Risk-Benefit Analysis and the Determination of Acceptable Risk

Rachel Dardis, University of Maryland
Julie Stremel, Arthur Young and Company
ABSTRACT - Consumer safety has been regarded as a major consumer issue for many years. In recent years, this has resulted in considerable debate concerning "how much risk" is acceptable. The objectives of this paper were to develop a methodology for assessing risk and to apply the methodology to a selected consumption activity. The results indicate the role of risk-benefit analysis in the identification of "high-risk" consumption activities. This type of screening is important in view of resource limitations which preclude the development of risk reduction strategies for all consumption activities.
[ to cite ]:
Rachel Dardis and Julie Stremel (1981) ,"Risk-Benefit Analysis and the Determination of Acceptable Risk", in NA - Advances in Consumer Research Volume 08, eds. Kent B. Monroe, Ann Abor, MI : Association for Consumer Research, Pages: 553-558.

Advances in Consumer Research Volume 8, 1981      Pages 553-558

RISK-BENEFIT ANALYSIS AND THE DETERMINATION OF ACCEPTABLE RISK

Rachel Dardis, University of Maryland

Julie Stremel, Arthur Young and Company

[University of Maryland Agricultural Experiment Station Article No. A-2828.]

ABSTRACT -

Consumer safety has been regarded as a major consumer issue for many years. In recent years, this has resulted in considerable debate concerning "how much risk" is acceptable. The objectives of this paper were to develop a methodology for assessing risk and to apply the methodology to a selected consumption activity. The results indicate the role of risk-benefit analysis in the identification of "high-risk" consumption activities. This type of screening is important in view of resource limitations which preclude the development of risk reduction strategies for all consumption activities.

INTRODUCTION

Consumer safety has been regarded as a major consumer issue for many years. While there is general agreement that consumer safety can and should be increased, there is less agreement concerning the level of safety that should be achieved. Proponents of zero risk argue that no risk is acceptable though the attainment of zero risk in a complex technological society such as the United States has been challenged with respect to both feasibility and desirability. The rejection of zero risk leads, in turn, to the question of "how much risk" is acceptable. Unfortunately, the frequently used argument that any risk is acceptable as long as the expected benefits are greater than the expected costs neglects the costs of developing risk reduction strategies for a wide spectrum of consumption activities. The costs of obtaining such information may be far greater than the potential benefits.

The objectives of this paper were to develop a methodology for assessing risk and to apply the methodology to a selected consumption activity in order to demonstrate the feasibility and utility of risk-benefit analysis. The results should provide policy makers with a method for assessing risk. Risk assessment may then be used a) to obtain information concerning the level of risk to which consumer are exposed and b) to rank activities with respect to risk.

NEED FOR RISK ASSESSMENT

Deaths and injuries involving consumer products pose serious social and economic problems in the U. S. It is estimated that approximately 20 million Americans are injured each year in the home as a result of accidents involving consumer products (National Commission on Product Safety 1970, National Safety Council 1978). The annual cost to the nation has been estimated at more than $5.5 million.

The Consumer Product Safety Act which was passed in October 1972 was designed to protect the consumer from unreasonable risk of injury. A Consumer Product Safety Commission (CPSC) was established with jurisdiction over more than 10,000 consumer products. The Commission was granted authority to "develop, promulgate and enforce safety standards for consumer products and to inform the public concerning significant hazards associated with consumer products".

The number of consumer products which are subject to regulatory activities presents problems as the Commission acknowledges.

"It is not difficult to know when the need exists to ban an obviously dangerous or hazardous product.

But it is difficult to determine which of more than 10,000 consumer products are of the greatest risk to consumers all over the country and which of these products are the likeliest candidates for safety standards without undue cost to the consumer." (U.S. Consumer Product Safety Commission 1973)

In view of the impossibility of dealing simultaneously with all consumer products, it is necessary to establish a basis for ranking products with respect to unreasonable risk. One method for ranking products was the hazard index developed by the Consumer Product Safety Commission in 1973 and revised in 1975. The index was based on frequency and severity of injury. Accident frequency was obtained from NEISS data while severity values were based on type of injury and necessity for hospitalization. Severity values in the 1975 and 1976 hazard indices are given in Table 1. (U.S. Consumer Product Safety Commission 1976).

TABLE 1

SEVERITY VALUES USED IN CPSC HAZARD INDEX

Categories 1 to 5 were increased by one severity value if the injury resulted in hospitalization. Similar values were used in the 1973 hazard index with the exception of category 8 where a value of 34,721 was assigned in place of 2,516.

A Frequency Severity Index (FSI) was obtained by multiplying the frequency of injuries in each severity category. by the corresponding severity values. The Age Adjusted Frequency Severity Index (AFSI) differs from the FSI in that injuries to children under 14 are first multiplied by 2.5 in order to provide special consideration for this age group. The AFSI is the hazard index employed by the CPSC. In issuing the hazard index, the Commission noted that "a product does not have to be high on this list to receive Commission action but those high on the index are more likely to receive early attention".

The hazard index was criticized with respect to the particular severity values since there was no indication that the values were related to the economic consequences of a particular type of injury. As a result, the hazard index was replaced in 1978 by numbers of injuries involving various products. While the use of a simple numerical index avoids the bias associated with an incorrect severity scale, it provides less information to the policy maker. The new numerical index also has many of the same limitations as the discontinued hazard index.

First, neither index considers the number of products in use nor the degree of product usage by the consumer. Thus, products with the same number and types of injuries could be viewed as equally hazardous though hours of exposure might differ to a considerable extent. Both Lowrance (1976) and Rowe (1977) have discussed the need for including hours of exposure in any risk assessment.

Second, consumption benefits are ignored. However, the need for including product benefits in any evaluation of risk has been pointed out by Starr (1968, 1972), Clark (1969), and Dickerson (1968). Starr used consumer expenditures for an activity as a measure of benefits and concluded that the public was willing to accept a relatively high level of "voluntary" risk. Clark examined the conditions under which a product slight be considered reasonably safe. His hazard index included injury severity, number of products used per year, costs of avoidable injury, costs of safety improvements and importance of product use. The first four terms in the hazard index represented the ratio of the costs of injuries avoided by a safety improvement to the costs of making the safety improvement. The fifth term - importance of product use -reflected the degree of product substitution since "a hazard is more grimly accepted as "reasonable" if there is little alternative to the use". Dickerson pointed out that "the mere fact that a product is dangerous is not enough; it must be unreasonably so". He also noted that it is not possible to achieve complete safety and that the need for and utility of the product should be considered in product regulation.

It is important, therefore, to consider both the costs and benefits from a particular consumption activity in evaluating risk. Risk-benefit analysis is a technique which may be employed to assess the costs and benefits of a given activity, which involves risk. Risk assessment includes both the probabilities of various outcomes and the consequences of such outcomes expressed in dollar terms. Benefit assessments measures the benefits to the individual and society from the given activity.

Interest in risk-benefit methodology as a decision-tool for evaluating risks with respect to public health and safety was evidenced in two conferences which were held in 1971 and 1975. The first conference was organized by the Committee on Public Engineering Policy, National Academy of Engineering who noted that there "had been neither extensive public discussion or formal attention devoted to the subject or the emerging technique of benefit-risk analysis" (National Academy of Engineering 1972). Topics discussed at the conference included data requirements for decision-making, the process of risk-benefit analysis and the problems of implementing good analyses.

The second conference was part of a National Science Foundation funded study at UCLA entitled "A General Evaluation Approach to Risk-Benefit for Large Technological Systems and Its Application to Nuclear Power" (Okrent 1975). The major emphasis of the conference was risk assessment and two large technology areas were considered - nuclear reactors and the shipment and storage of liquefied natural gas. Risk assessment utilized probabilities of various outcomes and the consequence of such outcomes to estimate the expected loss from a particular activity. Consideration was also given to the determination of acceptable levels of risk. Besuner and Gibson focused on risk per hour of exposure and concluded that the risk associated with a product or activity was acceptable if it fell at the lower end of the scale. Wiggins noted, however, that several factors might be responsible for discrepancies in risk per hour of exposure including benefit differences which would warrant a higher risk for some activities. In the case of man-made hazard, he emphasized that benefits must be weighed against the risks. This was also the opinion of Oberbacher who concluded that it was important to assess both the risks and benefits from a given technology.

Risk assessment was also the focus of a recent symposium which was held at the National Meeting of the American Chemical Society (American Chemical Society 1977). The introductory paper by Lowrance included a discussion of acceptable risk and procedures for determining acceptable risk. Lowrance defined safety as a judgement of the acceptability of risk but noted that "it is not often clear who should decide the acceptability of what risks, for whom and on what terms and why". Lowrance also stressed the need for "adequate socio-political decision-making tools" commenting that while the concept of "safety" is poorly understood, the concepts of "efficiency, cost, benefit, and other decisional attributes may be even less well understood".

The advantages of risk-benefit analysis with respect to risk evaluation are as follows:

a) It serves as a basis for comparing products since it includes both the costs and benefits of a consumption activity. In contrast, hazard analysis ignores the need for and the utility of the product to the consumer.

b) Risks and benefits pertain to the same period of time so that it is not necessary to estimate hours of exposure. Benefit analysis is based on consumer expenditures per unit of time (hour, day, week, year). Risk analysis is also based on the same unit of time. Thus, the number of hours of exposure per unit of time is immaterial. While risks and benefits pertain to product use per unit of time is not necessary to estimate hours of exposure since it affects both risks and benefits equally. In contrast, hazard analysis is confined to risk so that hours of exposure are important. Lack of information concerning consumer product usage may limit the employment of hazard analysis in risk evaluation.

RISK ASSESSMENT MODEL

The following assumptions were used for the risk assessment model proposed in this paper.

1. Accident probabilities and economic consequences of accidents vary by accident severity, e.g., fatal injury, major injury, minor injury.

2. Each product in a consumer product category (e.g., bicycles, ovens) has an equal probability of being involved in an accident of a given severity and will incur the same accident costs. However, this probability may vary between product categories.

3. The value attached to a consumption activity is equal to its ownership costs.

4. Risks and benefits from ownership pertain to the time of consumption, i.e., only risks and benefits are incurred during product use.

Relative risk is then given by w'x/v where x is a column vector of increases in various types of injuries due to the activity, w is column vector of the costs of injuries and v is the value attached to the activity by the consumer. The term w'x may be considered the expected loss from a particular consumption activity. Ownership costs reflect the willingness to pay for a consumption activity and include both acquisition and maintenance costs. [Market prices are used to represent the value attached to the last unit purchased of a good or service. Such prices reflect "value in exchange" as opposed to "value in use". However, the two are interrelated since the consumer's willingness to pay is based on the utility from the product or the perceived value in use.] In contrast to ownership costs which are incurred by all consumers, accident costs are borne by only a small group of consumers. Expected loss is a measure of this uncertain prospect.

In the derivation of risk-benefit ratios, the term w'x/v may be re-written as follows:

w'x/v = (PfCf + PnCn) /V   (1)

The above equation may be re-written as

EQUATION   (2)  and    (3)

The numerator in equation (3) is the total costs of accidents involving consumer products while the denominator is the total ownership costs associated with such products.

APPLICATION OF MODEL

The risk-benefit model was applied to seven categories of adult clothing. The seven items selected were based on their consideration as possible candidates for flammability standards. (Product Safety and Liability Reporter 1974, National Cotton Council of American 1974, Simon et al. 1975).

The seven categories examined were: a) women's robes and housecoats, b) women's nightgowns, c) women's pajamas, d) men's pajamas, e) women's dresses, f) men's shirts, and g) men's pants.

Ownership costs and accident costs were derived for purchases in 1976 and all cost estimates were expressed in constant 1976 dollars.

Accident Costs

The direct costs of accidents were the medical costs while the indirect costs were the earnings foregone due to accidents (Dardis 1978). The present value of foregone earnings was used in the case of death. Two discount rates -5% and 10% were used to obtain present value estimates.

This measure neglects the pain and suffering component associated with accidents. While some agencies such as DOT have used court awards for pain and suffering to reflect such costs, this measure has generally been regarded as unreliable due to the extreme variability of court awards and the fact that only a small proportion of accident cases are decided in court (Dardis 1978). The omission of pain and suffering costs need not invalidate the ranking of products with respect to acceptable risk if it is assessed that this cost component is proportional to the sum of direct and indirect costs. Under this assumption, all risk-benefit ratios will be increased by a constant factor leaving product rankings unchanged.

Ownership Costs

It was assumed that the various clothing categories were non-durable goods with an expected life of one year. Ownership costs were based on total consumer expenditures for the particular clothing category in view of lack of data concerning maintenance costs for clothing. However, maintenance costs for most clothing items are likely to be small in view of the importance of easy care clothing in 1976. In addition, specific clothing items comprise only a small portion of total clothing consumption so that their contribution to total maintenance costs is likely to be small.

Extension of product life does not have a significant impact on risk-benefit ratios. For example, if the expected life is assumed to equal two years, then a first approximation of the present value of costs from purchases in 1976 is given by

0.5C1976 + (0.5C1977) / (1+i)  (4)

If the number and severity of burn accidents is similar in 1976 and 1977, then the present value of costs from purchases in 1976 is given by

0.5C1976 + (0.5C1976) / (1+i), or 0.95C1976   (5)

for a 10% discount rate. The assumption of an expected life of one year when the average product life is greater than one year will thus serve to underestimate accident costs. However, such underestimation is likely to be small in the case of non-durable goods.

Provision for Imports

Consumer expenditure data were based on U.S. value of shipments. However, accident costs pertained to total U. S. consumption which included imported garments. Accident costs pertaining to U.S. shipments were based on the relative importance of U. S. shipments. According to a recent study by the Council on Wage and Price Controls (1978) domestic production accounted for approximately 90% of domestic consumption in 1976. Thus, 90% of U. S. accident costs were attributed to U. S. production.

RESULTS

Consumer expenditures on U. S. merchandise in 1976 were based on U. S. value of shipments and retail mark-ups for the different clothing categories (U.S. Department of Commerce 1977, National Cotton Council 1974). Expenditures ranged from $713.46 million for men's pajamas to $3,735.22 million for men's pants. Based on the assumption that garments have an expected life of one year and zero maintenance costs, total consumer expenditures in 1976 are equal to total ownership costs in 1976. Ownership costs are given in Table 2.

TABLE 2

RETAIL VALUE OF U. S. CLOTHING SHIPMENTS IN 1976 ($ MILLION)

Accident Costs, 1976

The number of injuries for each category was based on a total burn injury estimate of 60,000 for all ages in 1976 (U. S. Department of Health, Education and Welfare 1976). Data from the National Center for Health Statistics was then applied to this estimate to obtain the number of adult burn injuries, ages 13 and above (U. S. Department of Health, Education and Welfare 1977). The number of adult clothing burn injuries was based on a low estimate from the New York State Department of Public Health (1977) and a high estimate from the National Burn Information Exchange (U.S. Department of Health, Education and Welfare 1971). The resulting low and high adult clothing burn injury estimates range from 5,213 to 17,798. It should be noted that the high estimate from the National Burn Information Exchange (NBIE) reflects the fact that most cases from this data source are characterized by severe burns and hence are more likely to involve clothing. The New York State study includes all types of burn injuries and probably presents a more realistic picture of clothing involvement.

Adult clothing death estimates were based on data from the National Center for Health Statistics (U. S. Department of Health, Education and Welfare 1977) and the National Fire Protection Association (1969). The number of deaths ranged from a low of 459 to a high of 801.

The distribution of injuries and deaths for each adult clothing category was based on data from flammable fabrics investigations conducted by the Consumer Product Safety Commission (1975). The results are given in Table 3. The number of injuries is considerably higher than the number of deaths in all instances. The men's shirt category has the highest number of burn injuries while men's pajamas have the lowest number of injuries. The women's robes and housecoat category has the highest number of deaths while women's pajamas have the lowest number of deaths.

TABLE 3

NUMBER OF CLOTHING BURN INJURIES AND DEATHS ATTRIBUTED TO U. S. SHIPMENTS IN 1976

The direct costs of burn injuries and deaths were based on length of hospital stay, which varied according to burn severity. Length of hospital stay ranged from 32-124 days which included an allowance for readmission and retreatment while a weighted average of 17 days was used for deaths (Dardis 1978). Hospital costs for flammable fabrics victims were estimated at $333.00 per day and served as a basis for estimating total medical costs. Data from Rice (1966) and the Social Security Administration (Gibson and Mueller 1979) indicated that hospital costs accounted for 59% of total medical expenses for injuries.

The indirect costs of injuries were derived from the probability of working or housekeeping, number of days lost from working or housekeeping, and mean earnings. In the case of death, the present value of future lifetime earnings was obtained using 5% and 10% discount rates. Total direct and indirect costs of burn injuries and deaths are given in Table 4 for a 10% discount rate. Injury costs based on low injury estimates range from $4.35 million for men's pajamas to $27.94 million for men's shirts. Injury costs for high injury estimates range from $14.84 million to $95.41 million for the same two clothing categories.

TABLE 4

COSTS OF CLOTHING BURN INJURIES AND DEATHS ATTRIBUTED TO U.S. SHIPMENTS IN 1976 ($ MILLION)

The costs of burn deaths are considerably lower reflecting the small number of burn deaths relative to burn injuries. As a result, the use of a 5% discount rate in place of a 10% discount rate had little impact on the total costs of burn injuries and deaths.

Risk-Benefit Ratios: 1976

The ratio of total accident costs and total ownership costs is given in Table 5 for each of the seven categories. In the case of low cost estimates, the ratios range from 0.004 for men's shirts and pants to 0.049 for women's pajamas. This means that every dollar spent on men's shirts or men's pants has an expected loss of less than half a cent. In contrast, an expected loss of approximately 5 cents is associated with every dollar spent on women's pajamas. If one wishes to consider the expected loss as a type of risk tax, then the risk tax for sleepwear is the same order of magnitude as a sales tax while the risk tax for daywear is considerably lower. However, an essential difference between the risk tax and the sales tax should be emphasized. All consumers pay a sales tax while no consumer pays the specific risk tax. Some consumers (those involved in accidents) pay considerably more than the risk tax while other consumers pay nothing. For society as a whole, the risk tax is as meaningful as a sales tax since it indicates the total costs of accidence associated with a particular consumption activity.

TABLE 5

RISK-BENEFIT PATIOS FOR VARIOUS CLOTHING CATEGORIES

In the case of high cost estimates, the risk is higher as might be expected. The risk tax for sleepwear ranges from 7% to 16% while the risk tax for daywear ranges from 1% to 4%. However, risk taxes based on high cost estimates are probably overstated since they are based on data from the National Burn Information Exchange which are biased in favor of clothing burns.

It is also of interest to compare the data in Table 3 and 4. Absolute costs of accidents in the case of daywear are highest for men's shirts followed by men's pants and women's dresses. However, when risk-benefit ratios are examined, women's dresses emerge as the highest risk category followed by men's shirts and pants. Similar reversals occur with respect to sleepwear with women's pajamas replacing women's robes and housecoats as the highest risk category.

DISCUSSION

The results of this study indicate that considerable variations exist with respect to risk benefit ratios for the various adult clothing categories. Risk-benefit ratios for sleepwear are far higher than risk-benefit ratios for daywear. If low cost estimates are accepted, intervention in the case of man's shirts or pants is unnecessary. In contrast, the data for women's sleepwear provide support for intervention. This intervention may be a flammability standard similar to that used for children's sleepwear or it may be an educational program. Both types of intervention could be undertaken by the Consumer Product Safety Commission. An alternative strategy might be the voluntary marketing of more flame retardant garments by the apparel industry. Such marketing would have to be accompanied by an educational program since consumers, in many instances, are unaware of the hazards from flammable fabrics or the need for protection.

Other results of the study are also of interest. First, the proposed method for assessing risk appears feasible. The assignment of economic consequences to various types of injuries permits comparison between products with more than one type of injury. Ownership coat data may be obtained from consumer expenditure data, product life estimates, and maintenance cost data.

Second, risk-benefit analysis provides insight concerning the level of risk to which consumers are exposed. The issue of prevailing risk is of particular importance since a society that is not risk-free must decide what level of risk requires intervention. If there is no basis for determining when intervention is necessary then risk reduction activities may be ineffective, i.e., we may concentrate on low risk areas and ignore high risk areas.

Finally, it should be noted that risk-benefit analysis is only part of the risk reduction process. This type of analysis serves to indicate prevailing levels of risk and those areas where intervention may be desirable. Once the necessity for intervention is perceived, the selection of an appropriate strategy requires additional analysis. The main advantage of risk-benefit analysis pertains to its identification of risk levels and, in turn, those areas where intervention is necessary. This type of screening is important in view of resource limitation which preclude the development of risk reduction strategies for all consumption activities. Thus, in the case of adult clothing, alternative strategies should first be investigated for women's sleepwear. This investigation would, in turn, indicate if intervention were justified and the optimal intervention strategy.

REFERENCES

American Chemical Society (1977), "Symposium on Risk Assessment and Hazard Control," National ACS Meeting, New Orleans, Louisiana.

Clark, C. C., (1969), "When is a Product Reasonably Safe?" Appliance Engineer, 3, 17-19.

"CPSC Working Draft of Standard for Flammability of Fabric for Specified Apparel Items," (1974), Product Safety and Liability Reporter.

Council on Wage and Price Stability (1978), Textiles/ Apparel, Washington, D. C.

Dardis, R., et. al. (1978), Cost-Benefit Analysis of Consumer Product Safety Programs, NSF-RANN, Final Report.

Dickerson, R. Reed (1968), Product Safety in Household Goods, Indianapolis: Bobbs-Merrill Company.

Gibson, R. M. and M. S. Mueller (1979), "National Health Expenditures, Fiscal Year 1976," Social Security Bulletin, 49, 3-22.

Lowrance, W. W.(1976), Of Acceptable Risk, California: W. Kaufman, Inc.

National Academy of Engineering (1972), Perspectives on Benefit-Risk Decision-Making, Washington, D. C.

National Cotton Council of American (1974), Possible Economic Impact of Flammability Standard on Selected Items of Apparel.

National Safety Council (1978), Accident Facts.

New York State Department of Public Health (1977), An Epidemiologic Study of Burn Injuries and Strategies for Prevention.

National Fire Protection Association, Fire Record Department Study (1969), "The Single Fatality Fire," Fire Journal, 63, 34-35.

Okrent, D. (1975), "Risk-Benefit Methodology and Application: Some Papers Presented at the Engineering Foundation Workshop," California.

Rice, D. P. (1966), Estimating the Cost of Illness, U. S. Department of Health, Education and Welfare, Public Health

Service, Health Economics Series, Number 6.

Rowe, W. D. (1977), An Anatomy of Risk, New York: John Wiley and Sons.

Simon, S. R., et. al (1975), Study of the Feasibility of Applying Flammability Standards to Adult Nightwear and Other Apparel, Research Report, Batelle Columbus Laboratories.

Starr, C. (1968), "Social Benefit versus Technological Risk", Science, 165, 1232-1238.

Starr, C. (1972), "Benefit-Cost Studies in Sociotechnical Systems," in Perspective on Benefit-Risk Decision-Making, National Academy of Engineering, Washington, D. C.

U. S. Consumer Product Safety Commission (1973), "Consumer Product Hazard Index".

U. S. Consumer Product Safety Commission (1976),"Consumer Product Hazard Index".

U. S. Consumer Product Safety Commission (1975), Flammable Fabrics, 7th Annual Report.

U. S. Department of Commerce, Bureau of Census (1977), Current Industrial Reports -- Apparel, Series MA-23A(76)l.

U. S. Department of Health, Education and Welfare (1976), Burn Care, A Report to the Senate.

U. S. Department of Health, Education and Welfare (1971), Flammable Fabrics, 3rd Annual Report.

U. S. Department of Health, Education and Welfare, Public Health Service, National Center for Health Statistics, U. S. Hospital Discharge Survey, unpublished burn injury data.

----------------------------------------