in 1980.

The appendix also contains more specifics than the actual standard. For example, the standard requires that "horizontal surfaces shall be minimized," while the appendix indicates that the "suggested angle of repose is 60 degrees."^[33] This is another tactic in the search for a liability-proof standard. The hope is that through the appendix the generalities in 61B can be given meaning in a nonenforceable manner. The official NFPA position, set forth in every standard, is that the appendix "is not part of the requirements … but is included for informational purposes only." Of course, one of the stated purposes of NFPA's "requirements" is to provide information.

The difference is really a matter of wishful thinking. All NFPA "requirements" are informational until an "authority having jurisdiction" chooses to enforce them. Moreover, the authority can choose to enforce the appendix as well as the "requirements" proper. An NFPA member who has worked with this committee confirms that appendix material is often treated by state fire marshals as having equal weight to the "requirements." The wishful thinking paid off, however, in the case of grain elevators. Some OSHA citations were overturned by the Occupational Safety and Health Review Commission precisely because they relied on material contained in the appendix of 61B.^[34]

Technical Arguments, Surprising Results

This is not to say that 61B is all appendix and no substance. Nor is it to say that disagreements are foreign to the Agricultural Dusts committee. Certain provisions in 61B are specific and have been the source of repeated debate throughout the years. Should there be a vent between bins to allow dust generated during loading to settle into adjoining bins instead of blowing back up into the work area? To what extent is it possible or desirable to vent silos and elevator legs to relieve pressure from an explosion? Should compressed air be allowed, and under what circumstances, to "blow down" dust from walls and rafters?

The technical nature of these questions is not surprising. Private standards-setting is often described as technical. Esoteric questions such as those above seem to fit the bill. How the committee has handled these issues is more unusual. "Irrational," "unrealistic," and "unduly stringent"—criticisms often reserved for government regulation—are all strictures used by some NFPA members to describe the resolution of these issues.

For example, even though most insurance and grain company rep-

resentatives privately concede that the requirement is irrational, 61B prohibits interbin venting,^[35] and it is also accused of being unrealistic about explosion venting. (Explosion vents are movable panels designed to vent the pressure from an actual explosion; interbin vents, just discussed, are actual openings designed to vent airborne dust.)^[36]

NFPA 61B includes several other provisions that are considered by many in the grain industry to be too stringent. Removing layered dust with compressed air is prohibited unless other equipment is shut off. The concern is that "blowing down" can create a combustible cloud. There are no known cases in which an explosion has occurred this way, and one insurance representative expressed serious doubts about whether compressed air could create dust concentrations above the LEL. The only explanation for this provision is that it represents a compromise between those who are opposed to blowing down on principle and those who think that it is never dangerous.

Requirements incorporating the National Electric Code's classifications for environments with agricultural dust are also quite strict. The code, written by NFPA, specifies two classes for such environments: Divisions 1 and 2.^[37] Equipment certified for Division 1 must be tested under conditions in which dust concentrations regularly exceed the LEL. Equipment for Division 2 is tested under conditions in which such concentrations are "occasionally" encountered. Herein lies the controversy. Division 2 equipment is tested under extreme conditions compared to those normally encountered in those parts of a grain elevator where Division 2 equipment is required. The Division 2 test methods do not just represent the worst case, they represent an impossible case.^[38] Not only does this provision increase the price of grain-handling equipment, but in some cases the required equipment is not available. The same arguments apply to Division 1 equipment if it is required in areas without explosive dust clouds. The NFPA leaves these implementation questions to "the authority having jurisdiction."

NFPA 61B is also stringent with respect to ignition sources in bucket elevators, but this is a recent development. Several requirements for bucket elevators were added in 1980, including (1) monitoring devices that cut off the power to the drive and sound an alarm when the leg belt slows down, (2) magnets or separator devices to minimize metal objects entering the grain stream, and (3) closing devices to prevent flame propagation through idle spouts.^[39] These requirements did not generate the controvery produced by those discussed above. Neither were they really new; they were moved from the appendix to the body

of the standard. This unusual move is best understood by the politics of the moment. The changes were suggested around the time OSHA initiated its Advance Notice of Proposed Rulemaking. Some committee members privately acknowledge that these changes were an attempt to forestall OSHA regulation. The changes certainly strengthened the standard, at least for new construction. But they did not forestall OSHA.

The OSHA Standard

Grain elevator operators did not have a very merry Christmas in 1977. Three grain elevators and two feed mills exploded during the third week of December, killing fifty-nine people and injuring forty-eight others. The Department of Agriculture, hit particularly hard because thirteen federal grain inspectors were killed in the explosions, established a Special Office for Grain Elevator Safety. OSHA was put on the spot. There was intense political pressure to do something about grain elevator safety, but OSHA had little experience with the issue. Lacking its own safety standard, the agency had only its "general duty" clause to back up any enforcement actions. Citations based on this clause require, among other things, proof of a "recognized hazard."^[40] NFPA 61B came in handy, since one possible method of establishing a "recognized hazard" is to cite private standards. OSHA did so on numerous occasions after the 1977 explosions. Citations tend to evoke indignation, however, and these cases were no exception. Some of the citations lived up to the agency's reputation for unreasonableness—one firm was cited for an ungrounded coffee pot in an office building detached from the grain elevator, for example—but even those involving seemingly serious violations were met with hostility. Virtually every citation was contested, and OSHA lost almost every one on appeal: 61B was considered too vague in some instances; in others, the review commission that hears appeals from OSHA citations honored the "advisory" nature of the appendix. OSHA had to look elsewhere to advance its enforcement strategy.

To overcome its limited background in grain elevator safety, and in the hope of defusing an already politicized issue, OSHA contracted with the National Academy of Sciences to evaluate possible methods of preventing grain elevator explosions. The first of four NAS reports confirmed the inadequacies of OSHA's experience, criticizing the agency

for emphasizing citations instead of investigating explosions sufficiently to understand the problem.^[41]

OSHA adopted a cautious, if not timid, stance. In January 1980, the same month that the Department of Agriculture released its report "Prevention of Dust Explosions—An Achievable Goal," OSHA issued a "request for comments and information and notice of informal public meetings," which posed approximately two hundred questions for public comment.^[42] Although it was widely assumed that OSHA intended to propose a grain elevator safety standard, the agency did not start drafting one for two more years. It was waiting for support from the NAS. Testing the political waters in the meantime, OSHA held hearings a few months later in Wisconsin, Louisiana, and Missouri.

The water was hot. The hearings opened an acrimonious debate that shifted locations over the years but never lost its political intensity. Several hundred comments were received in response to OSHA's request, and over two thousand pages of testimony were received at three "informal public meetings." Most of it was extremely negative. This hostility hindered the NAS, which noted the "reluctance of elevator management to cooperate" with their study.^[43] The key report, "Prevention of Grain Elevator and Mill Explosions," was released in June 1982.^[44] It identified as the two most significant issues in grain elevator safety (1) ignition sources in bucket elevators and (2) dust concentrations throughout facilities. With this background, OSHA began drafting a standard for grain elevators. Within a month, the head of OSHA told a House subcommittee that grain elevator safety was one of the agency's two priority projects, and that a rule would be completed by January 1983.^[45] It took much longer than expected, but the agency eventually proposed a regulation, with fourteen basic requirements.^[46] Many of these, particularly those concerning ignition-source control in bucket elevators, came directly from the NAS recommendations. One received practically all of the attention: the "action level" for removing layered dust.^[47]

Hanging a Number on Dust Control

The action-level concept came from the NAS report, which proposed that corrective action should be taken whenever layered dust exceeded a specified depth (over a 200-square foot area). The NAS did what the NFPA would not: hang a number on dust control. "It was done to satisfy labor and other political groups who felt that without a definite

figure you do not have a club," explains an industry member on the NAS panel. The panel chose one sixty-fourth of an inch as its "club."^[48] The number was a compromise, plain and simple. Some panel members argued for one-hundredth of an inch, based on flame propagation experiments and without regard for economic feasibility. One-eighth of an inch, a figure already used in some proprietary standards, was also suggested. One sixty-fourth was somewhere in between.

OSHA engaged in its own search for an agreeable number. Again, the process was predominately political, not scientific. Under more pressure than NAS to take into account the economic impact of its proposals, OSHA considered a higher range of numbers. The argument was between one sixty-fourth and one-eighth. The former had new-found credibility thanks to the NAS; the latter had historical acceptance. OSHA chose historical acceptance.

The selected number appears to have been favorable to industry. The rejected figure (one sixty-fourth) was, after all, eight times stricter than the one chosen. One-eighth is considered by many to be not only lenient but downright dangerous. An elevator with an eighth of an inch of layered dust, according to an insurance representative, is "a bomb waiting to explode."

But appearances can be deceiving. Industry opposed one-eighth and one sixty-fourth with practically equal vigor. In reality, the disagreement was not about the best action levels; it was about whether to have action levels at all. OSHA wanted a number. Industry did not. "You've got to have a number for it to be enforceable," according to an OSHA official familiar with the agency's bad luck under the general duty clause. For related reasons, industry was opposed to any number. Although it cited the specter of dangerous facilities being in compliance with the rule—those with one-ninth of an inch of dust, for example—industry's real concern was the opposite: seemingly safe facilities being in violation. It is likely that even the best facilities would at some time, in some part of the facility, be in violation of the one-eighth-inch requirement. This would not be a problem if OSHA could be trusted to use good judgment in enforcing the rule. But most elevator operators did not have such confidence. As a result, unreasonable enforcement strategy was of much greater concern than the "false sense of security" some claimed would follow from an action-level approach.^[49] The paramount concerns of the National Grain and Feed Association when it later petitioned for a partial stay of the OSHA regulation were the enforcement directives for implementing the housekeeping require-

ments.^[50] Unfortunately, the enforcement strategy, including possible methods for heading off unreasonableness, was never seriously discussed during the rulemaking proceedings. Instead, the debate was dominated by a battle of the cost-benefit analysts. The issues they raised were important, but so were many of the ones they ignored.

The Battle of the Cost-Benefit Analysts

The battle itself was costly. First OSHA hired a well-known consulting firm (Arthur D. Little) to analyze its draft proposal. The resulting report was riddled with errors and questionable assumptions.^[51] It was so vulnerable to attack that OSHA hired a second firm (Booz, Allen & Hamilton) to massage the data. The NGFA subsequently hired a third consulting firm (G.E.M. Consultants) to attack the conclusions reached by the first two.

On the cost side, estimates prepared for OSHA by Booz, Allen pegged the total initial cost of the standard at $200 million, with annual recurring costs of approximately $137 million. These estimates reflected a series of assumptions about the cost of each of the fourteen provisions in the proposed rule.^[52] Industry accepted the Booz, Allen estimates for some of the minor provisions, such as permit systems for hot work (see table 5), but it took issue with most of the others. By

making marginally higher estimates for almost every subsidiary assumption, industry argued that the cost of many provisions were understated by a factor of two.^[53] Except for the housekeeping provisions, these differences of opinion did not add up to much. Hundreds of millions of dollars, however, separated the cost estimates for housekeeping. Some of these differences are impossible to evaluate.^[54] Other assumptions made by OSHA's consultant seem more reasonable than the industry's.^[55] Industry also offered some persuasive indictments of the OSHA estimates. The Booz, Allen study assumed that operators would purchase "dust-tight" vacuum cleaners, even though the National Electric Code requires the more expensive models certified for class (g) environments. In short, it seems safe to assume that the true cost of the rule would fall in between the estimates prepared by the two consultants.

On the benefit side, where estimates are normally subject to more uncertainty, the report prepared for OSHA took a realistic position. It was assumed that the housekeeping provision would prevent approximately 30–50 percent of all fires and 17–32 percent of all explosions. Through a series of calculations intended to gauge the "willingness-to-pay" to avoid property damage and personal injuries, these figures produced an estimated total benefit of approximately $286 million.^[56] Industry did not take issue with the specifics of these estimates. Although it is always arguable that things will not work out as well as planned, the estimates prepared for OSHA were not vulnerable to the charge of overoptimism.

The battle of the cost-benefit analysts was not won clearly by either side. Assuming that actual benefits would be between 50 and 90 percent of OSHA's (possibly high) estimates, the rule would yield from $143 million to $286 million in total annual benefits. Estimates of annualized cost, based on the figures presented in table 5 and discounted to present value, ranged from $113 million (OSHA) to $240 million (NGFA). The true cost is likely to be somewhere in between. Therefore, the OSHA standard may or may not generate more benefits in excess of costs—a plausible case can be made for both propositions. Either way, the costs appear to be in the same ballpark as the benefits.

Country Elevators and Distributional Effects

Aside from cost-effectiveness, the standard was controversial for distributional reasons. Industry and OMB argued that the rule dispropor-

tionately affected country elevators. Between 80 and 100 percent of export terminals and inland elevators were thought to be in compliance with most of the OSHA requirements (other than housekeeping). No more than 15 percent of country elevators were. Large facilities generally had dust-control programs; many country elevators did not. As a result, the costs of the proposal fell much more on country elevators than on other facilities. Yet, as OMB argued, the risks were more significant in larger facilities.^[57] OMB took the position that "big" facilities should be covered and "small country elevators" should not. OSHA did not want to exclude any facilities. As an OSHA official testified, "When workers are hurt or killed [we] don't particularly care about the size of the facility." Both positions were less than reasonable. OMB wanted to exempt all "country" elevators, even though some were, by any measure, as large as inland terminals and export facilities. But OSHA's position indicated complete disregard for marginal costs and benefits. The cost of bringing the smallest facilities into compliance would be far out of proportion to the losses likely to occur in those facilities.

A political dispute of surprising dimensions took shape. Both the vice president of the United States and the president of the AFL-CIO got involved when OMB held up the rule for months longer than the normal sixty-day review period. The housekeeping provision was the major bone of contention. In a compromise that pleased practically no one, the proposed rule was released for publication when OSHA, at OMB's insistence, added two alternative proposals for housekeeping: sweeping once per shift or installing pneumatic dust control.^[58] It was no secret that OSHA still favored the action-level alternative.

The added alternatives did not cool the political debate. The NGFA organized an impressive campaign to flood OSHA with "worksheets" from elevator operators opposed to all three alternatives.^[59] A few congressmen held hearings to allow elevator operators in their home districts to complain that "the people who wrote these standards have never been to a country elevator."^[60] OSHA continued to take heat from all sides, with the exception of labor. In what was termed an "unusual alliance," the AFL-CIO supported the proposed rule even though it would have preferred an action level stricter than one-eighth of an inch.^[61]

The OSHA Standard Prevails (with Minor Improvements)

The stalemate between OMB and OSHA lasted until there were two more serious explosions. In July 1987, following a fatal explosion in

Burlington, Iowa, Senator Tom Harkin (D-Iowa) complained to the OMB director about delays in releasing the rule. He complained again four months later, following another explosion. The rule was published on the last day of 1987.^[62]

The rule contains a major change, but not an actual softening, in OSHA's position on housekeeping. The standard retains the action level of one-eight of an inch for housekeeping—eliminating the ineffective "once-a-shift sweeping" option. OSHA modified the provision, however, by limiting its application to three "priority areas."^[63] This seems much more reasonable than the earlier version, which applied to any 200-square-foot area in the facility. It may even help reduce frivolous or unreasonable enforcement actions. But it does little to alter the basic housekeeping tasks required by the rule, since many of the steps used to minimize grain dust in "critical" areas will also minimize it elsewhere. OSHA also moderated the rule slightly to favor country elevators. Some country elevators are exempt from the requirements for alignment and motion detection devices.^[64] But none are exempt from the action-level requirement. Country elevators were given three extra years to come into compliance, however. The symbolism of these changes, all conciliatory, far exceeds their actual substantive impact.

The most substantial change in the final standard concerned feed mills, and again the adjustment was in the direction of leniency. In response to arguments that risks are lower in feed mills, OSHA exempted these facilities from the most powerful provisions in the standard: the housekeeping and bucket elevator requirements. OSHA defended this decision, calculating that "the risk of an employee death or injury resulting from an explosion is almost five times greater in grain elevators than in mills."^[65] OSHA rejected this line of argument in the case of country elevators, however, justifying the apparent discrepancy on the grounds that "the available data are not sufficient to permit an accurate estimate of the relative risks in large and small elevators."^[66] Although nobody was entirely pleased with the final standard, it certainly was reasonable in many respects. An editorial in the trade journal Feedstuffs, often an outspoken critic of OSHA, allowed that "grain elevators and feed mills can be much safer places to work merely as a result of OSHA's attention…. The rules on bin and entry, housekeeping plans, work permits, etc., are needed. A federal edict appears to be the best way to implement such rules."^[67] But industry was incensed at the housekeeping requirements and immediately challenged them in court. Labor countered, challenging the exemption for feed mills. The

Fifth Circuit Court of Appeals upheld all but the housekeeping provisions, remanding that issue for "reconsideration of the economic feasibility of the standard."^[68]

Summary Evaluation

The OSHA standard is a classic example of the time-consuming adversary process often attributed to government regulation. Ten years elapsed between the 1977 explosions and the final OSHA rule. The most controversial issue, housekeeping, remains unresolved after a court challenge, even though OSHA spent at least half a million dollars on consulting firms. Hearings held around the country produced volumes of testimony, mostly to the ill will of industry toward OSHA. The standard was loosened somewhat during this arduous administrative process. Whether these changes were a victory for business or a calculated bargaining strategy by OSHA depends on whom you ask. Whatever the reason, OSHA exempted small facilities from some requirements and extended the deadline for compliance with others. OSHA also restricted the final standard to bulk grain-handling facilities, exempting feed mills and other processing facilities included in the original proposal. Most important, OSHA concentrated the housekeeping provisions in three "critical" areas. These changes all made the standard more reasonable.

OSHA's final standard has fewer requirements but packs more punch than NFPA 61B. OSHA stood firm on the action-level approach to housekeeping, resisting strong political pressure to weaken the provision by allowing sweeping once per shift. The agency adopted an action level identical to that used in several proprietary standards, which, if anything, is too lenient (the NAS suggested a more stringent level). The standard is strict enough to cover the worst facilities, however. (These "bad apples" are estimated to constitute 10 to 30 percent of the facilities.) OSHA's housekeeping requirements also have the advantage of being performance-based, allowing firms to determine the most efficient method of compliance. The overall cost of compliance is difficult to estimate, but the range of estimated costs is lower than the range of estimated benefits.

In contrast, the NFPA standard bears out the common conception that private regulation is weaker than its public counterpart. To NFPA's credit, its committee meetings are not plagued by adversity or delays. Committee members are knowledgeable, bringing an array of practical

experience to the task. They meet regularly and adopt revisions in relatively short order. But the outcome (at least in the case of grain elevators) leaves much to be desired. NFPA 61B is relatively spineless. Many important topics, including housekeeping, are glossed over with empty generalities. This situation is due in part to concerns about liability. Seeking to avoid a standard that can be used against business in any post-explosion lawsuits, the Agricultural Dusts committee relegates many issues to the appendix, while making others intentionally vague. The committee also shares the sense that certain "management" issues should not be subject to industrywide regulation. A representative of the Grain Elevator and Processors Society disavows "any kind of declarative 'thou shalt do so-and-so'" standard.^[69]

Not all aspects of the NFPA standard are as flimsy as the housekeeping provisions: 61B is fairly strong on basic design issues. Because of the retroactivity clause, however, these provisions apply to fewer facilities every time the standard is upgraded. This technique minimizes political opposition and helps explain why upgrading is possible. Some provisions of general application are also strict, a few to the point of being unreasonable. Provisions such as the UL specifications for equipment used in hazardous locations are far more demanding than is warranted by the circumstances. These provisions require expensive electrical devices where normal equipment would pose no real hazard. Other requirements, aimed at reducing fire hazards more than explosion hazards, may actually be counterproductive—reducing the former at the expense of increasing the latter.

These unexpected tendencies are not major aspects of 61B; they are noteworthy because they are evidence of a "technical" decisionmaking realm characterized by influences and behavior not normally attributed to the private sector. These "technical" decisions reflect the professional norms of engineers. This engineering ethic is conservative in some respects and stringent in others. Engineers are loathe to get involved with political issues, including matters of management prerogative, but they are zealous advocates of safety on occasion. OSHA has no quarrel with the fire safety provisions in NFPA 61B. The extent to which private standards-setters are influenced by regulatory philosophy and professional ethics is explored in detail in chapter 7.

Four
Standards for Aviation Fire Safety

Unlike grain elevator safety, aviation safety is an entire field of regulation. While the major hazards in a grain elevator can be addressed in a single standard, there are clusters of standards around each of the major aviation hazards. Navigation, aircraft integrity, equipment maintenance, and fire safety each accounts for a myriad of standards, both public and private. Perhaps the most accessible standards, given their familiarity on the ground, are those for fire safety. Even here, the array of regulatory issues is overwhelming. There are at least four facets of fire safety—prevention, detection, suppression, and evacuation—and each accounts for a multiplicity of standards. Suppression, for example, involves standards for fire extinguishers as well as flammability requirements for seat cushions and building materials. The regulatory territory can also be divided by location in the aircraft. The galley, the cabin, the lavatories, and the baggage compartment all pose special conditions that occasionally merit separate standards.

Out of this complicated maze of potential topics for regulation, public standards tend to emerge from specific accidents. The "Airplane Cabin Protection Rule" examined in this chapter stems directly from the Air Canada fire in 1983. The standard covers fire extinguishers and smoke detectors. On the private side, standards evolve quite differently. The number and scope of those developed by the National Fire Protection Association is controlled by the organization's Standards Council in conjunction with the Correlating Committee on Aviation. NFPA

does not have a standard for aviation smoke detectors; but it does have a standard for aviation fire extinguishers. That standard is compared to the FAA regulation in this chapter.

Many of the propositions borne out by the standards for grain elevator safety are contradicted by the standards for aviation fire safety. In some respects, the NFPA standard is better than the FAA's. It is rooted in a technical understanding of fire detection and suppression, and it avoids some of the pitfalls of the FAA regulation. Surprisingly, the NFPA standard requires more fire extinguishers than its public counterpart. That does not necessarily make it better. It is doubtful whether the benefits of this standard exceed the costs. Most remarkable, given the common perception that concerns about cost dominate private standards-setting, is the absence of any arguments to this effect by NFPA committee members.

Compounding the curious juxtaposition of this demanding private standard and the lenient one for grain elevators is their common origin: both were written under the auspices of the National Fire Protection Association. Apparently, NFPA is capable of producing standards ranging from almost spineless to possibly too stringent. The NFPA standard examined in this chapter also changed dramatically over time. It was stringent when first adopted in 1956, languished and became outdated some twenty years later, but was revitalized in 1980 to its stringent, possibly unreasonable, form. These differences across NFPA standards and over time suggest the importance of influences beyond institutional design and administrative procedure. Three influences are identified through this case: (1) the professionalism of fire safety engineers, (2) the tangled web linking this standard and UL's generic standard for fire extinguishers, and (3) the peculiar political culture of aviation safety.

On the public side, the standard described in this chapter shatters the popular image of federal "notice and comment" rulemaking as burdensome and time-consuming. The FAA adopted it less than a year after it was proposed. There were no lengthy hearings, no subsequent judicial appeals. Although the FAA standard suffers from failure to take certain technical realities into account, it is unquestionably strict, like many aviation standards, and arguably unreasonable. These unexpected outcomes, both public and private, suggest that regulatory behavior can be quite issue-sensitive. Cultural attitudes about the risks of flying are, to say the least, peculiar. Apparently no risk is acceptable so long as regulation holds out the promise of improving safety. As elaborated below, this results in decisionmaking dynamics, both public and private, unlike

those attributed to other agencies charged with regulating health and safety.

Fire Extinguishers, Smoke Detectors, and the FAA

The FAA has an expansive regulatory mandate—to certify the air-worthiness of new aircraft, to control the nationwide air traffic control system, to regulate pilot certification, and to police airline operations and maintenance. The agency functions by leaving much to the discretionary judgment of the airlines and airframe manufacturers. Aviation technology is so complicated that the FAA realized years ago it did not have the resources or expertise to pass judgment independently on all new airframe designs. Instead, "designated engineering representatives," engineers employed by the airframe manufacturers, certify new airplanes for the agency. Matters of operation and maintenance are almost as complicated as aircraft design, involving highly technical systems and countless potential subjects for regulation. Airframe manufacturers employ hundreds of safety engineers and continually issue bulletins and warnings to their customers, the airlines. Airlines adopt their own safety routines and maintenance procedures as well. Confronted with the complexities of aviation safety, the FAA proceeds with a mixture of deference, prodding, and formal regulation. All three strategies are reflected in the evolution of the FAA's regulation of fire extinguishers and smoke detectors on commercial airlines.^[1]

Prodding takes the form of advisory circulars and general notices, both of which are "non-binding" but carry significant weight with the airlines. Smaller airlines with limited engineering capacity to develop their own standards are likely to adopt these recommendations. Larger carriers do their own evaluation, cognizant of the cost of ignoring government recommendations should something go wrong.^[2] Formal regulation usually follows from the most serious accidents and those that result in National Transportation Safety Board (NTSB) recommendations that have been made before. How these stages take shape depends on the seriousness of the accident and the nature of any prior experiences.

In the case of fire extinguishers and smoke detectors, the formal) standard was largely influenced by two tragic accidents: the Varig fire in 1973 and the Air Canada fire a decade later. Before these tragedies, however, the FAA long took a deferential approach toward fire extinguishers, mandating simply that "the type and quantity of extinguishing

agent must be suitable for the kinds of fires likely to occur in the compartment where the extinguisher is intended to be used."^[3] The airlines and airframe manufacturers were left to determine what was "suitable." The agency provided additional guidance with the advisory circular in 1965 that recommended up to three fire extinguishers per airplane and sanctioned the use of carbon dioxide extinguishers. ^[4] It contained no actual product specifications such as capacity, discharge time, or nozzle type. Nor did it speak to the placement of extinguishers or the training of personnel in their use.

Cabin fire safety attracted intense attention in 1973 when a terrible in-flight fire killed 124 people on a Varig Airlines Boeing 707 en route to Paris from Rio de Janiero. The fire, attributed to a discarded cigarette in a lavatory wastepaper disposal unit, broke out shortly before landing. Thick black smoke soon filled the cabin and cockpit. The pilots literally stuck their heads out the window in order to make a forced landing a few miles from the airport. Of the 135 people on board, only eleven survived; the remainder died of asphyxiation or from toxic gas. The NTSB focused on facilitating earlier detection and more effective suppression through smoke alarms and smoke masks for flight attendants respectively. The NTSB made numerous recommendations, including requirements for "a means of early detection of lavatory fires … such as smoke detectors or operating procedures for the frequent inspection of lavatories by cabin attendants."^[5] The FAA opted for more "No Smoking" signs, an appeal to the airlines for better monitoring by the crew, and a promise that the "FAA has begun a preregulatory study of the feasibility and justification for a requirement for smoke detectors in lavatories."^[6]

The advisory circular on fire extinguishers was completely out of date by the late 1970s, failing to take into account changes in airplane size and extinguishant technology. New jumbo jets with a seating capacity in excess of three hundred require many more extinguishers than the old planes. Virtually all airlines carried more extinguishers than suggested in the circular. Some also used the new Halon extinguishant, a liquid gas that extinguishes fire by chemically interrupting the combustion chain reaction (rather than by physically smothering it). Halon is a derivative of halgonated hydrocarbon. It is suitable for use in cold weather, leaves behind no chemical residue to contaminate or corrode aircraft parts, and can be three times more effective than carbon dioxide extinguishers of equal weight.

Boeing provided it on all new airplanes after 1979, but Halon still

was not mentioned in the FAA advisory circular until a series of "volatile liquid hijackings" raised concerns about the capability of current extinguishers. (The hijackers in these cases carried jars of gasoline, threatening to set the plane on fire.) The FAA's Office of Civil Aviation Security sponsored a study of hand-held fire extinguishers by the Factory Mutual Research Corporation. The FAA quickly revised the advisory circular on fire extinguishers to explain the advantages of Halon.^[7] The circular did not require Halon, however, and it provided no additional product specifications. The agency stepped up the pressure after tests at the FAA Technical Center revealed that Halon was better than carbon dioxide or dry chemicals at extinguishing volatile liquid fires. The agency's strategy, something between unsolicited advice and formal regulation, was to issue a general notice. G-NOTs, as they are known in the agency, are formal requests for "voluntary" compliance with a suggestion. They are the FAA's way of putting the airlines on the spot—requesting a specific safety improvement and requiring a formal response concerning the airlines' intentions. The agency accordingly issued a general notice on November 29, 1980, requesting all airlines to carry at least two Halon extinguishers.^[8]

About half of the carriers indicated that they had complied or would do so. A few expressed mild resentment at the FAA's refusal to take responsibility for the issue. If Halon was desirable, argued the vice president for engineering and quality control at Frontier Airlines, "there are regulatory means of requiring it rather than encouraging such devices be installed without providing sufficient justification."^[9] Several carriers worried about the possible toxic effects of Halon extinguishers. Some of this concern was prompted by warning labels. ("Use in an enclosed place may be fatal," warns the label on a Kidde hand-held Halon extinguisher.) But the evidence about the toxic effects of Halon in aviation use was limited and mixed. Boeing conducted live fire tests and endorsed the use of Halon, although the engineers refused to release their test data.^[10] American Airlines also conducted tests, and declined to use Halon because it left behind "a strong bromine smell that caused burning eyes and coughing."

The FAA relied on mild coercion until June 2, 1983, when an in-flight fire killed twenty-three people on an Air Canada flight near Cincinnati. The FAA was backed into a regulatory stance toward fire extinguishers and smoke detectors. As a rulemaking staff member put it, the agency gets "one free bite" in adopting regulatory strategies; that is, it has the most discretion in deciding how to address issues being raised

for the first time. But the FAA had already had one bite at these issues, The agency had tried gentle and coercive "advice" on fire extinguishers, and the NTSB had already recommended smoke detectors. New regulations would certainly follow. Their genesis is so clearly in the Air Canada disaster, however, that it is necessary to recount some details of that tragedy.

The Air Canada Fire

Stories from the survivors of the fire, which eventually killed twenty-three people, were sensational, frightening, and widely publicized. A fire broke out in the lavatory of the DC-9, and, as in the Varig fire, thick acrid smoke soon filled the plane. During its emergency descent to the Greater Cincinnati Airport, described by one passenger as "like an elevator ride," the pilot's chair was literally on fire.^[11] His vision was totally obscured by the time he landed the plane, tires exploding on impact. Flames burned through the roof of the fuselage, and scenes of firefighters combating the blaze with foam and water topped the national news.

The complete story, pieced together from separate reports by the National Transportation Safety Board and a special investigative unit of NFPA, is much less dramatic than the emergency landing and rescue effort.^[12] It all began with an electrical short in a lavatory pump. Three circuit breakers tripped in the cockpit, and a crew member tried without success to reset them. Eleven minutes later, a flight attendant detected smoke emanating from the lavatory. An attendant entered the lavatory, saw smoke coming from the wall liner, and discharged a carbon dioxide fire extinguisher in the general vicinity. Minutes later a second extinguisher was also discharged, and the smoke appeared to clear; but a flight attendant checking on the situation soon thereafter found the lavatory door so hot it was considered unsafe to open. The smoke reappeared and got worse. A master caution light in the cockpit signaled an electrical system failure. The pilot decided to make an emergency landing. It took ten minutes to land the plane. Visibility approached zero as the plane filled with smoke. Only those passengers who reached the emergency exits within about one minute of landing got out of the plane safely. The rest apparently succumbed to carbon monoxide poisoning.

Aviation accidents are almost never attributed to a single cause. Usually a combination of mechanical and human factors are involved.

With fires, the direct "cause" of any particular incident is obviously the ignition source. In that sense, a short circuit in the lavatory flushing pump caused the Air Canada fire. The consequences of ignition, however, depend on the combustibility of surrounding materials, the quality of fire detection and suppression, and various elements of the emergency response. Accordingly, the extensive damage from the Air Canada fire—the consequences of ignition—can be attributed to the delay before the crew detected and responded to the fire, the ineffective use of the fire extinguishers, the toxicity of the seat covers, and the difficulties encountered in evacuation.^[13]

Accident investigators tend to take an expansive approach when determining the "cause" of an accident. Aware that regulations are influenced by accident reports, investigators often seek to effect the greatest possible change. "It's better if you don't find the exact cause because then only one thing will get fixed," according to an NTSB investigator. Instead, for every serious accident the NTSB recommends a laundry list of changes in FAA regulations.

The Air Canada incident was no exception. Many culprits were identified. The NTSB was particularly critical of the crew for the delay between detecting the fire and deciding to land. A few independent experts consider this criticism unfair but concur that the crew was ineffective in its use of fire extinguishers. The NFPA report emphasized the problems of smoke and toxic gas, suggesting the need for more research on flammable materials, particularly seat covers. The NTSB wanted several things "fixed" as a result of the fire. It immediately recommended that the FAA inspect lavatory flushing pumps and establish a procedure for verifying whether the circuitry had been damaged over time. Three months later it recommended, among other things, the installation of smoke detectors and the use of Halon fire extinguishers.^[14]

Political Pressure and a Prompt Proposal

Every fatal airplane accident generates extensive media attention and strong political pressures. Congressional committees have an almost limitless inclination to investigate airplane accidents. In his study of policy analysis in the FAA, Steven Rhoads notes that "it would be difficult to overestimate the seriousness with which Congress views commercial air crashes." Within two months of the accident, three separate congressional committees held hearings on the Air Canada

fire.^[15] Accused of "footdragging" and indifference bordering on callousness, the FAA soon became Congress's scapegoat for the incident. Congressman Dan Burton (R-Ind.) made the exaggerated claim that "had [Halon extinguishers] been on-board Air Canada nobody would have died."^[16] Several congressmen introduced bills to mandate the NTSB's proposals on smoke detectors and Halon fire extinguishers.^[17] Congress was unlikely to regulate the matter by statute, however, since it had neither the inclination nor the resources to address such technical questions. (In any case, as one congressman put it, "I would not want to fly in a plane designed by Congressional committee.")^[18] Instead Congress strengthened the oversight process. One committee required the FAA to file monthly progress reports on the implementation of various NTSB recommendations. The pressure to adopt new regulations, including ones concerning fire extinguishers and smoke detectors, was intense.

The FAA knew it had to respond to the Air Canada fire with a regulation and was prepared to do so quickly. In short order the agency drew on its previous experience and drafted a simple standard. First, smoke detectors would be required in airplane lavatories and galleys. No effort was made to define the technical specifications for these devices. Second, a built-in fire extinguisher would be required in the towel disposal receptacle of each lavatory. (This had been suggested by the NTSB after both the Varig and Air Canada fires.) Finally, the number of fire extinguishers required would be increased, and at least two Halon 1211 extinguishers would be required on every plane. The choice of Halon 1211, and the exclusion of Halon 1301, was apparently based on the general notice sent out after the gasoline hijackings. The increase in the number of extinguishers required was simply an incremental guess. "We took a look at the wide bodies," a rulemaking staff member stated, "and said 'we need another extinguisher for every one hundred people.'" Otherwise, to a large extent, the NTSB essentially drafted the regulation.

Objections to the proposal were meek. The American Transport Association, which represents most major airlines and is considered by many to be a powerful lobbying organization, did not oppose "the basic thrust" of the proposal, only the requirement for smoke detectors in galley areas. The few other objections to the rule were technical. Some engineers argued that the standard should permit Halon 1301 as well as Halon 1211 (the numbers denote differences in chemical structure). Others alleged that household smoke detectors would not necessarily be reliable in airplanes. The effects of vibration were cited by Underwriters

Laboratories as one of several possibly significant aspects of aviation use that might impede performance. Unusual air currents might also be important, particularly in airplane lavatories, where the air moves down and out through the toilet bowl.

These claims might have been self-serving—UL was basically advocating that the FAA require laboratory certification of smoke detectors for aviation use—but they were also well founded. Devices designed specifically to endure the rigors of the aviation environment are far more sophisticated than a standard household detector. Detectors for airplane cargo holds, produced in accordance with an FAA Technical Standards Order, cost $800 to $1,300 each.^[19] When the Regulatory Analysis Division in the FAA first analyzed the proposed rule, they used cost estimates based on these devices and arrived at a benefit-cost ratio of less than one. The critical but unanswered question is to what extent accuracy and reliability are sacrificed by allowing the basic dime store model instead of the sophisticated aviation model. Household detectors are not particularly sturdy either. Given the rigors of aviation use, they might not work in time of need. (Liability concerns of this nature prompted at least one major manufacturer to decline an airline's recent order for 1,500 detectors.) Missing batteries, a problem noted by carriers that experimented with smoke detectors, could also incapacitate the smoke detector. Petty theft would not be the only motive. Sabotage by smokers breaking what is widely thought to be one of the most ignored FAA prohibitions—against smoking in the lavatory—is another possibility.

Another possible consequence, argued an official of the British Civil Aviation Authority, is false alarms, which would "soon give rise to a loss of faith in the detection system." Household detectors are sensitive to changes in air flows, something that occurs regularly in flight but seldom at home. Smoke is also common in galley areas, partly as a by-product of food preparation. Several airlines feared that the panic caused by an activated alarm might be worse than the possibility of a fire going undetected without a smoke detector.

Technical Objections and a Final Rule

The other technical objections to the proposed FAA standard concerned either the fire extinguishant or the nozzle configuration. Several manufacturers argued that Halon 1301 should be permitted in addition to

Halon 1211.^[20] Halon 1211 discharges in a more liquid state than Halon 1301 and has better range and direction in use. But Halon 1301 is considerably less toxic than 1211, so it may be advantageous in small areas such as the cockpit. The use of Halon 1301 in hand-held fire extinguishers is relatively recent. There were no hand-held Halon 1301 extinguishers when the FAA issued its general notice in 1980. Three years later, Metalcraft, Inc., received Factory Mutual Research Corporation's first approval for such a product. UL still did not list any. So there was limited information about the merits of this technical question.

The FAA balked and refused to consider the Halon 1301 alternative because, according to an FAA rulemaking staff member, "1301 is not rated for a Class A fire." The reference is to the UL method for rating fire extinguishers by type of fire and extinguisher capacity.^[21] But the reasoning is flawed because UL's Class A fire is simply too big. "Halon is effective on an 'A' fire," notes a fire protection engineer, "but the smallest 'A' fire [that UL builds] takes about nine pounds of agent." Hand-held extinguishers usually have three to five pounds. There may be technical reasons for restricting the use of Halon 1301, particularly in large cabin spaces, where it might be less effective than Halon 1211, but the reason offered by the FAA indicates no understanding of these issues.

A few commenters suggested that the FAA require flexible discharge hoses on all fire extinguishers. This would be particularly helpful in battling fires in concealed spaces, overhead, or under the seats. Fixed nozzles, which are supposed to be operated in an upright position, are difficult to operate under such circumstances. Tests conducted by one airline suggested that flexible nozzles increased effectiveness by an incredible magnitude of ten. The FAA showed little interest in this issue. The rule was published in final form less than four months after the comment period closed. There was only one significant change: smoke detectors would not be required in the galley. This eliminated the airlines' strongest objection without compromising on anything the staff considered critical. The safety record with galley fires was far more reassuring than the record for lavatories. There has never been a catastrophic in-flight fire, domestic or foreign, that originated in the galley. While this requirement was easily dropped, addressing the other objections raised in the comment period was not so easy. They called for changes in specific requirements and would require additional analysis. The FAA had no patience for these arguments. The agency wanted to placate Congress and get the rule published. The rulemaking staff had

no intention of letting public comments delay the process. The candid explanation of a staff member about comments concerning the value of requiring a flexible hose on Halon extinguishers illustrates the attitude: "There probably is a lot of benefit in a flexible hose. But we have a certain practical limitation here: we can't just go ahead in the middle of a rule action and change our minds and say that when the rule comes out we want twenty-five hundred airplanes to be equipped with flexible hoses. That throws everything into a cocked hat…. We had to make a decision and we made it." The rule was finalized in record time. It was promulgated on March 29, 1985, less than ten months after it was first proposed. Three years later, the FAA banned smoking on all domestic flights of under two hours. This may be the most cost-effective move the FAA could take toward reducing the risk of a cabin fire. The ban was adopted for health reasons, however, not for reasons of fire safety.

Evaluating the FAA Standard

Whether the FAA's new regulation would have prevented the Air Canada disaster or, more important, another catastrophe in the future is difficult to determine, partly because serious accidents are so rare. There were no deaths on U.S. commercial aircraft in 1980, for example, and only four in both 1981 and 1984. Even 1985, considered by many "the worst year ever" for aviation safety, was remarkably safe by comparison to other modes of transportation (particularly if deaths due to terrorist acts are subtracted from the total).^[22] Accidents are so rare, according to an actuary Metropolitan Life, that "it is almost pure chance as to which [major commercial airline] has a total loss" in any given year.^[23] The chances of a serious in-flight fire are even more remote. Most aviation injuries and fatalities are caused by impact, not fire. As an airline safety engineer put it, the fatal in-flight fire is "a rare animal of a rare breed." The first recorded fatalities from an in-flight cabin fire in the United States were in the Air Canada calamity.

The almost random nature of accidents creates the first paradox of developing (or analyzing) aviation safety regulations. Accidents precipitate strong political pressures for regulatory change, but they provide little factual basis for making meaningful improvements. Preventing random events is practically an impossible task. Instead of trying to anticipate the unknown, attention tends to get focused on preventing any repetition of what has already happened, regardless of the likelihood that it will happen again.

The politics of aviation safety also conflict with broader political trends concerning government regulation. Many members of Congress openly challenge the notion that aviation safety regulations should be undertaken only if benefits are at least commensurate with costs. Testifying before a House committee about several recent proposals for upgrading cabin safety, the chairman of the NTSB said he "would hate to see [their] implementation delayed for a cost-benefit analysis."^[24] Nevertheless, the FAA is bound by the same executive orders that require all agencies to conduct an economic analysis of proposed regulations.^[25]

This raises the second dilemma of aviation regulation: the FAA must justify in economic terms regulations that sometimes can only be justified on other grounds. This is not to say that the FAA's Regulatory Analysis Group engages in trickery or deception. The agency has refined the use of economic analysis over the years through its capital improvement projects for airport facilities. A detailed FAA manual spells out the procedures and many specific values (including the always controversial value of a life) to use in cost-benefit analysis.^[26] This simplifies and standardizes the agency's analysis. Nevertheless, this case demonstrates how the use of favorable assumptions can make a regulation of questionable economic benefit look economically desirable.

The FAA's Regulatory Analysis Group wrote a forty-three-page cost-benefit analysis of the proposal for smoke detectors and fire extinguishers. The analysis is systematic, comprehensive, and prominently featured in the Federal Register notice. The bottom line, according to the analysis, is that "total expected benefits equal $42.8 million and total costs equal $13.8 million, resulting in total expected benefit-cost ratio of 3.1 and a total expected net benefit of $29.0 million."^[27] There is much to take issue with in this analysis. For example, smoke detectors, assumed to cost $50 each, will very likely cost much more if airlines choose even a few of the features justified by the aviation environment (for example, tamper-proof battery packs, better vibration tolerance).

More important, there are several reasons to call into question not only the magnitude of estimated benefits but whether they actually exceed expected costs. One problem concerns predicting catastrophic in-flight fires. Recognizing the difficulty in predicting the distribution of essentially random numbers, the FAA staff utilized an elegant statistical solution: the Poisson distribution.^[28] The elegance of the solution masks the significance of one critical underlying assumption: the "expected mean value" of two catastrophic cabin fires in the next ten years. This

figure is purportedly based on "historical data." The Varig fire occurred in July 1973 and the Air Canada fire in June 1983. But the Varig fire was neither on a domestic carrier nor a domestic accident. The Air Canada fire caused the first fatalities in the United States from an in-flight cabin fire. To expect two similar fires every ten years seems overly gloomy, especially since there have been improvements in aviation fire safety since the Varig incident. After that fire, several airlines installed heat-sensitive fire extinguishers in trash receptacles. Moreover, there is nothing magic about a ten-year period. Thomas Hopkins, an economist at the University of Maryland's School of Public Affairs, notes in connection with the FAA's analysis of Floor Proximity Emergency Lighting—another proposal linked to the Air Canada fire—that "if the past five years are considered more representative of what lies ahead, there were no pertinent fatalities and so no plausible benefits."^[29] One problem with the agency's cost-benefit analysis, then, is that it does not test the sensitivity of this important assumption. Assuming a mean of one catastrophic fire (instead of two) every ten years could cut the potential benefits of the rule in half.

Other assumptions of vital importance to the analysis include the estimated "coefficients of effectiveness" for smoke detectors and fire extinguishers. Here the analysis purports to incorporate "conservative" estimates. These estimates are not as unambiguously "conservative" as the FAA staff suggests. One assumption is that a smoke detector could avert 50 percent of catastrophic lavatory fires. The staff provides no basis for this estimate or for labeling it "conservative." Stated another way, the coefficient is based on two assumptions: first, that in 50 percent of the lavatory fires that become catastrophic, quicker detection would prevent the catastrophe; and second, that off-the-shelf smoke detectors would actually provide quicker detection. Both assumptions are doubtful. Speed of detection certainly was not the problem in the Air Canada fire. The short-circuit alarm probably alerted the pilots faster than a smoke alarm would have. Delays and uncertainty in responding to the alarm—something as likely to occur with smoke detectors as with circuit breakers—were the main problems. Second, the detection capability of household smoke detectors in an aviation environment, as already mentioned, is quite uncertain. False alarms are likely to be a problem; some fires may go undetected. Combining these considerations, it appears that 25 percent, or maybe even 10 percent, is as reasonable an estimate as 50 percent. Similar arguments apply to the coefficients used for calculating the benefits of trash receptacle extinguishers.

A third problem with the economic analysis is that it overstates the benefits by failing to take into account existing "compliance" with the proposed rule. At least half of the commercial carriers voluntarily complied with the FAA's 1980 request to carry Halon extinguishers (a few even installed smoke detectors without any FAA advice). The benefits attributable to the proposed rule should reflect only the incremental benefit of adding Halon extinguishers to the remaining carriers. Instead, the FAA adjusted the cost figure to reflect the marginal cost of Halon extinguishers but not the marginal benefits.^[30]

The cost-benefit analysis also did not take into account the special training necessary to use a Halon extinguisher effectively. This means either overstated benefits or understated costs, depending on whether the airlines voluntarily improve existing training programs. Most airline employees receive cursory instruction in fire fighting. Only two airlines provide actual "hands on" training with fire extinguishers.^[31] "Hands on" training is particularly important with Halon extinguishers, which discharge in a liquid stream that must be carefully applied because it lasts less than ten seconds. Those not properly trained in the use of these sophisticated extinguishers may actually be less effective than with other extinguishers. Even with less-sophisticated extinguishers, untrained operators are generally able to extinguish only about half as big a fire as those with training.^[32] The carbon dioxide extinguishers aboard the Air Canada jet were discharged without effect in the early stage of that disaster. As an engineer with Factory Mutual Research Corporation put it, "It wouldn't have made any difference if you had given them another thirty extinguishers." The cost-benefit analysis never considered this issue. In fact, the provision for Halon extinguishers was given only superficial treatment. The staff considered these extinguishers "clearly cost-beneficial" because they are lighter than older extinguishers, so "fuel savings alone are expected to pay for this proposal." In short, they were assumed to be more effective. The projected $2.9 million in "pure safety benefit" might be more than outweighed, however, by the cost of training flight attendants adequately to ensure effective use.

Finally, even if the FAA were predisposed to write its own standard, the rulemaking staff should have been aware that the private standard NFPA 408 existed. At least one FAA employee has been on the NFPA's Aircraft Rescue and Firefighting Committee, charged with standards for aviation fire extinguishers, since NFPA 408 was first adopted in 1956. Two FAA representatives were on the committee when the standard

was revised in 1980 and later when the FAA decided to draft and adopt its own standard.^[33] One reason the rulemaking staff was unaware of NFPA 408 is organizational. Neither of the FAA employees were from the Airworthiness Division—the division that wrote the FAA's standard for fire extinguishers—and neither took a particular interest in NFPA 408. They were more interested in other aircraft rescue and firefighting standards.^[34] Beyond these peculiar circumstances, however, the fact that the rulemaking staff did not think to check for the existence of an NFPA standard is evidence of the low profile that currently characterizes these standards.

None of these issues was raised by those commenting on the proposed rule or, surprisingly, by the Office of Management and Budget in its review of the cost-benefit analysis. Politics, it seems, loomed larger than economics. "This was a motherhood issue," explains an FAA rulemaking staff member. "Who is going to argue about fire extinguishers in airplanes?" Indeed, the FAA received hundreds of handwritten letters and postcards from individual citizens in favor of the proposed rule.

A Little-known and Surprisingly Strict Private Standard: NFPA 408

One of the comments the FAA never formally responded to was a suggestion that the agency adopt a private standard for hand-held aircraft fire extinguishers, NFPA 408. That the FAA did not do so is not surprising—government regulators often look on private standards with disfavor. What seems unusual is that the FAA (at least the staff member in charge of drafting the standard on fire extinguishers) was unaware of the existence of NFPA 408 until the agency received the suggestion (which, incidentally, came from NFPA). Ignorance in this instance was a function of poor communication within the FAA and, more broadly, of the waning influence of private, industrywide aviation safety standards.

As with several other areas of regulation, aviation safety used to be addressed entirely by the private sector. When NFPA first got involved in the subject, government regulation was minimal, and the future of private aviation regulation looked promising. Responding to requests from the National Aircraft Underwriters' Association, UL formed an Aviation Department in 1920, and two years later it started offering a service that the FAA would later take over: certifying the airworthiness of aircraft. Insurance groups, interested in standards to use in making

underwriting decisions, asked the NFPA to develop various aviation standards.

NFPA is a membership organization similar to the American Society for Testing and Materials. "Volunteer" committees write the 260 NFPA codes and standards, and the membership at large votes on various standards at semiannual meetings. NFPA has over thirty-two thousand members, including architects, engineers, firemen, and representatives of manufacturers, insurance interests, labor, and government.

One of the standards NFPA developed in the aviation area was NFPA 408,^[35] which contained recommendations concerning the "type, capacity, location and quantity of aircraft hand fire extinguishers and accessory equipment provided essentially for the protection of aircraft compartments occupied by passengers and crew.^[36] It was drafted between 1947 and 1955 by a technical subcommittee of NFPA's Aircraft Rescue and Firefighting Committee. The group had twenty-four members, including seven from government agencies in the United States and Canada, six from commercial airlines, two from academia, and one each from UL and UL of Canada. NFPA 408, with appendices, was less than six pages long. It provided information on the two most prevalent types of extinguishers (dry chemical and water) and mandated that airlines carry one small extinguisher in the cockpit and, depending on occupancy, between one and three in the passenger compartment. The standard also set forth a suggested training outline on the use of fire extinguishers. Unfortunately, NFPA has no record of how these specific provisions were developed. (Only in recent years have comments and committee minutes routinely been retained.)

Over time, interest in private, industrywide aviation safety standards diminished—at least in this country. The Civil Aeronautics Board, predecessor to the FAA, displaced UL's entire Aviation Department. Insurers began using compliance with the FAA's airworthiness standards as a condition of insurance. As technology became more complicated, airframe manufacturers (for example, Boeing, Lockheed, McDonnell-Douglas) assumed much of the responsibility earlier undertaken by commercial carriers. Currently, "most carriers will accept what the airframe manufacturer offers," according to the fire protection engineer at the only major airline to employ one.^[37]

This apathy affected NFPA 408. Although reissued in 1965, 1970, and 1973, the standard was largely unchanged from its original version. The main reason, suspects a current member of the Aircraft Rescue and Firefighting Committee, is expressed by the adage about letting sleeping

dogs lie: "There had never been a demonstrated problem warranting attention." There also was little interest in such standards among domestic airlines and airframe manufacturers. The in-house standards at Boeing were more important than NFPA 408. The primary interest in NFPA 408, and in many of NFPA's other aviation safety standards, was from foreign countries. As one committee member put it, "There is no FAA in Greece." Many foreign governments thus look to these standards for guidance. But the foreign contingent on the committee did not attend meetings regularly, and many lacked sufficient technical background to suggest improvements. Moreover, given the limited demand for the standard, NFPA 408 generated almost no income for NFPA. (The sale of publications accounts for two-thirds of NFPA's income.) But NFPA 408 is one of many NFPA standards offered more as a public service than as a money-making proposition, and at times these standards suffer from lack of attention.

NFPA 408 languished in the late 1970s when changes in technology rendered it out of date (it did not take into account the new jumbo jets, which could hold over three hundred passengers). The relevant provision in the 1973 version (for occupancies "over 61 passengers") called for three fire extinguishers. "An airframe manufacturer would never provide so few extinguishers [for a jumbo jet]," notes an NFPA committee member. Halgonated extinguishing agents also came into use in the 1970s, but NFPA 408 made only a passing reference to them.^[38] The 1973 version still allowed carbon dioxide extinguishers, which had long since fallen out of favor with most fire protection engineers because of the damage they can do to electrical equipment.

The 1980 Revival of NFPA 408

The Standards Council, the general oversight group within NFPA, recognized the problem in 1980, when nine aviation safety standards, including NFPA 408, were overdue for revision and reissue. The chairman of the Technical Committee on Aircraft Rescue and Firefighting considered 408 so inadequate that he proposed that NFPA withdraw it and start over from scratch. Withdrawing the standard as outdated was not in the interest of the Standards Council, however, which seeks to protect NFPA's reputation and is aware that the organization obtains its income largely from the sale of standards.^[39] Even though the income from this standard is minimal, it is in NFPA's general interest to keep its standards available. The Standards Council instructed the committee to

expedite the process of bringing the standard up to date. A technical subcommittee met several times in the following year and drafted a new version. The new standard included changes in the number of extinguishers required, the type of extinguishers, and the nature of employee training. All of these changes were in the direction of being more stringent, although a few were left as suggestions rather than stated as requirements. Most significant, however, the basic requirements of NFPA 408 were more demanding than what the FAA eventually required.

The most significant and costly provision in NFPA 408 specifies the number of extinguishers required. The revised version requires more than twice the number specified in the old version—and more if necessary, to ensure that there is an extinguisher within thirty feet of any passenger.^[40] How did the subcommittee choose these numbers? By doing for aviation safety what the 61B committee would not do for grain elevator safety: a combination of guesswork and fire protection rules of thumb. "If you are asking whether it is like Newton's law, where we can categorically support the conclusion, the answer is no," explained an engineer on the committee. There was surprisingly little disagreement among committee members, however, concerning the specific numbers chosen. Most members, even those representing the airlines, took the general view that the standard should err on the side of safety. In this respect, NFPA 408 reflects the professional norms of aviation safety engineers, who frequently rely on significant margins of safety. The margin of safety in the revised version is more than adequate. This standard is unlikely to produce benefits in excess of costs, however. A candid NFPA staff member admitted that this is probably true of all NFPA aviation safety standards.

The revised version of NFPA 408 also takes into account recent changes in extinguishant technology. The standard prohibits carbon dioxide extinguishers. (The FAA still allows them.) NFPA 408 also specifically requires, for the first time, the use of Halon 1211 extinguishers. Several key committee members knew that Halon 1211 is an extremely effective extinguishant. The toxicity question, raised by several airlines in response to the FAA's general notice, was not a sticking point. Most members consider these concerns exaggerated and inappropriate. "You have to put out the fire before you start worrying about toxicity," explained one member. There was minor disagreement about which Halon agent to require, but the idea of Halon was endorsed largely on the recommendation of the committee's engineers. Represen-

tatives of companies that manufacture such extinguishers naturally supported the idea as well, but those firms make all types of extinguishers and have no particular stake in Halon.

Flexible Nozzles and Special Training

The Technical Committee on Aircraft Rescue and Firefighting considered two other issues to be important—flexible nozzles for extinguishers and special training for the use of Halon—but they acted in a markedly less decisive manner on both. On flexible nozzles, the fire protection engineer from a major airline made a convincing case that extinguishers would be much more effective with this design change. He conducted tests that indicated that models with a flexible hose could be almost ten times more effective than current models in aviation use. The committee was largely unmoved, however, and simply changed the standard to permit flexible hoses but not require them. An appendix section advises that "for access to underseat, overhead, and other difficult to reach locations consideration should be given to using extinguishers with a discharge hose."^[41]

This seemingly timid approach is a product of the tangled web between installation standards (such as NFPA 408) and product standards (such as UL's). NFPA does not write product standards per se. It does, however, specify some performance characteristics for products. This creates an awkward relationship with the product standards written by such organizations as UL. Sometimes performance characteristics are closely linked to basic product specifications. For example, requiring high enough temperature tolerances for a chimney necessitates that it be made of metal, not masonry. In some instances, NFPA's requirements seem to drive UL's standards; in others, the UL standard appears to control the NFPA standard. The situation is often compared to the proverbial chicken-and-egg problem. In the case of fire extinguishers, however, it is clear which came first: UL did.

Realistically, NFPA can require something different from UL only when it is sure that UL will change accordingly. This is practically assured when the NFPA standard affects a substantial share of the certification market.^[42] But in the case of aircraft fire extinguishers, the NFPA standard affects a minuscule portion of the market regulated by UL. Not only does UL feel little pressure to change its fire extinguisher standard to satisfy the special concerns of aviation use; it foresees a

limited reward for the effort as well. The market for aviation fire extinguishers is too small.

The generic UL standard for fire extinguishers continues to take precedence over any NFPA requirements in 408. UL tests extinguishers under specific conditions and certifies them with different ratings.^[43] NFPA 408 depends on these standards to define the capabilities of the fire extinguishers required by NFPA standards. What UL requires is not necessarily what NFPA would choose for aviation use. UL does not require a flexible hose on small hand-held extinguishers, for example. Nor does it require a discharge time of more than eight seconds for the typical small Halon extinguisher. A United Airlines engineer thinks that twelve to fourteen seconds would be much more desirable. And flexible nozzles are clearly a major improvement over fixed nozzles for aviation use.

The other technical issue considered, but skirted, by NFPA involved the training requirements for using Halon extinguishers. The committee settled on a vague requirement that "training shall provide classroom instruction and manipulative skills training." The appendix removes the teeth from this provision, however, by adding that "it is highly recommended that live fire training on representative aircraft fires be conducted … [but this is] not required by this standard."^[44]

The mild-mannered approach to this provision also stands in contrast to the other more stringent provisions of 408. It reflects in part NFPA's reluctance to specify training requirements and in part the resistance of the airlines to a significant and recurring expense. The representative of Factory Mutual, who had recently conducted a study of hand-held fire extinguishers under contract to the FAA, recalls that the importance of training costs was repeatedly mentioned by those he surveyed. Fire safety experts agreed that training in realistic test situations would be costly. NFPA has a general position against addressing training or other seemingly managerial tasks. The Aircraft Rescue and Firefighting Committee managed to include more specific statements about the nature of recommended training procedures than are contained in most NFPA standards. But even that language is weak.

The committee members approved the proposed changes in May 1983, a full year before the FAA proposed its own rule. A public comment period followed release of the document to the NFPA membership, and when the committee met the following November it was faced with a total of fourteen comments from only four individuals. UL submitted the most detailed comments, most of them definitional, demon-

strating a better understanding of fire extinguishers than the committee as a whole. There was a minor spat about whether a competing Halon agent—lower in toxicity, but also less effective—should be permitted. The committee rejected those proposals on the grounds that they were not supported by accompanying technical data.^[45] NFPA 408 was approved without comment or question by the general membership at the organization's 1984 annual meeting. It became effective on July 5, 1984, a little more than two years after the Standards Council instructed the committee to expedite the revision process.

Summary Evaluation

Deciding whether either the FAA or the NFPA standard is desirable depends on the choice one makes between the economic and political views of aviation safety. In economic terms, it is unlikely that either standard produces benefits in excess of costs. Aviation safety experts in both sectors confirm that almost no recent proposals for improved aviation safety can be justified on economic grounds. This does not render these standards unpopular, however. The political culture of aviation safety is characterized by an extreme "no-risk" perspective that apparently cuts across public and private boundaries.^[46] Congressman Norman Mineta (D-Calif.) recently allowed that "no rational risk analysis or cost-benefit analysis would conclude that the next increment of safety improvements needed in this country is in aviation." He went on to argue for precisely such expenditures.^[47] A similar view prevails at NFPA. Apparently, all that matters in either sector is whether new regulations might make the skies safer—how much safer and at what cost are of little interest.

"Safe enough," notes an NTSB official, sounding a popular chord, "means safer every year." In those terms, both the FAA and NFPA standards are probably desirable. That is, they would contribute, however minimally, to improvements in aviation safety. But even that conclusion is unsure, particularly for the FAA standard. One aviation safety expert considers the FAA's response to the Air Canada fire purely "cosmetic." While virtually everyone connected with the smoke detector business had serious doubts about putting household detectors in airplane lavatories, the FAA was unconcerned. A rulemaking staff member demurred that the UL standard for household detectors "is a very impressive document," implying that it provides sufficient requirements for aviation use. UL disagreed, detailing in a letter to the FAA's public docket numerous reasons why household detectors are inappropriate

for aviation use. The height of FAA hubris is summed up in a staff member's conclusion that, although there might be operational difficulties, "we say in the rule that [the airlines] are expected to keep [the smoke detectors] working."

A lack of technical understanding, similar to that expressed by OSHA in the grain elevator proceedings, was evident in how the FAA dealt with fire extinguishers. The rulemaking staff did not appreciate the possible significance of flexible nozzles. Nor did they apparently understand the need for special training in the use of Halon extinguishers. These points were called to the FAA's attention, however, and the failure of the agency to respond demonstrates the extent to which FAA rulemaking is driven by political pressures. After the Air Canada fire, the FAA was painfully aware that Congress wanted the agency to enact a rule, any rule. And that is precisely what the FAA did—placate an impatient Congress with a quick, but half-baked, safety standard.

In sum, the FAA displayed a surprising range of regulatory behavior over time. The agency relied on quiet advice for years, deferring largely to private decisions. That tactic is more effective than it is often portrayed. The airframe manufacturers probably do more to advance aviation safety than the FAA does. But the FAA is quick to regulate when it is subject to strong congressional pressure, usually triggered by the NTSB recommendations that follow every calamity. The FAA's response to the Air Canada fire demonstrates that public standards-setting need not get bogged down in procedural requirements. This encouraging note is tempered, however, by the realization that important technical issues were overlooked in the rush to regulate.

The private sector, on the other hand, was slower and more sensitive to technical issues. Expressing the kind of technical knowledge also present in the grain elevator proceedings, the NFPA committee recognized the benefits of flexible nozzles and proper training. Showing the same regulatory philosophy present in the grain elevator proceedings, the committee declined to adopt specific requirements in areas deemed "managerial." The committee made "non-binding" recommendations instead. But the NFPA standard is actually more demanding than the FAA's in its requirements for fire extinguishers. That is probably because professional fire safety engineers played a central role in the development of NFPA 408. These engineers do not profess to balance costs and benefits in the pursuit of fire safety. Rather, there is a powerful professional tendency, even in the private sector, to "favor fire protection for the sake of fire protection."

The NFPA standard does not address smoke detectors; they are not within the formal jurisdiction of the Aircraft Rescue and Firefighting Committee. While this lapse is not necessarily bad, it suggests a possible shortcoming of private standards-setting. The specialization of committees threatens to overlook broader regulatory issues and ignore some of the connections between standards. To address these problems, there is a proposal within NFPA to create a "Cabin Fire Protection" committee. An active participant in NFPA 408 calls the proposal controversial and political. "It would require much broader expertise than anyone has. But everyone would want to be on that committee."

The tangled connection between so-called installation standards (such as NFPA 408) and product standards (such as UL's standards for fire extinguishers) also affected the quality of NFPA 408. The standard relies on UL's generic standards for fire extinguishers. But the UL standard is not geared to the aviation environment. Technological changes that would be appropriate to aviation standards—such as flexible nozzles, longer discharge time, or a Class A rating for five-pound extinguishers—are therefore not incorporated into NFPA 408. Unfortunately, the cause and possible cure of this disjuncture between installation standards and product standards is not clear from this case alone. This chicken-and-egg problem confounds various cases of overlapping private regulation.

There was also a marked change in private behavior over time, but for reasons unrelated to the Air Canada fire. NFPA 408 languished in the 1970s, when, as an NFPA officer quaintly explains, there was "an attendance problem." The Standards Council intervened in 1980, and the standard was revitalized shortly before the Air Canada fire in 1983. The changing fortunes of NFPA 408 reflect the importance of the demand for private standards. Standards are demanded for a host of reasons: for example, to provide technical information, to lend credibility, and to minimize exposure to liability. The first two of these influences waned as airframe manufacturers seized the initiative for most safety issues. These manufacturers obviously have the technical capability, and they apparently have attained the political credibility to engage effectively in self-regulation. (Liability law does not appear to be an important influence in aviation standards, since the law assesses liability practically without regard to fault.) As the demand for NFPA 408 waned, so did attendance at the Rescue and Firefighting Committee. Some foreign air carriers still sought regulatory information, but the domestic demand was minimal. The correspondingly low revenue gen-

erated by this standard probably helps explain NFPA's acquiescence in this lapse. The evolution of NFPA 408 suggests that minimal demand can beget minimal standards. Attempting to reverse this trend, the NFPA Standards Council pushed its aviation committees to update their standards in 1980. This restored the standard to the substantive status it had enjoyed before falling into disuse. But the demand is artificial. There is no evidence that air carriers or airframe manufacturers actually use this standard. It seems likely, therefore, that it will languish again, NFPA's gallant efforts to sell more standards notwithstanding.

Five
Safety Standards and Labeling Requirements for Woodstoves

In the first two case studies, the public and private standards were developed largely in isolation from each other: by choice with grain elevators, through ignorance in aviation fire safety. These standards were essentially independent. They were not aimed at each other, so they are conducive to being analyzed alone. The final two case studies are more complicated. Both involve the Consumer Product Safety Commission, an agency that has packed considerable experience with private standards into its short history. The CPSC was born of controversy about the desirability of private standards and has since experimented with various strategies for interacting with the private sector, including the ill-fated offeror process.^[1] Currently, the agency is required by statute to consider the desirability of private standards before proposing any government regulation. This relationship—where instead of being isolated, standards are deliberately intertwined—raises important questions about the interaction between public and private standards. Both of the remaining paired cases suggest that the secondary effects of safety standards—that is, how they affect each other—is as important as how they regulate behavior directly.

In the case of woodstoves, the effects of the CPSC regulation on the private standard are perhaps the only saving grace of an otherwise disappointing regulatory effort. As with so much government regulation, the process of developing the woodstove rule was long-drawn-out. The final CPSC standard addresses the labeling issue but avoids

the more serious problem of creosote fires. The prospect of judicial "second-guessing" has apparently intimidated the CPSC in much the same way that liability law is thought to stifle innovation in the private sector. Fortunately, the private standard improved significantly while the CPSC standard was being developed.

This case also highlights some important and unexpected differences in the patterns of compliance with public and private standards. The private sector achieved much better compliance with its standard than did the CPSC. Apparently, the force of law behind "mandatory" public standards is not necessarily stronger than the forces that induce compliance with "voluntary" private ones. Although the CPSC refuses to recognize it, the institutional arrangements attendant to UL's standards—a function of being in the business of product testing—are sometimes superior to command and control regulation in attaining high levels of compliance.

In addition to dispelling the notion that private standards are "voluntary," the UL standard for woodstoves partially contradicts the idea that private standards are lax. The standard can be faulted for avoiding certain hazards (particularly creosote fires) but several of its requirements are rather stringent. As with the private standards already examined, this standard was developed long before the threat of government regulation arose. Many of its provisions are not supported by specific scientific evidence; rather, they are the product of engineering judgment and educated guesses. These judgments apparently command respect, and there have been minimal objections to the UL standard. This suggests that UL has greater credibility or, perhaps, greater clout than the CPSC. A strong demand for compliance with private standards, emanating in this case from building codes and product liability insurers, may bolster the effectiveness of private standards-setting.

An Overview of Woodstove Safety

Americans rediscovered the woodstove in the 1970s, almost 250 years after Benjamin Franklin designed the first model intended primarily for heating rather than cooking. Sales of woodstoves tripled between 1974 and 1978. With increased popularity came improvements in technology. Franklin's design was altered first by the addition of doors—the Franklin stove was simply a firebox with an open front—and more recently by airtight construction that makes stoves burn hotter and more efficiently.

Woodstove-related fires also became common in the 1970s. Woodstoves were mentioned frequently in fire incident data analyzed by the Center for Fire Research at the National Bureau of Standards. Whether these hazards should he addressed through safety standards is disputed. It is widely agreed, however, that woodstoves are potentially dangerous. "Making wood heat an effective alternative to conventional heating requires the ultimate in careful planning," warns a popular consumer magazine.^[2] In the absence of proper precautions, woodstoves pose four general hazards: creosote fires, ignition of nearby combustibles by radiant heat, escaping sparks or fire, and surface burns.

The Creosote Problem

Creosote fires account for approximately 60 percent of woodstove-related fires. Creosote is formed when the moisture expelled from burning wood combines with unburnt combustible gases in the flue. A tarry substance builds up on the flue lining, eventually becoming brittle and highly flammable. If the chimney is not cleaned in time, high temperatures will start a fire that can spread to the surrounding structure through radiant heat or, if the fire is hot enough, by burning through the chimney.

The process of burning wood inevitably produces creosote. The amount depends on the type of wood, its moisture content, and, most important, the temperature of the fire. Greener wood creates more creosote. So do low burning temperatures, such as those obtained when the stove damper is adjusted for overnight burning. Product design also affects creosote production. New high-efficiency stoves burn hotter and create less creosote than most traditional models. Catalytic combustors, an even newer technology, reduce creosote production through a complex chemical interaction between wood smoke and noble metals such as platinum that enables the smoke to release more heat before going up the chimney.

Nearby Combustibles and Other Hazards

Woodstoves radiate heat that is deceptive in its capacity for igniting nearby walls or furniture. As a rule of thumb, stoves should be installed at least thirty-six inches from the wall and surrounding furniture. This varies both by stove and by building material, but installing a stove even a foot too close to a wall can eventually result in a fire. There is a similar problem with chimney connectors and other piping, which require

about an eighteen-inch clearance from the wall and ceiling but are often installed closer, particularly when passing through a wall or ceiling (see figure 2). Approximately 20 percent of woodstove-related fires are thought to be caused by insufficient clearance to combustibles. Some experts consider the connection to the chimney—rather than the distance from the wall—to be the most dangerous aspect of woodstove installation.

The third hazard, least significant in occurrence, is that fire will escape from the stove. This includes fires caused when (1) sparks escape from the air inlets, (2) coals or flames escape through the stove door (often because it is open), and (3) the fire actually burns through the firebox.

The final hazard associated with woodstoves is the most common, the least severe, and is almost impossible to control through regulation. It is surface burns caused by contact with a hot stove.

Are Woodstoves a Serious Problem?

Woodstoves appear to pose a sizable safety problem. They are second only to careless smoking as the leading cause of residential fires. The CPSC estimates that solid-fuel heating equipment was involved in 140,000 fires in 1985, causing approximately 280 deaths and over $300 million in property damage.^[3] There are several reasons to discount the significance of these numbers, however. First, they stem from dubious extrapolation techniques.^[4] The CPSC's sample is limited and nonrandom. The data do not distinguish between woodstoves and fireplace inserts. They are both lumped together under "solid-fuel appliances," leaving it unclear how much of the problem is actually attributable to woodstoves. Moreover, reports compiled by local fire departments are usually sketchy and sometimes inaccurate in assessing causes.^[5]

Second, these national estimates conceal the fact that most woodstove-related fires are minor. According to the National Bureau of Standards, "of the 11,534 residential solid fuel related fire incidents reported in the [eleven-state] data base … the loss was under $1,000 in seventy-two percent of the fires."^[6] Finally, and most important, it is not clear whether any of these hazards, whatever their frequency, can be reduced by product standards. There is a report in the CPSC files, for example, of an injury caused when an adult tried to retrieve an aerosol can from his woodstove. Obviously, no product standard could prevent

Figure 2.  Woodstove and Wall Pass-through System

this kind of injury. This is not necessarily true of all, or even most, fires commonly attributed to the consumer. "Injuries," as one independent consultant put it, "are often caused by an unfortunate combination of design, installation, and use." The number of fires directly attributable to the product itself is probably very small.^[7]

Even with all these uncertainties, both a public and a private organization chose to write standards for woodstoves. UL officially proposed a draft standard for woodstove safety in January 1978—seven months after the CPSC received a petition requesting that the government regulate woodstoves. But forestalling government regulation was not, as it might appear to have been, UL's motive. The UL standard (in "unpublished form," as explained later) actually predates the petition to the CPSC by several decades. Moreover, there is minimal overlap between the CPSC's standard and the UL standard. The former addresses only labeling; the latter aims to be comprehensive and includes performance tests and design requirements as well as labeling requirements.

Underwriters Laboratories

UL is the premier product testing lab in the country. Building codes and other use and installation codes inevitably require that various products be certified by a "nationally recognized testing laboratory such as Underwriters Labs." In fact, there is no other organization quite like it. This has led to charges that UL is a monopoly—a charge that UL officials privately admit is true in many product areas. It is not true, however, in the case of woodstoves.

Founded in 1894, the year after the Palace of Electricity astonished visitors at the great Chicago Exposition, UL was established by an electrical investigator hired earlier by the Chicago Board of Fire Underwriters to assist exposition officials. Aware of the burgeoning market for electrical equipment and the lack of available safety standards, William Merrill started a business with the still-popular UL motto: Testing for Public Safety. The original idea was to provide information to insurance companies on the fire risks attendant to various electrical equipment. The organization (known then as Underwriters Electrical Bureau) soon became affiliated with the National Board of Fire Underwriters. Incorporated as Underwriters Laboratories in 1901, UL began offering its testing services directly to manufacturers. A manufacturer would submit its product to UL for testing, pay a testing fee, and receive permission to display the UL label (also for a fee) if the product was approved. This third-party certification service, as it is often called, was soon expanded

to include a "follow-up service," now a major part of UL's function. Currently, the "listing" of a product is contingent upon the use of UL's follow-up service, whereby representatives of UL make periodic inspections of the products at the factory and possibly from the open market to determine compliance with UL requirements. This comprehensive inspection system is a form of quality control, meant to assure that proper tolerances are kept in the manufacturing process. Subscribers to UL's listing service are visited approximately four times per year.

UL has always prided itself on its independence from manufacturers' interests. Although originally affiliated with the insurance industry, UL became legally independent in 1936, when it incorporated as a nonprofit corporation in Delaware. The Internal Revenue Service challenged the tax exemption, however, and a federal court eventually agreed that UL "may be good business, but it is not charity."^[8] The tax exemption was restored by Congress, and UL continues to be viewed as the most independent of the private standards-setters. UL's reputation is excellent. It has been remarkably free of scandals or horror stories. Staff members at the Federal Trade Commission, which proposed to regulate private standards and certification in 1977, concur that UL is usually above reproach.

UL has branched out considerably over the years from its best-known area, electrical devices. There are over three thousand employees (a third of whom are engineers) and six major departments—burglary protection, casualty and chemical, fire protection, heating and refrigeration, marine, and electrical—handling literally thousands of product categories. In 1984 over 2.5 billion labels bearing the UL mark were used at 35,381 manufacturing sites in sixty-four countries.^[9] Other testing labs compete with UL, but only in a limited sense. The competition often uses UL's standards. Some labs provide certification at a lower cost than UL, in part because they do not have to bear the expense of developing standards. They also do not have to assume the same risk of being held liable in a product liability suit, since they bear no responsibility for the content of the standards. These labs are not always considered "nationally recognized" for the purposes of regulation, however, and some insurance companies charge higher premiums for product liability insurance when the manufacturer uses a testing lab other than UL.

UL 1482: Room Heaters, Solid-Fuel Type

At first blush, UL's standard for coal and wood-burning stoves appears complicated and highly technical. The current version, incorporating

minor changes made four years after the first edition was published in 1979, is sixty-one pages long and covers both the construction and performance of stoves. Some provisions specify design requirements. For example, cast-iron stoves must not be thinner than 3.17mm (one-eighth of an inch).^[10] Other provisions involve mechanical tests. The "strength test" for chimney connectors calls for a twenty-pound sandbag to be swung on a pendulum against the chimney connector.^[11] The stove itself is subject to various stability tests, including a "drop test" in which the stove is dropped ten times from a height of one inch. Other provisions are more general and subjective, such as the prohibition against "edges, corners, or projections that present a risk of a cut or puncture-type injury to persons."^[12] Twenty-six pages of UL 1482 are devoted to electrical blowers, an optional item not included on many stoves, which help circulate heat throughout the room. Given UL's orientation toward electrical devices, this section is extremely detailed and refers to many other UL standards for motors, wire, switches, and component parts.

But the fire tests are the guts of UL 1482. They subject the stove to both intense "flash fire" conditions and more stable, long-term burning conditions. The temperature at designated points on the stove's surface and surrounding walls must stay within specified tolerances during three separate tests. The manufacturer specifies the distance the stove should be from the wall during testing. If the stove passes, the subsequent label indicates the precise certified clearance (or "distance to combustibles").

From Proposal to Publication

There are two versions of the making of UL 1482: official and unofficial. Like textbook descriptions of the legislative process, the official version misses most of the important subtleties. But it describes the framework within which the standard was written and is necessary to understanding the unofficial story that follows, which is based more on interviews than on official documents.

The official story begins in January 1978, when UL "proposed" the standard to its Fire Council and to manufacturers of wood-burning appliances, circulating the standard for written comments. Five engineering councils composed of outside safety experts provide input to UL engineers in standards development. Council members may not have any formal relationship with manufacturers or distributors. Councils

meet only every other year, however, so most business is conducted by letter ballot. The standard was revised and circulated for comments in August. Three months later, UL hosted a two-day meeting for stove manufacturers and trade association representatives to discuss the proposed standard with UL's engineers. Based on comments received at those meetings, another revised version of the standard was circulated for comments in January 1979. By this time, the wording of the warning label was the only provision at issue. A revised warning label proposed by UL in April 1979 received sufficient support from industry for UL to consider the standard acceptable and publish it.

Successive drafts of the proposed standard indicate how it changed during this process. Unfortunately, this paper trail reveals little about the motivation for these changes. And UL employees are reluctant to discuss the deliberations that go into a standard.^[13] Even the so-called rationale statements recently added to UL standards are often brief and general, providing little insight into the trade-offs and underlying issues.

In the case of UL 1482, most of the changes involved the finer points of test methods. For example, the proposed standard described the charcoal briquettes to be used in testing by reference to the product manufactured by the Kingsford Chemical Company. The revised version adopted a more generic approach, specifying the size, weight, and moisture content of acceptable briquettes.^[14] The aspects of the standard that were changed most significantly during this process and appear to have generated the most disagreement were the warning labels and installation instructions. As originally proposed, UL 1482 required limited installation instructions and a warning label—or "caution mark," in UL terminology—on the stove.^[15] The Wood Energy Institute took strong objection to the warning label and convinced UL of Canada,^[16] but not UL, to drop the requirement. UL revised the labeling requirement, however, in response to this opposition. The minimum letter size for the warning label was reduced and, for stoves with glass doors, the label did not have to appear on the front of the stove.^[17] The requirements concerning installation instructions were less controversial. They were expanded during the revision process without substantial opposition.

Canvassing for Consensus

Once satisfied that industry representatives had no major objections to the standard, UL began a separate process to obtain acceptance from the American National Standards Institute that UL 1482 was a "na-

tionally recognized consensus standard." ANSI aims to be a central clearinghouse and general overseer of nongovernment standards-writers. However, the organization has limited resources, no technical staff, and no information collection system. It depends on the voluntary cooperation of standards-writers. Although UL routinely submits its published standards to ANSI for such approval, the gesture is largely a matter of courtesy without any practical significance for the recognition or use of UL standards.^[18]

There are three separate procedures for gaining ANSI approval. UL uses the one subject to the most criticism: the canvass process.^[19] Under this procedure, UL developed (and ANSI reviewed and approved) a "canvass list" of parties thought to be interested in reviewing the woodstove standard. A professor at Auburn University who had conducted research on woodstoves for the CPSC was the only individual on a list of twenty-six. The rest represented organizations ranging from the Alliance of American Insurers to UL's Consumer Advisory Council and the National Safety Council. The standard was distributed to those on the list with a request for comments and an affirmative or negative vote. ANSI also solicited "public review comments" through a notice in its newsletter, Standards Action .

Under ANSI's canvass method, public comments and a compilation of canvass votes are submitted to ANSI's Board of Standards Review (BSR), whose job is to certify "consensus." They do so mainly by examining any "unresolved negatives" that emerge from the process. In the terminology of the BSR, UL 1482 was a "clean case." It was submitted to ANSI with no "unresolved negatives." No comments were received from the general public during ANSI's own comment period. On the basis of that information, the BSR approved UL 1482 without discussion on January 28, 1981.

What appeared to be a "clean case" to the BSR was not nearly so straightforward for UL. The canvass process took UL over twenty months—longer than it took from official proposal to publication of the standard—and it generated numerous negative comments along the way. Negative ballots were cast (at least initially) by members of five organizations on the canvass list, including NFPA, the National Bureau of Standards, the International Conference of Building Officials (ICBO), and the American Society of Heating, Refrigeration and Air-Conditioning Engineers (ASHRAE). Others voted affirmatively but registered negative comments.

The objections were to the scope and severity of the standard. Rep-

resentatives from NFPA and NBS objected to the lack of any provisions concerning the tendency of stoves to produce creosote.^[20] Representatives from ICBO and ASHRAE objected to the stringency of various test methods. The former considered an aspect of the fire tests too weak, while the latter argued that the impact test for glazing was unreasonably stringent.^[21]

UL responded to each party that cast a negative vote, seeking to elicit a vote in favor. The Standards Department at UL is charged with this task. The department, with almost fifty employees, acts as an intermediary between commenters and UL engineers. The department aims to defuse opposition and gain acceptance for UL's standards by informing commenters of the reasons for the provision in question and, if this fails, by trying to convince them that improvements can be made in the future. Commenters often agree to vote in favor of a standard on the assurance of the Standards Department that the issue will be addressed in the future. This tactic changed all of the votes against UL 1482 but that of the International Conference of Building Officials. UL had to request a six-month extension from ANSI in order to bring ICBO around.

Several of the objections to UL 1482 were dismissed by UL in a rather perfunctory fashion. Those who objected to the lack of provisions governing creosote formation were told that ongoing research in the scientific community was expected to provide a data base upon which a future test method might be developed. This response appears less than forthright given the events preceding the canvass. There, in response to questions about how (and whether) to deal with creosote formation in chimneys, the matter was dropped because, according to UL's report of the meeting with industry representatives, "the discussion indicated that it was primarily an installation consideration and not a product construction or performance requirement for inclusion within the body of UL 1482." An internal UL memo dated March 19, 1985, indicates that UL is still "not in a position to include any creosote tests."^[22] And a UL engineer confirms that without a request and financial support from industry, there are no plans to develop such a test.

Other objections to UL 1482 were also "resolved" in an unresponsive manner. A building code official objected that the glazing impact test would "increase the cost of glass doors without producing a safer product." UL's response: the test was intended "to provide the assembly with impact conditions under various modes." UL offered no evidence that the product would actually be safer as a result.

The Unofficial Story and UL's Unpublished Standards

Missing from the official version of the adoption of UL 1482 are any details of what turns out to be the most critical step in the process: developing the "proposed" standard. Although the proposed standard was modified through UL's review and comment process, the changes were not significantly related to the safety implications of the standard. The most important provisions in UL 1482, at least from the point of view of safety, came from the original proposal and were not affected by the formal process that followed. The official story, then, picks up where the unofficial story ends—with the "proposed first edition" of UL 1482.

It is misleading to describe the standard circulated by UL in January 1978 as a "proposal." It was, in fact, both a proposal and a working (but unpublished) standard. UL had actually been certifying woodstoves for over twenty-five years when it circulated as a "proposal" the requirements it had been using. These provisions were contained in an unpublished UL standard (also referred to as a "desk standard"). Unpublished standards, the foundation of all UL standards, are not well known or understood. When the CPSC first began investigating woodstove safety, it operated for several months under the mistaken impression that there was no UL standard. In fact, there was an unpublished standard, just not a published one.

Unpublished standards and their method of development are one of the only blemishes on UL's otherwise excellent reputation. These standards are written entirely in-house by UL engineers and, true to their name, are not generally available to outsiders. Those who evaluate standards-writing by the extent to which the procedures are public and accessible are most critical of UL's unpublished standards.^[23] Leaving the merits of that debate aside until later, the fact is that unpublished standards are an integral part of UL and its standard for woodstoves. Relevant to the present discussion is how these standards take shape.

The Importance of Precedents

The process begins when a manufacturer of a product for which there is no UL standard asks UL to test and certify the product's safety. Lacking a standard with which to test the product, UL creates a rough draft (an unpublished standard), which is refined through use. Then, if there is sufficient demand, a formal "proposal" and publication follow.

Precedent is a guiding principle at UL, and it explains the direct origin of most provisions in UL 1482. The test method for the three fire tests, for example, came from UL's standard for fireplace stoves, UL 737. (Fireplace stoves are basically woodstoves without doors, and many of the provisions in UL 737 provided the basis for UL 1482.) Similarly, the provisions concerning metal thickness came from UL's standard for oil-burning appliances, according to a UL engineer in the Fire Department. That standard was also cited by UL as a precedent for the warning label eventually included in UL 1482.^[24]

Precedent says something about the source, but not the underlying rationale, of these provisions. The same question lurks behind each precedent: Where did the original provision come from? Take, for example, UL 737, which accounts for many provisions in UL 1482. Where did the provisions for UL 737 come from? Unfortunately, UL's own records do not document the complicated regulatory genealogy that covers ninety years, hundreds of published standards, and active testing in thousands of product categories. Even if it did, the original question would still remain: How does UL formulate requirements when there is no UL precedent? The answer seems to vary by type of requirement. Design requirements, performance standards, and labeling provisions seem to evolve differently. Like all UL standards, UL 1482 contains all three types of requirements. Deciding the appropriate combination of these approaches turns out to have safety implications as significant as the eventual content of each type of requirement.

Design Standards and Delegation

Design standards are disfavored by economists and standards purists, but not by UL engineers. Many of the requirements in UL 1482, and other UL standards, specify design parameters, such as the minimum metal thickness for cast-iron stoves. Economists favor performance standards—for example, a "burn through" test instead of a specific metal thickness—because such standards are less likely to constrain technological innovation.^[25] However, all of UL's design standards are qualified by an "equivalency statement" that, in theory, allows for technological innovation. This reads:

A product employing materials or having forms of construction differing from those detailed in the requirements of this Standard may be examined and tested in accordance with the intent of these requirements and, if found to be substantially equivalent, may be judged to comply with the Standard.

This clause is invoked "at least once a week," according to a UL attorney, who would not disclose any details about whether or when it has been applied to woodstoves (or any other specific product).

UL uses design standards for two reasons that are often overlooked in the discussion of performance and design standards. First, design standards are much cheaper from the point of view of testing. The thickness of cast iron can easily be measured. A performance test intended to simulate "burn through" would be much more complicated and expensive. Second, design standards—at least the way in which they are used by UL—allow certain matters to be "delegated" to the manufacturer. This is particularly helpful when field data or relevant research do not provide an adequate basis for a performance standard.

Many of UL's design standards are actually codifications of generally accepted business practice. The requirements for minimum metal thicknesses do not constrain woodstove manufacturers; they choose them in the first place. Design standards of this origin appear to contradict the notion that UL's standards are an "independent" test of safety. In UL's defense, reliance on industry practice, if done critically and selectively, helps keep down the cost of testing, while ensuring that products beneath the accepted minimum are not certified.

Performance Requirements and Educated Guesses

The design requirements in UL 1482 are not the most important from the point of view of safety; the performance requirements are. The fire tests, for example, specify that during testing the stove must not cause temperatures on exposed walls to rise more than 117° F above ambient temperature. Similarly, "a chimney connector furnished as part of the assembly shall not break, disassemble, or become damaged to the extent that it is unacceptable for further use after being subjected to a longitudinal force of 100 pounds."^[26] Unlike design specifications, which often leave the decisionmaking to industry, performance requirements are created entirely by UL's engineers. UL decides which aspects of performance to test and how to go about doing so.

On the first score, UL 1482 is quite comprehensive. It tests several aspects of normal stove operation as well as performance following various mishaps (for example, a heavy impact to the chimney connector). On the other hand, 1482 is far from scientific. According to UL, "sound engineering principles, research, records of tests and field experience … [and] information obtained from manufacturers, users, and

others having special experience" form the basis for its standard.^[27] Most test methods in UL 1482 reflect two factors not included in this list: educated guesses and concessions to the practicalities of product testing.

Guesswork abounds in UL's standards, although it is usually disguised by the exactitudes of scientific language. For woodstoves with glass doors, for example, the impact test for glazing calls for a steel sphere of 1.18 pounds mass and 50.8mm diameter to be dropped against the glass from a pendulum arc with a height of 16.25 inches.^[28] Research results certainly do not form the basis for this test. The explanation provided by a UL engineer is that a two-inch ball bearing—the steel sphere described more "scientifically" above—swung from around 45 degrees "seemed about right." The idea, according to this engineer, was to simulate an accidental jab against the glass with a fireplace poker. A larger ball would, of course, simulate a more serious jab. But absent any information on real-world experience with jabs and related mishaps, this educated guess lives up mainly to the second half of its name. A representative of a major glass manufacturer takes exception to UL's explanation. "We could never get [UL] to tell us why [the falling ball test] has these specific requirements." This engineer believes that the requirements came from some other UL standards. In fact, rather than simulating a specific hazard for woodstoves, the two-inch ball bearing test would more accurately be described as UL's generic impact test. It appears in similar form in a host of product standards, including the standard for portable video recording systems.^[29]

Even the most basic requirement of the three fire tests—that the temperature not rise more than 117° F on exposed surfaces and 90° F on unexposed ones^[30] —is largely guesswork. "There is some evidence to support these temperatures," observes a widely respected woodstove expert, but "they are still doubtful. Lots of different temperatures are plausible." A challenge to the rationale for these specific temperature tolerances was raised in the canvass process, but UL was unable to provide a specific response.^[31]

The practical necessities attendant to laboratory product testing help explain why guesswork often takes the place of science. In writing performance requirements, UL's engineers are sensitive to the cost of the endeavor. UL will not undertake costly scientific experiments where guesswork seems satisfactory. The engineers also try to keep the cost of the test itself within reasonable limits. For example, in UL 1482 a photoelectric method for measuring wisps of smoke was abandoned in

favor of less accurate visual observation tests because the former was considered too costly. Similarly, since walls and ceilings are made of a variety of materials, the most realistic test method would include separate tests for each material. Striking a balance that favors economy over realism, the tests are done with only one building material.

Product Certification and the Real World

The business of product testing also requires that test methods be reproducible. For example, myriad factors other than the woodstove itself affect the extent to which temperatures on surrounding walls will increase when the stove is used at a specified distance. The type of fuel, the type of chimney, the insulation in the room, not to mention the finer points of product testing (for example, type and placement of thermocouples), are all significant. A test method must control these influences sufficiently to ensure that the same test will yield similar results over time. Unfortunately, the more "controlled" the test, the less likely it is to bear a relationship to reality.

This is a recurring problem in product testing. The more inexpensive and reproducible the test, the more tenuous its relationship to what happens to the product in the real world. Performance standards require that numerous assumptions be made about how the product will be used in the real world. These assumptions can take many forms. Different sections of UL 1482 appear to be based on different notions of the relationship between test methods and reality. In some sections, UL takes its lead from the NFPA standards for chimneys and venting. The stove is tested, in other words, on the assumption that it will be installed in the method recommended by the NFPA. This assumption is on the utopian side of optimistic. An NBS engineer recounts that when the Bureau contracted to have a regulation NFPA 211 masonry chimney built for testing purposes, the mason exclaimed that the chimney was unusually sturdy. "We never build them like this," he told the engineer.

A particularly difficult question for UL is to what extent it should "test for" consumer misuse. Or, as a UL engineer puts it, "How forgiving should products be?" The answer implicit in UL 1482 is mixed. Several of the tests for structural integrity in 1482 simulate what might be considered average mishaps (for example, bumping the stove door or chimney connector). Other portions come much closer to simulating "worst case" conditions. The flash fire test, for example, is supposed to simulate the kind of overfiring that would occur if, say, a Christmas tree

or cardboard box were burned in the stove. The fire used in this test burns unusually hot and much longer than in a typical overfiring situation. Chimney height affects, among other things, the tendency of the stove to emit sparks into the room. The chimney used in all fire tests simulates a kind of worst condition. According to an NBS engineer, the chimney specified by UL has an unusual propensity to emit sparks. (This is not necessarily why UL chose the chimney, however, and neither is this particular chimney required if the manufacturer specifies otherwise.)

The portions of UL 1482 that are most controversial are those in which real-world experience diverges most from testing conditions. UL generally assumes that consumers are literate, obedient, and only occasionally clumsy. Most of the test methods in UL 1482 presume that the user will follow the manufacturer's instructions. Therefore, if the manufacturer's instructions say that a grate should not be placed inside the stove, UL tests without a grate. Since the use of grates is widespread and results in higher burning temperatures, an engineer with the International Conference of Building Officials thinks that "if a conventional grate [can] be placed in a heater, then it should be tested with one."^[32] UL disagrees.

This difference in testing philosophy explains why Consumer's Union (CU), publisher of Consumer Reports, graded down a stove that UL considered acceptable.^[33] CU considered the stove dangerous because of its tendency to emit sparks through the air inlets. UL, which tests stoves in accordance with NFPA's installation instructions—including, among other things, use of a floor protector—does not recognize the problem as a legitimate testing concern. That stove poses no danger—assuming it is used with a floor protector.

Evaluating UL 1482

Assessing the quality of UL 1482 as a safety standard is difficult because information about actual fires is limited and there may be significant variations in how the standard is applied.^[34] How test results are evaluated may also vary by laboratory. UL 1482 leaves many decisions to the discretion of the testing laboratory.^[35] Differences in test methods might allow a woodstove to pass the fire test at, say, twenty-eight inches at one lab and thirty-six inches at another. But since the UL standard itself is based on a combination of educated guesses, an agreed-upon margin of safety, and limited test data, it is likely that stoves certified for thirty-six inches would actually be safe at closer distances. In other

product areas, test data have frequently revealed that UL standards contain large, some say too large, margins of safety. Moreover, manufacturers often build in a margin of safety (actually, a margin of error) when they have their stoves tested. On the basis of stove design alone, engineers cannot calculate the distance at which the stove will pass the UL tests. Therefore, the manufacturer, who specifies the distance at which the stove is to be tested, must estimate the clearance, knowing that an overly optimistic estimate will result in expensive retesting. Accordingly, a stove certified for thirty-six inches might pass the test at a closer distance. In short, differences in product testing might lead to inconsistent applications of UL 1482, but the inconsistencies do not seem very important. As an NBS engineer put it, "The standard could be tightened up a great deal, but I don't think that it would make much difference in safety."

Several government studies of fire incident data have attempted, with mixed success, to quantify the nature and extent of fire damage related to woodstoves. All of these studies support the proposition that the consumer is responsible in one way or another—usually by improper installation or maintenance—for almost all such fires. A UL spokesman cites this as evidence that the standard is effective. While the conventional wisdom appears to hold true in a numerical sense—most studies blame the woodstove itself for only a few percent of all woodstove-related fires—it is not necessarily true in a policy sense. Even if the universe of injuries preventable by product improvements is small, it might still be easier to eliminate some of those injuries than to achieve a similar reduction in fires caused by consumer behavior. Moreover, some of the injuries "caused" by consumers could nevertheless be prevented by changes in the product.

Unfortunately, it is impossible to ascertain from recent studies of fire incident data whether certified stoves are actually any safer than uncertified ones. The data are both too general and unreliable. According to an NBS engineer, "Fire incident reports are sometimes detailed enough to indicate whether the fire originated in the appliance or the chimney, but they almost never indicate the appliance type or how far it was from the wall." None of the existing data bases even differentiates certified from uncertified stoves.

The Nagging Creosote Problem

UL 1482 enjoys wide support. Unlike UL's standards for fireplace inserts and metal chimneys, it has not been controversial within UL.^[36]

Insurance companies and various independent consultants endorse the standard without qualification. Even the CPSC concluded, at least unofficially, that the standard is adequate in all respects save some minor labeling provisions. A few manufacturers have voiced opposition to UL 1482, but their complaint, contrary to what might be expected, is that the standard is too lax. Anticompetitive motives apparently explain the opposition in one case.^[37] More typical of those who do not actively endorse UL 1482 is the view of a woodstove consultant who thinks that "listing under 1482 has little or nothing to do with safety." This is not an indictment of UL's standard so much as an expression that standards per se have little effect on woodstove safety.

While that may be true in general, one of the only substantial complaints about UL 1482 is that it should cover a problem it largely ignores: creosote. The standard has no provisions for evaluating the tendency of a woodstove to create creosote. In fact, the existing test procedures intentionally "control for" the two major causes of creosote formation: low burning temperatures and high-moisture wood. All of the fire tests are conducted when the stove is burning its hottest, and the "test wood" is much drier than seasoned firewood, often with a moisture content approaching zero. It has been argued that these performance tests may actually encourage greater creosote formation in stoves, at least when they are used under normal conditions. In other words, design changes that might help bring a stove into compliance with UL's temperature limits can, according to several woodstove consultants, also increase the stove's tendency to produce creosote in the real world.^[38] Whether the standard has actually prompted such design changes is not known.

UL claims that it cannot test for creosote. "We would test for [it] if there was a way to do it," according to a UL engineer familiar with the myriad factors affecting creosote formation. To be sure, just switching to wood with normal moisture content would pose big problems. It is much harder to standardize wood at a moisture content of, say, 20 percent than at percentages approaching zero. Creosote formation is also a slow process, raising the specter of long and involved test procedures. Tests for compliance with UL 1482 can be accomplished in a few days. The NBS spent several months of constant firing for its measurements of creosote formation.

Although these difficulties are not trivial, neither are they insurmountable or of a nature very different from the problems attendant to most product testing. There is a way to test for creosote formation. One

independent laboratory already conducts such tests (not for certification purposes, but as a consultative service). Similarly, there is an established test procedure for measuring woodstove efficiency—a factor directly related to creosote formation. The test uses high-moisture wood and is being performed without difficulty by the same lab.

So why doesn't UL do it? The short answer is that there is insufficient demand. Inside UL there is a feeling that creosote formation is a user problem, not a product problem. UL argues that the tendency to create creosote is irrelevant if the user avoids wet wood, monitors burning temperatures, and has the chimney cleaned whenever indicated through regular inspections. This philosophical position is reinforced by practical considerations. Including creosote tests in the standard might invite product liability suits. There have been very few woodstove product liability suits, and UL has never been named in one. However, given that creosote is the second leading cause of woodstove-related fires, that situation would likely change if UL 1482 treated creosote as a product problem. Setting an acceptable limit on creosote would also be difficult to justify, but no more difficult than the educated guesses behind many provisions in UL standards. Unlike other provisions, however, it would arouse strong opposition from some manufacturers. A UL engineer, implicitly acknowledging that these considerations outweigh any technical concerns, allowed that UL would test for creosote "if industry came to us and asked for a test and would pay for it." But that hasn't happened and isn't likely to.

The certification business for woodstoves, at least in 1979, when UL 1482 was formally adopted, was neither comprehensive nor consistent. The market for third-party certification was unusual in several respects. On the demand side, UL is often thought to have leverage over industry because the demand for its services is inelastic. In countless product areas, such as microwave ovens and television sets, everything on the market is UL-listed. Firms often need UL approval to meet requirements incorporated into law or contract. This puts UL in a position to demand various safety measures. However, in the mid 1970s there were only scattered local requirements that woodstoves be listed "by a nationally recognized testing laboratory," and the demand for certification was correspondingly weak. No more than 10 to 25 percent of the woodstoves were certified by independent labs.

On the supply side, UL is often in a monopoly position. This allows UL to invest in standards-related research without the concern that other labs will capitalize on the effort. With woodstoves, however, there

has always been considerable competition from small testing labs. A CPSC survey conducted in 1981 concluded that UL had only 28 percent of the market. The competitors use UL 1482 as the standard for certification, but, as mentioned earlier, studies by the National Bureau of Standards confirm that actual test methods vary and some labs are far less demanding than others in judging woodstoves.^[39]

The government played no role regulating woodstoves; that is, until a man in Midland, Michigan, wrote a letter to his congressman.

The CPSC Labeling Rule

Adam Paul Banner, a retired chemist turned woodstove salesman, thought the labeling on most woodstoves sold in 1976 and 1977 was inadequate. Of particular concern to Mr. Banner were stoves that did not specify "minimum clearance to combustibles and type of chimney required." Mr. Banner put his concerns on paper. Familiar with the world of private standards through an earlier association with the American Society of Chemical Engineers, he sent copies of his letter to the NFPA, the National Bureau of Standards, the CPSC, his congressman, and the governor of Michigan. The CPSC took the letter seriously, classifying it as a "petition" and assigning the staff to analyze it. (NFPA and NBS took no formal action.)

Under statutory provisions that have subsequently been amended, the CPSC was supposed to rule on the petition within ninety days.^[40] That meant gathering all information currently available, analyzing it, and presenting a recommendation to the commission. An informal ground rule provided that the staff would not attempt to generate new information in handling petitions. Experience proved the ninety-day period unrealistic. According to one program manager, "Four to six months is more realistic." In the case of the Banner petition, the process took almost two years.

Staff members attribute the delay to the "low priority" status of the petition. Unlike unvented gas-fired space heaters, for example, woodstove safety had not been singled out by the CPSC for special attention. Ignorance compounded indifference as the staff discovered how little it (or the commission) knew (or could easily find out) about woodstoves. The commission was briefed three times in the two years after the petition was received. Each time the commissioners requested more information from the staff. By the third occasion the staff voiced annoyance at the commission's seeming inability to make a decision.

Part of the problem was turnover on the commission. Three of the five commissioners were new when the staff conducted its March 1979 briefing, twenty-one months after receiving the petition. One commissioner asked what "overfire" means. (It means what it sounds like, firing the stove beyond its capacity, as, for example, in burning a dry Christmas tree.) Another dwelled on questions about the toxicity of artificial logs—something irrelevant to woodstoves, but apparently the subject of a Washington Post article. The scarcity of helpful data available to the staff further hindered the commission. The staff was unable to find out, for example, what percentage of woodstoves were installed by consumers versus professionals. (They eventually turned to the Washington Post for this as well, citing an estimate from a then-recent article.) The staff was also unable to determine the percentage of fires related to installation. "I don't have any quantitative feel," a staff member eventually told the commission, "but installation is mentioned a lot in the in-depth investigations."

The situation changed when the commission learned in May 1979 that the National Bureau of Standards had been studying woodstoves for almost two years under contract with the Department of Energy. Before receiving the NBS report, the commission expressed confidence that it would provide an authoritative basis for CPSC action. The report did not actually answer the previously unanswered questions, but it confirmed that installation was a major culprit in woodstove fires.^[41] The commission apparently agreed with a staff memo arguing that "we cannot state the specific reason for the fires" but "common sense indicates that there is a problem."^[42] With draft copies of the NBS report in hand, the commission granted the petition with little discussion on June 7, 1979.

Several reasons explain why the commission got involved in an issue it knew little about and that many people considered trivial. First, woodstoves were a trendy topic. The Department of Energy was studying them, and the issue tied the CPSC into a topic of national importance: energy use. Second, there was a virtual vacuum of private interests to oppose a CPSC regulation. Three separate trade associations claimed jurisdiction over woodstoves, and were barely more effective than no association at all.^[43] The commission sensed a natural (and easy) area for regulation. Finally, the idea of a labeling rule, as opposed to a full-fledged product standard, was attractive to the commission and the staff. It seemed simple and likely to keep the agency out of the kind of complex technical arguments that bogged down the CPSC's lawn mower and chain saw proceedings.^[44]

The Staff Drafts a Rule

Following a directive from the commission, the staff set out to develop a labeling rule. The desirability and effectiveness of mandating information disclosure was not considered during this process; it was taken for granted. Since the commission had already endorsed the notion of requiring information on minimum clearances, the staff concentrated on two other questions. First, what problems other than the "safe distance to combustibles" should be addressed by the labeling rule? Second, in relation to each problem, what specific information should the rule require?

The staff took an expansive approach in determining the scope of the labeling rule. Almost every hazard scenario that might be associated with woodstoves was considered an appropriate subject for the warning label. Many proposed warnings addressing the obvious—cautioning, for example, that stove surfaces are "hot during operation" and that hot ashes should not be placed in cardboard boxes. Others seemed more practical, such as disclosing the conditions that signal overfiring and stating how often a chimney should be cleaned and inspected. Deciding on the scope of the labeling requirements was easier for the staff than determining what specific information should be required on the label. In the case of information about clearances to combustibles, the staff had to confront the same intricacies of test methods faced by UL. The safest minimum clearance between a woodstove and a combustible wall—the information considered most important by the commission—is neither readily apparent nor easily measured. It is akin to the gas consumption of an automobile. Just as "your mileage may vary" with different driving conditions, the safest minimum clearance for woodstoves varies by such factors as fuel type, chimney size, and, most of all, by type of wall materials.

The staff recognized the problem and attempted to sidestep it. They knew that the commission did not want to get into the business of testing stoves. The CPSC had neither the budget nor the necessary technical skills. Part of the attraction of the woodstove labeling rule was its seemingly low cost and simplicity—the agency could accomplish something without facing difficult, technical issues. The commissioners also wanted to avoid the kind of criticism EPA had received over automobile mileage standards, so the development of a CPSC test method was out of the question.

Instead, the staff sought to put the burden on industry, reasoning

that "because the staff cannot know all the conditions for which manufacturers may recommend or promote their appliances, the determination of the appropriate information is the responsibility of the manufacturer."^[45] How the "appropriateness" of test methods would be reviewed and the extent to which different methods would be comparable were questions left unanswered, at least in the first draft of the regulation.

The staff took the same approach with other provisions. Instead of trying to figure out how often a chimney should be cleaned, they put the burden on the manufacturer to specify how often. Similarly, the staff proposed that labels "indicate the conditions which signal overfiring" rather than setting forth an accepted statement. Stated in terms most favorable to the staff, the rule was a paragon of flexibility; in a less favorable light, this flexibility disguised the inability of the staff to write its own standard.

Analysis or Post Hoc Rationalization?

Several of the CPSC directorates—functional divisions within the agency—analyzed the proposed rule. The Epidemiology Directorate revised its national injury estimates and conducted in-depth investigations of specific incidents. The Economics Directorate used this information in a preliminary and final "economic impact statement." The significance of this analysis, later cited in the Federal Register as justification for the rule, is questionable, because the concept of a labeling rule was endorsed by the commission before these analyses were done.

The Epidemiology Directorate analyzed the hazards associated with woodstoves with two goals in mind: estimating the total national losses (death, injury, property damage) related to woodstoves, and determining the most common hazard scenarios. CPSC's national injury estimates have long been subject to criticism. They are based on a reporting system from seventy-one hospital emergency rooms. These estimates are practically blind to cause, and they pick up only certain kinds of injuries. The injury data for woodstoves suffer from both problems. On the one hand, the estimates indicate a dramatic increase in injuries associated with woodstoves (546 percent from 1974 to 1978) without revealing that almost all of these injuries were caused by touching or falling against the stove—something that could not possibly be eliminated by regulatory action.^[46] On the other hand, these estimates do not reflect incidents involving property damage, however significant, but not in-

juries requiring emergency treatment—the case with most residential fires.

In order to better understand the scenarios in which woodstoves resulted in injuries, CPSC field representatives conducted "in-depth investigations," following up on incidents reported by consumers or collected from hospitals or the CPSC's newspaper clipping service. Approximately 150 in-depth investigations were conducted on woodstove-related incidents in 1980–81. These reports verified the hazard scenarios that the staff had in mind when considering the scope of the rule. Unfortunately, most of the "in-depth investigations" do not provide information that would be particularly helpful in evaluating the rule. For example, few indicate whether the stove was certified or whether the owner read or followed the instruction manual. Some of these omissions were due to the limited training these investigators had in fire incidents. In other cases, the investigators faced uncooperative or hostile subjects.

Minimal Costs, Doubtful Benefits

Armed with these assorted injury data, the Economics Directorate attempted to analyze the costs and benefits of the proposed rule. As is so often the case, the costs of the rule were easier to estimate than the benefits. The staff concluded that the rule would cost approximately $2.80 per stove, or $3.6 million annually. The estimate is probably low. It was based on several optimistic, but questionable, assumptions.^[47] Although a higher estimate of costs would have been more realistic, the CPSC's estimates were not unreasonable. The same cannot be said about the estimated benefits.

CPSC's economic analysis did not seriously consider whether the benefits of the rule were likely to exceed the costs. Instead, the costs were presented alongside the national injury estimates with the assertion that "any reduction in these injuries and deaths would result in significant benefits to consumers."^[48] When the rule was published, the Federal Register notice included the statement that "the Commission is unable to estimate the degree to which the rule may reduce fire incidents, [but] a reduction of seven percent (or possibly less) … would offset the total yearly cost of the rule."^[49] The likelihood of this happening was never discussed, and although it would be very costly to conduct tests to ascertain the probable effects, there are two reasons to think that a reduction of 7 percent is extremely unlikely.

First, consumer information and education campaigns are notoriously unsuccessful. A former CPSC commissioner argues in a recent analysis of three such efforts that they frequently fail.^[50] A CPSC staff member in the Human Factors Directorate agrees that measured results "around 2 percent are about all you can expect." Second, the estimated impact must be adjusted to reflect the percentage of stoves already labeled. Since over 70 percent of the stoves on the market already bore labels substantially equivalent to those proposed by the CPSC, the marginal benefit of the CPSC rule would, at best, constitute 30 percent of the estimated potential of labeling. (As explained later, this figure is optimistic, since the CPSC has not even matched the performance of the private sector in achieving compliance.) Moreover, by the CPSC's own estimates, injuries did not decrease at all during the years in which product certification, and hence product labeling, increased significantly.^[51]

Comments, Some Changes, and a Deferred Decision

The woodstove labeling rule was "kind of a small potatoes rule," according to a CPSC economist. "It was very low visibility." It drew few comments from the public and generated little controversy. Seventy-six comments were received after the rule was published in the Federal Register: thirty-seven favored the rule, twenty-four were against it, and fifteen stated no position. Only seven people took the opportunity to testify before the commissioners at a hearing in Washington, D.C. Dissatisfaction, to the extent that it was expressed, mainly concerned issues peripheral to the content of the labeling rule. A consumer group complained that the rule should also cover fireplace inserts.^[52]

Manufacturers complained that the proposed rule would take effect before they had an adequate opportunity to adapt and sell off existing inventories.

The relationship between the proposed government standard and the existing UL standard ended up being the most difficult issue for the CPSC to put to rest. Testing labs, aware of the intricacies and importance of test methods, pointed out that mandating the disclosure of clearance information would have little meaning without specifying what constitutes "appropriate" test methods. Lacking the resources, and possibly the knowledge, to develop test methods or criteria for evaluating them, the CPSC was forced to defer to the testing labs. The proposed rule was changed to indicate that UL 1482 was an "appropriate" method.

More troublesome to the adoption of the CPSC's labeling rule, particularly given the agency's practical endorsement of the test methods in UL 1482, were increases in private product testing. The percentage of stoves tested to the UL standard rose significantly in the two years after the CPSC received the Banner petition. By some estimates, 80 percent of the new stoves on the market were certified to UL 1482 when the CPSC finally published the proposed rule in November 1980. (In 1978, it was approximately 10 percent.) The UL standard also changed while the CPSC analyzed the Banner petition. In direct response to concerns expressed by the CPSC, UL changed its labeling requirements to conform, in all but a few minor respects, to the proposed CPSC standard.^[53]

Manufacturers argued that a federal rule was unnecessary and possibly counterproductive. With compliance levels near 80 percent, the rationale for a federal rule was limited to whatever benefits the government could generate by affecting the remaining 20 percent. Unfortunately, the CPSC staff did not acknowledge the concept of marginal benefit. None of its analyses pointed out that most of the benefits of labeling, if indeed there were any, were already captured by the UL standard. Moreover, as discussed below, there was little reason, then or now, to believe that a federal rule could achieve any additional benefit.

A more difficult question for the CPSC was whether a federal rule would have an adverse effect on the laboratory certification business. The testing labs argued that the existence of a mandatory federal rule would decrease the use of independent testing and, by implication, compliance with their standards. In other words, manufacturers, content in the knowledge that they satisfy all federal requirements, might stop meeting the "voluntary" requirements, which cover much more than just labeling. Since woodstoves generated approximately $350,000 in income for UL alone in 1982, the economic implications of a drop in business were at least as significant as the possible safety implications. These two concerns were so troubling to the commission that adoption of the proposed rule was deferred in order to address these issues.

In a May 1981 briefing paper, the staff informed the commission that an estimated 70 to 85 percent of new stoves were certified to UL 1482. It was not known how many of the remainder were nevertheless built in conformance with the standard. In light of this new private sector initiative, the staff was hard pressed to recommend adopting the proposed rule. Neither did they want to abandon it, particularly in the absence of total compliance by the private sector.

The commission opted to delay its decision so that a market survey

could confirm the extent of third-party certification.^[54] The survey confirmed that approximately 80 percent of new stoves were certified to UL 1482; however, some of the laboratories competing with UL apparently were not as demanding in applying the labeling requirements.^[55] This compounded the staff's already considerable skepticism about voluntary compliance. The generally high levels of voluntary compliance confirmed in the market survey did not lessen the reluctance of the staff to abandon the proposed rule.

The commission agreed and adopted the rule, using as the main justification the differences between the informational requirements in UL 1482 and the proposed rule. The differences were slight. The CPSC rule demanded more detailed instructions on installing chimney connectors and cleaning the chimney.^[56] Whether the type of detail envisioned by the standard would be forthcoming depended on the process of implementation; the requirements themselves were not specific. On the other hand, a staff memo (not mentioned in the Federal Register notice) identified several respects in which the UL standard was stricter" than the CPSC rule.^[57]

The arguments advanced by the CPSC in the Federal Register rang hollow. Information disclosure is generally considered capable of affecting only a few percent of all product injuries, if wildly successful. And the commission was fiddling with the difference between using the word "furniture" as opposed to "combustibles" on the warning label. The real reason for the CPSC rule was that levels of compliance with the UL standard—estimated at 70 to 85 percent—were considered too low. Whether or not such levels of compliance should be considered acceptable, the important, but unstated, premise behind the CPSC rule was that a government standard would result in higher levels of compliance.

Presumed Benefits, Implementation Problems

The CPSC's experience has not been as rosy as the staff's expectations. Stated in terms most favorable to the agency, approximately 70 percent of the woodstoves on the market comply with the CPSC standard—less than the percentage in compliance with the UL standard. Using assumptions least favorable to the agency, compliance may be well below 50 percent.^[58]

One reason for this relatively poor performance is that the agency has had trouble reaching many of the small, family-sized businesses that manufacture woodstoves. A more important reason is the loose word-

ing of the CPSC rule. Flexible requirements, while desirable in theory, can be difficult to enforce in reality. Most of the woodstove labeling requirements are purposely open-ended. The commission wanted to appear flexible, and the staff had no desire to decide (and later defend) such matters as the minimum size for lettering on labels or the best way to describe how to pass a flue pipe through a combustible wall. The CPSC had already lost a similar battle in court over proposed warnings for swimming pool slides.^[59]

Some firms, in the view of the Enforcement Directorate, are subverting the rule by taking advantage of its vagueness. The rule mandates, for example, that a warning label must be "legible," "conspicuous," and "readily visible."^[60] To the CPSC this means understandable; to many manufacturers it means capable of being understood. The difference, something not addressed in the rule, entails how well the information is conveyed. A few firms stamp the warning label into an aluminum plate, for example. The practice does not clearly violate the rule, but, as a staff member in the Enforcement Directorate put it, "I defy anyone to read it." Print size poses a similar problem. The lettering on many labels is small and difficult to read. It is not, however, illegible.

A more widespread problem, at least in relation to the installation manual mandated by the rule, is that firms do not understand what the rule requires. Vagueness can be confusing. Manufacturers are supposed to provide "step-by-step installation instructions." The rule says nothing about how detailed these instructions should be. Practically no firms are providing all of the details deemed appropriate by the Enforcement Directorate. The agency's response provides an ironic conclusion to the tale of its involvement in woodstove safety regulation: after justifying its entry into woodstove safety regulation on the grounds that independent laboratories were not doing an adequate job, the CPSC is now turning to the same labs for assistance in correcting the problems with the agency's rule. The Enforcement Directorate is trying to persuade these labs to ensure a level of detail in installation manuals beyond that specified by the CPSC rule.^[61]

Summary Evaluation

"A smidgen of good and no harm" is how a former CPSC commissioner describes the effects of the woodstove labeling rule. That assessment might be apt for UL 1482, but for the CPSC rule it is overstated. Given the modest scope of the rule and the low level of compliance with it,

there is little doubt that the rule cost more than it is worth. It is not a very burdensome rule, however, so the loss itself might be only a smidgen. But adopting the standard did not demonstrate good judgment on the part of the commission. The performance of the staff was equally disappointing, as they were unable to provide the commission with answers to many of the questions that arose during the proceedings. They shied away from technical issues at all possible junctures. Although the CPSC regulation apparently did little to improve safety directly, it may have done so indirectly. The CPSC's proposal certainly prompted UL to modify its own labeling requirements, and those changes might have produced minor benefits. The CPSC's action may also have contributed to the dramatic increase in the percentage of stoves certified to UL 1482 in the years after the Banner petition was received.

The major fault with the CPSC rule is not what it covers, but in what it does not. By writing only a labeling rule, the agency missed the opportunity to regulate a problem not covered by the UL standard: creosote production. CPSC staff members defend the decision on the grounds that the agency was responding solely to the Banner petition. This excuse is disingenuous in light of the broad reading given to other petitions. Mr. Banner's general concern was woodstove safety, and the agency certainly would not have exceeded its statutory authority by adopting a more comprehensive rule. Why, then, did the CPSC ignore the creosote problem? Two explanations are most likely. First, the agency misperceived the problem. It thought that most fires were caused by poor installation, when the data suggest that creosote in chimneys is in fact a much larger problem. Second, the staff sought to avoid technical issues. They realized that a product standard would be complicated and difficult to support in court. The agency had lost legal challenges to several earlier rules and, of late, had also lost substantial funding from Congress. A labeling rule seemed easier to defend in court and would be much less resource intensive for the agency. In short, the CPSC staff did not address creosote because they were looking for simple issues susceptible to simple solutions. A labeling rule fit the bill much better than a full-blown product standard, even though the latter might have had much more effect on safety.

UL's performance is harder to evaluate. UL 1482 has prompted some manufacturers to make design changes, while it codifies the existing design of others. Some of the changes dictated by the standard have improved the safety of the product; many have made little difference.

This suggests an element of unreasonableness in the standard. Although some firms express displeasure with UL 1482 for this reason, the sentiment is not widely held. The woodstove manufacturers' trade association endorses the standard almost without qualification. The main benefit of UL 1482 is the fire tests, which provide helpful, standardized information on clearances to combustibles. There are legitimate questions about the assumptions built into the test methods, and there are only limited data to support the specific performance criteria, but no one familiar with UL 1482 considers it to be grossly inappropriate. The standard appears to do a good job of determining safe clearances—something the CPSC knew it could not accomplish.

To the extent that labeling matters, UL can be faulted for allowing limited warnings and installation instructions for so many years. These provisions were upgraded when the standard was formally published, probably in response to the CPSC. The UL standard also avoids the creosote problem. It does not test for creosote production and, contrary to UL's assurances to canvass participants, the organization has no intention of addressing the issue. This appears to reflect a philosophical position that UL has against addressing problems caused by consumer misuse or neglect. The causes and implications of this philosophy are explored in chapter 7.

Six
Safety Standards for Unvented Gas-Fired Space Heaters

The final paired case study is the most interactive. The CPSC essentially made an "optional" provision of a private standard mandatory (and then later revoked the standard). The CPSC standard has been considered favorable even by some of the agency's strongest critics. The obvious question is why the private sector was unwilling to upgrade its standard sufficiently to fend off government regulation. The mystery deepens with the realization that the private standard, written by the American Gas Association Laboratory, was otherwise strict and had been upgraded significantly over the past few years. The answer reveals much about regulatory philosophy. This case strikes at the core of some fundamental differences between public and private approaches to safety regulation. The inexpensive oxygen depletion sensor (ODS)—required by the CPSC, but made "optional" in the private standard—became a lightning rod for concerns about paternalistic regulation, liability law, and antitrust law. The case highlights significant differences between the regulatory environment in the public and private sectors.

The CPSC may have been given more credit than it deserves. The rule was slow and painful in coming. Only reluctantly did the CPSC drop its original idea of banning space heaters entirely. Moreover, the agency's information is so poor that it is impossible to be sure whether the standard has generated the hypothesized benefits. Finally, in a bizarre twist that emphasizes the importance of federalism issues, the CPSC

rescinded the rule after it was deluged with petitions for exemption from preemption. In the final analysis, the case demonstrates several shortcomings of CPSC standards-setting in specific and federal regulation in general.

By contrast, the performance of the private sector was surprisingly good. The standard for unvented gas space heaters, developed decades before the CPSC was created, provided an ample degree of safety for basic operations. Moreover, the standard was strengthened over time. But AGA Labs refused to require the ODS before the CPSC made it mandatory. This position demonstrates both the significant role of regulatory philosophy and the subtle influences of product liability law. Those factors are elaborated below following an introduction to the unvented heater.

The Unvented Gas Space Heater

Space heaters are inexpensive, convenient, and popular with energy-conscious consumers. They are also illegal in many jurisdictions—at least certain types are. Unvented space heaters fueled by gas or kerosene are the most controversial because, unlike electric heaters and various vented heaters (including woodstoves), they discharge their combustants indoors. The long-term effects of these pollutants are unknown. The acute effects—carbon monoxide poisoning—are better understood and, under the worst circumstances, can be fatal. Owing to the differences in fuel, gas-fired heaters pose a more serious threat of carbon monoxide poisoning than those fueled by kerosene. An unvented gas-fired space heater, along with the controversial oxygen depletion sensor (discussed later in this chapter), is pictured in figure 3. In the interest of brevity, this product will be referred to as the "unvented heater" for the remainder of this chapter.

The unvented heater is a primary heating source for poor families in southern states, where many homes do not have central heating. With installation and operating costs considerably lower than those of other space heaters, and fuel efficiency of nearly 100 percent, these appliances are popular with cost- and energy-conscious consumers. They are disfavored in northern states, however, where homes tend to have central heating and better insulation. Three major producers currently account for almost all of the 180,000 units sold annually in the United States. (There were thirteen manufacturers and much higher sales before the CPSC proposed a ban on space heaters in 1975.)

Figure 3.  Unvented Gas Space Heater and Oxygen Depletion Sensor

Two acute safety problems are posed by the gas space heater. The first, fire and burn hazards, is common to all heating equipment but is somewhat more serious with gas or liquid fuel sources. The primary fire and burn hazards are contact burns and clothing ignition. The burn hazard is easily mistaken for the hot radiator problem. In fact, temperatures on certain parts of the unvented heater can exceed 500° F, so the

resulting injury can be extremely serious. Lesser problems include the clearance between the heater and combustible materials, and gas explosions—less serious because they are so rare.^[1]

The Invisible Hazard: Carbon Monoxide Poisoning

The second acute hazard—carbon monoxide poisoning—is the most significant. This hazard is not common to all heating equipment. It is unique to unvented (or improperly vented) space heaters fueled by gas (or, to a lesser extent, kerosene). Carbon monoxide poisoning can occur if the heater's burner is maladjusted or if the heater is used without adequate ventilation. The hazard is particularly ominous because carbon monoxide is odorless and colorless—you can be poisoned without knowing it, especially if you are asleep. Most deaths appear to be caused by inadequate ventilation, and apparently most of those are in bedrooms. It seems counterproductive to many consumers to open a window while heating a room, but ventilation is required to ensure safe use.^[2]

The carbon monoxide hazard is well recognized but poorly documented. Industry trade associations do not collect injury data in any systematic way. Representatives on the relevant AGA committees are aware of the common hazard scenarios, sometimes as a result of newspaper clippings sent to the committee.^[3] But the problem is not widely discussed, and anecdotes are not freely exchanged during the standards setting process.

The CPSC is the only organization that has attempted to quantify these hazards, relying on files of death certificates, reports from hospital emergency rooms, and its own in-depth investigations. The CPSC placed the unvented heater fifteenth on a list of over 350 hazardous consumer products. Except for possible publicity value, however, this designation is of dubious assistance to the agency's regulatory effort.^[4] More pertinent than the total number of injuries is the character and seriousness of various components of the injury problem. How many cases of carbon monoxide poisoning occur? What are the causes and possible solutions?

Epidemiologists at CPSC have reached two relevant conclusions. The first is widely accepted: contact burns are the most common hazard associated with the unvented heater. Four-fifths of reported injuries are contact burns; the majority of these are children and elderly people who fall against the heater.^[5] The second conclusion is more controversial:

unvented heaters are responsible for some seventy carbon monoxide deaths per year. This number has appeared repeatedly in briefing papers and newspaper articles over the past ten years, although its origin and accuracy are unclear.^[6] Still, the number emerged with what Max Singer has dubbed "the vitality of mythical numbers.^[7]

AGA Labs: The UL of Gas Appliances

AGA Labs is the premier certifier of gas appliances in the country. Best known for its Blue Star Certification Seal—which appears on furnaces, ranges, and virtually all other types of gas appliances—AGA Labs tests these products for compliance with so-called voluntary safety standards. These standards are not actually voluntary, since many jurisdictions require compliance with them through building or related installation codes.^[8] AGA certification is also required by many utility companies as a condition of providing service. As a practical matter, then, it would be almost impossible to market a gas appliance in this country, at least one requiring professional installation, without complying with the "voluntary" AGA standard.

Other labs test gas appliances for compliance with the same standards as AGA Labs, but AGA plays a unique role in this private regulatory scheme. It is both the oldest and the best-recognized organization in the field. Some people refer to it as the "UL of gas appliances." The description accurately conveys AGA's stature, but not its mission or method for writing standards. AGA Labs is devoted solely to testing gas appliances, while UL's interests have branched out considerably, and somewhat controversially, over time.^[9] As the creation of a trade association, AGA Labs is also looked upon with more suspicion than UL by those concerned about anticompetitive motives. The "trade association mentality," as a government attorney describes it, fosters in organizations such as AGA Labs a "bias toward protecting members of the trade" that does not, in his view, characterize "the purists, like UL."

AGA Labs stresses that it does not actually write the standards it uses to test products. "Our job at the Labs is policeman, not judge," explains the Labs' director for product certification. The safety standards that AGA Labs uses to test gas appliances, including the unvented heater, are the product of the Z21 committee of the American National Standards Institute. The unvented gas-fired space heater is a distinct species in the American Gas Association's taxonomy of gas appliances. It has its own safety standard, numbered Z21.11.2—the "Z21" denotes gas appli-

ances, the "11" is for room heaters, and the "2" signifies unvented equipment. The Z21 committee, with approximately thirty-five members, is "balanced" in accordance with ANSI requirements. Nine members represent AGA, nine represent the Gas Appliance Manufacturers Association (GAMA), and one or two members come from each of eighteen additional organizations, including the American Insurance Association, Consumers Union, the National Safety Council, the Southern Building Code Conference, UL, the General Services Administration, and the CPSC. The Z21 committee meets annually and has jurisdiction over the forty-seven standards. Some of these standards cover gas appliances; others, component parts.

AGA Labs is an unusual policeman; it is directly involved in writing the "law" it enforces. AGA Labs is the official "secretariat" of the ANSI Z21 committee. The arrangement is confusing, particularly to those who know that AGA wrote these standards before the American Standards Association (predecessor to ANSI) was formed in 1930. As recently as 1969, the year ANSI was formed, AGA Labs was recognized as the author of these standards. Many existing Z21 standards, including the unvented heater standard, were first developed under the old system. Since the power of precedent carries significant weight in most standards-writing schemes, the trade association of old may really be the co-author of any long-standing AGA standards.

Parties disagree about the extent to which standards are influenced by AGA Labs under the current ANSI system. Officially, AGA prepares meeting agendas and minutes, circulates draft standards for comments, and occasionally provides "technical input" to the subcommittees. These subcommittees, consisting primarily of utility representatives and appliance manufacturers, are responsible for proposing the actual standards and any revisions to the parent committee, Z21. There is a technical subcommittee for every standard.

The Standards Department at AGA Labs provides the secretary for each subcommittee. Other lab personnel attend subcommittee meetings to provide "technical input" and, in some instances, to draft language or proposals. Lab personnel, however, are not voting members of these subcommittees. The only decisions made explicitly by lab personnel concern the date when a standard will take effect. Decisions about content are made in the first instance by the subcommittees. Those decisions must be ratified by the Z21 committee. "If somebody wants to grumble," as an engineer at AGA Labs puts it, "they can go to the Z21 committee." Evidence of conformance to ANSI's procedural cri-

teria is subsequently reviewed by the ANSI Board of Standards Review in the same cursory fashion as most UL and NFPA standards.^[10]

The ANSI Z21 committee is not a meddlesome parent. The subcommittees normally send what one AGA staff member termed "very polished documents" to the parent committee for "rubber stamping." The Z21 committee almost never overrules subcommittee actions. Most of the Z21 committee decisions are made by letter ballot. When the committee meets in person, very few people come to grumble. Occasionally the Z21 committee raises concerns or requests "further consideration" by a subcommittee. Like many parental pleas, however, these requests are often discounted, if not entirely disregarded.

The Z21 standards differ from other ANSI standards in one important respect: AGA, as secretariat, has its own elaborate procedural requirements for approving standards at the subcommittee level. These requirements are similar to ANSI's public review process.^[11] Actually, they steal the show. AGA circulates "review and comment texts" of all proposed changes in standards to a diverse group of manufacturers, gas companies, and state and local officials, as well as to interested federal agency representatives. This process generates far more interest and participation than when it is publicized later by ANSI. In March 1982, for example, AGA Labs sent draft revisions of Z21.11.2 to 7 gas appliance manufacturers, 4 manufacturers of decorative appliances, 185 gas companies, and 220 state and local officials and other miscellaneous organizations. Numerous comments were sent back to AGA. When ANSI repeated the process, notification was through Standards Action, and only one party responded.

AGA Labs maintains that these gas appliance standards are ANSI standards, not AGA standards. The organization frequently asserts its "independence" from the standards-writing process.^[12] This position is weakened by admissions that AGA provides financial support for the standards-writing activity. (The official line is that standards-writing is financed solely by fees for service.) Critics go much further, charging that Z21 standards are simply the product of two large trade associations—AGA and GAMA, organizations that were one and the same before an antitrust decision in 1935.

Neither "independence" nor "collusion" seems an appropriate characterization in the case of the unvented space heater. The AGA and GAMA did not agree, let alone collude, on the content of the standard. The AGA at large was split over the issue, and there were disagreements within AGA Labs.

ANSI Z21.11.2

The first standards for gas space heaters were adopted more than sixty years ago by gas utilities concerned about the safety of gas appliances. The Pacific Coast Gas Association published a standard for gas space heaters (vented and unvented) in 1924. The newly created AGA Labs followed suit two years later. In 1930 the AGA "Approval Requirements Committee" became "Sectional Committee Z21" of the American Standards Association, predecessor to ANSI. A technical subcommittee oversaw fourteen editions of the standard for gas-fired room heaters in the next thirty years. Records from this time are not readily available, but in all likelihood the standard was changed in only minor respects during these years. There were no major controversies, according to a recently retired committee member. "Things rolled along fairly smoothly for many years."

The Z21 subcommittees for gas heating appliances were reorganized in 1959, and separate committees were given jurisdiction over vented and unvented room heaters. The first standard specifically for unvented gas heaters—American Standards Association Z21.11.2—was published in 1962. It was sixteen pages long and contained both construction and performance requirements. Many of these provisions are still in effect today. Some of the construction requirements were design standards. For example, "Sheet metal air shutters shall be of a thickness not less than 0.0304 inch."^[13] Others were performance-based. "Orifice spuds and orifice spud holders shall be made of metal melting at not less than 1000 F."^[14] Many of the construction requirements were of a more subjective nature, however. "Burners shall be easily removable for repair and cleaning."^[15] "Gas valves shall be located or constructed so that they will not be subject to accidental change of setting."^[16] The performance requirements included a host of tests covering burner and pilot operation, wall and surface temperatures, valves, thermostats, combustion, and safety guards.

The standard assured that certain basic elements of safety were achieved. In other words, a heater certified to Z21.11.2 could safely be connected to a gas line and, under the right conditions, be used without incident. The standard could be faulted, however, for being less than forthcoming with warnings about the dangers that still existed. These early versions of the standard did not require any information about the need to ventilate the room in order to prevent carbon monoxide poisoning. The prevailing attitude of many manufacturers about disclosing

Map 1.  Statewide Distribution of Unvented Gas Residential Space Heaters and Bans

information seems callous by today's standards. Warning labels and instructions were generally disfavored. The space heater standard required, for example, that "instructions provided with the heater shall include information to adequately cover cleaning of the heater." But a manufacturer took the position in 1964 that it did not have to include such instructions because the requirement was optional—in other words, the provision applied only if the manufacturer chose to provide instructions. The subcommittee agreed.^[17]

The fire and burn hazards associated with the gas space heater also received short shrift in the early versions of Z21.11.2. A performance test was supposed to ensure that clothing "could not readily come in contact with flames or with parts of the heater which would easily cause ignition or charring." The test method involved exposing a piece of twenty-pound bond paper to the front of the heater for ten seconds.^[18] For reasons explained later, this test bears little relationship to reality and, surprisingly, is not very effective in detecting the tendency of a heater to ignite clothing. The subcommittee did not take this hazard very seriously. The problem was periodically raised and routinely referred to a working group of some sort, where the issue languished until the cycle began anew. The triggering event tended to come from outside the committee. In 1962, for example, the NFPA forwarded a newspaper clipping concerning accidents resulting from contact with open-front heaters. AGA appointed a working group that concluded further study was necessary!^[19] The same conclusion had been reached by AGA working groups in 1953 and 1955. In 1970 the secretary of commerce sent the committee details of six clothing ignition cases. Another working group got its marching orders. No changes were made in the standard.

This early version of Z21.11.2 also did little to address the problem of carbon monoxide poisoning. The combustion tests included a threshold limit for carbon monoxide, but under conditions in which there is seldom a problem—when the heater is properly cleaned and adjusted. Thus, while the standard assured that the appliance was capable of safe use, it made no provision for the possibility of unsafe conditions such as poor ventilation. This is not to say that the subcommittee was unaware of the problem. The gas space heater had long been a controversial product, and there are oblique references to the carbon monoxide problem in the minutes of various meetings. The controversy predates Ralph Nader, the expansion of the tort law, and the Presidential Commission on Product Safety. The danger of carbon monoxide poisoning, particularly in northern states, prompted many utilities to oppose the product

in the 1960s. Gas space heaters were banned in Columbus, Ohio, in 1958. Several states later banned them, many others have partial bans (see map 1). Federal "minimum property standards," applicable to 15 percent of new housing in this country (but recently revoked by HUD), also prohibit the unvented heater.^[20] A Philadelphia utility company argued recently it is "immoral" to allow use of the unvented heater.^[21]

Whither the Unvented Heater?

The Z21.11.2 subcommittee could not ignore these developments. More than once, the subcommittee addressed the same existential question: Should Z21.11.2 exist? The answer was not necessarily a foregone conclusion. Within the AGA, sentiment at large was (and still is) divided over the unvented heater. AGA represents the distributors of gas. These companies have different interests because some are "gas-only," while others sell gas and electricity. Some do a great deal of marketing; others do not. Their interest in protecting a product such as the unvented heater varies accordingly.

Privately, some manufacturers were also less than enthusiastic about the appliance, leading one CPSC commissioner to conclude that the industry "didn't really care" about whether the heaters were banned. Some of the engineers at the AGA Labs even supported the notion of a ban. Not surprisingly, however, the committee's position has always been in favor of the gas space heater.

The stated reasons have been less than compelling. The standard should exist, the subcommittee has stated on several occasions, because without it space heaters will continue to be sold, but not necessarily with the safety levels maintained by Z21.11.2. In other words, economically speaking, the standard creates a marginal benefit. This benefit is real, however, only if the product would still be marketed without AGA approval. This is a critical and dubious assumption. AGA approval is practically a prerequisite to marketing a gas appliance. There are strong reasons to believe that the lack of an AGA standard would eliminate the product from the market, and that if it did not, manufacturers would continue to satisfy the requirements in the current standard.

The forces that motivate AGA certification would not disappear with the enactment of a limited government standard. There is no reason to think that utilities who refuse to install or service appliances without AGA certification would change their behavior in light of a federal requirement that does not even address some of their most serious

concerns (explosions, fire hazards). Moreover, pressures created by product liability law would remain unchanged. A manufacturer who chose to forgo AGA certification would run serious liability risks in relation to those hazards covered only by the AGA standard.

Relying on the distinction between product standards and installation standards, AGA Labs took the position that the existential question should be answered by someone else. Accordingly, the preface to Z21.11.2 warns: "Safe operation of a gas-fired unvented room heater depends to a great extent upon its proper installation, and it should be installed in accordance with the National Fuel Gas Code, ANSI Z223.1, manufacturers' installation instructions, and local municipal building codes."^[22] In short, AGA Labs took the same position UL often does with respect to "banning" a product: it claimed to defer to use and installation codes. If NFPA will not allow an appliance, then AGA will not list it. This is the case with a cousin of the gas space heater, the so-called cabinet heater. Cabinet heaters are unvented heaters without a gas line connection. They use a gas cylinder instead. These heaters are not permitted under NFPA 58, so AGA Labs does not list them.

With the space heater, however, the most important installation code—the National Fuel Gas Code—is written by AGA Labs, further complicating the symbiotic relationship between installation codes and product standards.^[23] The Fuel Gas Code recognizes the seriousness of the hazards connected with the space heater but takes the ineffectual position of "prohibiting" the appliance in sleeping quarters, sanitariums, and certain other institutions. Although an unqualified prohibition would probably keep the product off the market, the use restriction leaves enforcement largely in the hands of the consumer, who has probably never heard of the National Fuel Gas Code. The code notwithstanding, most carbon monoxide deaths apparently occur in sleeping quarters.

In reality, the existential question caused the subcommittee no existential distress. They were committed to the unvented heater. This assured certification business for AGA Labs and an imprimatur for manufacturers. As an AGA engineer remarked in response to suggestions that space heaters be redesigned to lower surface temperatures: "There is no point in pricing them out of the market if your intent is to keep them in the market." Utility representatives did not object to this unofficial intention, because even if the heaters were on the market, they could keep them out of specific gas distribution networks by refusing to hook them up and service them. Those who objected to the gas space

heater politely abstained from the discussion. A representative from Peoples Natural Gas Company, for example, responded to a proposed change in Z21.11.2 by commenting: "Peoples does not approve the installation or use of unvented heaters by our customers. I am sure that you will understand that the decision to abstain [on these revisions] was dictated by these factors and does not reflect on the fine work of the Subcommittee."^[24]

The Overhaul of Z21.11.2

Z21.11.2 may never have been in danger of being scrapped, but a major revision was inevitable. Space heaters were getting a bad name, and worse yet, the federal government, for the first time ever, was considering regulating a gas appliance. The results were significant. Between 1960 and 1980 the committee added requirements for an automatic ignition system, a pilot regulator, lower surface temperatures, and improved shielding. The average retail price of the appliances more than tripled as a result, from around $50 to approximately $180. How beneficial were these requirements? Opinions vary, and the data are inconclusive.

Surface temperatures have come down slightly over the years, reducing the burn hazard to some extent. Organizations such as Consumers Union argue for further reductions. This would not prevent accidents, of course, but it would provide a longer response time before a burn injury becomes serious.^[25] How that would translate into the real world of product injuries is impossible to say. There are no good estimates of the number of injuries involving surface burns, let alone how changes in surface temperature might affect them. Significantly lower surface temperatures probably cannot be obtained, however, without substantial changes in heater design. The cost and feasibility of this task cannot be estimated without research and development work. If, as manufacturers argue, surface temperatures cannot be reduced without dramatic changes in design, then it is likely that the cost would not justify the fractional change in response time to accidents. Similar uncertainty confounds any analysis of the other performance tests in Z21.11.2. The clothing ignition test, long criticized for using twenty-pound paper, was revised to incorporate the use of terry cloth. This makes the test more realistic,^[26] but the real-world significance of the revised test method is impossible to ascertain given existing data. And without access to con-

fidential certification records, it is impossible to determine whether the new test resulted in actual product changes.

Perhaps the most dramatic result of these changes was elimination of a model of small bathroom heaters known to be connected with many injuries.^[27] This "drop-out ban," as a trade association representative dubbed it, apparently did not improve the image of the unvented heater enough to forestall government intervention. But the industry claims that the overhaul of Z21.11.2 improved the safety of the product sufficiently to eliminate almost all carbon monoxide deaths. The CPSC disputes this claim.

The CPSC and the ODS

The CPSC first got involved with space heaters in 1974 when it received a petition from the Missouri Public Interest Research Group to ban all "space heaters." The request was apparently prompted by a tragic fire, ignited by an electric space heater, that claimed the lives of several children in Missouri. At the time, the commission did not understand the differences between the various types of space heaters (that is, electric, gas-fired, kerosene, and wood-fueled), let alone differences in styles and models. Choosing not to read the petition narrowly, and with a predisposition to grant it, the commission approved the petition "in substance" and directed the staff to figure out which space heaters should be banned.^[28]

It took the CPSC several years to become familiar with the world of heating appliances. The education process was terribly frustrating for AGA Labs and GAMA. "They just didn't understand the equipment," complains a GAMA staff member who tried, with mixed success, to point out the difference between vented and unvented gas space heaters.^[29] The agency's uncertainty stemmed largely from ignorance, but it was also indicative of a larger problem. Gas appliances, like most consumer products, are diverse and difficult to analyze with specificity. The CPSC's injury surveillance reporting system did not have a product code for unvented gas space heaters. Few of the newspaper reports collected through the Injury Surveillance Desk specified whether or not the appliance was vented. Even the CPSC's own "in-depth investigations," intended to compensate for data problems elsewhere, were sketchy and sometimes of questionable accuracy. Thus, even when they tried to find out specifically about unvented heaters, the staff had trouble obtaining helpful information.^[30]

Eventually the CPSC singled out the unvented gas space heater, deciding that it should be banned and that vented gas heaters and other types of space heaters need not be regulated at all.^[31] The overriding concern about the unvented gas space heater was carbon monoxide poisoning. "The thing that motivated the commission most over the years," according to a former commissioner, "was the actual experience of consumers as manifest by death and injury statistics." In this case, people were dying in their sleep—as many as seventy a year, according to the widely cited CPSC estimate. Industry representatives took issue with the figure. In addition to echoing familiar criticisms of the CPSC's information system, they argued that changes in the product had rendered it much safer in recent years. A representative of the Gas Appliance Manufacturers Association boldly claimed that there had been no carbon monoxide deaths involving heaters built after 1978. There is no way to prove or disprove this allegation with existing data.

Criticisms of the government's injury estimates masked a larger complaint about its motives. The unvented heater was a "politically targeted product," charged one industry representative.^[32] "It isn't nearly as bad as the kerosene heater," charged another, who manufactures only gas equipment. In fact, political forces favoring regulation of the gas space heater were in motion long before the commission received the petition from Missouri. The unvented gas space heater was investigated by the FDA in the early 1970s, and by the Public Health Service before that. It was also mentioned by the Presidential Commission on Product Safety, precursor to the CPSC.

If the unvented heater was politically targeted, the CPSC missed the mark—sort of. The agency was set to ban the product when word arrived of a technological answer to carbon monoxide poisoning. The ideal solution, a carbon monoxide sensor, had always been considered far too expensive to be practical, but a second-best solution—an oxygen depletion sensor (ODS)—was touted by a European manufacturer, Sourdillion, at the CPSC hearings in 1978. Similar to a pilot flame, the ODS consists of a Bunsen burner that utilizes a synthetic ruby orifice and other precision parts to achieve uniform control of aeration and flame characteristics (see figure 3). The ODS is a delicate device; it relies on an unstable flame, also known as a metastable flame. It is stable with normal levels of oxygen but less stable as oxygen levels decrease, to a point where the flame literally lifts off and shuts down the heater.^[33]

Since the CPSC was prohibited by statute from banning products for which there was a viable private standard, Sourdillion's presence at the

hearings demanded attention. Even those commissioners who personally favored a ban felt obligated to examine the ODS option. The CPSC's in-house engineers examined the device, and the agency contracted with NBS for additional research.

The most startling thing about the CPSC's analysis of the ODS is how late it occurred in the discussion of unvented heater safety. The device was introduced in 1961 in Europe. In 1972 the Z21 committee appointed a subcommittee to "follow the development of such devices and, if warranted, to develop revisions" for existing standards.^[34] (They had not yet done so when the CPSC first got involved in the issue.) Remarkably, it was not until 1978, four years after receiving the petition to ban space heaters, that the CPSC acknowledged the existence of these devices. Even then, in a January 1978 briefing package, the staff informed the commission that although the device "might be technically feasible [it is] economically impractical."^[35]

The European manufacturer, which boasted worldwide sales of thirty million units since 1961, begged to differ. But the success of the ODS in Europe must be put in context. European fuel gases differ from those widely available in this country, so technology that works in Europe will not necessarily work in the United States. To Sourdillion, however, adapting the ODS to this country did not pose anything more than a good engineering challenge. Still, there were two important respects in which the reliability of the ODS was called into question. First, the device does not measure carbon monoxide; it measures oxygen. Whether the relationship between oxygen and carbon monoxide was sufficiently predictable for the ODS to provide reliable protection against specified levels of carbon monoxide was a legitimate concern. Second, the metastable flame can be inhibited by dirt and lint. The dirtier the pilot, the more stable the flame and the less reliable the shutoff. How reliably the shutoff device would work in the field was a serious concern. (It was not an equivalent concern with European equipment, which requires the pilot light to be lit with each use, minimizing the amount of dirt and lint.)

Optional Equipment: AGA's Uneasy Solution

These concerns had been raised years earlier when a manufacturer requested AGA certification for a heater with an ODS device. The existing standard did not address these devices, so the engineers on the testing

floor had no guidance. AGA was on the spot. Its engineers did not know enough about the ODS to assess its reliability, let alone its desirability. The device had no track record in this country. On that basis, they decided to make the ODS an "optional requirement." For the short run, this sounds reasonable (even if oxymoronic). The solution was an uneasy one, though, because "optional requirements" appear to contradict two cornerstones of AGA/ANSI standards-writing: uniform requirements and open participation.

The certification process, both at AGA and UL, is built on a binary approach to safety regulations. Either a product receives the AGA label or it does not. AGA Labs does not want to get into the business of rating degrees of safety, putting asterisks on labels, or otherwise differentiating between products. First, the manufacturers—the lab's clients—would object.^[36] Second, it might diminish the use of AGA standards. The influence of "voluntary" standards, and hence the demand for certification, is due at least in part to their ease of use. The certification process simplifies the job of building officials. No label, no approval. If asterisks are placed on labels, putting conditions or qualifications on the meaning of approval, or if safety features are labeled "optional," then the standard becomes harder to use. It cannot be applied without additional judgments about the desirability of the optional equipment or the significance of the asterisk. Once this occurs, deference to the remaining portions of the standard may also diminish. Why rely on AGA's judgment about the need for, say, automatic ignition when you are already making your own judgments about the ODS device?

Optional requirements also contradict the procedural premises of standards-writing under ANSI's committee method. AGA constantly maintains that it does not write standards, the Z21 subcommittees do. The ODS, as an optional requirement, is a clear exception. The Labs drafted the "AGA requirement" for the ODS. Such provisions are akin to UL's "desk standards." They are written by staff and not reviewed by the public. In time "AGA requirements" are usually submitted as formal revisions to standards, thereby subjecting them to the normal process of public review. This did not happen very quickly, however, with the ODS. The manufacturer who originally requested approval of the ODS device never actually marketed it. The reasons are disputed: some claim that there was no interest among manufacturers, others say that the device was not adequately proven. In either case, the ODS issue remained a theoretical one for the AGA Labs.

The CPSC Investigates the ODS

Sourdillion's presence at a Washington, D.C., hearing put tangible pressure on the CPSC. The firm apparently considered a mandatory standard the best way to create an American market for its product. The commission, committed to banning space heaters, was forced to reverse its position and examine the ODS (about which it knew practically nothing). The obvious technological question was whether the device would operate reliably under U.S. conditions. There were secondary technological questions as well. The cost of such technology was of prime concern to industry, but apparently less so to government.

To answer the basic technological question, the CPSC contracted with the National Bureau of Standards. Simple questions do not always elicit simple answers, however, particularly when a regulatory agency is asking and NBS is answering. In this instance, NBS gave a more complicated answer than the CPSC wanted. Yes, the ODS works, it concluded, but the shutoff level in the "optional AGA requirement"—an 18 percent oxygen level-might not be appropriate.^[37] A higher shutoff level, such as that used in the French standard, would be more protective, since the corresponding levels of carbon monoxide would be lower. The problem is that higher shutoff levels are also more likely to cause "nuisance shutoffs" (where the heater shuts off accidentally because the cutoff point is so close to normal levels of oxygen).

The origin of the 18 percent level remains something of a mystery. Minutes of the Z21 subcommittee meetings at which the ODS was first discussed do not reflect any discussion of the adequacy of the 18 percent level. An NBS engineer who analyzed these documents for the CPSC concluded that the figure most likely came from the American Conference of Government Industrial Hygienists, a nongovernment standards-writing organization (contrary to its name).^[38] Since the relevant health concern is carbon monoxide, however, and not oxygen, it would be strange indeed if the shutoff level for the ODS was selected without considering the estimated corresponding levels of carbon monoxide. More likely, members of the Z21 committee had a rough idea of the relationship between oxygen and carbon monoxide and made their decision without explicitly discussing the details. According to one frequent guest at the subcommittee meetings, "We knew what the curves would look like."

Whatever its origins, 18 percent appears to provide substantial margins of safety. The NBS tests revealed that ODS devices set for 18 percent

actually shut down at higher oxygen levels (18.2 to 20.4 percent). The corresponding carbon monoxide levels in those conditions ranged from 7 to 98 ppm, with a mean concentration of 37 ppm. A 1971 OSHA standard permits normal concentrations in the workplace of up to 50 ppm. At four times that level, the expected medical response, according to the CPSC's Directorate for Epidemiology, is a "possible mild frontal headache in two to three hours." Another twofold increase, and nausea is likely; twice again higher, unconsciousness.^[39]

The CPSC staff divided on the question of whether the shutoff level in the AGA standard was adequate. The Directorate for Engineering, sensitive to the "nuisance shutoff" problem concluded that 18 percent was "adequate" and that higher levels were probably "overly conservative." The Directorate for Epidemiology disagreed, urging reconsideration of the 18 percent level because it "appears to be based largely on technical feasibility rather than on possible health effects."^[40]

Whether the ODS device was commercially feasible was not a question that much interested the Program Management staff. A negative answer would have been humiliating. The proposed ban had already been tabled because of the ODS. Frustration grew as NBS engineers questioned the appropriateness of the 18 percent cutoff and industry raised the specter of nuisance shutoffs at higher levels. "We had to get a consistent story," explains one staff member. "First we said ban them, then we said no. We had to get our story straight." The story, then, was that the ODS was technically feasible and would work well with an 18 percent shutoff.

The issue never reached the commission. "Staff assured us that 18 percent was all right," recalls one commissioner. Actually, the staff was more pragmatic than it was satisfied. "Nineteen percent would have meant a five-year fight," recalls a staff member who favored the stricter level but was even more concerned about keeping things as simple and uncontroversial as possible. Eighteen percent had history on its side.

Playing Poker with the Private Sector

Once the CPSC was convinced that the ODS technology was sufficiently feasible to favor a product standard over the proposed ban, the obvious question was whether the private standard could do the job. AGA was not about to ban the space heater, but it might regulate it to the extent deemed necessary by the CPSC. The AGA standard already had an

"optional requirement" for the ODS, along with provisions covering many other hazards. If the Z21 committee would simply require the ODS device, the "voluntary" standard could eliminate the need for a government one.

The matter was not so simple from the perspective of the AGA. Industry was largely opposed to the ODS device for reasons that were never stated explicitly. Some of the resistance no doubt was based on the cost of the device. Others objected to requiring a device that was only available from one supplier—and a foreign one at that. Whatever the reason, an AGA vice president wrote the commission in the summer of 1978 that a "strong statement from the Commission as to the value of the ODS and experience with it would be needed" in order to prompt revisions in the voluntary standard.^[41] There was not enough support for the idea in the private sector, even given the specter of government regulation. Changing a voluntary standard generally takes more than a year, but, warned the AGA, it takes "much longer when there is any controversy." There was in this case. To the surprise of several commissioners, the AGA seemed to be suggesting government regulation.

Assured that the device was desirable and technically feasible, the CPSC drafted a proposed mandatory standard. The action did not push the private sector too far; but the AGA made one apparently significant change. The Z21 committee added the "optional AGA requirement" to the standard at large. In other words, the ODS was now a "mandatory" part of the so-called voluntary standard. But there was a catch: the "mandatory" requirement would take effect only when AGA determined that a "suitable and certifiable device [was] available."^[42]

In the eyes of the CPSC staff, it would take an additional push to make this happen. "As long as unvented heaters without ODS devices can be sold for an indefinite period," the CPSC staff argued in a memo, "the industry is not compelled to develop and test the device in a short time." Therefore, the CPSC went ahead with its proposed rule duplicating almost exactly the provisions in the AGA standard, with the addition of a specific effective date—December 30, 1980.^[43] "We called their bluff," boasts one CPSC staff member.

The staff could press for an early effective date without worrying that inability to meet it would mean a de facto ban. That was what staff wanted in the first place! "Fortunately or unfortunately," as a CPSC program manager put it, "we were working with a product that was not absolutely needed by the public." In other words, the staff did not care

whether the requirement would result in a short-term ban. Apparently blind to the commercial interests of space heater manufacturers, the Economics Directorate concluded that an interruption in production would cause "no serious economic impacts."^[44]

Aware that concerns about meeting the deadline would not get a sympathetic hearing within the CPSC, private interests argued against the CPSC proposal on the grounds that a government standard would displace the nongovernment one, worsening safety in areas regulated only by the private standard. These areas were significant. The AGA standard addressed surface temperatures, sharp edges, clothing ignition, and a host of other hazards unrelated to carbon monoxide. The proposed CPSC standard covered only the ODS and related labeling. But the argument against CPSC involvement depended on whether manufacturers would actually change their behavior in the event of a federal standard and forgo AGA certification on items not required by the government. That was highly unlikely, since the forces for complying with the AGA standard were unaffected by the existence of a CPSC standard. The argument was really about turf, not safety. AGA's concern was that the CPSC was embarking on a venture that might, if extended, cut into its business. And industry's concern was more about the implications for the future than about the implications of a single requirement.

There was one potential hazard that neither the CPSC nor AGA Labs addressed: chronic health effects. The issue came up during the CPSC proceedings, but the staff had no idea which elements of combustion, if any, posed potential long-term health problems. "It would have created a furor to try to address chronic hazards," recalls a former CPSC program manager. Subsequent research at the Lawrence Berkeley Laboratory, under contract to the CPSC, suggests that one of the six tested pollutants—nitrogen dioxide—might pose a serious long-term health hazard. The CPSC has called for further studies. For now, AGA Labs is deferring to the government. "We don't have the expertise or experience to address long-term health effects," notes an AGA Labs engineer. Neither does the CPSC for that matter, ensuring that the unvented heater story is not over.

Petitions and Confusion Follow the CPSC Rule

The CPSC was determined to go ahead with the standard even though AGA Labs had all but required the ODS device. Some staff members

earnestly argued that a federal standard was desirable because the CPSC had a better warning label than AGA. But the differences were, by any measure, trivial.^[45] Others felt that without a federal rule, the ODS would never be commercially adopted. One of the commissioners was determined to get back at an industry that had dared to start producing its product immediately after the CPSC lifted its proposed ban. This distrust was shared by many CPSC staff members. The commission eventually approved the rule but extended the effective date in a gesture of accommodation.

The gesture was lost on industry. Less than two weeks after the CPSC published a Federal Register notice adopting the space heater rule, GAMA petitioned the agency to revoke it. This was only the first of a plethora of petitions and other postadoption posturing that confused the commission and, years later, resulted in revocation of the ODS rule. The GAMA petition was a foregone conclusion, dependent more on the CPSC's presence than on its program. This was the federal government's first foray into regulating gas appliances. Although the rule was one that GAMA now agrees that "industry could live with," some segments of the industry could not live with the idea of government regulation. An association spokesman says that it was inconceivable that the commission could have written a rule that GAMA would not have petitioned to revoke.

Petitioning so soon after the rule was adopted was probably a tactical error on GAMA's part. To grant the petition, the CPSC would have to admit that it had made a mistake—not enough time had elapsed to argue that circumstances had changed. Positions had also hardened as a result of the just-completed rulemaking process. It was doubtful whether the CPSC would take a hard look at the petition. In fact, the staff responded to the petition with a brief memo prepared in a matter of weeks, instead of with the conventional briefing package prepared over a course of months.

The petition brought forth unexpected disagreements within the commission and within industry. The CPSC Office of Program Management could not reach a consensus. Representatives from the Engineering Directorate favored granting the petition, those in Epidemiology and Compliance opposed it, and the economists said that it would make little difference either way. Industry was similarly divided. GAMA did not want the federal government to regulate gas appliances, but some major retailers saw an advantage in federal regulation. The "carrot of preemption," as CPSC staff members often refer to it, holds

out the promise of preempting state and local regulations. For the space heater this means preempting bans in many jurisdictions, opening up a possible national market. (That is, unless the states and localities petition for an exemption from preemption—something that GAMA predicted, and that soon came to pass.)

Unwilling to revoke what it had just enacted, the commission rejected the GAMA petition. Petitions for exemption from preemption soon followed. Officials at AGA Labs who questioned the desirability of the unvented space heater persuaded the president of AGA, without the approval or knowledge of those departments concerned with marketing, to provide information to jurisdictions interested in reinstituting their bans by obtaining an exemption from preemption. Petitions poured in from cities such as Victorville, California, which argued that a ban "provides a significantly higher degree of protection from the risk of carbon monoxide than does the Commission's standard."^[46]

The legal issues concerning exemption were cloudy.^[47] And the practical implications of these petitions were formidable—the CPSC apparently had to rule on each petition individually. By mid 1983 there were more than two dozen petitions. On October 5, 1983, the commission proposed what looked like an easy way out: revoke the rule. Sufficient time had passed since enactment of the rule to argue that circumstances had changed. The private sector had achieved important gains; by now, the ODS was standard equipment on all space heaters. To GAMA, it was a wish come true. To some of GAMA's members, however, it meant losing the carrot. Both Atlanta Stove Works and Birmingham Stove and Range Company opposed revocation as an unfair impediment to market expansion plans undertaken after the rule was first adopted.^[48]

The issue was emotionally charged within the commission, but the reasons seemed symbolic, not substantive. Since the AGA/ANSI standard required the ODS, the only practical effect of revocation was that many states and localities would reinstitute bans that might have been granted anyway through the exemption process. There were no arguable detrimental effects on safety. Still, two of the five commissioners voted against revocation. One wrote a twenty-nine-page dissent that castigated the commission for taking "such an extreme and unparalleled action."^[49] The same commissioner objected so strongly to wording in the proposed revocation notice that intimated CPSC endorsement of the voluntary standard that the notice was eventually reworded. The

CPSC had (finally) deferred to the private sector, but it was unwilling to say so directly.

Summary Evaluation

Even among CPSC critics, of whom there are many, the ODS rule is widely considered to have been a desirable government action.^[50] The adverse effects of the rule predicted by some did not come to pass. The ODS has become standard equipment on unvented gas space heaters, and a trade association representative confirms that there has not been a significant problem with nuisance shutoffs or consumer tampering. Moreover, there was no decrease in AGA certification after the CPSC rule went into effect. In short, there is good reason to believe that as a result of the ODS device required by the CPSC, fewer people are dying in their sleep. That is sufficient reason for the CPSC staff to conclude that the standard was beneficial. Unfortunately, there is not enough information to determine whether the averted harm is small, as industry maintained all along, or whether it is substantial (that is, as many as seventy fatalities per year), as maintained by the CPSC.

In general, the CPSC's performance was mixed. Many of the issues, such as the appropriate shutoff level, were highly technical, and the agency was not well equipped to analyze them. Delay gave way to deferral, and the technical content of the CPSC standard was eventually borrowed almost entirely from the private sector. The CPSC staff changed only the warning label and effective date of the requirement. On the former, the CPSC was insensitive to industry's concerns—product liability and adverse consumer reaction to shrill warnings—and petty in its insistence that the wording itself was an important safety issue.

The CPSC was most controversial and probably most effective in setting the date for the regulation to take effect. The staff thinks that setting an early date was a bold move that forced industry to make the ODS commercially feasible. Some industry representatives still maintain that universal use of the ODS would have been forthcoming without government intervention. That may be true, but it certainly would not have happened as quickly. Moreover, several AGA staff members speculate privately that the ODS would never have been required by AGA Labs but for the CPSC.

Outcomes aside, the process leading up to the CPSC standard was not commendable. Early in that process, the CPSC relied on political instincts

instead of information to make its decisions; and its instincts were not very good. Granting "in substance" a petition to ban all space heaters was ill informed and needlessly antagonistic. Although the CPSC did not end up banning any kind of space heater, it expended considerable goodwill in the process of reaching that decision. Beyond insensitivity, there was a streak of anti-business sentiment among the commissioners throughout the space heater proceedings. That sentiment was most evident when the commission decided to revoke its standard in favor of the private one. Publication of the revocation notice was delayed while two commissioners argued for changes in wording that would erase any affirmative statement that the CPSC endorsed the AGA standard (even though it obviously did so by implication).

On the private side, the performance of AGA Labs and ANSI was also mixed. Z21.11.2 long provided an ample degree of safety for basic operations. It also got stricter over time. Surface temperatures were reduced, and the small bathroom heater was eliminated. But before the CPSC intervened, the standard did little about the carbon monoxide problem. For years the subcommittee maintained that the ODS device should not be required because it was "not a proven technology" and had "no track record" in this country. That explanation was reasonable in 1971 when the issue first arose. Eight years later, the same explanation looked more like a lame excuse. The explanations suggested in private interviews are more convincing than those found in committee minutes. Several segments of the private sector never really believed that carbon monoxide poisoning was a product problem. Since it only occurred through misuse of some sort, they considered it a "consumer problem." Some utility company representatives and AGA Lab employees disagreed, but they respectfully kept their view outside of the standards-setting process. But the standard did address some misuse problems," including surface burns and clothing ignition, so there must have been other reasons for eschewing the ODS device. The most likely reason is product liability law. According to a representative from a gas control manufacturer, "nobody wanted to be put in the position of selling a safety device that was a safety device when you sold it but didn't necessarily stay that way." Translation: if the ODS device failed, the manufacturer would get sued. Without the device, carbon monoxide poisonings would probably be considered the consumer's fault—a rare event in the world of product safety law.^[51] The perverse result, in the eyes of manufacturers, was that adding the device would increase exposure to product liability even though it would decrease product-

related hazards. This view may not hold in a court of law, but it certainly prevailed in the Z21 subcommittee.^[52] Only the desire to keep the federal government from regulating gas appliance safety was strong enough to overcome the opposition of manufacturers who were unsympathetic to "the problem" and afraid of "the solution."