Operationalizing and testing normal accident theory in petrochemical plants and refineries

Production and Operations Management, Fall 2001 by Wolf, Frederick G

OPERATIONALIZING AND TESTING NORMAL ACCIDENT THEORY IN PETROCHEMICAL PLANTS AND REFINERIES*

Since "Normal Accidents: Living With High-Risk Technologies" was first published in 1984, there have been very few tests of its validity or its application to different classes of technology or organizations. This paper describes a robust, parametric test of Normal Accident Theory applied to a class of high-risk systems: petrochemical plants and refineries. The theory has important implications for scholars interested in the relationship of organizations to the natural environment and managers responsible for safe, socially responsible, and environmentally appropriate operation of high-consequence systems.

(NORMAL ACCIDENTS; TECHNOLOGY; RISK; REFINERIES)

Introduction

This paper describes a test of the validity of Normal Accident Theory as it is applied to a sample of petroleum refineries located in the Western United States. The sample consisted of 36 refineries and covered the period 1993-1997. Normal Accident Theory provides a framework for making sense of such events and understanding the risks associated with technical systems. When the risks associated with a technical system are realized, decisions can be made in a context that enhances rather than degrades system safety. Despite the efforts of industry and government to improve safety, major industrial accidents are as likely today as they were 10 years ago. Since 1982, there has been no reduction in the fatality rate or major accident rate in U.S. or European industry (Sellers 1993; National Safety Council 1998). Normal Accident Theory may provide the insight required to understand this disturbing trend.

Until very recently no serious attention was paid to what happened inside the gates of chemical plants and refineries. As Davidson reported in 1970:

Most managers assert they are deeply concerned with the health and safety of their employees. Most of them are undoubtedly sincere. But for a manager to take really effective action to protect his employees puts him squarely into conflict with his basic role in the profit system. Production-and the achievement of more and more production-is the duty of operating managers who direct most work forces. (Davidson 1970, p. 120)

He pointed out ". . . management has an incentive to avoid fires which destroy extensive sections of a plant and interrupt production, or disasters which kill or maim large numbers of people and thereby soil the company's public image" (Davidson 1970, p. 121). By 1990, little had changed. Following the fire and explosion at the Phillips 66 chemicals complex located in Pasadena, Texas, which occurred on October 23, 1989, and the subsequent explosion at the ARCO Chemicals facility in Channelview, Texas, 9 months later, Ainsworth and Hanson wrote:

... during the late 1980s when chemical sales and profits were booming, many chemical plants were running above nameplate, stressing the plant to crank out product at a rate well beyond its original design. And to further maximize output . . . companies may have let their maintenance schedules slip a bit, further stressing the plant. And there's a lag effect to that: the failures and explosions are happening several years or months after the maintenance was delayed. (Ainsworth and Hanson 1990, p. 12)

In May 1994, Johnson and Burke reported ". . . some companies have learned to live with a slim margin for error. There is an acceptance of a certain level of risk when you downsize. Management knows this. . . . It's going to take a major accident before someone says Hold it!" (Johnson and Burke 1994, p. 32). For many years, as long as manufacturing output was not seriously disrupted, process-related fires and small explosions were accepted as routine, and chemical spills were all but ignored.

As long as the production continued, management had little cause for concern. Government only became interested in their chemical accidents, fires, and explosions after communities (and the public) began to perceive that they were at risk from chemical and petroleum manufacturing operations after the catastrophic events at Flixborough in 1974 and Bhopal 10 years later in 1984. These industrial accidents and several in the United States, including the October 1989 Phillips Petroleum Company fire and explosion that killed 23 and injured 132, led to a series of incremental Congressional and regulatory actions. These actions included the Superfund Amendments and Reauthorization Act (SARA) Title III: Emergency Planning and Community Right to Know Act (EPCRA), which was passed in 1986, and the Occupational Safety and Health Administration's (OSHA) Final Standard, "Process Safety Management of Highly Hazardous Chemicals" published in the Federal Register in February 1992. Industry groups also took action. For example, the Chemical Manufacturers Association formulated its Community Awareness and Emergency Response (CAER) initiative at the same time.

Despite the growing awareness that chemical manufacturing plants, refineries, and industrial operations involving hazardous substances must be operated with greater sensitivity and concern for their safety and risk, almost no management research occurred in this important area of social and corporate responsibility. This is particularly surprising since a theoretical framework for investigating operations that had high-consequence potential existed. In 1984, Perrow published "Normal Accidents: Living with High Risk Technologies," which offered a system-level explanation of why some processes are riskier than others. According to the theory, risk is determined by two characteristics associated with the organization and its technology, interactive complexity and coupling. According to Perrow, interactive complexity arises from tight spacing of production equipment, proximate production steps, commonmode interconnections, limited possible isolation of failed components, personnel specialization, limited substitution of supplies and materials, unfamiliar and unintended feedback loops, many control parameters with possible interactions, indirect and inferential information sources, and limited understanding of production transformations. Coupling, according to the theory, is determined by the degree of time dependence in a process, flexibility in process sequencing, single or multiple paths to a goal, and the availability of slack resources (Perrow 1999).