Impact of Extended Red Clearance Intervals on Left-Turn Red-Light Violations: A Time Series Analysis

Abstract

Red-light running (RLR) incidents at signalized intersections contribute substantially to intersection-related crashes. They are increasing in number, posing a significant safety challenge. This study examines the impact of signal timing parameters, specifically red clearance intervals, on left-turn RLR violations and assesses effective countermeasures. Adjustments to red clearance and yellow intervals, in line with the ITE 2020 guidelines, were made at signalized intersections in major urban areas within the City of Phoenix. The measurements for left-turn widths at the selected study sites ranged from 60 to 125 ft, with ITE calculations resulting in red clearance intervals between 3 and 5 s. Smart sensors were deployed to collect comprehensive data on signal timings, traffic counts, and RLR violations before and after implementing the updated intervals. Analysis of the pre-and post-intervention data revealed a striking 112% increase in the frequency of RLR violations for left-turn movements following the extension of red clearance intervals. Additionally, the synergistic impact of red clearance and yellow intervals showed that the negative impact of the increased red clearance intervals had offset the positive effect of the extended yellow interval. Further analysis using the interrupted time series method confirmed a significant rise in RLR violations associated with extended red clearance intervals. These findings suggest that increasing the duration of red clearance intervals significantly increases the frequency of RLR violations for left-turn movements. While the study is focused on the City of Phoenix, its implications are relevant to urban areas with similar traffic characteristics. This underscores the importance of thoughtful consideration when modifying signal timings to address RLR behavior effectively.

Keywords

traffic signal change interval clearance interval signalized intersection safety red light running

Red-light running (RLR) is recognized as one of the most hazardous violations at signalized intersections, resulting in severe right-angle, rear-end, and vehicle-to-pedestrian crashes, thereby significantly contributing to a substantial toll of fatalities and serious injuries. According to data collected by the Insurance Institute for Highway Safety (IIHS), RLR violations directly contributed to 1,109 fatalities in the United States in 2021 ( 1 ). Additionally, an estimated 127,000 individuals suffered injuries from crashes related to RLR. This represents an increase compared with 928 deaths and an estimated 116,000 injuries reported in 2020 ( 1 ). At signalized intersections, drivers encounter a pivotal decision point when approaching a yellow signal, necessitating them to choose between proceeding or stopping immediately ( 2 ). When drivers think they have adequate time to cross the intersection safely, they often choose to continue their trajectory. Conversely, they opt to stop if they think they have insufficient time ( 2 ). Nonetheless, drivers are counted as red-light runners if they opt to proceed and the signal transitions to red before they cross the stop bar or exit the intersection entirely. The criteria for identifying individuals as red-light violators can differ from one state to another, primarily based on the unique legal definitions within each jurisdiction. For instance, in Arizona, a red-light runner is characterized as a vehicle that crosses the point where curb lines intersect while the traffic signal displays a red signal ( 3 , 4 ). If a driver decides to stop despite being close enough to proceed safely, while the following driver expects them to continue, this discrepancy in decisions can result in a rear-end crash, where the following vehicle runs into the back of the stopping vehicle ( 5 ). Conversely, red-light running increases the likelihood of right-angle crashes, which occur when vehicles from intersecting approaches collide as they proceed through the intersection ( 6 , 7 ). While rear-end crashes are the most common type at signalized intersections, right-angle crashes tend to be more severe ( 7 , 8 ).

RLR can be categorized into two distinct types: avoidable and unavoidable. Avoidable RLR occurs when a driver deliberately ignores a red light despite having the opportunity to stop safely. On the other hand, unavoidable RLR arises in scenarios where a driver either feels it is unsafe to stop in time and chooses to continue or fails to recognize the critical need to stop ( 9 ). The majority of RLR violations occur under conditions where drivers perceive it as unsafe to stop before the red light or fail to recognize the urgency to stop. Extensive research has examined the effects of environmental factors, road geometry, and traffic density on driver compliance. Findings suggest RLR violations increase with heavier traffic flow ( 10 – 12 ). Additionally, the decision to stop or proceed at intersections is strongly influenced by the driver’s distance from the intersection and their speed at the onset of the yellow light ( 5 , 13 , 14 ). Drivers are also more likely to run a red light if they are following a lead vehicle that decides to proceed during a yellow light ( 5 ). Other critical factors include the vehicle’s position within a group, lane position, and the spacing between adjacent vehicles. These are all key determinants in a driver’s decision to stop or continue at a red light ( 15 ).

The countermeasures used to reduce RLR violations can be divided into two main categories: engineering and enforcement. Enforcement countermeasures, such as employing red-light cameras and issuing citations, aim to reduce avoidable RLR violations where drivers could have stopped safely but proceeded intentionally through the intersection. However, enforcement countermeasures are less effective with regard to RLR violations where drivers entered the intersection just after the signal turned red because of their inability to stop or a lack of awareness with regard to the need to stop. In such instances, engineering countermeasures, including modifications to traffic signal parameters, have shown greater efficacy ( 16 – 18 ). Modifying yellow and red clearance intervals is among the engineering countermeasures to reduce the RLR violations and improve safety at signalized intersections. The red clearance interval is when all signals at an intersection display a red phase to allow vehicles that entered during the yellow interval to clear the intersection safely before the conflicting phase turns green. Despite several decades of investigation and study, a consensus on the most suitable approach for establishing these intervals has yet to be achieved ( 19 – 21 ).

While extending yellow has generally been effective in reducing RLR violations, findings on the impact of red clearance intervals alone are mixed ( 22 – 25 ). For instance, Bonneson and Zimmerman ( 22 ) demonstrated that lengthening the yellow interval could significantly decrease RLR violations. Similarly, Retting et al. ( 23 ) reported a 36% reduction in RLR occurrences following an increase in yellow intervals. However, the effectiveness of extending red clearance intervals in curbing RLR violations is less clear. Retting and Greene ( 24 ) found no reduction in RLR violations from extending red clearance intervals, contrasting with Datta et al. ( 25 ), who noted a significant drop in right-angle crashes after such as increase in red clearance intervals. Moreover, Noyce et al. ( 26 ) found that implementing red clearance intervals would reduce traffic signal violations. However, the study evaluated two red clearance interval durations, 1.5 s and 5.0 s, and found no statistically significant evidence to determine which duration was more effective ( 26 ). Although increasing red clearance intervals is a cost-effective and straightforward strategy, it may reduce the capacity of intersections and increase delays ( 18 ). While much of the existing literature has concentrated on the effects of yellow intervals on RLR violations, fewer studies have explored the consequences of altering red clearance intervals. This area remains important for understanding the broader impacts of traffic signal timing on road safety. Moreover, there is a gap in the literature with regard to the combined effects of updated red clearance and yellow intervals. Furthermore, to the best of the authors’ knowledge, the effectiveness of red clearance intervals in reducing RLR violations for left-turn movements has not been studied before. This study seeks to analyze the effectiveness of adopting the ITE 2020 guidelines for red clearance intervals in reducing left-turn RLR violations by conducting a thorough before-and-after analysis at eight intersections within Phoenix, following a prior study that exclusively examined the impact of yellow intervals. Additionally, this study explores the synergistic impact of concurrently updating red clearance and yellow intervals on left-turning RLR violations behavior.

Current Practices and Guidelines

The red clearance interval is intended so that a reasonable driver who approaches the intersection before the yellow interval ends has enough time to clear the intersection before conflicting traffic enters the intersection. The ITE 2020–recommended equation for calculating the red clearance intervals is shown below:

R = [\frac{W + L}{1.47 V_{E}}] - t_{s}

(1)

where

$R$ = red clearance interval (s),

$V_{E}$ = intersection entry speed (mph); the speed at which a “reasonable” driver is assumed to cross the stop line of the intersection,

$L$ = length of vehicle (ft.); 20 ft is often used as the representative length for vehicles entering the intersection,

$W$ = distance to traverse the intersection (width); stop line to far side no-conflict point along the vehicle path (ft.), and

$t_{s}$ = conflicting vehicular movement start-up delay (s); an optional parameter with an initial value set at 0.0 s.

Values may be used based on engineering judgment or a study’s support.

In this study, the start-up delay for conflicting vehicular movements ( $t_{s}$ ) was set at 0.0 s. The aerial images from Google Maps were used to measure the intersection width for left-turn movements. The intersection entry speed ( $V_{E}$ ) for left-turn movements is assumed to be 20 mph in accordance with the ITE 2020 guidelines. The newly calculated intervals are presented in Table 1 alongside the previously implemented red clearance intervals. Comparing the City of Phoenix policies and newly measured intervals, the differences range from 1.5 to 4.4 s.

Table 1.

Calculated Red Clearance Intervals for Left-Turn Movements

Intersections	Red clearance interval (s) (ITE 2020)		Red clearance interval (s) (COP Policy)		Delta $Δ$ (s) (ITE—COP)
Intersections	EB/WB	NB/SB	EB/WB	NB/SB	EB/WB	NB/SB
#1	3.9	-^*	1	-^*	2.9	-^*
#2	3.1	3.2	1	1	2.1	2.2
#3	5	4.9	1	3.3	4	1.6
#4	3.7	3.8	1	1	2.7	2.8
#5	3.8	3.8	1	1	2.8	2.8
#6	4.2	4.4	1	1	3.2	3.4
#7	3.1	3.3	1	1	2.1	2.3
#8	4.2	4.2	1	1	3.2	3.2

Note: COP = City of Phoenix; EB = eastbound; NB = northbound; WB = westbound; SB = southbound.

No left-turn phase.

Study Sites and Data Collection

Using crash data collected from 2016 to 2021, along with the following ranking criteria, the 12 study intersections were selected:

1) Intersection prioritization: Signalized intersections were ranked based on various factors, including crash frequency, severity, and crash type.

2) Selection of highest-ranked intersections: Among the ranked intersections, those with the highest ranking were given preference for inclusion in the study.

3) Geographic diversity: To ensure a representative sample, selected intersections were distributed across the city to capture geographic diversity.

4) Infrastructure compatibility: Intersections that necessitated minimal infrastructure upgrades to accommodate sensor installation were favored; this was mainly considered as a result of conduit limitations for installing new sensors.

It is important to emphasize that this study builds on previous research on the effects of yellow intervals on RLR violations ( 27 ). Among the 12 selected study sites, two intersections featured exclusively permissive left-turn phases, making them unsuitable for inclusion in this left-turn study. Consequently, intersections with permissive-protected or protected-only left-turn phases were chosen to collect left-turn RLR data, ensuring the study’s relevance and accuracy. Additionally, two intersections were specifically kept to investigate the long-term effect of updated yellow intervals, with no modifications made to the red clearance intervals. As a result, this research will focus only on the eight remaining intersections, analyzing the impact of red clearance intervals on the frequency of RLR violations both individually and in combination with the updated yellow intervals. Figure 1 shows the locations of the study sites within the City of Phoenix. Table 2 provides descriptive characteristics of the selected study sites, including the posted speed limits and the intersection widths for left-turn movements. The speed limits at the study sites range from 35 to 45 mph. The widths for left-turn movements, measured according to ITE 2020 guidelines, range from 60 to 125 ft.

Table 2.

Study Sites Characteristics

Study sites	Speed limit (mph)				Left-turn width (ft.)				Number of through lanes				Number of left-turn lanes
Study sites	EB	NB	WB	SB	EB	NB	WB	SB	EB	NB	WB	SB	EB	NB	WB	SB
Intersection #1	35	35	35	35	85	-	93	-	2	2	2	2	1	1	1	1
Intersection #2	35	35	35	35	62	72	69	60	3	1	2	1	1	1	1	1
Intersection #3	35	35	35	35	116	123	125	115	2	2	3	2	1	1	1	2
Intersection #4	45	40	45	40	80	89	87	82	2	2	2	2	1	1	1	1
Intersection #5	45	40	45	40	86	90	90	91	2	3	2	3	1	1	1	1
Intersection #6	45	40	45	40	103	108	97	105	3	3	3	3	1	1	1	1
Intersection #7	40	40	40	40	71	72	71	76	2	3	3	2	1	1	1	1
Intersection #8	40	40	40	40	103	101	102	96	3	3	3	3	1	1	1	1

Note: EB = eastbound; NB = northbound; WB = westbound; SB = southbound.

To provide additional insights into the geometry and scale of the study sites, site layouts are presented in Figure 2. This figure illustrates the approach speed limits and movement directions.

Data Collection, Preprocessing, and Outlier Filtering

Data collection was conducted by deploying Miovision^© smart sensors at the study sites. Miovision^© SmartView 360 sensors provide a wide array of capabilities, including collecting traffic counts, signal timing data, and instances of RLR violations. Leveraging video-based technology, these sensors also detect vehicles as they traverse the stop bar. A video-detection algorithm is employed to precisely record the presence of a vehicle at the stop bar, thus enabling the identification of the frequency of RLR violations when the signal status transitions to red at that specific timestamp.

To enhance the integrity and reliability of the results, outliers were identified and removed from all data sets. Initially, data points linked to prolonged signal timing intervals, such as unusually extended cycle lengths, were excluded. These outliers are often attributed to signal communication disruptions, transitional phases, or instances involving emergency vehicle preemption. Next, instances of deep RLR violations, defined as those occurring late into the red phase, were filtered out. Finally, the moving interquartile range (IQR) algorithm was employed for further outlier removal. The IQR is calculated as the difference between the first quartile ( $Q_{1}$ ) and the third quartile ( $Q_{3}$ ). In the realm of outlier-filtering techniques based on the IQR, data points falling outside the range of [ $Q_{1} - 1.5 \times IQR, Q_{3} + 1.5 \times IQR$ ] are identified as outliers. We adopted the moving IQR outlier-filtering methodology to eliminate erroneous data sharing similar temporal and spatial characteristics. Our approach involved aggregating the data into 15-min intervals and applying the moving IQR outlier-filtering technique to parameters such as delay, signal timing, and traffic flow at each intersection. Within each 15-min time bin, the IQR was computed for each of these variables, and the outlier-filtering procedure was employed to remove values that fell outside the [ $Q_{1} - 1.5 \times IQR, Q_{3} + 1.5 \times IQR$ ] range ( 12 , 28 ).

Methodology

Experimental Design

The selected study sites were categorized into “control” and “treatment” sites. No alterations were made to the yellow and red clearance intervals at the “control” sites. In our statistical analysis, the “control” sites served to mitigate the effects of traffic volume and pattern variations, especially during holiday seasons. The “treatment” sites were then segmented into five groups to investigate the impact of updating red clearance intervals on signal violations for left-turn movements. Each group was distinguished by the signal timing protocols used during different data collection periods (six periods in total). Detailed information with regard to the specific yellow and red clearance intervals implemented is presented in Table 3.

Table 3.

Implemented Experimental Design

The baseline is implemented City of Phoenix policy intervals.

The colors texts are used for distinguishing different Intersection Types.

This study extends our previous research, which focused solely on the influence of yellow intervals on driver compliance ( 27 ). Figure 3 illustrates the proposed timeline for the study design. The yellow intervals were reverted to City of Phoenix policy (referred to as baseline condition in Table 3) at intersections #1, #2, #3, and #6 during Periods 1 and 2. To explore the short-term effects of updating red clearance intervals on signal violations, red clearance intervals were increased incrementally during Periods 3 and 4, followed by a gradual reduction in Periods 5 and 6 at intersections #3 and #6. To assess the long-term individual effects of red clearance intervals on RLR violation behavior, the yellow intervals were reinstated to the City of Phoenix policy at intersections #1 and #2 while updating the red clearance intervals to align with the ITE 2020 guidelines from Period 3 through Period 6. Additionally, to evaluate the synergistic impact of both updated yellow and red clearance intervals, the ITE 2020 guidelines for both yellow and red clearance intervals were implemented at intersections #4 and #7 in Periods 3, 4, 5, and 6.

Table 4 provides a detailed overview of the implemented red clearance intervals for left-turn movements in each period. Control sites (Intersections #5 and #8) maintained their baseline intervals, while treatment sites experienced changes based on the experimental design. For instance, Intersections #1 and #2 adopted the ITE 2020 guidelines from Period 3 onward, while Intersections #3 and #6 saw incremental adjustments to the red clearance intervals. Intersections #4 and #7 had both yellow and red clearance intervals updated in line with the ITE 2020 guidelines throughout Periods 3 to 6, allowing for the evaluation of the combined effect of updated yellow and red clearance intervals.

Table 4.

Implemented Red Clearance Intervals (s) for Left-Turn Movement

Note: EB = eastbound; NB = northbound; WB = westbound; SB = southbound.

The colors texts are used for distinguishing different Intersection Types.

Before-and-After Analysis

This study used Welch’s t-test to determine the statistical significance of differences in the average rates of RLR violations across various periods. The Welch’s t-test is particularly suitable when comparing two groups with unequal sizes or variances, providing a robust method for such conditions ( 29 – 31 ). The formula for Welch’s t-test calculates the t statistic as follows:

t = \frac{\bar{x_{1}} - \bar{x_{2}}}{\sqrt{\frac{s_{1}^{2}}{N_{1}} + \frac{s_{2}^{2}}{N_{2}}}}

(2)

where

$x_{i}$ represents the sample means,

$s_{i}$ denotes the corresponding standard deviations, and

$N_{i}$ indicates the respective sample sizes.

To quantify the effectiveness of the interventions, an odds ratio (OR) was employed, which compares the improvements observed in the “treatment” sites to those in the “control” sites over the same period. The OR measures how strongly an event is associated with a specific factor in before-and-after studies. The OR compares the odds of the event occurring after the implementation of an intervention to the odds of the event occurring before the intervention. It helps identify how likely it is that the intervention will lead to a specific outcome. A larger OR indicates higher odds of the event occurring after the intervention. Conversely, an OR smaller than 1 implies that the event is less likely to happen following the intervention ( 32 , 33 ). The inclusion of “control” sites in this analysis is essential, as it helps account for any underlying general trends that might influence the frequency of RLR violations across the study sites. This method enhances the precision and depth of our evaluation of the intervention’s effectiveness, strengthening our research findings’ validity and reliability. Furthermore, it acts as a corrective measure to mitigate the influence of external factors on the observed outcomes ( 34 , 35 ). To calculate the OR, the equation suggested by Hauer ( 36 ) was employed:

OR = \frac{K N}{L M}

(3)

where

K represents the post-RLR frequency for the treatment sites,

L denotes the pre-RLR frequency for the treatment sites,

M represents the post-RLR frequency for the control sites, and

N denotes the pre-RLR frequency for the control sites.

The 95% confidence interval (CI) can be derived as:

C I_{95 %} = \exp [\ln (OR) \pm 1.96 s]

(4)

where $s$ is the standard deviation of $\ln (OR)$ , and it can be calculated as:

S^{2} = \frac{1}{K} + \frac{1}{L} + \frac{1}{M} + \frac{1}{N}

(5)

Interrupted Time Series Analysis (ITSA)

Interrupted time series analysis (ITSA) was conducted to provide trends in RLR frequency changes and intuitive visualizations by comparing the periods before and after the implementation of updated red clearance intervals. ITSA—widely applied in medical, public health, and public policy research—has proven its efficacy as a robust tool for longitudinal intervention evaluations ( 37 , 38 ). This method not only evaluates intervention effectiveness but also unveils changes in measures of effectiveness (MOE) before and after interventions ( 38 , 39 ).

Despite the prevalent use of time series analytical methods in traffic flow modeling and prediction, their application within the domain of traffic safety has been comparatively underexplored ( 37 – 42 ). The interrupted time series design requires collecting RLR violations at regular periods before and after the implementation of updated yellow and red clearance intervals. This design enables the observation of potential disruptions in RLR frequency stemming from alterations in signal timing parameters. The interrupted time series model assesses post-intervention outcomes by comparing collected RLR violation data before and after adjusting yellow and red clearance intervals. ITSA offers several advantages in evaluating treatment effects. Firstly, it effectively controls for long-term trends in the outcome measure over time, a feature lacking in simple before–after designs. Additionally, besides estimating the overall effectiveness of the intervention pre- and post-implementation, ITSA provides detailed insights into temporal changes. Lastly, ITSA generates clear and easily comprehensible graphical representations, enhancing the accessibility of its analytical findings ( 37 , 38 , 43 ).

To analyze the impact of extending red clearance intervals on left-turn RLR violations, a regression model was formulated for ITSA. This model was constructed to collect data both before and after the implementation of new signal timings:

Y_{t} = β_{0} + β_{1} T_{t} + β_{2} I_{t} + β_{3} P_{t}

(6)

where

$Y_{t}$ : the aggregated outcome variable (RLR violations rate per day) of the treated group, measured at time t,

$T_{t}$ : the time in days since the start of the study for the observation at the time $t$ ,

$I_{t}$ : the interventions (increasing red clearance intervals) implemented or not (0,1), and

$P_{t}$ : = the time in days after the intervention (i.e., increasing red clearance intervals) for the observation at the time $t$

The coefficients used to measure the impact of updating red clearance intervals on RLR violations are as follows: $β_{0}$ represents the intercept; $β_{1}$ represents the slope of the dependent variable since the start of the study period; $β_{2}$ represents the immediate change observed after implementing the new signal timings; and $β_{3}$ represents the effect of the intervention over time.

Results

Preliminary Data Analysis

Table 5 contains the average daily RLR frequency for the eight study sites. The table has been color-coded to help visually detect patterns and trends, with lighter shades representing cells with lower values and bolder shades indicating higher values. Table 5 shows that the RLR frequency increased with the increase in red clearance intervals, highlighting a correlation between these two variables. Specifically, increasing the red clearance intervals in Period 3 at intersections #1 and #2 led to a rise in RLR frequency. Additionally, intersections #3 and #6 exhibited higher RLR frequency during Periods 3 and 4, when longer red clearance intervals were implemented, compared with the baseline and Periods 5 and 6, when the red clearance intervals were shorter. Conversely, no specific pattern in RLR frequency was observed at intersections #5 and #8, where signal timing remained unchanged.

Table 5.

Average of RLR Frequency Per Day^*

The tables have been color-coded to help visually detect patterns and trends, with lighter shades representing cells containing lower values and bolder shades indicating cells with higher values.

Data not available.

The colors texts are used for distinguishing different Intersection Types.

Accounting for traffic volume is essential to ensure a fair and unbiased analysis. Figure 4 illustrates the traffic flow pattern for the eight selected study sites during each study period. Based on Figure 4, consistent traffic flow trends can be observed for each intersection across different periods. While there are some variations in traffic flow between different times of the day, these differences are not statistically significant.

Traffic volume data were used to normalize the RLR frequency to achieve accurate and unbiased results. By calculating the RLR rate (RLR per 1,000 vehicles per day), we minimized the impact of varying traffic volumes on the study’s outcomes. Statistical t-tests were then conducted on these normalized RLR rates to provide a robust comparison of before-and-after scenarios. For different study periods and sites, RLR rates are provided in Table 6. As the yellow intervals during Periods 1 and 2 had not yet returned to baseline, the RLR rate showed a reduction compared with the baseline. However, this reduction in the RLR rate disappeared once the yellow intervals incrementally returned to baseline conditions in Period 3.

Table 6.

RLR Rate (RLR Per 1,000 Vehicles Per Day)

Note: RLR = red light running.

A statistically significant increase in the RLR rate compared with the baseline at a 95% confidence level.

No data available.

The colors texts are used for distinguishing different Intersection Types.

Based on the results in Table 6, increasing red clearance intervals in Period 3 led to an increase in the RLR rate at Intersections #1, #2, #3, and #6. The increase at Intersections #1 and #3 was not statistically significant, whereas the increase at Intersections #2 and #6 was statistically significant compared with the baseline.

Furthermore, comparing Periods 5 and 6 with the baseline, it is evident that increasing red clearance intervals led to a statistically significant increase in the RLR rate at Intersections #1 and #2. Additionally, analyzing the results from Intersection #6 shows a notable increase in the RLR rate with the incremental extension of red clearance intervals up to Periods 5 and 6. These findings strongly suggest that adjusting red clearance intervals according to ITE 2020 guidelines contributed significantly to increased RLR violations for left-turn movements.

Although the results of our previous study showed that increasing the yellow change interval according to new ITE guidelines could significantly decrease the RLR rate, our current examination of the synergistic effect of changes in yellow and red clearance intervals presents less promising results. Specifically, Intersections #4 and #7 experienced an increase in the RLR rate after the red clearance intervals were extended. Comparing the most recent periods with the baseline reveals that the negative impact of the increased red clearance intervals has offset the positive impact of the extended yellow light, thus eliminating the reduction in the RLR rate previously observed.

ITSA Results

Table 7 provides a comprehensive overview of the ITSA model coefficients and parameters for each intersection, investigating the impact of increasing red clearance intervals both individually and in conjunction with longer yellow intervals on RLR violations. The intersections designated for analyzing the incremental impact of red clearance intervals (i.e., Intersections #3 and #6) were not considered in this analysis because of periodic changes in signal timings, which prevent the implementation of ITSA methods.

Table 7.

ITSA Regression Results

Variables	Intersection #1				Intersection #2
Variables	Coefficient	95% CI		P-value	Coefficient	95% CI		P-value
Intercept	1.35^*	0.80	1.90	0.00	6.10^*	4.11	8.09	0.00
Time	−0.02	−0.04	0.01	0.26	0.01	−0.10	0.12	0.81
Intervention	0.83^*	0.10	1.56	0.03	12.32^*	9.65	15.00	0.00
Time after intervention	0.01^*	0.00	0.03	0.05	−0.07^*	−0.13	−0.01	0.01
	Intersection #4				Intersection #7
Variables	Coefficient	95% CI		P-value	Coefficient	95% CI		P-value
Intercept	1.28^*	0.03	2.53	0.04	1.17^*	0.53	1.81	0.00
Time	0.04	−0.04	0.12	0.30	0.01	−0.03	0.04	0.79
Intervention	0.34	−1.25	1.94	0.67	2.04^*	1.23	2.85	0.00
Time after intervention	0.02	−0.02	0.06	0.33	−0.01	−0.03	0.01	0.19

Note: ITSA = interrupted time series analysis; CI = confidence interval.

A statistically significant at a 95% confidence level.

The coefficient estimates for “Intervention” reveal the immediate impact on the RLR rates. A positive sign for the intervention coefficient indicates that the addition of the intervention will increase the RLR rate. Based on Table 7, statistically significant increases are observed at intersections #1, #2, and #7, with coefficients of 0.83 (95% CI [0.10, 1.56]), 12.32 (95% CI [9.65, 15.00]), and 2.04 (95% CI [1.23, 2.85]), respectively. Intersection #4 also showed an increase in RLR rates after the intervention, although this result was not statistically significant, with a coefficient of 0.34 (95% CI [−1.25, 1.94]).

Examining the “Time” variable, the results reveal a discernible impact on RLR rates (the coefficients are not statistically significant). However, the “Time after Intervention” results show varied trends. At intersections #1 and #4, there is a marginal increase in RLR violations over time following the implementation of new red clearance intervals, with coefficients of 0.01 (95% CI [0.00, 0.03]) and 0.02 (95% CI [−0.02, 0.06]), respectively. Conversely, intersections #2 and #7 exhibit a slight reduction in RLR rates over time following the implementation of extended red clearance intervals, with coefficients of −0.07 (95% CI [−0.13, −0.01]) and −0.01 (95% CI [−0.03, 0.01]), respectively.

To better examine the results of Table 7, the ITSA regression coefficients are illustrated in Figure 5. In this figure, the solid lines represent the trend derived from actual RLR violation data points, while the dashed lines illustrate the predicted counterfactual trend, assuming the RLR rates would follow the pre-intervention trend. Comparing the actual trend with the counterfactual trend shows that increasing the red clearance time significantly increases the RLR rate. This increase has a positive slope at Intersections #1 and #4 and a negative slope at Intersections #2 and #7.

Random-Effects ITSA Results

To account for the dependencies in RLR rates, the basic ITSA model was extended by incorporating random effects, which allows for unobserved heterogeneity at the intersection-approach level. Random-effects models have been extensively applied in prior studies to examine various traffic-related topics, including crash frequency and severity ( 44 – 46 ), traffic flow predictions ( 47 ), and operating speed analysis ( 48 ). Since RLR violations at each intersection approach may be influenced by unobserved factors unique to each specific location, adding intersection-approach random effects to the ITSA model provides a more accurate reflection of the data. Given the different experimental designs used for Intersections #1 and #2 versus Intersections #4 and #7, the random-effects ITSA model was applied separately to each group, and the results are presented in Table 8.

Table 8.

Random Effects ITSA Regression Results

Variables	Intersections #1 and #2				Intersections #4 and #7
Variables	Coefficient	95% CI		P-value	Coefficient	95% CI		P-value
Intercept	2.954	−2.715	8.624	0.30	2.257^*	0.859	3.654	0.00
Time	0.036^*	0.015	0.056	0.00	−0.001	−0.018	0.016	0.91
Intervention	5.395^*	3.706	7.084	0.00	1.12^*	0.18	2.06	0.02
Time after intervention	−0.074^*	−0.109	−0.039	0.00	0.015	−0.011	0.042	0.26
Random effect	Variance		SD		Variance		SD
Intersection-approach	48.12		6.937		2.807		1.675
Residuals	21.26		4.611		5.765		2.401

Note: ITSA = interrupted time series analysis; CI = confidence interval; SD = standard deviation.

A statistically significant at a 95% confidence level.

For Intersections #1 and #2, the results indicate that “Time,”“Intervention,” and “Time after Intervention” were all statistically significant at the 95% confidence level. The “Time” variable showed a significant increasing trend in RLR violations over the entire study period, with a coefficient of 0.036 (95% CI [0.015, 0.056]). Additionally, the implementation of extended red clearance intervals resulted in a significant increase in RLR violations, with a coefficient of 5.395 (95% CI [3.706, 7.084]). However, after the intervention, a significant decrease in the RLR violations trend was observed, as indicated by the “Time after Intervention” coefficient of −0.074 (95% CI [−0.109, −0.039]). In contrast, for Intersections #4 and #7, “Time” and “Time after Intervention” were not significant, with coefficients of −0.001 (95% CI [−0.018, 0.016]) and 0.015 (95% CI [−0.011, 0.042]), respectively. However, similar to Intersections #1 and #2, the extended red clearance intervals led to a significant increase in RLR violations, with a coefficient of 1.12 (95% CI [0.18, 2.06]). Overall, the findings found that extending the red clearance intervals increases RLR violations. Nevertheless, the lack of consistent significance for the “Time” and “Time after Intervention” variables across sites prevents us from drawing a robust conclusion. This inconsistency may be a result of the limited data collection period and the short duration of the study.

With regard to the random effects, the variance and standard deviation for the intersection-approach random effects were 48.12 and 6.937 for Intersections #1 and #2, and 2.807 and 1.675 for Intersections #4 and #7, respectively. The larger variance and standard deviation for Intersections #1 and #2 show greater variability in RLR violations across the intersection approaches in this group, possibly because of site-specific factors that were not directly measured in the model. In contrast, the lower variance and standard deviation for Intersections #4 and #7 indicate less variability in RLR violations, showing more consistent behavior across these sites. These random-effect estimates highlight the importance of accounting for intersection-approach specific characteristics in the analysis, as they contribute significantly to the observed variability in RLR behavior.

Odds Ratio Calculation

To evaluate the effectiveness of the new signal timing implementations, ORs were calculated to compare the outcomes at treatment sites with those at control sites over the same period. This approach measures the likelihood of RLR violations occurring under updated red clearance intervals versus without the updated timings, providing clearer insights into its effectiveness. The ORs were calculated by examining the average daily frequency of RLR events at both treatment and control sites before and after implementing the ITE 2020 guidelines for red clearance intervals (Table 9). For treatment sites, the average daily RLR violations increased from 14 to 32 after updated red clearance intervals were implemented. The average daily RLR violations for control sites increased slightly from 202 to 217. The analysis indicates that extending red clearance intervals results in a 112% increase in the likelihood of left-turning RLR violations. This significant increase suggests a strong correlation between longer red clearance intervals and a higher probability of RLR violations for left-turn movements. The CI of 1.10 to 4.09 indicates that we can be 95% confident that the true OR lies within this range, further reinforcing the robustness of the findings.

Table 9.

Results of the Before-and-After Study of Increasing Red Clearance Intervals

Study sites type	Before	After	Odds ratio (95% CI)
Treatment sites	14	32	2.12 (1.10–4.09)
Control sites	202	217	2.12 (1.10–4.09)

Note: CI = confidence interval.

Conclusion

Red-light running (RLR) behavior is a significant concern at signalized intersections, leading to a high number of intersection-related crashes. To address this pressing issue and enhance safety at signalized intersections, it is crucial to thoroughly investigate RLR behavior locally. This study’s primary goal is to determine whether adopting the updated ITE 2020 guidelines, released in March 2020, for setting red clearance intervals can effectively reduce left-turn RLR violations at signalized intersections. To assess the impact of extended red clearance intervals on left-turn RLR violations, we conducted a comprehensive before-and-after analysis at eight intersections in Phoenix. The selected sites have speed limits ranging from 35 to 45 mph. Left-turn widths, measured according to ITE 2020 guidelines, vary between 60 and 125 ft. The site with the 125-ft left-turn width is located along a corridor where light rail operates in both northbound and southbound directions. Using the ITE 2020 calculations, red clearance intervals were determined to range from 3 to 5 s. At each study site, we collected data, including signal timing information, turning movement counts, high-resolution event-based data, and RLR violations, using video-based smart sensors. The study incorporated multiple 2-week periods, with a new red clearance interval implemented during each period. Some sites were designated as control sites where no changes were made to the red clearance intervals.

This study’s findings showed that increasing the red clearance intervals for left-turn movements resulted in a higher occurrence of RLR violations. The likelihood of an increase in the average RLR rate was 2.12 times higher after extending the red clearance intervals. In light of the empirical findings, we recommend not aligning the red clearance intervals for left-turn movements with the ITE 2020 guidelines, as our study demonstrated a significant rise in RLR violations resulting from this change. However, it is essential to acknowledge that determining the optimal red clearance intervals for all intersections is a complex challenge, given their intricate interplay with factors such as driver behavior and demographics, traffic patterns, and intersection geometry.

Although increasing red clearance intervals could help reduce crashes by allowing vehicles that have already entered the intersection more time to clear the intersection, this study exclusively explores the impact of extending red clearance intervals on driver compliance behavior. It does not consider other potential factors such as geometric design, cycle length, green interval, and environmental conditions. The omission of these variables may have affected the observed effects. Therefore, further research is needed to comprehensively assess all relevant factors and understand their influence on driver compliance and intervention effectiveness. Moreover, our study had inherent limitations because of time constraints, including data collection duration and potential prolonged effects. Future research should expand the scope to include a more diverse range of study sites and collect data over an extended period including additional factors such as demographic characteristics, GPS or waypoint data, and potential variables. Furthermore, this study did not separately analyze protected-permissive left-turn signal phasing because of constraints in the experimental design and limited data availability. Future research could explicitly examine signal phasing and timing to provide more comprehensive and robust insights into the impact of protected and permissive left-turn phases on RLR violations. Additionally, investigating the impact of driver demographics, such as age and gender, on signal light compliance and assessing the influence of law enforcement measures, like RLR camera enforcement, on driver behavior could provide valuable insights into this critical issue.

Footnotes

Acknowledgements

The authors would like to thank the City of Phoenix for their invaluable support in recognizing the importance of addressing the issue of red-light running (RLR). Special thanks to Eric Hernandez for his assistance in facilitating this research. The authors are grateful to Faith Victoria Williamson for her help in proofreading.

Authors’ Contributions

The authors confirm contribution to the paper as follows: study conception and design: Pouya Jalali Khalilabadi and Abolfazl Karimpour; data collection: Pouya Jalali Khalilabadi, Abolfazl Karimpour, and Simon T Ramos; analysis and interpretation of results: Pouya Jalali Khalilabadi, and Abolfazl Karimpour; draft manuscript preparation: Pouya Jalali Khalilabadi, Abolfazl Karimpour, Yao-Jan Wu, and Simon T Ramos. All authors reviewed the results and approved the final version of the manuscript.

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article but wish to declare that Yao-Jan Wu is a member of Transportation Research Record’s editorial board.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

ORCID iDs

Pouya Jalali Khalilabadi

Abolfazl Karimpour

Yao-Jan Wu

References

IIHS. Red Light Running. 2022. https://www.iihs.org/topics/red-light-running.

Bonneson

J. A.

Brewer

Zimmerman

Review and Evaluation of Factors that Affect the Frequency of Red-Light-Running. Vol. 4027. Texas Transportation Institute, Texas A&M University System, 2001.

City of Scottsdale. Photo Enforcement Program. The City of Scottsdale has Used Photo Enforcement Technology since 1996 to Help Reduce Red-Light Running and Speeding. https://www.scottsdaleaz.gov/police/police-units/photo-enforcement.

Smith

Red light ticket?

Where’s the Line? 2016. https://www.kgun9.com/news/local-news/red-light-ticket-wheres-the-line#:∼:text=The%20law%20says%20youre,line%20from%20each%20curb%20line.

Elmitiny

Yan

Radwan

Russo

Nashar

Classification Analysis of Driver’s Stop/Go Decision and Red-Light Running Violation. Accident Analysis and Prevention, Vol. 42, No. 1, 2010, pp. 101–111.

Polders

Daniels

Hermans

Brijs

Wets

Crash Patterns at Signalized Intersections. Transportation Research Record: Journal of the Transportation Research Board, 2015. 2514: 105–116.

Wang

Abdel-Aty

Right-Angle Crash Occurrence at Signalized Intersections. Transportation Research Record: Journal of the Transportation Research Board, 2007. 2019: 156–168.

Wang

Abdel-Aty

Temporal and Spatial Analyses of Rear-End Crashes at Signalized Intersections. Accident Analysis & Prevention, Vol. 38, No. 6, 2006, pp. 1137–1150.

Bonneson

J. A.

Son

H. J.

Prediction of Expected Red-Light-Running Frequency at Urban Intersections. Transportation Research Record: Journal of the Transportation Research Board, 2003. 1830: 38–47.

10.

Baguley

‘Running the Red’at Signals on High-Speed Roads. Traffic Engineering and Control, Vol. 29, No. 7–8, 1988, pp. 415–420.

11.

Porter

B. E.

England

K. J.

Predicting Red-Light Running Behavior: A Traffic Safety Study in Three Urban Settings. Journal of Safety Research, Vol. 31, No. 1, 2000, pp. 1–8.

12.

Karimpour

Jalali Khalilabadi

Homan

Y. J.

Swartz

D. L.

Modeling Red-Light Running Behavior Using High-Resolution Event-Based Data: A Finite Mixture Modeling Approach. Journal of Intelligent Transportation Systems, Vol. 28, 2023, pp. 1–16.

13.

Jahangiri

Rakha

Dingus

T. A.

Red-Light Running Violation Prediction Using Observational and Simulator Data. Accident Analysis and Prevention, Vol. 96, 2016, pp. 316–328.

14.

Vall

N. D.

Liu

H. X.

Cheng

Jia

Analysis of Drivers’ Stop-or-Run Behavior at Signalized Intersections with High-Resolution Traffic and Signal Event Data. Transportation Research Record: Journal of the Transportation Research Board, 2013. 2365: 99–108.

15.

Chen

Ren

Estimation of Red-Light Running Frequency Using High-Resolution Traffic and Signal Data. Accident Analysis and Prevention, Vol. 102, 2017, pp. 235–247.

16.

Elias

Ghafurian

Samuel

Effectiveness of Red-Light Running Countermeasures: A Systematic Review. Proc., 11th International Conference on Automotive User Interfaces and Interactive Vehicular Applications, Utrecht, Netherlands, 2019.

17.

Gates

T. J.

Noyce

D. A.

A conceptual Framework for Dynamic Extension of the Red Clearance Interval as a Countermeasure for Red-Light-Running. Accident Analysis & Prevention, Vol. 96, 2016, pp. 341–350.

18.

Hallmark

Oneyear

McDonald

Toolbox of Countermeasures to Reduce Red Light Running. Iowa State University. Center for Transportation Research and Education, 2012.

19.

Noble

D. E.

Guidelines for Determining Traffic Signal Change and Clearance Intervals. ITE Journal, Vol. 90, No. 3, 2020, pp. 28–31.

20.

Muller

T. H.

Dijker

Furth

P. G.

Red Clearance Intervals: Theory and Practice. Transportation Research Record: Journal of the Transportation Research Board, 2004. 1867: 132–143.

21.

Souleyrette

R. R.

O’Brien

M. M.

McDonald

Preston

Storm

Effectiveness of All-Red Clearance Interval on Intersection Crashes (No. MN/RC-2004-26). 2004.

22.

Bonneson

J. A.

Zimmerman

K. H.

Effect of Yellow-Interval Timing on the Frequency of Red-Light Violations at Urban Intersections. Transportation Research Record: Journal of the Transportation Research Board, 2004. 1865: 20–27.

23.

Retting

R. A.

Ferguson

S. A.

Farmer

C. M.

Reducing Red Light Running Through Longer Yellow Signal Timing and Red Light Camera Enforcement: Results of a Field Investigation. Accident Analysis and Prevention, Vol. 40, No. 1, 2008, pp. 327–333.

24.

Retting

R. A.

Greene

M. A.

Influence of Traffic Signal Timing on Red-Light Running and Potential Vehicle Conflicts at Urban Intersections. Transportation Research Record: Journal of the Transportation Research Board, 1997. 1595: 1–7.

25.

Datta

T. K.

Schattler

Datta

Red Light Violations and Crashes at Urban Intersections. Transportation Research Record: Journal of the Transportation Research Board, 2000. 1734: 52–58.

26.

Noyce

D. A.

Bill

A. R.

Claros

B. R.

Chitturi

M. V.

Michael

Francis

Hurwitz

D. S.

Peter

Brehmer

C. L.

Evaluation of Change and Clearance Intervals Prior to the Flashing Yellow Arrow Permissive Left-Turn Indication (No. NCHRP Project 03-125). 2024.

27.

Khalilabadi

P. J.

Karimpour

Y.-J.

Ramos

S. T.

Understanding Drivers’ Compliance Behavior: Data-Driven Assessment of Longer Yellow Intervals. Transportation Research Record: Journal of the Transportation Research Board, 2024. 03611981241263827.

28.

Jalali Khalilabadi

Karimpour

Y.-J.

Severity Analysis of Red-Light–Running Behavior at Signalized Intersections. Journal of Transportation Safety & Security, Vol. 16, No. 5, 2024, pp. 482–506.

29.

Gong

Yang

X. T.

Impact of COVID-19 on Traffic Safety from the “Lockdown” to the “New Normal”: A Case Study of Utah. Accident Analysis & Prevention, Vol. 184, 2023, p. 106995.

30.

Welch

B. L.

The Generalization of ‘STUDENT’S’problem when Several Different Population Varlances are Involved. Biometrika, Vol. 34, No. 1–2, 1947, pp. 28–35.

31.

Kadeha

Alluri

Sando

Quantifying the Mobility Benefits of Express Lanes Using Real-Time Traffic Data. Transportation Research Record: Journal of the Transportation Research Board, 2020. 2674: 414–423.

32.

Szumilas

Explaining Odds Ratios. Journal of the Canadian Academy of Child and Adolescent Psychiatry, Vol. 19, No. 3, 2010, p. 227.

33.

Andrade

Understanding Relative Risk, Odds Ratio, and Related Terms: As Simple as it Can Get. The Journal of Clinical Psychiatry, Vol. 76, No. 7, 2015, p. 21865.

34.

Mayora

J. M. P.

Piña

R. J.

An Assessment of the Skid Resistance Effect on Traffic Safety Under Wet-Pavement Conditions. Accident Analysis & Prevention, Vol. 41, No. 4, 2009, pp. 881–886.

35.

Daniels

Nuyts

Wets

The Effects of Roundabouts on Traffic Safety for Bicyclists: An Observational Study. Accident Analysis & Prevention, Vol. 40, No. 2, 2008, pp. 518–526.

36.

Hauer

Observational Before/After Studies in Road Safety. Estimating the Effect of Highway and Traffic Engineering Measures on Road Safety. Emerald Group Publishing Limited, Bingley, 1997.

37.

Song

Noyce

Effects of Transit Signal Priority on Traffic Safety: Interrupted Time Series Analysis of Portland, Oregon, Implementations. Accident Analysis & Prevention, Vol. 123, 2019, pp. 291–302.

38.

Penfold

R. B.

Zhang

Use of Interrupted Time Series Analysis in Evaluating Health Care Quality Improvements. Academic Pediatrics, Vol. 13, Vol. 6, 2013, pp. S38–S44.

39.

Khan

M. A.

Etminani-Ghasrodashti

Kermanshachi

Rosenberger

J. M.

Pan

Foss

Do Ridesharing Transportation Services Alleviate Traffic Crashes? A Time Series Analysis. Traffic Injury Prevention, Vol. 23, No. 6, 2022, pp. 333–338.

40.

Ishak

Al-Deek

Performance Evaluation of Short-Term Time-Series Traffic Prediction Model. Journal of Transportation Engineering, Vol. 128, No. 6, 2002, pp. 490–498.

41.

Ghosh

Basu

O’Mahony

Bayesian Time-Series Model for Short-Term Traffic Flow Forecasting. Journal of Transportation Engineering, Vol. 133, No. 3, 2007, pp. 180–189.

42.

Bernal

J. L.

Cummins

Gasparrini

Interrupted Time Series Regression for the Evaluation of Public Health Interventions: A Tutorial. International Journal of Epidemiology, Vol. 46, No. 1, 2017, pp. 348–355.

43.

Huitema

B. E.

Van Houten

Manal

Time-Series Intervention Analysis of Pedestrian Countdown Timer Effects. Accident Analysis & Prevention, Vol. 72, 2014, pp. 23–31.

44.

Chin

H. C.

Quddus

M. A.

Applying the Random Effect Negative Binomial Model to Examine Traffic Accident Occurrence at Signalized Intersections. Accident Analysis & Prevention, Vol. 35, No. 2, 2003, pp. 253–259.

45.

Kidando

Moses

Ozguven

E. E.

Sando

Incorporating Travel Time Reliability in Predicting the Likelihood of Severe Crashes on Arterial Highways Using Non-Parametric Random-Effect Regression. Journal of Traffic and Transportation Engineering (English Edition), Vol. 6, No. 5, 2019, pp. 470–481.

46.

Abdel-Aty

Ahmed

Bayesian Random Effect Models Incorporating Real-Time Weather and Traffic Data to Investigate Mountainous Freeway Hazardous Factors. Accident Analysis & Prevention, Vol. 50, 2013, pp. 371–376.

47.

Y.-J.

Chen

C. T.

Yang

Urban Traffic Flow Prediction Using a Spatio-Temporal Random Effects Model. Journal of Intelligent Transportation Systems, Vol. 20, No. 3, 2016, pp. 282–293.

48.

Bassani

Cirillo

Molinari

Tremblay

J. M.

Random Effect Models to Predict Operating Speed Distribution on Rural Two-Lane Highways. Journal of Transportation Engineering, Vol. 142, No. 6, 2016, p. 04016019.