2. Automotive safety

As vehicle injuries and death became more prevalent with ever-increasing power and speed in the early and mid-1900s, automakers introduced many safety items, including emergency brakes, safety glasses, and padded dashboards. However, the focus on vehicle safety became a hot issue when Ralph Nader published his 1965 critique of the Corvair and automobile safety: Unsafe at Any Speed (Nader, Reference Nader1965). Unlike most domestic cars at the time, which had engines in the front and solid beam axles in the rear, the Corvair’s motor is in the back which requires an independent rear suspension. Due to its rear suspension design, the Corvair had a greater tendency to lose outside rear grip when negotiating a turn. The faster the cornering speed, the greater the propensity to lose control. The Corvair’s manufacturer, General Motors, recommended low tire pressure for the front tires and high tire pressures for the rear tires to counteract this tendency. Critiques, including Nader, noted under-inflation causes additional stress, uneven wear, and, after repeated use while under-inflated, an increase in the likelihood of tread separation and sudden failure.

Nader and others sought to require new safety equipment, including tire pressure monitors, to reduce accidents, injuries, and fatalities (Williams, Reference Williams2015). His testimony persuaded Congress to regulate passenger vehicle safety and design and establish the National High Traffic Safety Administration (NHTSA). Many of Nader’s safety equipment demands, such as two-piece steering columns (Standard No. 204, 1968), safety belts (Standard No. 208, 1968), and door bars (Standard No. 214, 1973) became required by law. However, others, like Nader’s demand for tire pressure monitoring systems, did not.

3. What is needed to become a regulation?

To investigate why tire pressure monitoring systems were not required by the early 1970s, while other safety measures were mandatory, requires a theory of regulatory action. Luckily, Bruce Yandle’s (Reference YandleYandle1983) “Bootlegger’s and Baptist’s” Theory of regulation provides such insight. Building on earlier theories of regulation, most notably the “Public Interest Theory” and Stigler’s (Reference Stigler1971) “Economic Theory of Regulation,” Yandle (Reference YandleYandle1983) notes that regulatory bodies have several ways to generate a preferred outcome. These methods can be categorized into two broad categories:

1. 1. Command and Control Standards

2. 2. Performance Standards

Under command and control standards, government agencies require the use of a specific technology or process. Failure to use the specific technology or process results in fine or other punishment. Under performance standards, government agencies create a minimum standard that must be met or exceeded. Any legal methods that meet the standard are acceptable; however, failure to comply with the standard results in a fine or punishment.

Yandle (Reference YandleYandle1983) then explores why agencies regularly choose command and control methods when performance standards provide more incentive to innovate and are often less costly.

To answer his question, Yandle recounts a story of living in a county where alcohol sales are illegal. He notes that Baptists seek to restrict or prohibit alcohol consumption for moral reasons. At the same time, Bootleggers benefit when such alcohol restrictions are in place. Command and control prohibitions are often successful because moral (Baptist) and economic (bootleggers) interest groups support the policy (Smith & Yandle, Reference Smith and Yandle2014; Yandle, Reference YandleYandle1983). The result: bootleggers gain the market share, and Baptists have met their moral responsibility.

Yandle’s Bootlegger and Baptist Theory has been used to explain many unlikely pairings: Archer Daniels Midland and environmentalists seeking biofuel subsidies (Vanderkam, Reference Vanderkam2007), anti-smoking organizations and cigarette makers finding ways to limit the use of electronic cigarettes (Adler et al., Reference Adler, Meiners, Morriss and Yandle2016), and the promotion of Accountable Care Organizations as part of the Affordable Care Act by physicians, hospitals, and other medical providers, along with health rights advocates (Smith, Reference Smith2019). And, yes, even the combined efforts of religious organizations and liquor store owners in wet Arkansas counties to keep neighboring counties dry (Horpedahl, Reference Horpedahl2021).

However, in the late 1960s, no bootleggers were present to support the Baptists’ desire to require TPMS. That is, in the late 1960s, bootleggers could profit from requiring seat belts and two-piece steering columns, but TPMS had not yet been invented, and thus, no TPMS bootleggers were present.

4. Introduction of the TPMS

In 1978, the Society of Automotive Engineers investigated real-world tire inflation and discovered that 27 % of the vehicles they analyzed had tires under-inflated by 4–16 pounds (Viergutz et al., Reference Viergutz, Wakeley and Dowers1978) and submitted their findings to the NHTSA. The NHTSA initially said it would require low-pressure warning systems, but ultimately nixed the idea (House of Representatives, 2000a).Footnote ¹

In the early 1980s, Brookstone began producing warning systems that could be placed on the end of each tire’s valve stem so that drivers could quickly and easily see if a tire was under-inflated before entering their vehicle. By the mid-1980s, TPMS that alerted drivers via dash lights were ready for retail customers. Porsche was the first to include factory-installed TPMS on their 1986 959 models. In 1991, Corvettes became the first U.S. vehicle equipped with TPMS. In 1998, Bartec began selling tools that enabled tire shops to read and test TPMS systems (Bartec, 2018). In 2000, Renault launched the first high-volume passenger vehicle to include TPMS: the Laguna II.

Over time, two systems became the most prevalent: the direct system and the in-direct systems. Direct systems, initially developed by Schrader Electronics, require a pressure sensor inside the wheel at the base of the valve stem. The sensor periodically measures the tire pressure and sends a wireless signal to the vehicle’s electronic control unit (ECU). If the tire pressure falls below a prescribed amount, the sensor alerts the vehicle’s ECU, which triggers a light on the dash. Indirect systems do not directly measure tire pressure. Instead, they measure the speed of tire rotation. An under-inflated tire has a smaller circumference than a properly inflated tire. Therefore, the tire must rotate faster than when properly inflated. The indirect system uses the vehicle’s stability control system to recognize when one tire spins more quickly than another. If this takes place, the TPMS alerts the driver.

By 2000, evidence began to mount that tread on BF Goodrich tires on Ford Explorers were abruptly failing and causing rollovers and fatalities. In February of 2000, numerous complaints were submitted to regulators after Houston’s KHOU TV station aired a segment on sudden tread separation (Tompkins, Reference Tompkins2002). In May 2000, after accumulating 90 complaints involving four deaths, the NHTSA opened a formal investigation (Pinedo et al., Reference Pinedo, Seshadri and Zemel2002). On August 9, 2000, Bridgestone-Firestone announced a recall of 6.5 million Firestone Wilderness, AT, ATX, and ATX II P235/75R15 tires (15″ tires). These tires were often, though not exclusively, standard equipment on many Ford Motor Company vehicles, including the Ford Explorer from 1991–2000. “By September 2000, there were 2,200 complaints involving 103 deaths and more than 400 injuries.” (Pinedo et al., Reference Pinedo, Seshadri and Zemel2002, Reference Pinedo, Seshadri and Zemel1). A vast number of articles were published decrying low tire pressure risks (Bradsher, Reference Bradsher2000a, Reference Bradsher2000b; CNN Money, 2000; Simison et al., Reference Simison, Lundegaard, Shirouzu and Heller2000; Spurgeon, Reference Spurgeon2000).

In September, Congress held hearings on the Firestone Recall and Ford Explorer. The hearings provided evidence that BF Goodrich tires experienced more failures than others. “For example, the 23575R15 tire, which amounted to only 6 % of Firestone production of these tires, nevertheless were 36 % of the total separations in 1 year alone in 1999” (House of Representatives 2000b, 3). They also discovered that the Decatur plant, which produced 17–18 % of the tires in question and experienced contentious labor disputes, accounted for 57 % of the total separations in 1999 (House of Representatives, 2000b; Krueger & Mas, Reference Krueger and Mas2004).

Much of the Congressional discussion surrounded proper tire inflation and rollover risk. BF Goodrich had tested their tires at 32 pounds per square inch (psi), while Ford recommended 26 psi for its Explorer. Ford recommended lower tire pressure to lower the vehicle’s ride height and the probability of rollover. However, the lower tire pressure increased tire stress and the risk of sudden failure.

In November, the U.S. Federal government enacted legislation that ultimately required tire pressure monitoring systems on all new passenger vehicles (H.R. 5164). Explicitly, Section 13 of the 2000 TREAD Act states,

Shortly after the NHTSA filed its notice of proposed rulemaking in July of 2001, Bootleggers and Baptists filed comments in support of the direct tire pressure sensor system. Direct tire pressure sensor producers, such as Schrader Electronics, suggested the NHTSA rule should require TPMS to “provide [a] warning when any number of tires, from one to four tires, is significantly under-inflated” (McClelland, Reference McClelland2001, Reference McClellandp.1). Though this suggestion appears reasonable on its face, it also effectively precludes systems that use indirect, differential tire rotation speed methods, because two similarly under-inflated tires will not trigger a warning. Therefore, Schrader was supporting a rule that required automobile manufacturers to buy sensors they produced for all wheels mounted to the vehicle, unlike an indirect system that uses the anti-lock braking system already available on many vehicles and the indirect/direct combination that uses sensors on two wheels in combination with the anti-lock braking system.

Baptists, including Public Citizen, Consumer Federation of America, Trauma Foundation, Consumers Union, and Advocates for Highway and Auto Safety, also objected to the NHTSA’s decision to allow automobile manufacturers to use existing antilock brake technology to design indirect TPMS (Claybrook, Reference Claybrook2001). These objections became more intense when, in February of 2002, John D. Graham of the Office of Information and Regulatory Affairs (OIRA) returned the proposed rule to the NHTSA due to cost concerns. Graham suggested the NHTSA allow automobile manufacturers to install less expensive, less accurate, and less informative types of indirect TPMS. In response to this return letter, Public Citizen President Joan Claybrook chastised Graham for blocking “an overdue, lifesaving rule required by Congress in the wake of the nation’s most publicized tire safety disaster because, in your view, NHTSA must permit industry to install a marginally cheaper, but far less accurate and beneficial, type of tire pressure monitoring system” (Claybrook, Reference Claybrook2002, p. 1).

Ultimately, Public Citizen, the New York Public Interest Research Group, and the Center for Auto Safety petitioned for a review of the Final Rule on Tire Pressure Monitoring Systems (Public Citizen, Inc. v. Mineta 2003). In their petition, they noted that initial indirect systems assess tire inflation by comparing the wheel speeds in diagonally opposed tires. Therefore, this indirect method is unable to detect incorrect inflation if it occurs simultaneously in all four of the vehicle’s tires, two tires on the same side of the vehicle, or two tires on the same axle of the vehicle. Moreover, indirect systems require a vehicle in motion, whereas direct systems can warn of incorrect inflation before a vehicle is moving. The court sided with the plaintiffs and the final command and control type rule required a specific technology: direct sensors. The alternatives included either two direct sensors along with an indirect method or four direct sensors, much to the delight of direct sensor manufacturers, such as Schrader Electronics and others.

5. Estimated ex-ante costs by the NHTSA versus ex-post costs

In 2002 the NHTSA issued an estimate of the costs and benefits of installing TPMS in passenger vehicles. They investigated two tire pressure options: Alternative 1 alerted drivers when tire pressure fell 20 % below the recommended psi. Alternative 2 was a 25 % decline. (US NHTSA, 2002). However, the NHTSA noted that it did not include the maintenance costs most direct systems, and some indirect systems, require (NHTSA 2001, VI-9). After public comments, the OMB returned the rule to NHTSA for reconsideration. Ultimately, after two court cases brought by Public Citizen and others, the NHTSA issued the final rule and regulatory impact assessment (RIA) in March of 2005. The updated 2005 RIA assumed 17 million vehicles sold annually, added a measure of maintenance costs, the cost of refilling tires, the benefits of less property damage, and reduced travel time from fewer accidents (US NHTSA, 2005).

The NHTSA’s 2005 RIA investigated three compliance options: a direct system that supplies a continuous readout of tire pressures, a direct system that alerts the driver with a warning light if one or more tires are under-inflated, and a hybrid system that includes two direct sensors and an indirect system with a warning light for the driver. The 2005 assessment estimates that a direct system (option 2) is $66.08 per vehicle ($2001). The RIA also notes that direct systems require batteries that will likely need replacement after 8 years. Therefore, the maintenance costs are estimated to be as high as $40.40 per vehicle. Table 1, row 1, replicates row 2 of Table VII-4 (b) of the 2005 RIA to show the total annual cost for 17 million cars and trucks (in millions $2001 at a 7 % discount rate). Notably, the 2005 assessment finds that the yearly cost of a direct system (option 2), including the cost of sensor/battery maintenance and tire inflation minus the benefits of improved fuel mileage, less tire wear, less time in traffic, and less property damage is $1,479 million.

Table 1. Total annual costs for 17 million vehicles at 7%

Mulholland (Reference Mulholland, Meiners and Haeffele2020) noted that NHTSA’s 2005 ex-ante RIA was overly optimistic about the cost of installing and maintaining a direct TPMS. First, the installation value of $66.08 is much lower than the teardown cost found by Simons (2017), on behalf of the NHTSA. Simons (Reference Simons2017) finds that the price of TPMS required by rule FMVSS 138 is $162.13 ($2012) for passenger vehicles or $125.25 ($2001). This finding is $59.17 ($2001) higher than the $66.08 ($2001) used in the Final 2005 RIA.

Second, the NHTSA assumes that each replacement sensor, excluding installation costs, will cost the owner $22.50 ($2001). However, Mulholland (Reference Mulholland, Meiners and Haeffele2020) uses the 2018 retail prices of the 159 sensors listed for sale on tirerack.com to show that the median price of a sensor is $49 each (Mulholland Reference Mulholland, Meiners and Haeffele2020). Following the NHTSA 2005 methodology but updating their estimated price of a sensor and the updated installation cost from Simons (Reference Simons2017), Mulholland (Reference Mulholland, Meiners and Haeffele2020) reports an annual loss of up to $2,813 million ($2001). This value is shown in row 2 of Table 1. Mulholland also adds the additional costs faced by those in northern states who use a dedicated set of wheels and winter tires. Including winter tires and wheels for northern states, the result is $1,915 to $2,987 ($2001) million in losses.

6. Estimated ex-ante benefits by the NHTSA versus ex-post costs

Combining both fatalities and injuries by estimated severity, the NHTSA’s initial 2002 RIA conducted for the proposed rule estimated the number of equivalent fatalities to be reduced by 300 per year under Alternative 1 (a 20 % decline in psi) and 184 per year under Alternative 2 (a 25 % decline in psi; US NHTSA, 2002). Adjusting for improved fuel mileage and reduced tread wear, the NHTSA estimates $369 million ($2001) annually for Alternative 1, or $1.9 million ($2001) per equivalent life saved (at a 7 % discount rate), and $138 million ($2001) annually, or $1.1 million ($2001) per equivalent life saved, for Alternative 2 vehicle (US NHTSA, 2002). Assuming $3.5 million per statistical life, this results in a net benefit to society of $480 million (300*$1.6 million) under Alternative 1 and $441.6 (184*$2.4 million) under Alternative 2. However, in the NHTSA’s revised 2005 estimates, the agency estimated that 161 equivalent lives would be saved using Option 2. Table 2, row 1, shows that this resulted in a cost-per-adjusted life saved between $4.90 and $9.20 million ($2001). Mulholland (Reference Mulholland, Meiners and Haeffele2020) finds a cost of anywhere from $11.16 to 18.56 million per adjusted life saved. Table 3 shows the range of estimated net benefits assuming a statistical life is valued at $3.5 million ($2001), In the 2005 RIA, the NHTSA reversed its positive net benefit finding from its 2002 RIA and estimated a net benefit of −$226 to −$915 million. Mulholland (Reference Mulholland, Meiners and Haeffele2020) compares the expected benefits in the NHTSA’s 2005 RIA and finds a net cost to society of $1,233 to $2,249 million. Including dedicated winter tire and wheel costs, results in a net loss to society of $1,352 to −$2,424 million ($2001) or a loss that is 2.5 higher than those estimated by the Final Rule’s 2005 RIA (Mulholland, Reference Mulholland, Meiners and Haeffele2020).

Table 2. Net cost per equivalent life saved

Table 3. Net benefits with a value of $3.5 million per statistical life

7. Estimated ex-post benefits versus ex-post costs

However, the estimates above are ex-ante benefits versus ex-post costs. Because it has been 15 years since the final rule became law, we can now determine whether TPMS is associated with a decline in fatalities, ex-post. Although the TREAD Act required all post-2007 models to have TPMS, the initial year of availability of TPMS varies by model. Figure 1 shows the fraction of models with factory-installed TPMS by model year. When the TREAD Act was signed into law in 2000, only about 3 % of models came with TPMS. By 2005 this had increased to 21.4 %. By 2007, this had risen to just under 42 %.

Figure 1. Share of models with TPMS by year.

Taking advantage of this variation in the model-year rollout, I perform a difference-in-difference estimation to determine whether the introduction of TPMS is associated with declines in fatalities. The estimating equation,

(1) $$ {y}_{\mathrm{model},\ \mathrm{model}-\mathrm{year}}={\displaystyle \begin{array}{l}{\alpha}_{\mathbf{0}}+{\alpha}_{\mathbf{1}}{\mathrm{tpms}}_{\mathrm{model},\mathrm{model}-\mathrm{year}}+{\alpha}_{\mathbf{2}}{\mathrm{sales}}_{\mathrm{model},\ \mathrm{model}-\mathrm{year}}\\ {}+\hskip2px {\gamma}_{1\mathrm{model}-\mathrm{year}}+{\gamma}_{2\mathrm{model}}+{\gamma}_3\ast {\mathrm{trend}}_{\mathrm{model}}+{\varepsilon}_{\mathrm{model},\ \mathrm{model}-\mathrm{year}},\end{array}} $$

investigates whether fewer fatalities from vehicle crashes, y _{model, model-year}, are associated with the introduction of TPMS, tpms_{model, model-year}, by model-year. Because the number of passengers potentially exposed to fatal crashes is a function of the number of vehicles sold, I also include the number of units sold for each model for each year, sales_{model, model-year} (carsalesbase.com). The estimation also includes model-year fixed effects, model-specific fixed effects, and, for some specifications, model-specific trends. All errors are clustered by model.

The number of deaths is measured in three ways. First, I investigate the total number of vehicle fatalities. However, because the TREAD Act calls for several vehicle and tire safety regulations, in addition to TPMS, looking at the overall number of deaths will likely overestimate the association between the introduction of TPMS and fatalities. Moreover, because the addition of TPMS was often part of safety updates and, at times, complete model redesigns with new, safety-enhancing features, the coefficient α₁ will also capture declines in fatalities associated with other safety features introduced simultaneously and will therefore be negatively biased.Footnote ² To partially address this bias, I focus on the number of vehicle deaths due to tire failure and the number of vehicle deaths due to tire misinflation. Although this alternative measure will also be negatively biased because of tire wear and longevity improvements, using these more precise measures of tire-related fatalities enables me to focus more directly on TPMS’s role in possibly reducing deaths.Footnote ³

8. Data

To estimate Equation 1, I combine the NHTSA’s Fatal Analysis Reporting System (FARS) data with TPMS model-year introduction data from BARTEC, a producer of tire pressure sensor readers, and vehicle sales data from CarSalesBase.com. Vehicle fatalities come from the NHTSA’s Fatality Analysis Reporting System (FARS) (NHTSA, 2019). The FARS data includes information on the model (or models) involved, the model year of the vehicle(s), the date of the accident, the number of fatalities, and what law enforcement reports as the reason or reasons for the fatal crash. BARTEC provides information on the model-years when tire pressure monitoring systems are standard equipment (BARTEC, 2018). Because the number of fatalities is a function of the number of vehicles sold for a particular model-year, I include the number of model-year units sold provided by CarSalesBase.com.

Table 4 shows the summary statistics for the model years before and after the introduction of TPMS. For pre-TPMS model-years, the average number of vehicle fatalities for a model year is 124 with a maximum of 1929. The mean number of tire failure-related deaths is 0.5 with a maximum of 67. The mean number of tire inflation deaths is 0.003, with a maximum of 2. For model-years after the introduction of TPMS, the mean number of fatalities for a model-year is just under 28, with a maximum of 874. Tire failure-related fatalities have a mean just under 0.02 with a maximum of seven. There are no post-TPMS reported tire inflation deaths.

Table 4. Summary statistics

9. Estimations and ex-post net-benefits

Table 5 shows the estimated coefficients from Equation 1 when looking at total deaths, tire failure-related deaths, and misinflation-related deaths separately. The first three columns do not include model-specific trends; columns 4–6 do. Focusing on the model-specific trends specifications, the introduction of TPMS is associated with 47 to 95 fewer vehicle fatalities per model-year. For tire-failure deaths, TPMS is associated with 0.27 to 0.62 fewer deaths. For tire inflation deaths, the estimate without a model-specific trend reports 0.004 fewer deaths. The specification with model-specific trends is 0.0009, but it is imprecisely estimated. These estimates suggest that the introduction of TPMS was associated with a reduction in fatalities overall and tire failure related fatalities.

Table 5. Change in deaths by model with introduction in TPMS

Standard errors in parentheses.

*** p < 0.01.

** p < 0.05.

* p < 0.1.

This method compares old versus new, but it also compares older pre-TPMS vehicles that have been on the road for many more years, with more modern, TPMS vehicles that have been on the road for fewer years. Because these new vehicles have been on the road fewer years, the estimated coefficient on TPMS is likely negatively biased. Therefore, to compare apples to apples, I truncate the sample to only include fatalities in vehicles that have been on the road for 5 or fewer years. This approach enables me to focus on the first 5 years of ownership when vehicles less likely to experience potential mechanical failures. It also allows me to address the concerns that newer vehicles are often driven by more prudent and healthier drivers (Blows et al., Reference Blows, Ivers, Woodward, Connor, Ameratunga and Norton2003; Farmer & Lund, Reference Farmer and Lund2006; Martin & Lenguerrand, Reference Martin and Lenguerrand2008; Sheehan-Connor, Reference Sheehan‐Connor2022).Footnote ⁴ Most importantly, with this methodology, I can estimate the change in the number of fatal crashes across 5 years so that I can calculate the change in the average annual number of fatalities.

Table 6 shows the estimates from the five-year sub-sample. The introduction of TPMS is associated with 7.1 to 21.9 fewer deaths for the first 5 years of operation per model-year. Using the model-year specific trends specification coefficient of 7.1 and recognizing that this is a reduction in fatalities over 5 years, the estimate suggests an average of 1.42 fewer deaths per year per model-year. The sample includes 467 different models. Therefore, assuming all types of fatalities forgone are a result of TPMS, the introduction of TPMS is associated with about 663 fewer deaths per year (or 1.42*467).

Table 6. Change in deaths by model with introduction of TPMS (with first 5 years of operation)

Standard errors in parentheses.

*** p < 0.01.

** p < 0.05.

* p < 0.1.

Table 7 then reports the net benefit of these 663 fewer deaths by assuming 17 million cars are purchased each year, as was the case in 2016, $3.5 ($2001) million per statistical life, and using the costs estimates from the NHTSA and Mulholland (Reference Mulholland, Meiners and Haeffele2020). Table 7 shows that the net benefit of 663 fewer deaths per year ranges from $405 million ($2001) to $1,530 million if we assume that the entire reduction in vehicle deaths is due to the introduction of TPMS alone and there are no maintenance costs. If the maximum maintenance costs are included, the net benefit ranges from −$667 to $841 million.

Table 7. Total annual costs for 17 million vehicles (assuming present year benefits and $3.5 million per statistical life; millions of 2001 dollars)

However, as noted above, some of this decline in fatalities is likely due to other new features introduced along with TPMS in updated models. Therefore, I re-perform the analysis above using the reduction in deaths associated with only tire failures. Taking the results from Table 6 that shows that each post-TPMS model year witnesses 0.12 fewer tire-related fatalities and again given the 467 models included, this is a reduction of 56 deaths over 5 years or just over 11 fewer deaths per year.

Table 8 reports the three possible estimated net benefits of TPMS by using the same cost estimates in Table 7 combined with the benefit of 11 fewer tire failure-related fatalities witnessed after the introduction of TPMS. As shown in Table 8, the net benefit ranges from −$752 million to −$1,876 million without maintenance costs included. The net benefit declines further from −$1,440 to −$2,949 million when the maximum maintenance costs are included. Thus, I find a net loss with the introduction of TPMS.Footnote ⁵

Table 8. Total annual costs for 17 million vehicles (assuming present year benefits and $3.5 million per statistical life): tire failures (millions of 2001 dollars)

10. Alternate ex-post benefits: Varying the number of years on the road

The use of a five-year window is somewhat arbitrary. To determine whether the 5-year window affects the results, I repeat the estimation above using vehicles that have been on the road for (a) 3 years or less and (b) 10 years or less. Table 9 shows the summary statistics for both samples. As was the case in the full sample, the number of vehicle fatalities is much larger pre-TPMS than post-TPMS. For the three-year sample, the mean number of deaths is almost three times greater before TPMS; for the 10-year sample, deaths are 3.4 times greater before TPMS. For the three-year sample, tire failure fatalities are 10 times greater in the pre-TPMS period; for the 10-year sample, tire failure fatalities are almost 22 times higher before the introduction of TPMS.

Table 9. Summary statistics for 3- and 10-year estimations

Table 10 shows the results from the 3-year estimation. Over 3 years, the introduction of TPMS is associated with 4.5 fewer fatalities overall. Fatalities from tire failures and tire inflation are not precisely estimated, suggesting that the introduction of TPMS did not alter the number of tire failure and tire inflation-related fatalities. Using the model-year specific trends specification coefficient of 4.5 and recognizing that this is a reduction in deaths over 3 years, suggests 1.5 fewer deaths per year per the 449 models included in the 3-year sample. Therefore, assuming the TPMS reduces all types of vehicle fatalities, the introduction of TPMS is associated with about 651 fewer deaths per year (or 1.45*449). This estimate is similar to the 663 deaths estimated when using the five-year window.

Table 10. Change in deaths by model with introduction of TPMS (with first 3 years of operation)

Standard errors in parentheses.

*** p < 0.01.

** p < 0.05.

* p < 0.1.

Table 11 repeats the exercise above with the 10 years or less sample. When including a model year-specific time trend, I find 23 fewer fatalities overall per model and 0.26 fewer tire failure-related fatalities per model. Given the 10-year time frame, this results in 2.3 fewer deaths per year and model and 0.026 fewer tire failure-related fatalities per year and model. Because there are 494 models, TPMS is associated with 1136 fewer vehicle fatalities of all types per year. This estimate is almost twice the size of the 663 fewer deaths found when using the 5-year window. If TPMS is responsible for reducing all types of fatalities, then the net benefit to society would be between $2.1 and 43.2 trillion ($2001) per year (without maintenance costs). However, if we focus on tire failure-related fatalities alone, the 12.8 fewer deaths per year would result in a net loss of $746 million to $1.9 trillion (2001$) per year (without maintenance costs). This loss is similar to the loss of $752 million to $1.9 trillion (2001$) per year found when using the five-year time frame.

Table 11. Change in deaths by model with introduction of TPMS (with first 10 years of operation)

Standard errors in parentheses.

*** p < 0.01.

** p < 0.05.

* p < 0.1.