The practice of performing work-zone activities at night has been around since at least the 1960s. Early attempts at night work were initiated because officials considered it impractical to close traffic lanes on certain high-volume roadways during normal daylight hours. Early experiences indicated that the concept did indeed reduce the impact of work on the traveling public, and lower traffic volumes meant fewer conflicts between traffic and construction vehicles, making it easier to get materials to and from the work site.
As recent research has documented, the decision to perform highway work at night requires the consideration of a number of interrelated factors. In essence, the benefits of doing road work at night—reduced congestion, cooler temperatures, longer allowable work windows—need to be balanced against the added costs and consequences of doing so, such as more difficult material-supply logistics, additional traffic-control costs, noise, safety and health concerns.
As this recent research effort and those previously have encountered, data on the actual safety consequences of working at night is difficult to come by. As a result, it is difficult for practitioners to accurately capture and account for the true safety consequences of doing work at night.
Safe at night?
Generally speaking, past research suggests that night work productivity and quality can be comparable to daytime work. However, the implications on safety are less clear. From the perspective of the highway workers out on the road, studies in non-work-zone environments strongly suggest that night work increases the amount of stress placed on the body, negatively affects the amount and quality of sleep that workers obtain and can significantly alter physiological characteristics such as appetite. Night work also tends to adversely affect the social and domestic aspects of a worker’s life. Perhaps more importantly, it has been shown that shift work in general also can impair overall worker alertness, reaction times and even motor skills.
Assessments of the implications of night-work activity upon traffic safety are even more difficult to come by in the literature. There have been several researchers who investigated the effects of long-term construction work zones on crashes occurring at night. Some of these concluded that nighttime crashes increase substantially in work zones, whereas other studies concluded the opposite.
However, none of these were able to isolate any effects that active night work itself may have had on the crashes, since information regarding whether a work zone is currently active at the time of a nighttime crash is not recorded on most crash report forms and was therefore not included in their analysis. Consequently, the effects of actual night work in existing crash databases confound with any additional crashes that occur at work zones that are inactive during the nighttime hours due to temporary restricted geometrics or work-zone traffic-control device deficiencies.
One study performed on several construction projects in California in the 1980s examined the impacts that performing work at night had upon highway crashes. The author concluded that the nighttime crash rate on those sections of roadway where work activity was being performed increased 87% over the normal nighttime crash rates at those locations. The data also suggested that the crash rate during nights when lane closures were required was 75% higher than during nights of activity when no lane closures were required.
Another study of night work lane closures in Virginia found similar trends. As the author of the California study noted, the significant increase in crash rates during night work did not necessarily imply that working at night was undesirable from an overall safety perspective. Rather, the overall frequency of crashes at night might still have been lower than expected if the work had been performed during the day due to higher vehicle exposure levels present during the day.
Unfortunately, direct comparisons of identical work-zone activities performed at night and during the day at a given location are generally not possible. The reason that work is being done at night in the first place at such locations is to avoid significant disruptions to high-volume traffic during daytime travel periods.
The objective of the Texas study was to assess the crash consequences of seven highway construction projects performed in Texas that involved some degree of night work activity. Researchers attempted to determine the change in crash likelihood during periods of active night work, active day work (if applicable) and during times of work inactivity.
Picking projects
Through contact with TxDOT personnel, TTI researchers identified a number of potential projects that involved nighttime activities. Two types of construction projects were identified. The first of these were labeled as hybrid projects, where certain work activities occurred during the day off of the travel lanes and active night work occurred whenever travel lanes needed to be closed. These projects tended to be fairly large roadway widening and major reconstruction efforts. The nighttime activities that required lane closures were typically to allow construction materials and equipment to be lifted across portable barriers into the work area, to change traffic-control alignment between construction phases and to put a final overlay over both new and existing pavement at the end of the project.
One of the key features of this type of work zone was that there typically exist geometric restrictions such as shoulder closures, lane-width reductions, lane shifts and other temporary changes in horizontal alignment that can influence traffic-crash likelihood at the location whether or not work activity is occurring.
The other type of construction projects considered in this analysis was those associated with pavement resurfacing. These projects were done exclusively at night because all of the required work activities involved the temporary closure of travel lanes. Furthermore, the nature of the work did not require temporary geometric restrictions as were present in the hybrid projects.
Researchers limited the analysis to projects located on interstates or controlled-access facilities. In addition, researchers had to limit their search to projects occurring during the 1999-2001 calendar years because of the lack of available timely crash data. This limitation hampered project-identification efforts somewhat, as many districts had already archived project information from that far in the past.
Researchers eventually identified seven projects for analysis. Two of these projects involved roadway resurfacing, whereas the other five projects were major roadway rehabilitation or reconstruction projects that involved predominantly daytime work adjacent to the travel lanes, but with occasional periods of active night work when performing activities in the travel lanes. To maintain confidentiality of the crash data, researchers opted to avoid project descriptors that specifically identify these projects. Rather, researchers assigned each a project code to use.
Collecting the info
Researchers first traveled to each district having jurisdiction over one of the project locations and reviewed the inspectors’ diaries for that project. They recorded the hours of work activity each day or night, type of work being performed if noted (on the hybrid projects, there were often several activities going on at the same time within the project limits), number of workers and equipment present, and location and length of lane closures if noted. Overall, researchers obtained data from 4,300 days and nights of project diary entries. Researchers then reduced these data to a set of dates and times corresponding to one of the following conditions:
- Daytime active, when work occurred (analyzed only for the hybrid projects);
- Daytime inactive, when no work occurred (also only for hybrid projects);
- Nighttime active; and
- Nighttime inactive.
Once project data were collected, researchers obtained crash records from the official statewide database maintained by the Texas Department of Public Safety (DPS). All crashes occurring within the limits of the project were obtained for the duration of that project and for three years prior. Due to the incomplete nature of the project diary data regarding location and direction of work activities, researchers decided to consider all crashes occurring in both directions of travel within the project limits. Researchers also theorized that even if direction of work activity was known, potential distraction effects due to driver rubbernecking could be present in the crashes and also should be included in the analysis.
For each project, researchers identified a comparison segment on the same roadway or on a nearby facility. This comparison segment allowed the researchers to perform a before-during analysis with a control group at each site. Researchers also extracted the crash data from each comparison segment for the same during period as its corresponding project and for the three years prior.
Gathering and analyzing
Researchers first performed a check for comparability between the work zone and comparison segment to ensure that the use of that comparison segment was appropriate. Once researchers verified that crash trends in the comparison and project segments were indeed comparable before the start of work activities, researchers divided the crashes occurring during project and its corresponding comparison segment into one of the four categories and conditions.
Researchers calculated simple cross-product ratios between the before and during time periods in both the project and comparison segments to estimate the extent to which crashes in each period exceeded the expected number of crashes for that particular period.
In analyzing the data gathered, researchers first summarized the total number of crashes that occurred at the five hybrid project locations for all times and work conditions together, the number of crashes that were expected to have occurred and the corresponding percent change between actual and expected crash frequencies.
For comparison purposes, researchers also calculated the crashes per unit time at each work zone before and during the project. Accurate traffic volume data by time of day before and during the projects were not available for calculations of crashes per vehicle-miles traveled. Researchers then determined the overall weighted average change in crashes across all five projects combined.
Comparing the total crashes that occurred at each project location to the expected number of crashes, researchers found the increase in crash rates at four of the five hybrid projects to range between 30% and 40%, values that are generally consistent with past work-zone crash studies.
The weighted average of overall crash increases was 31.5%, also very close to previous studies of work-zone crash increases at freeway reconstruction projects in Texas using this type of analysis.
The before-during comparison of crash rates at the project locations also suggested significant increases at four of the five projects (although the project without the significant increase was not the same as for the other analysis approach). However, the estimated overall increase using this simpler analysis method was a smaller 17.1%. Although lower, this increase was still significant.
Next, researchers analyzed each project by day and night and work activity or inactivity and once again performed the crash comparisons. Researchers believe the comparison of crashes during the inactive time periods reflect the influence of temporary work-zone geometric changes in the project location as well as temporary signing and other features left in place during the project.
In contrast, the comparison of crashes during work activity represents the influence of these conditions plus the impact of actual work activities within the project limits. The work activities during the daytime were essentially done out of the travel lanes in the median, in the separation between freeway main lanes and the frontage road lanes or in the space beyond the frontage roads.
Consequently, any effect of daytime activity would likely represent only the influence of construction vehicles entering and exiting the work area or distractions that the work created for passing motorists. In contrast, work performed at night involved lane closures. In addition to the work activity influences, researchers expected that the effect of capacity restrictions and required lane changes for the closures to have some effect on traffic crashes.
Researchers found little impact of the work zone upon daytime periods when no work was occurring. Compared with the expected number of crashes for these periods, actual crashes ranged between an 11.7% decrease and a 38% increase, neither of which were statistically significant.
Overall, crashes actually occurring at these projects during the 6 a.m. to 7 p.m. time period were 14% higher than expected, an increase that also was not significant. When only the crash rates before and during each project for this time period were compared, during rates were only slightly affected. Differences in these rates ranged between a 14% decrease and a 14% increase. The overall estimated change in these rates was an insignificant 2.6% decrease.
In contrast to the minimal changes in crashes during the daytime inactive period, researchers found more substantial changes in nighttime inactive crashes. Actual crashes during the nighttime inactive period were 49-63% higher than expected at four of the five hybrid projects. Overall, actual crashes during this period were 48.7% higher than expected. When the crash rates at the project locations themselves were compared for this time period, the magnitude of the changes was somewhat lower. The overall increase in crash rates at the hybrid projects was significant during the nighttime inactive period at nearly 26%.
The comparison of expected versus actual crashes at the hybrid and resurfacing projects during the nighttime active time period yielded highly variable results ranging from a 23% decrease at one site to an almost 500% increase at another. The sample sizes involved in this portion of the analysis were fairly small for most of the projects, making it difficult to detect highly significant changes. However, the overall weighted average of the changes across the five projects was found to be a significant 102.2% increase.
The percentage increase in crashes at the two resurfacing projects also was fairly substantial, although not statistically significant due to the small samples. The expected versus actual crash comparison yielded a weighted average increase of 55.4%, and the before-during comparison of crash rates yielded an average increase of 35.5%.
Conditions and crashes
The small sample sizes available from some of the projects did not allow the same types of statistical comparisons to be performed on any subcategories of the crashes. Therefore, to assess whether crash severity was affected by work activity and time period, researchers consolidated the crash data for all projects and computed the percent of crashes that were severe (non-property-damage-only crashes) as a function of the time period and work activity categories. The percent of severe crashes at the hybrid projects was slightly greater during the days of work activity as compared to the days of inactivity and to the before condition.
At night, the percent of severe crashes was slightly less overall during the project on both nights of activity and nights without activity, as compared to the before condition.
With regards to the resurfacing projects, the percentage of severe crashes was somewhat higher on nights of activity relative to the before condition, while the percentage of severe crashes on nights of inactivity during the project were actually lower than in the before condition.
Researchers hypothesize that the nights of inactivity at these resurfacing projects may have correlated with nights of poorer driving conditions, as pavement overlays and other resurfacing activities would likely not be performed on nights when rain or wet pavement were present. Such conditions typically reduce travel speeds somewhat and may have attributed to the lower-than-normal level of severe crashes on those nights. Adverse weather conditions also may increase the number of non-severe crashes that occur on those nights of inactivity, and therefore dilute the proportion of severe crashes existing in the dataset, even if the number of severe crashes itself remained constant over time.
The wide variation in site conditions from work zone to work zone often compound with the factors of interest and make it difficult to obtain with certainty an accurate measure of that factor’s influence.
Such was the case in assessing the relative safety effects of active night work. Although the analysis did find crashes at active night work zones to be higher (sometimes significantly) than expected at the location if the work zone was not present, researchers also found that crashes increased in some instances even when work activity was not present or during daytime work periods when travel lanes were not closed.
