H.1.3 Rolling Work and Sleep Schedules

H.1.3.2  Results of Rolling Work and Sleep Schedule Analysis<

The total change in time of day between the beginning and end of a working week schedule at which an OTR driver begins his or her sleeping break period was calculated. For example, if a driver begins a route at 8 am on the first day of the route and ends at 4 am on the last day of the route before an extended (multiple-day) break, the schedule is calculated to have rolled backwards by four hours. The threshold at which a schedule qualifies as having work-sleep cycles that rolled is a difference of at least two hours over a route for a backward-rolling schedule and three hours over a route for a forward-rolling schedule. Preliminary analysis suggested that less than two-hour change in sleeping time over a driving route was insufficient to result in a significant change in modeled schedules. Because initial tests of the model indicated that a schedule that rolls forward adds about two-thirds of the incremental crash probability as a schedule rolling backwards, the forward-rolling threshold was set to an equivalent to three hours. The likelihood of rolling was analyzed separately for the pre-2003 rules relative to full compliance and the 2003 rules, and for regional and LH long-haul; generally >150 mi. from base for property carriers operations. The number of hours a schedule rolls was also measured separately in order to down-weight schedules that roll for fewer hours than modeled.

For the current rule fully enforced, the analysis found three of 13 regional schedules and seven of 11 LH long-haul; generally >150 mi. from base for property carriers to shift. The majority of these were rolling backwards. The regional schedules rolled on average two hours and the LH long-haul; generally >150 mi. from base for property carriers over ten hours over the driving period. These driving periods varied in their length depending on the limits for each proposal and trip lengths for the driver types. FMCSA Federal Motor Carrier Safety Administration (The agency proposing the EOBR Electronic on-Board Recorder (A device attached to commercial motor vehicles that tracks the number of hours drivers spend on the road) rule) assumed that the proportion of OTR drivers is split evenly between regional and LH long-haul; generally >150 mi. from base for property carriers companies and combine the calculated proportion of regional and LH long-haul; generally >150 mi. from base for property carriers drivers whose schedules roll into an overall weighted average. That is, even though it is not expected that all of the drivers with sleeping times that roll to shift by a total of 10 hours, the modeling uses a weighted proportion that would be equivalent to the proportion whose sleep periods would shift backwards by ten hours. For the current rule fully enforced, the result is a weighted average of 34 percent of drivers rolling backwards an average of 10 hours (given a five-day route). For the 2003 rules, two of eight regional schedules and seven of ten LH long-haul; generally >150 mi. from base for property carriers roll backwards.⁶⁰

Table 48: Generation of Proportion of Schedules that Roll

This section discusses the issues related to the truck driver labor supply and the methodology used for the labor cost changes. Compliance with HOS Hours of service (Regulations issued by FMCSA that limit the number of daily and weekly hours a CMV driver may drive) rules are expected to result in changes in labor productivity for truck drivers, leading to changes in driver labor demand. The analysis uses the changes in labor productivity obtained from simulations of trucking routes for the various options and translates them to dollar impacts based on the labor supply relationships for truck drivers.

Issues related to the truck driver wage equation, as a function of job and employee characteristics are discussed first, followed by a discussion of the labor supply elasticity for truck drivers. Components of the indirect labor costs that are associated with driver wage costs are also analyzed to complete the discussion on various aspects of the labor cost changes.

H.2.1.1 Estimate of Driver Wage Function

To analyze the labor costs of the different options, the Agency examined the relationship between hours worked and wages earned for truck drivers. The issue of individual driver wages is important because it is one dimension of the cost of the HOS Hours of service (Regulations issued by FMCSA that limit the number of daily and weekly hours a CMV driver may drive) regulations. As hours of work are shifted from drivers who currently work very long hours to newly hired drivers, the cost implications for carriers depends on the employment costs per additional hour of work by existing drivers relative to the costs for new hires.

H.2.1.1.1 Data, Methodology, and Results

The primary data source for the analyses carried out in the following section is the Current Population Survey (CPS), a household based survey conducted by the BLS every month. This 2003 HOS Hours of service (Regulations issued by FMCSA that limit the number of daily and weekly hours a CMV driver may drive) RIA Regulatory Impact Analysis uses annual CPS data compiled by the National Bureau of Economic Research (NBER) that give earnings and hours worked for a randomly chosen sample. FMCSA Federal Motor Carrier Safety Administration (The agency proposing the EOBR Electronic on-Board Recorder (A device attached to commercial motor vehicles that tracks the number of hours drivers spend on the road) rule) combine annual data from 1995 to 2000 and model the wage equation for non-union truck drivers only.

The wage equation was estimated for truck drivers based on their demographic information and job characteristics. It was hypothesized that the wage earned by truck drivers depends on their hours worked⁶¹, along with their occupational experience, and dummy variables to capture whether they are high school graduates, married, sex, race, whether they are in the for-hire industry, as well as dummies to control for year and regional effects. The details of all the variables used in the regression, including descriptive statistics and the estimated coefficients are presented in table 49.

Table 49: Regression Results and Descriptive Statistics for Truck Driver Wage Relationship – 1995-2000

Notes: t-statistics in parentheses

Number of observations = 11,017

Regression model includes region and year control dummies

All the variables of interest have the expected signs and most are statistically significant (except for some of the region and year control dummies)⁶². Of particular importance are the coefficients associated with the two hours worked variables and the distribution of wages implied.

Table 50 presents predicted wages for different levels of weekly hours worked for the sample on non-union truck drivers only. Based on the total wage relationship, the model predicts that the average 50 hour/week driver earns $28,307, a 60 hour/week driver makes $33,588 and a 70 hour/week driver makes $38,022 annually.

Table 50: Predicted Annual Wages for Different Hours/Week

Chart 3 shows the implied relationship for the total annual wages and Chart 4 shows the average and marginal wages as a function of the hours worked.

Chart 3: Total Wage Curve for Non-Union Drivers

Chart 4: Marginal and Average Wage Curves for Non-union Drivers

The results indicate that the total annual wages for drivers is an increasing function of hours worked that increases at a decreasing rate. This implies that the marginal cost to the firm of an additional hour of driver labor diminishes constantly as the hours of work increases. The specific shape of the total wage curve also ensures that the average wage curve is below the marginal wage for a significant part of the distribution.

The downward slope of the marginal cost curve implies that as the number of hours worked by drivers under the HOS Hours of service (Regulations issued by FMCSA that limit the number of daily and weekly hours a CMV driver may drive) rules are curtailed, the cost savings to companies from cutting down hours of service Regulations issued by FMCSA that limit the number of daily and weekly hours a CMV driver may drive from drivers is less than the increase in cost due to the hiring of new drivers, that is, giving an hour of work to a new driver costs more than the savings from taking an hour of work away from an over utilized driver who would exceed the HOS Hours of service (Regulations issued by FMCSA that limit the number of daily and weekly hours a CMV driver may drive) limits. Another implication of the slope of the marginal wage curve is that every hour of driver labor does not cost the same for the trucking company. This is because of the “non-standard” labor-leisure choice faced by truck drivers. While they are on the road, they are willing to work an extra hour for a lower marginal wage (and cost to the firm), to maximize their earnings potential, in part because the value of leisure time out on the road is low. As drivers work more and more hours, the shape of the marginal wage curve implies that the cost to the firm for the extra hours declines gradually.

H.2.1.2 Supply of Drivers

Another aspect of the labor costs of the different HOS Hours of service (Regulations issued by FMCSA that limit the number of daily and weekly hours a CMV driver may drive) options is related to the issue of the labor supply curve in the market for truck drivers. The shape of the labor supply curve determines the impact that changes in labor demand would have on the wage rates for truck drivers, and this is expressed as the elasticity of labor supply. As discussed, there is evidence in the literature to suggest that trucking is a very competitive industry with relatively free entry and exit and that its market labor supply curve is quite elastic. This is because trucking is considered a low-skill job with relatively low fixed costs. A small change in labor demand from additional drivers needed for full compliance with the HOS Hours of service (Regulations issued by FMCSA that limit the number of daily and weekly hours a CMV driver may drive) rules will not lead to any substantial changes in wage rates.

H.2.1.2.1 Previous Studies

Rose (1987) contends that the truck driver labor supply curve, especially for the non-union TL sector should be highly elastic. This is because “truck driving is a low-skill occupation with considerable turnover.” She also discusses the fact that there is a large pool of drivers outside of the regulated interstate trucking industry who perform the same type of job – owner-operators, private carriage drivers and delivery drivers. She argues thus that the labor supply curve should be highly elastic for this occupation as a whole. Hirsch (1988) also makes the same argument that truck driver labor supply is likely to be highly elastic, given the fact that it is considered a low-skilled job. Engel⁶³ (1998) argues that the high turnover rate in trucking, especially in the TL sector, indicates that this occupation has a highly elastic labor supply curve and provides an easy entry to new truck drivers. The author further argues that these high rates of turnover also indicate that trucking is a job that is difficult to perform over extended periods. There is evidence in the literature that the trucking industry, especially the TL sector, suffers from significant driver shortage. The details about how this can potentially impact the issue of market labor supply are discussed below.

Other studies that looked at the issue of labor supply elasticity in general (not for trucking only) have come up with estimates ranging between 2 and 5. (See for e.g., Lettau (1994), Eberts and Stone (1992)). These studies introduce a spatial dimension to the analysis by looking at local labor markets and therefore are not directly comparable to the analysis here. Nevertheless, these estimates provide a “benchmark” for labor supply elasticity values. Lettau⁶⁴, for example, argues that empirical studies that look at local area labor demand-labor supply relationships, find that “an area’s elasticity of labor supply is between 2.0 and 5.0.” Eberts and Stone⁶⁵ use a recursive model to identify the labor supply and demand relationships in local labor markets using CPS data. They find a labor supply elasticity of 4.9 using a five-period lag structure.

H.2.1.2.2 Evidence from Industry Data

Analysis of historical employment data on truck drivers confirms the view held by experts on this industry that the market labor supply for truck drivers is relatively elastic. Table 51 shows the pattern of employment and annual earnings of truck drivers in the economy from 1983 to 2000, based on CPS data. Although driver employment has grown close to a million from 1983 to 2000 (a growth rate of about 40 percent), growth in real wages for drivers has not been nearly that dramatic. In fact, real wages, in 2000 dollars, have grown less than one-half of a percent during the same period.

Table 51: Economy-Wide Truck Driver Employment and Real Wage Levels

These data support the idea that employment growth for this occupation has not been “significantly impeded by wage movements”.

Wage elasticity for new drivers should ideally be considered in the context of potential truck drivers. Since hiring new drivers would mean shifting or attracting workers from other competing sectors to trucking, the Agency analyzed the existing labor pool for blue-collar workers to see where the new drivers could come from. Table 52 gives the number of blue-collar workers in some of the industries that could supply additional truck drivers needed to comply with the new rules.

Table 52: Employment Levels for Blue-Collar Occupations (thousands)

Source: Current Population Survey

The data in the above table indicate that truck drivers account for about 5 to 6 percent of a total blue-collar population.⁶⁶ Most of the occupations listed above can be considered similar to truck driving in terms of attracting people. This is one reason to believe that there is a large labor pool of blue-collar workers from which to attract potential new truck drivers as a result of the HOS Hours of service (Regulations issued by FMCSA that limit the number of daily and weekly hours a CMV driver may drive) options. This fact, coupled with the historical trends on wage movements, suggests that changes in labor demand would not lead to substantial wage effects due to the high labor supply elasticity.

Some analysts believe the high turnover rates in this industry are not driven by a shortage of drivers. According to a study done by the Upper Great Plains Transportation Institute (UGPTI)⁶⁷ in 1990, and quoted in a report on “What Matters to Drivers”⁶⁸ published in 1997, the trucking industry does not suffer from a “shortage of drivers” to hire from. The study claims the fact that this industry has been able to sustain such high driver turnover rates over the years is an indication that the problem is not one of labor shortage, but a lack of human resource strategy to take advantage of the available driver pool.

Also, the truck driver population in the U.S. under the current conditions is predominantly middle-aged white males. The average age of drivers in the CPS sample is 39 years (for both males and females), with 96 percent of the population being male. However, according to a study done by The Gallup Organization⁶⁹, females, non-whites (or minorities) and those that have less than 15 years of experience most likely see trucking as a good occupational choice. There is a growing segment of the labor force that has remained untapped to increase the pool of drivers. Improving the working conditions of drivers and making their job characteristics consistent with other competing occupations would be one way to attract this previously unused portion of the labor force.

Another issue that is related to this and could have a potential impact on labor supply is that of turnover. Evidence suggests that this industry, particularly the TL sector, has been plagued by very high rates of turnover.⁷⁰

The Gallup study argues that between 1994 and 2005, the industry will need to hire an additional 403,000 drivers/year (even before new HOS). Of these, about 320,000 (or 80 percent) would be because of “churning” or internal turnover, drivers leaving one company to go to another, because of their dissatisfaction with the present job and pay. Another 34,000 (or 8 percent) would be needed to account for growth in the industry. And the remaining 48,000 (or 12 percent) would be needed because of attrition, retirement and external turnover.

The study also notes five specific job attributes that can predict overall job satisfaction for truck drivers:⁷¹

Any improvement in the work schedules of drivers that makes it comparable to other competing occupations could reduce driver turnover. Pay structure is also important. Since pay is also listed as an important reason for the lack of satisfaction, any changes in the rules that results in a reduction in pay for drivers could increase driver turnover. Thus the net effect of better work schedules and lower overall wages is unknown.

Based on the issues discussed above and the evidence from previous literature and data on truck driver labor supply, the Agency assume a labor supply elasticity of 5 to measure the impact on wages as a result of a change in demand for drivers. An elasticity of 5 is consistent with the view held by industry analysts that trucking is a fairly low-skill, easy entry job. Although there seem to be very limited research on truck driver’s market labor supply models that are directly relevant for the 2003 HOS Hours of service (Regulations issued by FMCSA that limit the number of daily and weekly hours a CMV driver may drive) analysis, an elasticity measure of 5 is reasonable. Changes in labor demand due to compliance with HOS Hours of service (Regulations issued by FMCSA that limit the number of daily and weekly hours a CMV driver may drive) rules should not have large impacts on wage rates.

Using the methodology discussed above, labor costs of compliance with HOS Hours of service (Regulations issued by FMCSA that limit the number of daily and weekly hours a CMV driver may drive) rules are calculated separately for LH long-haul; generally >150 mi. from base for property carriers and SH. Using the labor productivity changes for these segments as inputs, FMCSA Federal Motor Carrier Safety Administration (The agency proposing the EOBR Electronic on-Board Recorder (A device attached to commercial motor vehicles that tracks the number of hours drivers spend on the road) rule) calculated the changes in the driver population that would be needed to maintain the same VMT. Then, using the relationships derived from the labor supply curve for individual truck drivers, as well as for the market labor supply, the avoided (from reducing hours of over utilized drivers) and new (from giving hours to new drivers) labor costs and the overhead.

Table 53: Summary of Labor Assumptions

A change in the number of drivers required to conduct trucking activity requires a complementary change in the fleet A group of motor vehicles owned or leased by businesses or government agencies size and supporting infrastructure. FMCSA Federal Motor Carrier Safety Administration (The agency proposing the EOBR Electronic on-Board Recorder (A device attached to commercial motor vehicles that tracks the number of hours drivers spend on the road) rule) identified methods for estimating the cost of new power units, trailers, and parking spaces, and maintenance and insurance costs of this equipment based on a review of the literature and relevant databases, as well as on the conclusions of its industry experts. It was assumed that no new docking facilities or change in mileage-based costs occurs since no direct change in the number of deliveries or in VMT is assumed, that is the amount of total work conducted by the trucking industry is held as fixed.

The method used to estimate the change in the number of tractors and trailers incorporates two countervailing impacts from the change in labor productivity. The first is the obvious change in the number of trucks associated with the incremental change in the number of drivers. The second has to do with the fact that a change in the number of tractors and trailers, under an assumption of no direct change in overall fleet A group of motor vehicles owned or leased by businesses or government agencies VMT, changes the life of the entire fleet A group of motor vehicles owned or leased by businesses or government agencies of tractors and trailers, including the life of newly purchased trucks. For example, consider the case of lower labor productivity requiring more drivers and, hence, tractors and trailers. The existing fleet A group of motor vehicles owned or leased by businesses or government agencies is now driving less to maintain the same VMT, meaning that the average life of each truck is longer. On average, this will translate into lower vehicle replacement based on the change in the number of trucks relative to the initial fleet.

For purposes of illustration, the preceding example will be worked through a representative case in which 10,000 new drivers are hired by trucking companies in response to lower labor productivity, where the initial fleet A group of motor vehicles owned or leased by businesses or government agencies size is 1,500,000 drivers. Table 54 summarizes the key assumptions used in the analysis. FMCSA Federal Motor Carrier Safety Administration (The agency proposing the EOBR Electronic on-Board Recorder (A device attached to commercial motor vehicles that tracks the number of hours drivers spend on the road) rule) assessed and incorporated a number of assumptions made by National Economic Research Associates (NERA)⁷² (2001).

The change in motor vehicle expenditures has two implications for the economy that are estimated for purposes of conducting the regional economic analysis. First, the direct expenditures on equipment are estimated based on the starting year of the policy, the anticipated reaction time by trucking firms, and the anticipated life of assets adjusted for the change in fleet A group of motor vehicles owned or leased by businesses or government agencies size. Second, the assumption is made that firms will not simply bear the swings in capital costs in each year. Instead, firms will finance the costs over the amortization schedule at a reasonable weighted-average cost of capital.

The money to cover these transactions is assumed to come from personal consumption in the regional economic framework. In year 1 of an amortization schedule, consumers provide the cash to cover the change in purchases of new vehicles, net of the first year’s principal the amount that is borrowed, not counting the cost of interest or fees and interest payments. In subsequent years, consumers receive the remaining principal the amount that is borrowed, not counting the cost of interest or fees and interest payments associated with the original loan in the first year. The principal the amount that is borrowed, not counting the cost of interest or fees and interest costs represent an increase in the production costs facing trucking-related sectors in the economy.

Table 54: Assumptions Used to Model Motor Vehicle Equipment Costs

*Pre-tax revenue requirement based on weighted-average cost of capital assuming a debt-to-equity ratio of 1 (50% debt/50% equity), a pre-tax bond rate of 8%, and a 20% pre-tax equity rate

Returning to the example, the steps involved in estimating the changes in the demand for motor vehicle equipment are as follows (t is used to represent time by year and MM for millions):

A change in the number of tractor-trailer sets will require that additional parking spaces be available at terminals. The construction and maintenance of new parking spaces requires both an up-front capital expenditure in the first year followed by annual maintenance costs in subsequent years. The capital expenditures will be capitalized and amortized as a cost to the trucking sector with financing assumed to be substituted for personal consumption as with equipment expenditures. Unlike trucks, no off-setting change in the life of existing parking spaces occurs, because the life of a parking space is expected to be longer than the 10-year horizon under consideration in this analysis.

The assumptions used in the analysis are documented in table 55. Information on parking space requirements for tractor-trailer sets was taken from the National Association of Truck Stop Owners (NATSO) and for auto parking spaces from International Parking Institute (IPI). Maintenance costs for auto spaces were assumed to occur at the same ratio as truck space maintenance to capital cost.

Not all new tractor-trailer sets will be at the terminal at any given point in time, where Terminal_Max summarizes the proportion of additional spaces to sets. In addition to parking spaces for new tractor-trailer sets, additional spaces must be constructed for new drivers to park at truck stops, rest areas, or terminals while en route. The ratio of new drivers parking at work to new tractor-trailer sets is accounted for in variable Terminal_TD. The installation and maintenance costs are estimated based on the following steps.

Table 55: Assumptions Used to Model Parking Space Construction & Maintenance Costs

*Pre-tax revenue requirement based on weighted-average cost of capital assuming a debt-to-equity ratio of 1 (50% debt/50% equity), a pre-tax bond rate of 8%, and a 20% pre-tax equity rate

Additional tractor-trailer sets have value whether they are on the road or not. Though incremental insurance costs are predominantly associated with changes in the VMT, a portion of the insurance cost is associated with the intrinsic value of the change in the capital stock represented by the change in the number of tractor-trailer sets. NERA estimated a value of $2,549 per new driver per year in insurance costs based on ATA American Trucking Associations data, and the 2003 HOS Hours of service (Regulations issued by FMCSA that limit the number of daily and weekly hours a CMV driver may drive) rule analysis used this estimate.⁷⁸ Industry experts estimated that perhaps 25% of this cost is associated with the intrinsic value of the truck, and is therefore a fixed cost per CMV. The remainder is assumed to be variable with changes in VMT. FMCSA Federal Motor Carrier Safety Administration (The agency proposing the EOBR Electronic on-Board Recorder (A device attached to commercial motor vehicles that tracks the number of hours drivers spend on the road) rule) assumed that no direct change will occur in the variable portion of insurance costs since overall VMT is assumed to remain the same. The end result is a change of $637.25 per driver per year, or, $6.3725 million per year.

Analogous to the issue of insurance, additional tractor-trailer sets require some regular maintenance whether they are on the road or not. Though incremental maintenance costs are predominantly associated with changes in the VMT, a portion of the maintenance cost is associated with regular safety inspections and other routine, scheduled maintenance represented by the change in the number of tractor-trailer sets. Industry experts estimated a value of $8,500 per new tractor-trailer per year in maintenance costs and that perhaps 25% of this cost is associated with fixed maintenance costs per truck. The remainder is assumed to be variable with changes in VMT. The portion of the maintenance cost associated with new trucks is negated by changes in VMT in the rest of the fleet A group of motor vehicles owned or leased by businesses or government agencies to yield no direct net change in VMT, and, hence, no change in the variable portion of maintenance costs. The end result is a change of $2,125 per truck per year, or $21.25 million per year.

The need for more or fewer drivers will have an impact on recruitment costs associated with the hiring of new drivers. Rodriquez, et al. (1998),⁷⁹ surveyed 15 LH, for-hire trucking firms to determine the average costs associated with driver turnover, or churn. The study estimated an average cost to firms of $5,423, as summarized by cost category in the first three columns of table 56. The Agency excluded the costs and profits from idle equipment and the production loss due to churn since equipment costs are explicitly modeled and VMT is assumed to remain constant. It also assumed that 75% of the advertising and staff labor costs are based on fixed annual budgets, where only 25% is variable based on the change in number of new drivers to be recruited. Firms are likely to have exhausted the most important marketing channels to them given the high industry churn rate. Administrative support (Staff Labor) is assumed to be characterized by a high degree of automation, with the 25% assumption used to cover the fact that some back-office recruiting labor was included in this category by Rodriquez, et al. The rest of the costs are assumed to be fully variable with a change in the number of new drivers. The result is an incremental cost of $1,610 per hire as compared to the average of $5,423 based directly on the survey results.

Table 56: Cost of Driver Churn

*Based on survey average of 1,311 hires per firm from Rodriquez, et al. (1998).

The cost per driver is multiplied by the change in the number of drivers in the first year. In subsequent years, the cost per hire is multiplied by the change in the number of new drivers times the assumed churn rate for drivers. Estimates of churn rates vary from 25% to over 100% depending on the survey. The 2003 HOS Hours of service (Regulations issued by FMCSA that limit the number of daily and weekly hours a CMV driver may drive) RIA Regulatory Impact Analysis employed the churn rate of 25% assumed by NERA (2001).⁸⁰ The cost in Year 1 is $1,610 * 10,000 drivers, or $16.10 million. The cost in subsequent years is equal to $16.10 times the churn rate of 25%, or $4.02 million per year.

This section estimates the safety benefits from reducing fatigue in over-utilized drivers by bringing them into minimal compliance with the HOS Hours of service (Regulations issued by FMCSA that limit the number of daily and weekly hours a CMV driver may drive) rules.

This review draws on existing literature to describe the function of sleep and the established relationship between sleep, fatigue, shift work and performance, and CMV Commercial Motor Vechicles accidents. For the purposes of this review, fatigue is defined as the decreased ability to perform induced by a lack of adequate sleep, approximately 8 hours per 24-hour period. The review does not intend to duplicate other major literature reviews performed on the subject of sleep, fatigue and truck-involved accidents.⁸¹ Rather, it is a targeted summary of key issues related to the analysis conducted for this study.

Sleep is an integral part of human functioning and longevity. Not getting enough sleep leads to drowsiness and impaired concentration for the subsequent non-sleep period.⁸² Too little sleep also leads to impaired memory and physical performance and reduced ability to perform on cognitive tasks. Without sleep, neurons may become so depleted in energy or so polluted with byproducts of normal cellular activity that they begin to malfunction. During sleep, the body also increases its protein production, enabling the repair of damaged cells, damaged from such external elements as ultraviolet rays or stress. Sleep is crucial to this process of cell repair and to the promotion of uncompromised performance during non-sleep periods.

Until the 1950s, sleep was regarded as a dormant, passive part of daily life. After this time, however, sleep became to be recognized as a dynamic process with multiple states of brain activity. There are five different stages of sleep, as measured through brain activity. These five stages are:

The brain passes through stages 1, 2, 3 and 4 (non-REM sleep) and then into stage 5 (REM sleep). The time spent in each stage varies depending on the stage and depending on the number of sleep cycles (progression through stages 1, 2, 3, 4 and REM sleep) completed in the sleep period. Approximately 50 percent of total sleep time is spent in stage 2, 20 percent in REM sleep and 30 percent in the remaining stages.

Stage 1 is the shortest phase, comprised of drifting in and out of sleep. People are easily awakened during this phase. This is also the phase where one experiences “hypnic myoclonia” where the sensation of falling is often felt. In stage 1, the body has slow muscle activity and eye movement. In stage 2, the stage in which most sleep time is spent, eye movement stops. The brain’s electrical activity decreases and short bursts of rapid brain waves occasionally appear. Stage 2 is the first stage of the non-REM sleep stages. Stages 3 and 4 (called deep sleep) are also non-REM sleep stages and are characterized by very slow brain waves.

REM sleep is the next stage. This is often referred to as the “active” sleep stage. The slowed brain waves begin to accelerate, breathing becomes more rapid, irregular and shallow, eye movement begins, heart rate increases and blood pressure rises. This lighter stage of sleep is where most dreaming occurs. As the sleep period progresses, the REM sleep stage increases in length where towards the end of the sleep period, an individual may spend up to one hour in REM sleep and experience very involved dreams.

The complete sleep cycle usually takes 90 to 110 minutes. The first sleep cycles of each sleep period contain relatively short REM periods and long periods of deep sleep. As the night progresses, REM sleep periods increase in length while deep sleep (stages 3 and 4) decreases. By the end of a “normal” sleep period (defined here as approximately 8 hours), almost all sleep is spent in Stage 2 and REM sleep.

Most sleep experts agree that adults need between six and ten hours of sleep per 24-hour period, with most people requiring approximately 8 hours of sleep per day.⁸³ Sleep most naturally occurs at night, due to the human body’s circadian rhythm. Circadian, or daily, rhythms operate on approximately a 24-hour cycle and are responsible for natural peaks and lulls in hormonal secretions, a heightened sense of fatigue during different parts of the day – particularly in the early morning hours and the late afternoon – and the coordination and timing of other internal bodily functions, including body temperature and sleep. Sunlight and other time cues help to set and maintain circadian cycles.

Body temperature fluctuates in accordance with other bodily fluctuations of the circadian cycle and influences the timing of sleep and sleep onset. During a single day, the body’s temperature rises and falls a number of times. Body temperature rises in the early morning hours, declines in the late afternoon, rises in the evening and declines later at night. People prefer to go to bed during certain phases in the temperature cycle over others, preferring phases when the circadian temperature cycle is at the nadir (lowest point).⁸⁴ When body temperature is on the rise, the body has a greater propensity to awaken.⁸⁵ Body temperature is on the rise in the morning hours, when people on regular night sleep schedules tend to wake up. It logically follows that it is more difficult to fall asleep during these morning hours (because body temperature is rising, not falling).

The sleep/wake cycle shows that the degree of sleepiness depends on the oscillating circadian rhythm and declining linear function (increased degree of sleepiness) based on the length of time spent awake.⁸⁶

The timing of sleep matters. Sleep duration is greatest after evening bedtimes and shortest after morning bedtimes.⁸⁷ The duration of sleep has also been found to be shorter the later in the morning sleep begins.⁸⁸ The shorter sleep duration after a morning bedtime might seem somewhat counterintuitive as a morning bedtime is often the result of sleep postponed (i.e. longer period elapsed since last period of sleep). However, this decrease can be explained by the strong influence of the circadian rhythm on sleep duration, which makes it more difficult to sleep during daytime hours than it is during nighttime hours.

A number of studies have shown that duration of sleep is influenced by the time of day of sleep. A survey found that night workers, who by default must sleep in some part, or entirely, during the day, slept three hours less than the recommended eight hours required to prevent sleep debt.⁸⁹ Another study found that night workers (shift starting around 2200 or 0000) slept on average 3.3 hours less than their day-working counterparts (shift starting around 0800), sleeping 4.3 hours and 7.6 hours respectively.⁹⁰ These data demonstrate that people who rely on daytime sleep for a significant part of their rest are experiencing less total sleep.

Sleep deprivation occurs when an individual sleeps two or more hours less than the optimal amount during any one sleep episode, eight hours being the standard optimal amount subject to significant variation by individual. Sleep deprivation over a series of sleep periods leads to sleep debt, the accumulated sleep loss over the course of time.⁹¹ The discussion in this section presents the results of a number of studies and the implications of sleep deprivation and sleep debt on performance based on the available literature.

Sleep deprivation and sleep debt have a number of consequences for performance. Sleep deprivation over a couple of days leads to slower response times and decreased initiative.⁹² After one sleepless night, cognitive performance may decrease 25% as compared to the performance of non-sleep deprived individuals.⁹³ After the second sleepless night, performance on cognitive tasks may decrease to nearly 40% the potential level. A meta-analysis found that people who are chronically sleep deprived, that is, have substantial sleep debt, performed at the 9th percentile of non-sleep-deprived subjects.⁹⁴

Individuals switching from an irregular to regular schedule do not immediately achieve improved fatigue levels. Sleep deprived individuals with irregular sleep schedules (as could be the case with truck drivers) who regularized their sleep schedules but suffered sleep loss in the process experienced an increase in daytime sleepiness and a concomitant deterioration in concentration ratings, immediately after regularizing their sleep schedule.⁹⁵

These findings suggest that routine sleep schedules that allow the individual sleeping an adequate number of hours (approximately 8, varying by individual) during approximately the same time during a 24-hour period facilitate daily functioning at unimpaired performance levels.

Workers experience a number of different types of fatigue while on the job. The three major types of fatigue affecting work performance are industrial, cumulative and circadian.⁹⁶ These types of fatigue are described below, focusing on the literature relating to truck drivers.

Industrial fatigue results from working continuously over an extended period of time without proper rest, often referred to in the literature as fatigue resulting from time-on-task. For example, a truck driver who has been driving for six hours, without a break, might be subject to industrial fatigue. Some studies have shown performance to decrease as time on task increases.⁹⁷ Time-on-task problems could be exacerbated by sleep loss, even in the early stages of the task. One study concluded that for sleep deprived individuals, performance is compromised even at early stages of performance of a monotonous task if the situation is undemanding and boring. This study suggests that the effect of sleepiness becomes immediately evident in the form of reduced vigilance.^98,99

Cumulative fatigue arises from working for too many days on any protracted, repetitive task without any prolonged break. This fatigue results from a lack of alertness brought on by familiarity and boredom with the task at hand. A truck driver could experience cumulative fatigue after driving for 12 hours, taking eight hours off and then driving another 12 hours (driving a total of 24 hours in a 32 hour period).

Circadian fatigue is a function of the circadian rhythm. Fatigue is greatest when approaching or at the nadir of the circadian cycle, where the body is least vigilant. The truck accident rate is much higher during the early morning hours than during any other time of day,¹⁰⁰ supporting the circadian effect hypothesis that accidents are more likely to occur when the human body is least vigilant.¹⁰¹

Night and rotating shift workers are especially susceptible to being fatigued on the job.^102,103,104 Permanently assigned graveyard-shift workers sleep between 5.8 to 6.4 hours per day.¹⁰⁵ Rotating shift workers, such as many truck drivers, sleep even less when they work a night shift (5.25 to 5.5 hours). Shift workers experience disturbances in their circadian rhythm, as measured by changes in hormonal levels;¹⁰⁶ they are also less alert during nighttime shifts and perform less well on reasoning and non-stimulating tasks than non-shift workers.¹⁰⁷ Though nightshift work for many workers is regular (the same schedule is kept over time), truck drivers often have irregular schedules which can amplify the effects of circadian, cumulative and industrial fatigue and increase the risk of fatigue-related accidents.

Fatigue increases over the duration of trips, regardless of the driving schedule¹⁰⁸ and total driving time has a significant effect on crash risk though there is variation on the point at which crash risk increases significantly, depending on the study methodology.^109,110 A study of industrial fatigue in truck drivers found that in over 65% of cases, truck accidents took place during the second half of a trip, regardless of trip length.¹¹¹ An analysis of Bureau of Motor Carrier Safety data in the 1970s found that about twice as many accidents occurred during the second half of trips than during the first half, regardless of trip duration.¹¹² Another study found that the risk of accident increased after the fourth hour of driving and peaked after nine hours of driving.¹¹³ These studies are among many finding that industrial fatigue plays a role in predisposing truck drivers to accidents. Determining the magnitude of this effect, however, and ensuring that other factors (such as sleep history and time of day) have been factored out, is quite difficult.

Researchers have long asked how long a person can sustain work effort at different tasks without lengthy breaks, before his/her performance of those tasks becomes unacceptably degraded. There has always been a notion that by itself, sustained performance at a task (Time on Task or TOT) eventually results in a “fatiguing effect” manifesting itself in the form of slower response times or errors of omission. Below is a short literature review of five studies about the time-on-task effect on driving and some concluding remarks.

Jones and Stein (1987)¹¹⁴ attempted to provide “adjusted odds ratios” to different categories of “length of time in driving” (TOT), assigning a baseline value of 1.0 to the relative risk of the likelihood of crashes attributable to a driving time of from 0 to 2 hours; and they presented an increased odds ratio of 1.2 for driving times of from 2 to 5 hours and also 5 to 8 hours of driving time (TOT). The work of Jones and Stein says nothing about projecting odds ratios for driving more than 8 hours, something at the root question of the entire discussion of truck driver HOS.

Lin, Jovanis, and Yang (1993)¹¹⁵ introduce a time-dependent logistic regression model formulated to assess the safety of motor carrier A person providing motor vehicle transportation for compensation. The term includes a motor carrier’s agents, officers and employees operations. They describe their model as being flexible, allowing the inclusion of time-independent covariates, time main effects, and time-related interactions. The model estimates the probability of having a crash at time interval t, subject to surviving (not having a crash) before that time interval. Covariates tested in the model in this paper include consecutive driving time, multiday driving pattern over a 7-day period, driver age and experience, and hours off duty before the trip of interest. Although the work of Lin, Jovanis, and Yang has some appeal in the conduct of this study, their methods and modeling are of some concern in that they do not model beyond the 8-9 hours of driving incidents, something which is obviously needed to examine the HOS Hours of service (Regulations issued by FMCSA that limit the number of daily and weekly hours a CMV driver may drive) alternatives.

In their description of nine logistic regression modeling attempts Lin, Jovanis, and Yang state that driving time (TOT) has the strongest direct effect on accident risk. The first 4 hr consistently have the lowest crash risk and are indistinguishable from each other. Accident (crash) risk increases significantly after the fourth hour of driving, by approximately 50% or more, until the seventh hour. The 8th and 9th hours show a further increase, approximately 80% and 130% higher than the first 4 hours.

Campbell (1988)¹¹⁶ states that there is a steady increase in the probability of accident involvement with the number of hours driving. To look into this, Campbell used data from accident reports filed with the Office of Motor Carriers and extracted the time of day that the accident occurred, the number of hours driving at the time of the accident, and the intended driving period had the accident not occurred. The accidents that were coded as the driver having dozed at the time of the accident were used to determine the time-on-task effect. The problem with this is that not all of the crash data was included and crashes may have been caused by fatigue yet the driver was not dozing at the time. It was concluded that the crossover point in which the proportion of accidents in the latter hours of driving is more frequent occurs around four hours of driving.

O’Neill et al. (1999)¹¹⁷ studied the operating practices of CMV Commercial Motor Vechicles drivers, as well as the relationship of these practices to driver fatigue. Drivers worked a 14-hour on, 10 hour off schedule driving a simulator for a 5-day week. Two 30-minute breaks and a 45-minute lunch break were taken during the day at regularly scheduled times. The observed recovery effect of the breaks was rather striking. The effects of 6.5 hours of driving were virtually reduced to the starting levels by a 45-minute break (O’Neil et al., 1999). It is important to keep in mind that while this recovery effect is remarkable, it occurred under very strict, adhered to conditions. This effect took place under daytime driving conditions, the 14 hours on/10 hours off driving schedule that allowed for adequate rest, and scheduled breaks. It cannot be said with a reasonable degree of certainty that this recovery effect would occur in the same way under different conditions.

Wylie et al. (1996) ¹¹⁸studied four different driving conditions to test several driving fatigue questions: a 10-hour “baseline” daytime schedule, a 10-hour “operational” or rotating schedule, a 13-hour nighttime start schedule, and a 13-hour daytime start schedule. The authors concluded that hours-of-driving (TOT) was not a strong or consistent predictor of observed fatigue. Interestingly, there was a positive correlation between driver’s self-ratings of fatigue and the number of hours driving within a trip while objective performance did not indicate a positive correlation.

Based on the literature reviewed, the time-on-task effect was not quantified independent of and in addition to the circadian and recovery/decrement recovery factors. Therefore, the TOT effect was not used as a separate factor in the 2003 HOS Hours of service (Regulations issued by FMCSA that limit the number of daily and weekly hours a CMV driver may drive) rule analysis.

Another important and relevant factor is time of day and continuity of sleep. Numerous studies have found an increased crash risk for truck drivers associated with night-time driving.^119,120 In a study of a group of drivers involved in single-vehicle accidents, almost twice as many of their accidents occurred in the early morning hours between midnight and 0800 hours (66%) as during the rest of the day (34%).¹²¹ Accidents and workplace errors from studies concerning road, maritime and industrial operations show a peak at 0300.¹²² Additionally, the continuity of sleep is significant. An elevated fatal crash risk was identified for drivers that split the required 8 hours off-duty into two sessions in a sleeper berth.¹²³

An arduous work schedule has also been identified as increasing the risk of truck involved accidents.¹²⁴ One study found that drivers on a regular 13-hour daytime-start driving schedule slept for 5.1 hours while drivers on a 10-hour daytime start driving schedule slept 5.4 hours.¹²⁵ While this study only looked at daytime-start schedules, the relationship between time-off duty and time spent asleep is remarkable. Drivers with 11 hours off spent 5.1 hours asleep (and an additional .4 hours in bed) while drivers with 14 hours off spent 5.4 hours asleep (and an additional .4 hours in bed). The study cites the fact that drivers on the 13-hour schedule were within 10 minutes of their sleep laboratory and thus may have been able to get more sleep than otherwise. The sleep numbers for both groups are likely to be high because each were able to obtain their principal the amount that is borrowed, not counting the cost of interest or fees sleep during optimal times of the day (in accordance with the circadian rhythm), starting late in the evening and ending early in the morning. It is possible that given the same schedule durations, these drivers could have slept less if conditions were different (e.g., if the schedule necessitated nighttime driving, if the drivers lived (or the sleep center was) further from their daily terminating point).

Driving requires sustained attention; it is an inherently fatiguing task in its monotony and repetition.¹²⁶ For many commercial motor vehicle Any vehicle owned or used by a business drivers, the inherently fatiguing task of driving is compounded by fatigue caused by working long, irregular hours that conflict with natural circadian rhythms.¹²⁷ Because of the economic incentives for rapid goods transport, many drivers may be unable to obtain sufficient, sustained, restorative sleep and may subsequently experience sleep deprivation or accumulate a sizable sleep debt. Sleep deprivation and sleep debt, as shown through this review, can lead to an increase in the risk of accidents through impaired performance. This fact supports the need to provide CMV Commercial Motor Vechicles drivers with conditions that make it possible and likely for them to get sufficient sleep, though even ideal conditions could not eliminate all fatigue-related crashes.

The National Highway Traffic Safety Administration (NHTSANational Highway Transportation Safety Administration) Fatality Analysis Reporting System (FARSFatality Analysis Reporting System (a database created by the National Highway Transportation Safety Agency to track national data on fatal car crashes)) and General Estimates System (GESGeneral Estimates System (a database created by the National Highway Transportation Safety Administration to identify traffic safety problem areas, provide a basis for regulatory and consumer initiatives, and form the basis for cost and benefit analyses of traffic safety initiatives)) databases and the MCMIS Crash File were reviewed for the years 1997 through 2000. They provided the primary basis for crash estimates. Other databases including the MCMIS Census File, National Motor Carrier Directory (NMCD), and Bluebook were used to categorize crashes by motor carrier A person providing motor vehicle transportation for compensation. The term includes a motor carrier’s agents, officers and employees firm operations so that the resultant crash data could be linked to the industry profile and schedule/risk analyses used to evaluate the potential effects of proposed changes to the hours of service Regulations issued by FMCSA that limit the number of daily and weekly hours a CMV driver may drive regulations.

The crash analysis began with an attempt to extract commercial motor vehicle Any vehicle owned or used by a business crashes from the three crash data files. Key variables included the state, and date of the crashes; vehicle type and configuration; motor carrier A person providing motor vehicle transportation for compensation. The term includes a motor carrier’s agents, officers and employees census number; total vehicles, occupants, injuries and fatalities; and driver, vehicle and environmental factors associated with the crashes. The goal was to be able to establish a profile of carriers/vehicles involved in crashes with particular attention placed on the apparent contributing factors or accident causes. There was an attempt made to eliminate “other driver” and environmental factors leading to the crash. This was done to extract truck crashes where the driver would probably be considered not “at fault” from the overall set of crashes. The key issue was to determine the extent to which CMV Commercial Motor Vechicles driver fatigue or associated factors could be reasonably established as a primary contributing factor in the crash.

In conducting such an analysis, it is essential that one recognize the potential weakness in using police accident reports (PARs) as the sole basis for attributing fatigue as a crash cause. The police officers who complete the reports rarely have specialized training in crash investigation or even in completing the forms. One should also note that completing the PAR is no greater than a third priority for officers who are involved in situation assessment, emergency response and victim assistance, and finally controlling and then restoring traffic flow around the crash scene. Additionally, PARs are believed to under- rather than overestimate fatigue involvement in large truck crashes.

Crashes where CMV Commercial Motor Vechicles driver fatigue is cited as a primary contributing factor should be viewed as the “minimum” number of crashes with fatigue as a cause. For this reason, the analysis was conducted to develop a more reasonable estimate of the total number of fatigue-related crashes. An analysis of data for crashes where driver inattention was cited within the PAR was used to apportion part of those crashes as fatigue-related. This conclusion was drawn from a comprehensive report of SH short-haul: generally, < 150 mi. from base for property carriers drivers that attributed more than 20 percent of all inattention crashes to driver fatigue.

In order to develop estimates of the total cost of truck crashes in recent years, the FARS, GES General Estimates System (a database created by the National Highway Transportation Safety Administration to identify traffic safety problem areas, provide a basis for regulatory and consumer initiatives, and form the basis for cost and benefit analyses of traffic safety initiatives) and MCMIS databases were reviewed to derive national summary totals of crashes by type, fatalities, and injuries. Crashes are defined by whether or not they involve fatalities suffered by vehicle passengers or non-occupants (pedestrians), reported injuries where no fatality was involved, or property damage with no fatalities or injuries (property damage only crashes.). The MCMIS database tends to contain more detailed information about the vehicle configuration or cargo carried and is especially useful for determining the identity of the motor carrier A person providing motor vehicle transportation for compensation. The term includes a motor carrier’s agents, officers and employees involved in a crash. Historically, there has been an undercount of truck crashes noted in the MCMIS database versus FARS. Comparably, there still seems to be a substantial undercount of injury and property damage only (PDO) crashes in MCMIS versus the national estimates derived in the NHTSA National Highway Transportation Safety Administration GES General Estimates System (a database created by the National Highway Transportation Safety Administration to identify traffic safety problem areas, provide a basis for regulatory and consumer initiatives, and form the basis for cost and benefit analyses of traffic safety initiatives) database. Part of the difference is because MCMIS only records PDO crashes that resulted in a vehicle being towed away, a subset of all PDO crashes.

The FARS Fatality Analysis Reporting System (a database created by the National Highway Transportation Safety Agency to track national data on fatal car crashes) database is considered the best source of fatal crash information since it is a census of all fatality involved motor vehicle crashes occurring within the United States. It was used to develop estimates of the total fatal crashes involving trucks, the total fatalities (broken down by truck, other vehicle, or non-occupants) and the numbers of combination and large single unit trucks involved. Data were reported for calendar years 1997 through 2000 and for the average over the four-year period.

National estimates of truck crashes that do not involve a fatal injury were derived from the GES General Estimates System (a database created by the National Highway Transportation Safety Administration to identify traffic safety problem areas, provide a basis for regulatory and consumer initiatives, and form the basis for cost and benefit analyses of traffic safety initiatives) files. Crashes, total injuries and trucks involved were reported for the injury crashes while total crashes and trucks involved were reported for the PDO crashes. The GES General Estimates System (a database created by the National Highway Transportation Safety Administration to identify traffic safety problem areas, provide a basis for regulatory and consumer initiatives, and form the basis for cost and benefit analyses of traffic safety initiatives) estimates are based on a stratified national sample where each crash is assigned a sampling weight according the stratum from which it is reported. The GES General Estimates System (a database created by the National Highway Transportation Safety Administration to identify traffic safety problem areas, provide a basis for regulatory and consumer initiatives, and form the basis for cost and benefit analyses of traffic safety initiatives) estimates are always rounded to the nearest thousand crashes, vehicles, or injuries. These national estimates are provided below.

Table 57: Large Truck Crashes

Source: FARS Fatality Analysis Reporting System (a database created by the National Highway Transportation Safety Agency to track national data on fatal car crashes) and GES General Estimates System (a database created by the National Highway Transportation Safety Administration to identify traffic safety problem areas, provide a basis for regulatory and consumer initiatives, and form the basis for cost and benefit analyses of traffic safety initiatives) databases.

In order to complete the baseline analysis, it was necessary to determine what proportion of truck crashes could be attributed to truck driver fatigue. The MCMIS Crash File for 1997 through 2000 was used. Bus and unknown vehicle type records were eliminated from the database and the “apparent driver condition” variable was used to code the data records for which “fatigue” or “asleep” had been cited as contributing factor in the crash. The “raw” fatigue/asleep crash estimates for 1997 through 1999 was approximately 1.31% of all truck crashes with the number dropping in 2000 to less than 1%. These very low values could seem to indicate a minimal fatigue rate for truck crashes. However, closer examination of the data and direct benchmarking to alternative data sources point to numerous deficiencies in such a simple analysis.

As a matter of policy, the “apparent driver condition” variable has been eliminated from the National Governor’s Association (NGANational Governors Association) required list of reportable data elements for commercial vehicle crash reports in the SAFETYNET 2000 (Version 2) reporting system. This was in large part due to historical under-reported and non-reported values for this variable. In 38% of the 2000 data records, the driver condition variable was missing compared to less than 10% of the time in earlier years. In the earlier years, the data field was reported as “unknown” rather than missing in about 7% of all crash records. A problem arises in that “appeared normal” and “unknown” are both coding options, but for analytic purposes, it is difficult to ascertain whether a blank value for this variable should be interpreted as “normal”, “unknown” or “not interpretable”. Previous estimates of driver fatigue associated with truck crashes have been hampered by this serious data quality problem.

A state-by-state examination of the data also showed several systematic problems in the reporting of the driver condition variable. Driver condition was not reported in truck crash data records from the States of Massachusetts, Oregon, South Carolina or Virginia in any of the years of data examined. Additionally, fatigue/asleep was never reported as a factor in truck crashes in the States of Colorado, Michigan, New Mexico or Wisconsin. These data reporting problems result in a logical inconsistency in calculating fatigue involvement rates for truck crashes. If it is impossible to add a fatigue event in the numerator of the national fatigue crash rate, then the data from these States should not be included in the denominator of the rate calculation. The problem grows worse in more recent years with many more States opting to not report apparent driver condition at all. For these reasons, the MCMIS database should not be used to derive estimates of fatigue-involved truck crashes.

As an alternative to using the MCMIS data, FARS Fatality Analysis Reporting System (a database created by the National Highway Transportation Safety Agency to track national data on fatal car crashes) truck crash data for the years 1997 through 2000 were reviewed. The FARS Fatality Analysis Reporting System (a database created by the National Highway Transportation Safety Agency to track national data on fatal car crashes) database contains information for crashes involving at least one associated fatality in the involved truck, in another vehicle, or a pedestrian. The FARS Fatality Analysis Reporting System (a database created by the National Highway Transportation Safety Agency to track national data on fatal car crashes) database has been used as a benchmark for the MCMIS database fatal crashes since the requirement for motor carrier self-reporting of crashes was ended in the early 1990’s. The FARS Fatality Analysis Reporting System (a database created by the National Highway Transportation Safety Agency to track national data on fatal car crashes) database has historically been held in high regard because of the NHTSA National Highway Transportation Safety Administration protocols for editing and coding the data elements within the data records.

In order to use the FARS Fatality Analysis Reporting System (a database created by the National Highway Transportation Safety Agency to track national data on fatal car crashes) database for analysis of fatigue related truck crashes, several key issues have to be examined. Is it reasonable to extend fatigue–related crash estimates from fatal crashes to injury and property damage only crashes? Since the FARS Fatality Analysis Reporting System (a database created by the National Highway Transportation Safety Agency to track national data on fatal car crashes) database contains data elements for reporting up to four driver factors, how should these multiple responses be handled? Are there data reporting issues for the FARS Fatality Analysis Reporting System (a database created by the National Highway Transportation Safety Agency to track national data on fatal car crashes) dataset comparable to those encountered in the MCMIS data set?

The FARS Fatality Analysis Reporting System (a database created by the National Highway Transportation Safety Agency to track national data on fatal car crashes) database is limited to crashes involving a traffic fatality. By definition, these crashes are more severe since the fatal outcome has a higher social or economic cost than would a comparable crash resulting in (perhaps minor) injuries or damages to property only. Fatal crashes certainly have other characteristics that separate them from injury only or PDO crashes, especially those factors associated with speed and type of impact. In truck crashes, the “other vehicle” occupant is almost six times more likely to die in the crash than a truck occupant, so it is not clear to what extent fatal truck crash characteristics can be reasonably generalized to injury or PDO crashes.

This question is difficult to answer with the data available. A review of MCMIS fatigue involved crashes by crash type reveal that there was an historical trend of fatigue being reported in fatal crashes more than in injury and property damage only crashes. The overall data reporting problems for MCMIS fatigue crash rates also present interpretation problems for this feature of the data. Data for the year 2000 should not be used since the apparent driver condition variable had already stopped being used. However, for 1999 and 1998, fatigue involvement in fatal crashes did exceed that reported in injury crashes or PDO crashes.

Another estimate of the relative prevalence of fatigue in the three types of crashes could be drawn from the GES General Estimates System (a database created by the National Highway Transportation Safety Administration to identify traffic safety problem areas, provide a basis for regulatory and consumer initiatives, and form the basis for cost and benefit analyses of traffic safety initiatives) data. The GES General Estimates System (a database created by the National Highway Transportation Safety Administration to identify traffic safety problem areas, provide a basis for regulatory and consumer initiatives, and form the basis for cost and benefit analyses of traffic safety initiatives) database serves as a very good source of national estimates of total crashes and for crashes with certain characteristics (such as number of occupants or injuries, or by vehicle type). However, some specific details of the crash cannot be estimated very reliably. Derived from state databases of police accident reports, the GES General Estimates System (a database created by the National Highway Transportation Safety Administration to identify traffic safety problem areas, provide a basis for regulatory and consumer initiatives, and form the basis for cost and benefit analyses of traffic safety initiatives) suffers from some of the same faults as the MCMIS. There may be only one (if any) of the driver condition variables included in the reports and fatigue may not always be a coded or reported factor. In the 2000 database, however, fatigue was cited as a contributing factor in 1.46% of all the fatal crashes, 0.94% of the injury only crashes and 0.65% of the PDO crashes. These percentages were drawn from the raw non-weighted sample. Additionally, the data were not edited for missing values or adjusted for any other factors related to the data reporting in the files.

With such small percentages reported in the MCMIS and GES General Estimates System (a database created by the National Highway Transportation Safety Administration to identify traffic safety problem areas, provide a basis for regulatory and consumer initiatives, and form the basis for cost and benefit analyses of traffic safety initiatives) databases, there is some uncertainty in concluding that FARS Fatality Analysis Reporting System (a database created by the National Highway Transportation Safety Agency to track national data on fatal car crashes) data for fatigue involvement in crashes can be extended to the injury and PDO crashes. In real terms, the differences are negligible. In percentage terms the differences could be viewed as substantial. However, the FARS Fatality Analysis Reporting System (a database created by the National Highway Transportation Safety Agency to track national data on fatal car crashes) database contains the most detailed and highest quality data. There is also evidence that the historical underreporting of fatigue involvement in FARS Fatality Analysis Reporting System (a database created by the National Highway Transportation Safety Agency to track national data on fatal car crashes) would tend to provide conservative estimates regardless.

Since the FARS Fatality Analysis Reporting System (a database created by the National Highway Transportation Safety Agency to track national data on fatal car crashes) database contains four different fields for reporting driver factors contributing to the crash, there was some initial concern that this could introduce an upward bias in the reporting of fatigue involvement in crashes. For the years 1997 through 2000 this appears not to be the case. In most all of the cases where fatigue was cited as a factor and there were other factors cited, the fatigue code tended to occur first in the list. Additionally, it was reported with a factor that would not confound the results reported in this analysis. The two most common other factors reported were “ran off road” and “inattention”. These accounted for a large proportion of the multiple factors cases.

Finally, from the standpoint of data quality, some factors ought to be considered when viewing the FARS Fatality Analysis Reporting System (a database created by the National Highway Transportation Safety Agency to track national data on fatal car crashes) data to assure that cases are not included in the denominator of the rate calculations, unreasonably biasing the estimates downward. Examination of the individual data records indicated that there are several sets of crashes that it seems unreasonable to consider in calculating the fatigue involvement rate. For one set of crashes, many of the key variables (vehicle configuration, body type, harmful events, driver charges, impact details, etc.) are coded as “9’s”, which means unknown or unreported. In all of these cases the driver contributing factors are coded as “99”. In another set of data records, the individual crashes can be matched back to the MCMIS file where the contributing factors are missing because of state data reporting systems and procedures. These appear in the FARS Fatality Analysis Reporting System (a database created by the National Highway Transportation Safety Agency to track national data on fatal car crashes) database with driver factors coded as “0’s”. Eliminating these records from the analysis set was a prerequisite to calculating the proportion of crashes with motor carrier A person providing motor vehicle transportation for compensation. The term includes a motor carrier’s agents, officers and employees fatigue as a contributing factor.

The final step in completing this analysis was to examine each fatal crash involving a large truck and use the factors cited to determine whether “fault” was attributed in the crash. Crashes were categorized as whether the “other” (non-truck) driver was determined to be at fault and whether the truck driver was determined to be at fault. Two other values for these variables were also considered and reported below. If inclement weather was cited that is so reported. If no fault was assigned for the crash, that was also reported. One should note that since weather conditions and multiple drivers may interact to provide multiple responsible conditions or persons, the percentages for fault attribution could add up to more than 100%. In all of the years of reported data, that is the case. One important fact to note from these attribution data, is that the “other driver” was deemed to be at fault almost twice as often as the truck driver.

The fatigue attribution for the FARS Fatality Analysis Reporting System (a database created by the National Highway Transportation Safety Agency to track national data on fatal car crashes) crash data is shown in table 58. The edited data provide us with an estimate of between 6.60% (2000) and 8.21% (1999) for the period examined. These fatigue numbers were adjusted further to account for a systematic phenomenon noted by Hanowski in his study of SH short-haul: generally, < 150 mi. from base for property carriers drivers.¹²⁸ In that study, fatigue was determined to be a contributing factor in 20.8 percent of the incidents where the driver was judged to be at fault due to inattention. When at fault, their PERCLOS (percent eyelid closure) values prior to the incidents were significantly higher than for other types of critical incidents.

Table 58: Fatigue-Related Fatal Crashes

Table 58 also shows the proportion of crashes in which inattention was cited as a major contributing factor. The final fatigue figures provided use the total fatigue cited crashes plus 20.8% of the inattention caused crashes to establish the final estimate of crashes that can be reasonably regarded as due to truck driver fatigue.

To estimate the relative involvement of large trucks in crashes by operations type, crash records must contain specific information about the trucks involved. Crashes can be classified by the operations of the vehicles or firms involved using either the actual characteristics of the trip in which a crash occurred or the identity of the motor carrier A person providing motor vehicle transportation for compensation. The term includes a motor carrier’s agents, officers and employees involved, assuming that the firm identifying number can be matched to a data source in which motor carriers are classified by their operations.

The most commonly cited crash databases do not contain trip-specific characteristics. The ideal information for determining whether a crash occurred in long- versus SH short-haul: generally, < 150 mi. from base for property carriers operations would be the starting and intended ending point of the trip in which the crash occurred. The calculated distance between those two points would provide with certainty the ability to classify the trip according to any specified cut-off point selected for differentiating between LH long-haul; generally >150 mi. from base for property carriers and SH short-haul: generally, < 150 mi. from base for property carriers operations.

A very good indicator for determining if a truck crash occurred in SH short-haul: generally, < 150 mi. from base for property carriers operations may be the vehicle or equipment type. Dump trucks, garbage trucks and concrete mixers are rarely involved in LH long-haul; generally >150 mi. from base for property carriers operations. To a lesser extent, cargo tanks are somewhat restricted to SH short-haul: generally, < 150 mi. from base for property carriers use. Flatbed trucks, straight trucks, and vans or enclosed boxes are widely used in both long- and SH short-haul: generally, < 150 mi. from base for property carriers operations. This group comprises a substantial population of the cargo body types noted in truck crashes.

An alternative method for determining the operations of the crash involved truck is to associate the individual vehicle with a firm and then research the primary operations type for the whole firm. The TTS Blue Book database of 2,681 motor carriers provides two key indicators for specifying operations type, and also provides the USDOT motor carrier A person providing motor vehicle transportation for compensation. The term includes a motor carrier’s agents, officers and employees identification number for these carriers. Average length of haul for trips made in the year and the freight revenues from long and short distance transportation are reported in the database. The average length of trip variable can be used directly to classify carriers. A calculated proportion of revenue derived from short distance operations could also be used to classify carriers. There are 2,481 firms in the TTS database with sufficient information to be able to classify them as long or SH. With these data, the motor carrier A person providing motor vehicle transportation for compensation. The term includes a motor carrier’s agents, officers and employees identification number can be used to match firm records to the MCMIS and (to a lesser extent) FARS Fatality Analysis Reporting System (a database created by the National Highway Transportation Safety Agency to track national data on fatal car crashes) crash files. There were 32,342 different firms with a fatal, injury or property damage only crash reported in the MCMIS database during 2000, so the TTS data alone cannot provide a very good match for ascertaining operations type. The 2,481 firms with data can account for 2,016 matched crash records in the year 2000.

Another possible criterion for establishing the operations of a motor carrier A person providing motor vehicle transportation for compensation. The term includes a motor carrier’s agents, officers and employees is the primary commodity being carried. The Blue Book database has one field for each carrier indicating the commodity hauled. The TTS National Motor Carrier Directory lists the top four commodities being carried, while the MCMIS motor carrier A person providing motor vehicle transportation for compensation. The term includes a motor carrier’s agents, officers and employees census lists up to 30 different commodities that each firm may carry. There are a number of commodities that are very clearly associated with SH short-haul: generally, < 150 mi. from base for property carriers carriage (cement, garbage, tank petroleum products, coal/coke, ores, grain, livestock, et al). These products are associated with special equipment used (as mentioned above), are hauled by train when moving long distances or have some other characteristic that makes long distance carriage untenable or uneconomical. The relative number of long and SH short-haul: generally, < 150 mi. from base for property carriers firms carrying any one of the 24 different commodity groups in the Blue Book data provides a very good benchmark for the potential classification of data in the motor carrier A person providing motor vehicle transportation for compensation. The term includes a motor carrier’s agents, officers and employees census. A discriminate analysis of the Blue Book data revealed that a substantial number of firms could be classified as long or SH short-haul: generally, < 150 mi. from base for property carriers based on the primary commodity that they carry. Petroleum tank products, dump trucks, agricultural commodities, film products and local cartage operations are predominantly listed by firms otherwise classified as SH. Firms handling refrigerated solids, refrigerated liquids and household goods were primarily classified as LH.

A final set of characteristics could be used to classify carriers by long and SH short-haul: generally, < 150 mi. from base for property carriers operations. The number of power units owned or leased or number of drivers employed divided by the total miles driven per year provides a measure of the utilization rate of these resources. Other data from the Blue Book indicate that the low mileage per power unit (or driver) firms tend to be involved in SH short-haul: generally, < 150 mi. from base for property carriers carriage. Low end and high end cut-off points were established for these variables and the carriers at the extremes were classified according to these utilization variables. Since mileage, drivers, and equipment details are available for a large proportion of the firms listed in the USDOT motor carrier A person providing motor vehicle transportation for compensation. The term includes a motor carrier’s agents, officers and employees census, this analysis was conducted for all 889,381 carriers in the database. Based on a review of the Blue Book data, a logical cut-off point of 30,000 or fewer miles per year per driver or power unit was established for defining SH short-haul: generally, < 150 mi. from base for property carriers trucking firms and a cut-off point of 60,000 or more miles per year per driver or power unit was established for defining LH long-haul; generally >150 mi. from base for property carriers trucking firms. Firms with average driver and vehicle usage between 30,000 and 60,000 miles per year were left unclassified.

The results of these three analyses were combined to develop the overall estimates of short and LH long-haul; generally >150 mi. from base for property carriers involvement in the 103,055 truck crashes in the 2000 MCMIS Crash database. The crash data cargo body criterion was given the highest priority in the classification, followed by the Blue Book average trip length, chief commodity, and utilization, finally incorporating the MCMIS Census file commodity and utilization information. Once a crash was classified by one of these methods the methods following it in the priority list were not used. For the MCMIS commodity and utilization tests, consistency checks were used to assure that there was no conflict in classification using the different methods. If a conflict was detected, then the carrier was left unclassified. Approximately 70% of the fatal crashes and 65% of the injury and property damage only crashes could be classified as long or SH short-haul: generally, < 150 mi. from base for property carriers using this procedure.

Table 59 shows the numbers and proportion of crashes in calendar year 2000 broken out by long- or SH short-haul: generally, < 150 mi. from base for property carriers firms. The raw percentages have been adjusted to reflect the relative involvement of long and SH short-haul: generally, < 150 mi. from base for property carriers operations noted. This allocation scheme assumes that the unclassified crashes should be distributed proportionally to the LH long-haul; generally >150 mi. from base for property carriers and SH short-haul: generally, < 150 mi. from base for property carriers groups. For 2000, there was an approximate 60% to 40% split between LH long-haul; generally >150 mi. from base for property carriers and SH short-haul: generally, < 150 mi. from base for property carriers operations involvement in fatal and property damage only crashes. LH long-haul; generally >150 mi. from base for property carriers operations were associated with approximately 55% of the injury only crashes. These estimates can be used with the baseline crash numbers derived from FARS Fatality Analysis Reporting System (a database created by the National Highway Transportation Safety Agency to track national data on fatal car crashes) and GES General Estimates System (a database created by the National Highway Transportation Safety Administration to identify traffic safety problem areas, provide a basis for regulatory and consumer initiatives, and form the basis for cost and benefit analyses of traffic safety initiatives) to establish the historical baseline of crash involvement for the two different types of motor carrier A person providing motor vehicle transportation for compensation. The term includes a motor carrier’s agents, officers and employees operations. From that one can estimate the relative benefits of crash reduction due to differing hours of service proposals.

Table 59: Division of Crashes by Length of Haul

Sleep models are used to analyze the major processes underlying sleep regulation. They also provide a conceptual framework for the analysis of sleep data. As pointed out in the sleep literature, sleep regulation involves three processes: (1) the homeostatic process¹²⁹ which increases during wakefulness and decreases during sleep; (2) the circadian process which depends on the circadian oscillator controlling temperature and alertness rhythms; and (3) the ultradian process which determines the NREM/REM (Non Rapid/Rapid Eye Movement) periodicity.¹³⁰

Over the past two decades, quantitative models have been developed to describe human sleep regulation. Most current mathematical models of alertness include a homeostatic component and circadian component.¹³¹ Sleep models also account for the regulation of the alternation between non-REM sleep and REM sleep. In one class of models, an ultradian oscillator regulates the alternation of non-REM and REM sleep. In the second class of models the alternation between non-REM sleep and REM sleep is governed by homeostatic processes related to non-REM sleep and REM sleep itself.¹³²