Strict product liability and safety: evidence from the general aviation market.

Nelson, Randy A. ; Drews, James N.

I. INTRODUCTION

Beginning in the 1960s, the legal system in the United States began a gradual transition to a standard of strict product liability. The proponents of the strict liability standard argued that such a move would enhance both economic efficiency and safety by placing the burden of accident prevention on manufacturers, who presumably have lower marginal costs of preventing accidents than consumers. (1) In addition, it was argued that the adoption of a strict liability standard would make it easier for consumers to obtain financial compensation from manufacturers for any injuries that they suffered.

A key component of the strict liability standard is the notion of design defect, according to which a manufacturer can be held liable if the risk of product injuries exceeds the product's utility based on current design standards. (2) The manufacturers of short-lived goods deemed to be guilty of design defect may replace existing products with safer versions, passing on the higher costs associated with accident prevention and liability insurance to consumers in the form of price increases. (3) The manufacturers of long-lived products, however, may be physically or financially unable to retrofit previously sold items to bring them up to current design standards, thus exposing manufacturers to retroactive liability for goods sold in the past. In this situation, the manufacturers of long-lived products are faced with one of two choices: (a) absorb the increased costs in the form of lower profits or (b) attempt to pass on the liability costs for the previously sold units to new customers in the form of higher prices. This latter scenario may lead to a downward spiral in which sales fall, leading producers to further price increases in an effort to recoup the sunk liability costs from an ever-declining number of new units sold.

One of the industries hardest hit by the change in liability standards was the general aviation (GA) industry. Martin (1991), Pattillo (1998), Priest (1991), Stimpson (1988), and Viscusi (1991), among others, have discussed the devastating impact of strict product liability on the manufacturers of GA aircraft. (4) GA planes are an example of a durable good with extremely long service lives; in 2005, the average age of the GA fleet in the United States was 34 yr. Although 75%-85% of GA accidents are attributed to pilot error, 90% of all accidents involving a fatality or serious injury result in a lawsuit against aircraft manufacturers. (5) As a result of the increase in litigation, claims and other legal expenses for aircraft manufacturers and their suppliers increased from $24 million in 1976 to $210 million in 1986, an increase of 775%. (6) As manufacturers attempted to pass on the higher liability costs by increasing prices, sales of new aircraft over the same period declined by 90%, from 15,451 units in 1976 to 1,495 units in 1986 according to the General Aviation Manufacturers Association (2006).

Supporters of the move to a strict liability standard in the GA market have argued that it forced manufacturers to cease production of outmoded models of aircraft that suffered from design flaws and manufacturing defects, thus improving the level of safety. This claim, they argue, is consistent with the fact that GA accidents and fatalities declined following the transition to the strict liability standard. (7) Martin (1991, 86-87), however, argues that

 in formulating the rule of strict liability ... no one
 considered the distinction between products purchased
 and used as capital items with a long service
 life and those purchased as consumables ...
 General aviation airplanes are prime examples
 of a capital item that has a long service life. Each
 general aviation airplane in the active U.S. fleet
 that carries a manufacturer's data plate represents
 a separate product liability risk to that
 manufacturer. The cost of that risk, whether
 self-insured or under-written, must be paid each
 year. On the income side of the ledger, the manufacturer's
 selling price for the airplane is fixed
 when it is sold ... On the expense side, however,
 the account remains open for the life of the airplane
 under the rule of strict liability.... When
 one considers that the expense side of the ledger
 will remain open, on average, thirty to forty
 years, ... it is not surprising to find that an
 industry such as the manufacture of small airplanes
 can be self-liquidating.

The loss of small plane manufacturing in the United States may be a price worth paying if it led to improved safety in the GA industry. Some critics, however, have argued that the change in liability standards, and consequent near elimination of the new plane market, may have compromised safety for three reasons. (8) First, much of the new technology in the GA industry, including more reliable engines, better avionics, safer fuel systems, and so on, are introduced into the market imbedded in new planes; the near demise of the new plane market after 1980 greatly impeded the diffusion of these capital-embodied technologies. In this regard, the GA industry is similar to the automobile industry in that new innovations such as airbags, antilock breaks, and stability control systems are embedded in new cars. Had the sale on new cars declined by 90% after 1980, it is unlikely that any of these innovations would have made it to the market, thus reducing driving safety. This argument is consistent with the evidence presented by the Boeing Corporation (2006) in their analysis of commercial jet accidents over the period 19592005. First-generation jets were found to have accident rates significantly above those of second-generation planes, with the difference positively related to the number of years since introduction. In addition, second-generation aircraft and early wide-body planes were generally found to have higher accident rates than current-generation planes. Although it may be possible to retrofit some models of older planes with new technology, it may be difficult or impossible to retrofit the "orphan" planes manufactured by firms no longer in existence or models of aircraft that are no longer produced by existing manufacturers like Cessna who have sharply limited the number of different models of GA aircraft they produce.

Second, the sharp reduction in the sale of new planes after 1980 resulted in an increase in the average age of the GA fleet, forcing pilots to fly older planes that may have higher accident rates because they are more prone to age-related structural failures and mechanical breakdowns. The bulk of the current GA fleet was designed to Civil Aviation Regulations 3 standards that did not require any design life specifications for certification; current standards specify the service life of certain components. Even though it is possible to repower existing planes with new engines, older aircrafts are more susceptible to problems from corrosion and brittle electrical wiring. (9) In September of 2003, the Aircraft Owners and Pilots Association (AOPA), the Antique Airplane Association, the Experimental Aircraft Association (EAA), and the Federal Aviation Administration (FAA) issued their Best Practices Guide for Maintaining Aging General Aviation Airplanes, which stated that "The GA Fleet is being used well beyond the flight hours and years envisioned when the airplanes were designed. There is concern that continued airworthiness safety matters will become more common as the fleet ages." (10)

Finally, the number of amateur-built aircraft in the United States increased by approximately 674% between 1971 and 2000. Some industry analysts have attributed the growth of the amateur-built fleet, which have higher accident rates than planes built by traditional manufacturers, to the decrease in the sales of new GA aircraft. (11) Martin (1991) has argued that these factors may explain why the GA accident rate declined more rapidly prior to the change in liability standards (1950-1969) than after the adoption of strict liability standard (1970-1989).

The purpose of the present study is to determine what, if any, impact the adoption of a strict liability standard had on the accident rate in the GA industry. In Section II, we provide a brief overview of the industry, while Section III examines the relationship between the aggregate accident rate and the new plane shipments. We employ model-specific data on accident rates and average age to test the hypothesis that older planes have higher accident rates than newer planes in Section IV. In addition, in this section we test the hypothesis that the decline in the sales of new aircraft led to an increase in the average age of the GA fleet. In Section V, we explore the relationship between the prices of GA aircraft and the growth of the amateur-built fleet. A brief summary and conclusions are offered in Section VI.

II. INDUSTRY BACKGROUND

The GA industry in the United States covers all civilian aircraft with the exception of commercial air carriers. In 2005, the active GA fleet in the United States totaled 224,352 aircraft, 74.71% of which were powered by piston engines, 3.54% by turboprop, and 4.38% by turbojet; the remaining planes were classified as rotorcraft, gliders, lighter-than-air, and experimental. (12) The majority (67.49%) of GA aircraft are utilized for personal use, 11.38% for business use, 5.97% for instructional purposes, and 4.70% for corporate use; the remaining planes are employed for a variety of uses including aerial applications and observation, sightseeing, air tours and taxis, and medical transportation. Single-engine piston aircraft, the type most commonly flown by private pilots, accounted for 60.91% of all hours flown by GA aircraft in 2005.

The owners of GA aircraft are required by the FAA to conduct rigorous periodic inspections of their planes. In addition, engine manufacturers establish a time between overhaul that serves as a guideline for when owners should overhaul or rebuild the plane's engines. As a result, GA aircraft have very long service lives, frequently extending to three or four decades. In 2005, the average age of the entire GA fleet in the United States was 34 yr; the youngest planes are single-engine turboprops with an average age of 13 yr, and the oldest are single-engine piston planes with a capacity of eight or more, with an average age of 44 yr. The type of aircraft most commonly flown by private pilots, single-engine piston planes with one to three seats and four seats, have average ages of 37 and 35 yr, respectively.

The extended service lives of GA aircraft exacerbated the economic impact of the transition to a strict liability standard in the United States in the 1960s and 1970s. Manufacturers were suddenly strictly liable not only for the aircraft manufactured after the transition to the new standard but also for the tens of thousands of aircraft produced prior to this time.

As a result of the long liability tail and the increase in litigation, the insurance cost for each new plane sold increased significantly. According to Sontag (1987), total industry liability expenses in 1977 were estimated to be $24 million, which when divided by the number of new planes sold that year implies a liability expense of $1,420 per plane. Priest (1987) estimates that by 1986 liability costs added between $75,000 and $80,000 to the cost of each new plane produced. Beginning in 1985, underwriters began to withdraw product liability insurance coverage for the three largest U.S. manufacturers, Beech, Cessna, and Piper. (13) Martin (1991) states that by 1987 Beech and Cessna were self-insured for the first $50 million and $100 million in losses and legal expenses, respectively, and Piper was uninsured for product liability. (14)

The liability expenses associated with previously sold aircraft constitute a sunk cost and thus do not affect the marginal costs of producing new planes. As a result, the liability costs of previously sold planes should have no impact on the price of new planes. In spite of this fact, GA manufacturers attempted to recover the higher costs of liability insurance by sharply increasing the prices of new planes. (15) In an effort to assess the impact of liability costs on the prices of new planes, we trace the prices for the 11 models of GA (non-jet) aircraft produced continuously between 1970 and 1985 for which we could obtain complete data. We utilize the factory list prices for a standard configuration as reported in the Aircraft Bluebook Price Digest and shipment data for each model obtained from GAMA's Annual Summary of Shipments of General Aviation Airplanes to construct a weighted price index. The price index increased at an annual average rate of 5.56% between 1970 and 1978. Between 1978 and 1985, the rate of increase was 18.10% per year, implying that the price of a new plane would double every 4.17 yr. It thus appears that liability costs increased the inflation rate of new plane prices by 326% after 1978. (16)

[FIGURE 1 OMITTED]

The price increases resulting from the change in liability standards had a devastating impact on the sales of GA aircraft. In 1978, the two most popular models produced by both Cessna and Piper, the Cessna 152 and 172 and the Piper 28-161 and 28-181, sold 1,852, 1,810, 595, and 525 units, respectively; in 1985, these four models sold 92, 194, 101, and 93 units, respectively. By 1995, Cessna had ceased production of all but a single non-jet aircraft, of which it sold a total of 87 units; Piper sold a total of 165 planes, including eighteen 28-161s and thirty-seven 28-181s.

The decline in sales after 1978 was an industry-wide phenomenon as shown in Figure 1, which illustrates total unit sales for the GA industry in the United States between 1950 and 2005. Sales of GA aircraft reached a peak of 17,811 units in 1978 and then declined almost continuously thereafter, reaching a minimum of 928 units in 1994. In an effort to aid the GA industry, Congress passed the General Aviation Revitalization Act in 1994, limiting the liability of aircraft and parts manufacturers to planes less than 18 yr of age. In 1994, Cessna announced that it would resume production of the Cessna 172, 182, and 206, and in 1995, Piper emerged from bankruptcy proceedings as the New Piper Aircraft Company. In spite of these developments, by the year 2005 total industry sales totaled only 2,857 units.

In addition to the adverse impact on the sales of new aircraft, liability costs also had a negative impact on innovation in the GA industry. (17) Between 1950 and 1980 the three largest manufacturers of GA aircraft, Beech, Cessna, and Piper, introduced an average of 17.67 new models per decade but only seven new airplanes between 1980 and 1990. Furthermore, GA manufacturers build the fuselage and wings and assemble the finished aircraft but do not manufacture the engine, propeller, avionics, fuel systems, and so on. Rather than expose themselves to the liability associated with the use of their products in GA aircraft, Martin (1991) has argued that many parts manufacturers not only ceased developing new products for use in the GA industry but also stopped supplying parts to GA manufacturers in an attempt to limit their liability exposure.

III. THE AGGREGATE ACCIDENT RATE

Much of the debate surrounding the impact of the transition to strict producer liability on the level of safety in the GA market focuses on the aggregate accident rate before and after the change in liability standards. Proponents of strict producer liability argue that the accident rate continued to fall throughout the 1980s and 1990s, implying that the change in liability standards benefited aviation consumers. Opponents of strict producer liability argue that the rate at which the accident rate fell over time declined after the change in liability standards, resulting in a lower level of safety in the GA market. We address the issue of the change in liability standards and the aggregate accident rate in this section.

The National Transportation Safety Board (NTSB) defines an aircraft accident as events in which "as a result of the operation of an aircraft, any person (occupant or nonoccupant) receives fatal or serious injury or any aircraft receives substantial damage." Accident reports are entered into the NTSB's Aviation Accident Data System and then combined with flight hour data reported by the FAA to derive the accident rate per 100,000 h of flight time. The annual aggregate accident rate for all GA aircraft for the period 1950-2005 is shown in Figure 2. (18)

The accident rate per 100,000 h declined from a maximum of 46.6 in 1950 to a minimum of 6.49 in 2004, a decrease of 82.22%. The rate of decrease was not constant throughout the entire period, however, as the curve appears to flatten after 1978; between 1950 and 1978, the accident rate declined by a geometric average rate of 4.94%, but between 1979 and 2005, the accident rate declined by a geometric average rate of 1.18%. It is of interest to note that the decline in the rate of decrease in the accident rate after 1978 coincides with the decrease in the shipment of new GA aircraft, which reached a peak of 17,811 units in 1978 and declined almost continuously thereafter.

[FIGURE 2 OMITTED]

To explore the relationship between the change in liability standards and the aggregate accident rate, we estimate the following regression model:

(1) [ACCRATE.sub.t] = [alpha] + [[beta].sub.1][T.sub.t] + [[beta].sub.2][%PERSONAL.sub.t] + [[beta].sub.1][SHIPMENTS.sub.t] + [[epsilon].sub.t],

where [ACCRATE.sub.t] is the aggregate accident rate per 100,000 h of operation in year t. The variable [T.sub.t] is a time trend, the coefficient of which is expected to be negative. The accident rate may change over time because of changes in the quality of the pilots; in an effort to control for this possibility, we include [%PERSONAL.sub.t], defined as the ratio of hours flown for personal use relative to all GA hours flown. (19) Based on the assumption that pilots flying for personal use are less skilled than pilots flying for business and corporate use, aerial observation, sightseeing, and so on, we hypothesize that [[beta].sub.2] > 0. Opponents of the move to a strict liability standard have argued that the change was counterproductive because of its impact on the sales of new planes. The near elimination of the new plane market both increased the average age of the existing GA fleet and greatly reduced the flow of capital-embodied technology into the industry. (20) To test this hypothesis, we include the annual shipments of new planes as an explanatory variable, the coefficient of which is expected to be negative.

Equation (1) was estimated employing annual data for the period 1950-2005 using first differences to control for trends in the data. The estimated coefficients, together with their standard errors, are presented in Table 1. As expected, the estimated coefficient for the time trend is negative and significant at the 1% confidence level, while the estimated coefficient for %PERSONAL is positive but statistically insignificant. Finally, the estimated coefficient for [SHIPMENTS.sub.t] is negative and significant at the 5% level, implying that the accident rate would be expected to increase with a decrease in the shipments of new aircraft.

In an effort to assess the impact of the decrease in new plane shipments on the accident rate, we use the estimated parameters from Equation (1) together with the actual data on [T.sub.t], [%PERSONAL.sub.t], and [SHIPMENTS.sub.t] to derive predicted values of the accident rate in each year from 1951 to 2005. Next, in an attempt to determine what would have happened to the accident rate had the level of new plane shipments not declined after 1980, we derive new predictions of the accident rate assuming that [SHIPMENTS.sub.t] remained at 9,607 units per year, the annual average of new plane shipments for the period 1950-1980. The results indicate that the accident rate would have declined by an average of 35.15% over the period 1981-2005, from an average of 7.73 accidents per 100,000 h to an average of 5.01 accidents per 100,000 h, if new plane shipments had remained at their average level over the period 1950-1980. Multiplying the difference in the estimated accident rates by the number of hours flown each year yields an estimate of the number of accidents that would have been avoided had new plane shipments not declined after 1980. The results indicate that the decrease in new plane shipments resulted in an additional 901 accidents per year or a total of 22,534 additional accidents over the period 1981-2005. During 1981-2005, each GA accident resulted in an average of 0.35 fatalities, implying that the reduction in new plane shipments resulted in an additional 315 fatalities per year or a total of 7,887 fatalities for the period 1981-2005.

IV. ARE OLDER PLANES REALLY LESS SAFE?

The results from the previous section demonstrate a negative, statistically significant relationship between the aggregate accident rate and the shipments of new aircraft. Opponents of the change to the strict liability standard argue that this relationship exists for two reasons. First, it is frequently argued that older planes are more prone to mechanical breakdowns and structure failures than newer aircraft. Second, new planes frequently embody technological improvements such as more reliable engines, better avionics, safer fuel systems, and so on, which enhance the level of safety relative to older planes that embody less advanced technology. As a result of these age and vintage effects, older aircraft are generally assumed to be less safe than newer planes. (21) If this hypothesis is correct, and if the reduction in the sale of new planes following the change in liability standards led to an increase in the average age of the GA fleet, then a direct link would exist between the movement to the strict producer liability standard and the accident rate.

To test the first hypothesis, we estimated the following regression to examine the relationship between aircraft safety and age:

(2) [ACCIDENT.sub.i] = [[alpha].sub.0 + [[beta].sub.1][LIMC.sub.i] + [[beta].sub.2][LFRAME.sub.i] + [[beta].sub.3][LHORSE.sub.i] + [[beta].sub.4][LCLIMB.sub.i] + [[beta].sub.5][LCARRY.sub.i] + [[beta].sub.6][LLAND50.sub.i] + [[beta].sub.7][LTO50.sub.i] + [[epsilon].sub.I],

where L is the natural log operator. The dependent variable [ACCIDENT.sub.i] represents the accident rate per 100,000 h of operation for the ith model of aircraft. The AOPA examined the accident reports for every GA accident reported to the NTSB over the period 1982-1988. The Aircraft Owners and Pilots Association Air Safety Foundation (1991) study reported the total number of accidents for the 7-yr period for specific models of aircraft, breaking them down into one of three causal factors, pilot error, machine related, or other. Model-specific accident rates were then computed by dividing the total number of accidents for a given model by the total number of hours flown for each model during 1982-1988 period, as reported in the FAA's General Aviation Activity and Avionics Survey (GAAAS). Four different accident rates, total, pilot related, mechanical, and other, were reported for each model.

The accident rate for a given model is assumed to be a function of both its mechanical and performance characteristics and the environment in which it operated. The first explanatory variables [IMC.sub.i] represent the percentage of total hours flown in instrumental meteorological conditions. The sign of [IMC.sub.i] is ambiguous; flying with limited visibility increases the risk of an accident, implying that [IMC.sub.i] may have a positive coefficient, but pilots must be instrument rated to fly in [IMC.sub.i], implying that they are more experienced and are better trained, so the coefficient may be negative. Each year, the FAA surveys aircraft owners to collect data on the number of airframe hours, of the cumulative hours of operation, for a given plane or airframe. The survey responses are used to construct the average number of airframe hours for each model of GA aircraft; we employ [FRAME.sub.i] as our proxy for the average age of a given model of aircraft to capture the combined effects of aircraft age and technological vintage. The data for [IMC.sub.i] and [FRAME.sub.i] were all obtained from the FAA's GAAAS.

The Aircraft Bluebook Price Digest contains data on the mechanical and performance characteristics of various models of GA aircraft. The variables [HORSE.sub.i], [CLIMB.sub.i], and [CARRY.sub.i] represent the engine horsepower, the maximum rate of climb, and the maximum carrying weight, respectively, for each model. The variables LAND[50.sub.i] and TO[50.sub.i] represent the length of runway required to land or take off when clearing a 50-foot-tall obstacle at the end of the runway, respectively. The variables [IMC.sub.i] and [FRAME.sub.i] represent the average value over the period 1982-1988; the mechanical and performance attributes are constant throughout the period. It was possible to obtain complete data for 79 different models of GA aircraft.

Equation (2) was estimated separately for single-engine and twin-engine aircraft; the estimated coefficients together with their t statistics are presented in Table 2. The estimated coefficient of [IMC.sub.i] is generally negative and significant at the 10% level or better in three of the regressions, implying that instrument-rated pilots have lower accident rates than pilots certified to fly only during visual meteorological conditions. Only one other control variable, LCLIMB, was significant at the 5% level or better in more than one regression.

The coefficient of the variable of greatest interest, [FRAME.sub.i], is positive and significant at the 5% confidence level or better in all the estimated equations. The results are consistent with the argument that older planes have higher accident rates than planes that have been flown fewer hours. As discussed earlier, older planes may be less safe because of physical deterioration that leads to mechanical and structural failures or because of vintage effects resulting from the fact that older planes frequently embody less advanced technology. Support for the first argument is found in the regression results for mechanical-related accidents, in which a 10% increase in the number of airframe hours would increase the mechanical accident rate by 0.2022 and 0.1217 per 100,000 h for single-engine and twin-engine planes, an increase of 11.55% and 11.12%, respectively. If the vintage argument is correct, pilots forced to fly older planes with less enhanced technology should commit more errors than pilots flying newer planes with better technology. This hypothesis is consistent with the positive and significant coefficient of [FRAME.sub.i] in the pilot-related regressions; a 10% increase in the number of airframe hours would increase the number of pilot-related accidents by 0.5409 and 0.3664 per 100,000 h for single-engine and twin-engine planes, an increase of 5.37% and 7.95%, respectively. Finally, a 10% increase in the number of airframe hours would increase the total and other accident rates per 100,000 h by 0.8180 (6.31%) and .0749 (6.67%), respectively, for single-engine planes, and by 0.3664 (7.95%) and 0.0651 (8.23%), respectively, for twin-engine planes.

The results presented in Table 2 offer strong support for the hypothesis that older planes, as measured by airframe hours, have higher accident rates than newer planes. If the decrease in the sale of new planes resulting from the change in liability standards increased the average age of the GA fleet, then the level of safety in this industry would have declined as a result of the adoption of the strict liability rule. To test this hypothesis, data were collected on new plane shipments and average airframe hours for 31 different models of GA aircraft over the period 1977-1996. The new plane shipment data were obtained from the annual reports of the GAMA; annual estimates of average airframe hours for the 31 models were obtained from the FAA's GAAAS for the years 1977-1992 and the General Aviation and Air Taxi Activity and Avionics Survey for the years 1993-1996. The data set excludes models in a given year if the shipments equal zero; the final data set includes 436 observations.

To explore the relationship between average airframe hours and new plane shipments, we estimate the following fixed-effects panel model:

(3) [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII]

where [FRAME.sub.it] represents the average airframe hours for the ith model in year t, [M.sub.i] is a dummy variable for each model included to capture model-specific factors (including the date of introduction) that are constant over time, [SHIPMENTS.sub.it] is a 3-yr moving average of the shipments of model i over the period t - 2 through t, and L is the natural log operator. Parameter estimates, together with the corresponding t statistics computed using robust standard errors, are presented in the last column of Table 2.

The estimated coefficient for [SHIPMENTS.sub.it] is negative and significant at the 1% level, indicating that the reduction in the sales of new planes following the change in liability standards led to an increase in average airframe hours for GA aircraft. Utilizing the results presented in Table 2, it is possible to assess the impact of the reduction in new plane shipments on the accident rate as follows. We first compute the average level of new plane shipments for the models in our sample, [bar.[SHIPMENTS.sub.it]]. Between 1960-1980 and 1981-2000, industry-wide shipments of GA aircraft fell from an average of 11,943 units per year to 2,120 units per year, a decrease of 82.25%. Based on the assumption that the sales of the models in our sample declined at the same rate as industry-wide sales, we increase the average value accordingly to obtain an estimate of what average sales would have been in the absence of the change in liability rules, [bar.[SHIPMENTS.sup.*.sub.it]]. Using the parameter estimates from Equation (3) to obtain predicted values of [LFRAME.sub.it] for both [bar.[SHIPMENTS.sub.it]] and [bar.[SHIPMENTS.sup.*.sub.it]], we conclude that average airframe hours would have been 39.94% lower had new plane shipments remained at their average level for the period 1960-1980.

Next, we calculate the average (total) accident rate using the mean values of the variables on the right hand side of Equation (2) and the estimated parameters from Table 2. Finally, we decrease the average value of [LFRAME.sub.it] by 39.94% and recompute the predicted value of the accident rate. The results indicate that the single-engine total accident rate would have declined from 12.70 accidents per 100,000 h to 9.61 accidents per 100,000 h, a decrease of 24.29%; the estimated accident rate for pilot-related, other, and mechanical accidents declined by 21.51%, 22.95%, and 38.40%, respectively. The twin-engine total accident rate declined from 6.74 accidents per 100,000 h to 4.92 accidents per 100,000 h, a decrease of 27.09%; the estimated accident rate for pilot-related, other, and mechanical accidents declined by 27.82%, 29.18%, and 35.91%, respectively. Finally, we compute an overall reduction in the average total accident rate by weighting the reductions in the total single- and twin-engine accident rates by the percentage of the total hours flown by single- (84.36%) and twin-engine (15.65%) piston-powered GA aircraft over the period 1980-2005. We then multiply the weighted average reduction (24.73%) by the total number of GA accidents and fatalities in each year to estimate the reduction in the number of accidents and fatalities that could have been achieved if the average number of airframe hours had not increased. Our results indicate that the number of accidents and fatalities would have declined by an average of 563.83 and 197.60, respectively, over the period 1980-2005, implying a total of 14,660 accidents and 5,138 fatalities.

The results presented in this section are broadly consistent with those presented in Section II, although the two sets of results are based on different statistical models and different data sets. In both cases, the decrease in the shipment of new planes following the change in liability standards is associated with an increase in the accident rate. The results from Section II, based on aggregate time series data, indicate that the accident rate would have declined by 35% if new plane shipments had not declined after 1980, while the results in this section, based on model-specific panel data, imply that the accident rate would have fallen by 25% if new plane shipments had not declined. In either case, the empirical results indicate that the level of safety in this industry declined following the adoption of the strict liability standard.

V. THE MARKET FOR EXPERIMENTAL AIRCRAFT

The change in liability standards precipitated a chain reaction, the impact of which extended beyond the GA market. Although most industries or sectors, including flight schools, parts manufacturers, and so on, were negatively impacted by the change, the market for amateur-built aircraft benefited greatly from the decline of the GA industry. The FAA defines an experimental plane to be "an aircraft the major portion of which has been fabricated and assembled by persons who undertook the project solely for their own education or recreation." The majority of experimental aircraft are amateur-built planes constructed from kits supplied by various manufacturers; the FAA requires that the owner must assemble and construct at least 51% of the airplane. Several industry analysts have argued that the rapid increase in the price of GA aircraft in the late 1970s led many private pilots to seek out cheaper substitutes, resulting in a sharp increase in the construction and use of amateur-built aircraft in the United States. (22)

The increase in the amateur-built fleet may have adversely impacted the overall level of safety in civilian aviation, as amateur-built aircraft have higher accident rates than GA aircraft built by traditional manufacturers. Over the period 1995-1999, amateur-built aircraft were involved in an average of 32.34 accidents per 100,000 h flown, while the average accident rate for GA aircraft was 11.10; the accident rate for amateur-built aircraft was thus 2.91 times the accident rate for GA planes. (23) Of greater concern is the fact that amateur-built aircraft have a significantly higher fatal accident rate than traditional GA aircraft; between 1995 and 1999, the average number of fatal accidents per 100,000 h flown was 9.46 for amateur-built aircraft and 1.98 for GA aircraft.

The EAA maintains a registry of amateur-built aircraft constructed in the United States since 1971. (24) The number of aircraft listed on the EEA register increased from 2,865 planes in 1971 to 22,187 in 2000, an increase of 674%. Using the EEA registration data, it is possible to test the hypothesis that the increase in the price of GA aircraft after 1980 led to an increase in the home-built fleet. To test this hypothesis, we constructed a price index that represents the average price of the single-piston engine airplanes, the closest substitute for an experimental aircraft, sold in the United States during the period 1960-2000. The index was constructed using model-specific sales data provided by GAMA and the new factory list price for a specific model of plane with average equipment as reported in the Aircraft Bluebook. The average price for the planes in our sample increased from $19,326 in 1960 to $46,105 in 1978, increasing at an average geometric rate of 4.95% per year. Between 1978 and 1987, the average price increased from $46,105 to $181,445, implying an average geometric growth rate of 16.44% per year. Finally, the average price increased from $181,445 in 1987 to $263,014 in 2000, increasing at an average geometric rate of 2.90%; the average growth rate for the entire period 1978-2000 was 8.24%.

To explore the relationship between the registration of experimental aircraft and the average price of GA aircraft, we estimated the following model:

[LAMATEURBUILT.sub.t] = [alpha] + [[beta].sub.1] [T.sub.t] + [[beta].sub.2][LAVEPRICE.sub.t] + [[epsilon].sub.t], (4)

where [AMATEURBUILT.sub.t] represents the number of amateur-built aircraft listed on the EAA registry in year t, [T.sub.t] is the value of a time trend (T = 1 in 1971) in year t, [AVEPRICE.sub.t] is the average price of new single-piston engine aircraft sold in year t, and L is the natural log operator.

Parameter estimates for Equation (4), obtained using the Prais-Winsten procedure to correct for autocorrelation, are presented in Table 3. The estimated coefficient of the time trend is positive and significant at the 1% level, indicating that the number of amateur-built aircraft increased by 5.81% per year, ceterus paribus. The estimate of [[beta].sub.2] is positive and significant at the 2% level, indicating that the increase in the average price of GA aircraft after 1980 increased the number of amateur-built aircraft registered with the EEA.

Using the estimated parameters from Equation (4), we compute the expected number of amateur-built aircraft registrations assuming that the average price of a single-piston engine aircraft increased (a) at the actual rate during 1981-2000 and (b) at the average rate for the period 1960-1978, 4.95% over the period 1960-1980. The results indicate that the increase in the average price of new GA aircraft increased the number of amateur-built aircraft registered with the EAA by an average of 2,234 units per year, representing an average of 15% of the amateur-built fleet during this period.

The NTSB's Aviation Accident Database (available online at http://www.ntsb.gov/ ntsb/query.asp) contains data for every civil aviation accident from 1962 on, including those involving amateur-built aircraft. For each year during the period 1981-2000, we tabulate the total number of accidents and fatal accidents for amateur-built aircraft in the NTSB database. Dividing the number of accidents by the number of amateur-built aircraft registered with the EEA allows us to compute both the accident rate and the fatal accident rate per plane for every year between 1981 and 2000. Multiplying the estimated accident rate by the predicted increase in the amateur-built fleet each year allows us to compute the estimated increase in the number of accidents resulting from the decrease in the sales of new GA planes. Our results indicate that the number of accidents involving amateur-built aircraft increased by an average of 37.80 accidents per year or 756 additional accidents over the period 1981-2000; the number of fatal accidents increased by an average of 11.33 accidents per year or a total of 226.65 additional fatal accidents for the same period. During the period 1995-2000, each fatal accident involving an amateur-built aircraft resulted in an average of 1.31 fatalities, implying an additional 296.91 fatalities during 1981-2000 resulting from the decrease in the shipments of new GA planes.

VI. SUMMARY

This paper has examined the impact of strict product liability on the accident rate in the GA industry. In theory, the move to a strict liability standard should have provided manufacturers with incentives to produce safer planes, leading to a decrease in the accident rate over time. Unfortunately, the failure to indemnify manufacturers from the retroactive liability associated with previously sold units forced them to try and pass on these liability costs to the purchasers of new aircraft. Not surprisingly, sales of new planes plummeted, resulting in an increase in the average age of the GA fleet.

Using data on the aggregate annual accident rate for GA aircraft, we estimate that the accident rate would have declined by an average of 35.15% over the period 1981-2005, from 7.73 accidents per 100,000 h to 5.01 accidents per 100,000 h, if new plane shipments had remained at their average level over the period 1961-1980. The results indicate that the decrease in new plane shipments resulted in an additional 22,534 accidents and 7,887 fatalities over the period 1981-2005. Using model-specific data, we demonstrate that older planes (as measured by airframe hours) have higher accident rates and that the average number of airframe hours increased by 39.94% as a result of the decrease in new plane shipments. Combining the two results, we estimate that the total accident rate would have declined by 24.73%, increasing the number of GA accidents and fatalities by 14,660 and 5,138, respectively, between 1981 and 2000. Based on the results obtained using two completely different data sets and empirical models, we conclude that the move to a strict liability standard increased the accident rate in this industry by 25%-35% over the period 1981-2000.

In addition, we examine the relationship between the demise of the GA industry and the growth of the amateur-built fleet. Using data on the number of experimental aircraft registered with the EAA, we estimate that the experimental fleet increased by 15% as a result of the decrease in the shipments of new GA aircraft over the period. Given the higher accident and fatality rates for experimental aircraft, our results indicate that the number of accidents and fatalities in the experimental sector increased by 756 and 226.7, respectively, during 1981-2000 as a result of the decrease in the shipments of new GA planes.

Taken together, our results indicate that the adoption of a strict liability standard not only proved devastating to the manufacturers of GA aircraft but also reduced the level of safety in this industry. Depending on the model and set of assumptions employed, our results indicate that the decrease in the sales of new GA aircraft increased the number of accidents by 12,791-17,095 and the number of fatalities by 4,514-6,023 during the period 1981-2000.

ABBREVIATIONS

AOPA: Aircraft Owners and Pilots Association

EAA: Experimental Aircraft Association

FAA: U.S. Department of Transportation Federal Aviation Administration

GA: General Aviation

GAAAS: General Aviation Activity and Avionics Survey

GAMA: General Aviation Manufacturers Association

NTSB: National Transportation Safety Board

REFERENCES

Aircraft Bluebook Association. Aircraft Bluebook Price Digest. Overland Park, KS: Penton Media, various years.

Aircraft Owners and Pilots Association Air Safety Foundation. General Aviation Accident Analysis Book. Frederick, MD: Aircraft Owners and Pilots Association, 1991.

--. "Antique Airplane Association, Experimental Aircraft Association, and Federal Aviation Administration." Best Practice Guide for Maintaining Aging General Aviation Airplanes, September, 2003.

Barker v Lull En'g. Co., 20 Cal. 3d 413. 573 P.2d, 443, 143 Cal. Rptr, 1978.

Boeing Corporation. Statistical Summary of Commercial Jet Airplane Accidents. Worldwide Operations 1959-2005. Renton, WA, 2006.

Cook, M. "Corrosion: The Airplane Consumer." AOPA Pilot, February, 1993. Available at http://aopa.org/ members/files/pilot/anp9302.html

Danzon, P. "Tort Reform and the Role of Government in Private Insurance Markets." Journal of Legal Studies, 13, 1984, 517-49.

Ells, S. "Aging Aircraft: When Is your Airplane too Old to Fly Safely?" AOPA Pilot, June, 2000, 95-100.

Epstein, R. "Products Liability as an Insurance Market." Journal of Legal Studies, 14, 1985, 645-69.

General Aviation Manufacturers Association. General Aviation Statistical Databook. Washington, DC, 2006. Available at http://www.gama.aero/documentCenter/index.php

--. Summary of Shipments of General Aviation Airplanes. Washington, DC, various years.

Huber, P. Liability: The Legal Revolution and Its Consequences. New York: Basic Books, Inc, 1988.

Landes, W., and R. Posner. "A Positive Economic Analysis of Products Liability." Journal of Legal Studies, 14, 1985, 535-67.

--. The Economic Structure of Tort Law. Cambridge, MA: Harvard University Press, 1987.

Martin, R. "General Aviation Manufacturing: An Industry Under Siege," in The Liability Maze: The Impact of Liability Law and Safety and Innovation, edited by P. Huber and R. Litan. Washington, DC: The Brookings Institution, 1991, 478-99.

National Transportation Safety Board. Annual Review of Aircraft Safety Data U.S. General Aviation. Washington, DC, various years.

Nelson, R., and M. Caputo. "Price Changes, Maintenance, and the Rate of Depreciation." Review of Economics and Statistics, 79, 1997, 422-30.

Pattillo, D. A History in the Making: 80 Turbulent Years in the American General Aviation Industry. New York: McGraw-Hill, 1998.

Petersen, B. "General Aviation Engineering in a Product Liability Environment," in Product Liability and Innovation, edited by J. Hunziker and T. Jones. Washington, DC: National Academy of Engineering, 1994, 62-67.

Piper Flyer Association. "Aging Aircraft Summit Report." 30 March 2006. Available at http://piperflyer.org/ news_details.asp?ID=2

Priest, G. "The Current Insurance Crisis and Modern Tort Law." Yale Law Journal, 96, 1987, 1521-90.

--. "Products Liability Law and the Accident Rate," in Liability: Perspectives and Policies, edited by R. Litan and C. Winston, Washington, DC: Brookings Institution, 1988, 184-222.

--. "The Modern Expansion of Tort Liability: Its Sources, Its Effects, and Its Reform." Journal of Economic Perspectives, 5, 1991, 31-50.

Rea, S. Jr "Nonpecuniary Loss and Breach of Contract." Journal of Legal Studies, 11, 1982, 35-53.

Smith, T. "Upward Mobility: Liability Costs Drive Small-Plane Business Back Into Pilots' Barns." Wall Street Journal, 11, 1991, A1-10.

Sontag, F. Testimony in a Bill to Amend the Federal Aviation Act of 1958 Relating to General Aviation Accidents. Hearings before the Subcommittee on Commerce, Consumer Protection and Competitiveness of the House Committee on Energy and Commerce. 100th Congress first session. Washington, DC: Government Printing Office, 1987.

--. "Indirect Effects of Product Liability on a Corporation," in Product Liability and Innovation, edited by J. Hunziker and T. Jones. Washington, DC: National Academy of Engineering, 1994, 68-76.

Stimpson, E. "Product Liability: A Root Cause of the Aviation Industry's Decline." Legal Backgrounder, Washington Legal Foundation, 15, 1988.

Truitt, L. and S. Tarry. "The Rise and Fall of General Aviation: Product Liability, Market Structure, and Technological Innovation." Transportation Journal, 34, 1995, 52-70.

U.S. Department of Transportation Federal Aviation Administration. General Aviation and Air Taxi and Avionics Survey, Washington, DC: U.S. Government Printing Office, various issues from 1993 on.

--. General Aviation Activity and Avionics Survey. Washington, DC: U.S. Government Publishing Office, various issues through 1992.

--. Statistical Handbook of Aviation. Washington, DC: U.S. Government Publishing Office, various issues.

Viscusi, W. K. Reforming Products Liability. Cambridge, MA: Harvard University Press, 1991.

Viscusi, W. K., and M. J. Moore. "An Industrial Profile of the Links between Product Liability and Innovation, "" in The Liability Maze. The Impact of Liability Law and Safety and Innovation edited by P. W. Huber and R. E. Litan. Washington, DC: The Brookings Institution, 1991, 478-99.

Wade, J. "On the Nature of Strict Tort Liability for Products." Mississippi Law Journal, 44, 1973, 825-51.

Wolk, A. A. "Product Liability Works for Pilots." Aviation Consumer, June 1/15, 1991, 14.

(1.) Landes and Posner (1985) and Landes and Posner (1987) argue for the efficiency of the strict liability standard; Priest (1988) argues on empirical grounds that the adoption of the strict liability standard has had little or no impact on product safety.

(2.) The legal precedent was established in Barker v. Lull En'g. Co. (20 Cal. 3d 413, 573 P.2d 443, 143 Cal. Rptr.); Wade (1973) and Landes and Posner (1985, 1987) discuss the efficiency of the risk-utility test.

(3.) If nonmonetary losses exist, however, Rea (1982), Danzon (1984), Epstein (1985), and Priest (1987) have all argued that the strict liability standard will result in a level of compensation insurance that exceeds the desired level.

(4.) Truitt and Tarry (1995), however, argue that a variety of other factors were responsible for the decline in the GA industry in the 1980s.

(5.) According to Aircraft Owners and Pilots Association Air Safety Foundation (1991), 76.9% of all GA accidents over the period 1982-1988 were attributed to pilot error; Stimpson (1988) cites an estimate of 85%.

(6.) See Sontag (1987).

(7.) See Wolk (1991).

(8.) For example, see Martin (1991) or Viscusi (1991).

(9.) In discussing the impact of corrosion on the GA fleet, Cook (1993) states "How bad is it? According to several sources, where airplanes used to be junked due to crash damage of run-out powerplants, more and more are being sent to the boneyard with irreparable corrosion damage."

(10.) The FAA has continued to express concern regarding the safety of aging GA aircraft. The Piper Flyer Association (2006) in discussing the second conference on aging aircraft held in 2006 summarized the FAA's stance as "The aging General Aviation fleet poses a threat to aviation safety.... many aircraft should not operate any longer."

(11.) Smith (1991) and Pattillo (1998) examine the impact of the increase in the prices of GA aircraft on the growth of the homebuilt fleet.

(12.) The information in this paragraph and the succeeding paragraph were taken from GAMA's 2006 edition of the General Aviation Statistical Handbook.

(13.) Martin (1991, 483-84) states that "An aviation underwriter affiliated with Lloyd's of London, when explaining why his firm was withdrawing from the GA market, explained 'We are quite prepared to insure the risks of aviation, but not the risks of the American legal system.'"

(14.) In discussing the relationship between liability and innovation, Viscusi and Moore (1991, 103-04) state that "the role of product liability insurance coverage for the aircraft and parts industry had been all but eliminated by the mid 1980s. In effect, this industry lies outside the scope of conventional insurance markets. This result is not due to a low level of liability that makes insurance unattractive. Rather, liability costs are so excessive that these firms have been led to make alternative arrangements."

(15.) As Viscusi (1991, 39-40) states, "A fundamental misperception by the industry has been the assumption that the liability costs for planes already sold can be recouped in today's marketplace. The price of planes sold in an earlier liability era did not fully reflect their ultimate liability costs. Firms should treat these unanticipated liability costs as sunk costs. Efforts to price new planes to cover past liability costs cannot succeed, because current consumers do not benefit from a liability price tag that includes not only their own prospective liability costs but also a share of earlier awards."

(16.) It is possible that the rapid rate of price increase after 1978 reflects manufacturer's attempts to improve the quality of the new planes they sold. Nelson and Caputo (1997) estimated quality-adjusted price indices for GA aircraft, which were then used to compute the rate of price increase before and after 1975. The percentage increase in the growth rate before and after 1975 was approximately the same using quality-adjusted prices.

(17.) This issue is discussed further in Petersen (1994) and Sontag (1994).

(18.) The data from 1987 to 2005 were obtained from the NTSB's Web site www.ntsb.gov/Aviation/Table10. htm; data for the remaining years were obtained from various issues of the NTSB's Annual Review of Aircraft Accident Data--U.S. General Aviation.

(19.) The necessary data to construct %PERSONAL are reported in the FAA's Statistical Handbook of Aviation.

(20.) See Martin (1991) or Viscusi (1991).

(21.) For example, see Huber (1988), Martin (1991), or Viscusi (1991). In an effort to address this issue, the AOPA, the Antique Airplane Association, the EAA, and the FAA issued a jointly endorsed Best Practices Guide for Maintaining Aging General Aviation Airplanes in September of 2003. For additional information, see Ells (2000).

(22.) Both Smith (1991) and Pattillo (1998) discuss the impact of changes in the prices of GA aircraft on the development of the homebuilt sector.

(23.) The accident rate for amateur-built aircraft covers only aircraft classified as amateur-built and does not include experimental aircraft classified as exhibition or other. Amateur-built aircraft comprised 82% of the experimental fleet in 2000. The accident rate for GA aircraft covers only single-engine piston planes and rotorcraft conducting personal/business flights as these are the type of GA aircraft most comparable to amateur-built planes. This overstates the accident rate for all GA aircraft as GA planes powered by multiple piston engines, turboprop, or turbojet engines have significantly lower accident rates. See the NTSB's 1999 edition of the Annual Review of Aircraft Accident data for U.S. General Aviation for details.

(24.) The FAA did not begin reporting the size of the experimental or amateur-built fleet until 1993.

RANDY A, NELSON and JAMES N. DREWS, We are indebted to the Douglas Chair Fund at Colby College for financial assistance and to seminar participants at Colby and an anonymous referee for very helpful comments that substantially improved the final paper.

Nelson: Douglas Professor of Economics, Department of Economics, Colby College, Waterville, ME 04901. Phone 1-207-859-5238, Fax 1-207-859-5229, E-mail ranelson@colby.edu

Drews: Graduate Student, Tuck School of Business, Dartmouth College, Hanover, NH 03755. Phone 1-781223-0713, E-mail james.n.drews@dartmouth.edu

TABLE 1
Aggregate Accident Rate Regression Results

Variable Estimated Coefficient t Statistic

T -0.7233 3.19
SHIPMENTS -1.9207 2.02
%PERSONAL 0.1090 0.04
[bar.[R.sup.2]] .1722

TABLE 2
Accident/Frame Regression Results

 Total Pilot-Related Mechanical
Variable Accidents Accidents Accidents

Single-engine
planes
 Constant -46.8011 (0.90) -20.7362 (1.10) 0.9655 (0.09)
 SHIPMENTS -0.0007 (4.79)
 LIMC -0.8608 (1.40) -0.8183 (1.51)
 LFRAME 8.5826 (3.46) 5.6754 (2.60) 2.1216 (4.34)
 LHORSE 4.2903 (0.60) -3.7271 (0.59) 5.5534 (3.91)
 LCLIMB 8.7446 (1.82) 11.1159 (0.96) -1.3874 (3.61)
 LCARRY -12.9723 (1.82) -6.0371 (0.96) -5.0677 (3.61)
 LT050 3.1506 (0.60) 2.5956 (0.57) 1.2894 (1.25)
 LLAND50 -3.5998 (0.69) -2.7616 (0.96) -1.3976 (1.36)
 [bar.[R.
 sup.2]] .3989 .4050 .3704
Twin-engine
planes
 Constant -78.9281 (0.69) -62.9693 (1.17) -28.1246 (1.95)
 LIMC -8.5625 (2.61) -7.7151 (3.24) 0.3040 (0.48)
 LFRAME 5.0836 (2.40) 3.8446 (2.49) 1.2765 (3.09)
 LHORSE 0.8630 (0.09) 3.3225 (0.47) -2.853 (1.49)
 LCLIMB 4.0965 (0.63) 6.0465 (1.27) 1.4027 (1.10)
 LCARRY -1.1866 (0.11) -5.7855 (0.72) 2.4102 (1.15)
 LT050 -2.3184 (0.37) -2.6277 (0.57) 0.0110 (0.01)
 LLAND50 3.1941 (0.48) 3.5878 (0.74) 0.9069 (0.70)
 [bar.[R.
 sup.2]] .2173 .3153 .2141

 Other
Variable Accidents LFRAME *

Single-engine
planes
 Constant 2.9696 (0.38) 7.0122 (79.70)
 SHIPMENTS
 LIMC -0.0358 (0.30) -0.0066 (0.07)
 LFRAME 0.7857 (2.13)
 LHORSE 2.4639 (2.30)
 LCLIMB -0.9839 (1.76)
 LCARRY -1.8675 (1.76)
 LT050 -0.7344 (0.94)
 LLAND50 0.5594 (0.72)
 [bar.[R.
 sup.2]] .1002 .5471
Twin-engine
planes
 Constant -3.2861 (0.33)
 LIMC -1.1574 (2.63)
 LFRAME 0.6825 (2.40)
 LHORSE 0.0879 (0.07)
 LCLIMB -0.9509 (1.08)
 LCARRY 0.7002 (0.47)
 LT050 -1.1812 (1.39)
 LLAND50 0.9315 (1.04)
 [bar.[R.
 sup.2]] .3816

* The estimated coefficients for the thirty model-specific
dummy variables are not reported in an effort to conserve
space; a copy is available from the authors upon request.

TABLE 3
Amateur-Built Accident Regression

Variable Estimated Coefficient t Statistic

Constant 5.7025 6.44
T 0.0447 5.37
AVEPRICE 0.2378 2.70
[bar.[R.sup.2]] 0.9816
DW 1.5610