Child Restraint Use and Seating Position of Child Passengers in Motor Vehicles and Their Correlations: Application of a Random-Effects Bivariate Probit Model

Abstract

Motor vehicle crashes are a leading cause of death and injury for children aged under 8 years. In the U.S.A., while some states are showing increases in the proportion of child restraint device use, only around half of child passengers aged from 4 to 7 years are being properly restrained. This study was undertaken to identify the factors contributing to the proper child restraint use and child passenger’s seating position through the direct observation surveys of more than 10,000 child passengers in 2015 and 2018 in Michigan. A bivariate probit model with random effects was developed to identify simultaneously the contributing factors for the proper restraint use and seating position of child passengers, accounting for the unobserved heterogeneity in the data. The bivariate framework is able to account for the correlation of the two dependent variables in the study. The results show that the two dependent variables are positively correlated, and this correlation is strongly significant. The key factors simultaneously influencing proper child restraint use and appropriate seating position of the child passenger include the age of the child, the number of child passengers in the vehicle, driver belt use, driver gender, age, and race, vehicle type, stratum, weather, and the time of the day and week. However, factors such as county-specific population, income, and education, and the type of location did not have any significant association with either child restraint use or the seating position of the child passenger.

Keywords

safety occupant protection child restraints

Motor vehicle crashes are a leading cause of death and injury for children under 8 years of age in the U.S.A. ( 1 ). Nationwide, between 2014 and 2018, a total of 304,803 children under the age of 8 years died or suffered from vehicle occupant injuries ( 2 ). Based on the latest statistics, across the U.S.A. in 2019, the percentage of unrestrained children dying in a traffic crash with a known restraint use increased to more than 40% from an average of 34% between 2014 and 2018 ( 3 ). However, particularly for unrestrained child passengers less than 1 year old who died in a traffic crash, this percentage was much higher, almost 55% in 2019 ( 3 ). These statistics imply that lack of proper restraint use among child passengers continues to be a nationwide concern.

At the state level, in Michigan between 2014 and 2018, a total of 72,094 children aged less than 8 years were involved in motor vehicle crashes combining all severities. This was a sharp increase (greater than 70%) from the previous period between 2009 and 2013 ( 4 ). Among the child passengers under the age of 8 years for whom restraint use information was recorded in 2014–2018, less than 73% used child-specific restraint, either a child restraint device (CRD) or a belt-positioning booster seat ( 4 ). Among the child passengers aged less than 8 years who were either unrestrained or improperly restrained, almost 5% suffered fatal or incapacitating injuries ( 4 ). This was significantly higher than the children who sustained fatal or incapacitating injuries but were properly restrained (0.12%) ( 4 ).

Over the past two decades, Michigan has experienced increases in the use of CRD among children under 4 years of age, from 74.5% in 1997 to 98.2% in 2018 ( 5 ). Following the enactment of statewide legislation in Michigan in 2008, booster seat use was also found to increase substantially ( 5 ). However, despite these increases in child restraint use, even less than 55% of children aged between 4 and 7 years used booster seats properly ( 5 ). Research has also demonstrated that children between the ages of 4 and 8 years are the least likely to be protected in the appropriate restraint ( 6 ). There are several potential explanations for the low rate of booster seat use, including a lack of knowledge of the state law and of best practice on the benefits of child appropriate restraints compared with seatbelts alone, as well as differences in risk perception among parents ( 7 , 8 ).

Children should be strapped in appropriate restraints based on their age, weight, or height. Michigan’s Child Passenger Safety Law requires infant and convertible safety seats for children under the age of 4 years and booster seats for children from 4 years of age until they fit in a seatbelt, which is usually at the age of 8 or 9 years ( 9 ). Research has also shown that the appropriate use of CRDs and booster seats can significantly reduce the risk of serious injury and death for children involved in vehicle crashes. Child safety seats reduce fatal injury by 71% for infants (under 1 year old) and by 54% for toddlers (1 to 4 years old) in passenger vehicles. The risk of serious injury for children 1 to 4 years old is 80% lower for children seated in forward-facing CRDs than children restrained merely in safety belts ( 10 ). For infants and toddlers in light trucks, the corresponding reductions are 58% and 59%, respectively ( 11 ). Booster seat use also reduces the risk of serious injury by 45% for children aged between 4 and 8 years, when compared with seatbelt use ( 12 ). Additionally, the Federal Motor Vehicle Safety Standards: Occupant Crash Protection (FVMSS 208) standard provides guidance on the appropriateness of the lap belt of any seatbelt assembly for child passengers less than 6 years old that they are only based on the 50th percentile of the dimensions of the target population ( 13 ). This indicates that most probably seatbelts are not adequately designed for the children in that age group.

While the extant literature provides important insights into child restraint use and the safety benefits of proper restraint use, literature assessing proper use of restraints simultaneously with the seating position of the child passenger is scant. To this end, this study examines the factors that are associated simultaneously with appropriate child restraint use and proper seating position of the child passenger. The data was collected from 263 sites in 30 counties across Michigan in 2015 and 2018 as part of direct observation surveys. Data including the type of restraint use, drivers’ demographic characteristics, and vehicle type along with county-level sociodemographic information were obtained in the process of data collection. The appropriateness of the restraint use was defined based on the child’s age and the corresponding restraint use type. Similarly, the seating position is defined as appropriate if the child was placed in a rear seat.

Literature Review

Previous research identified a myriad of factors that may affect roadway safety ( 14 – 16 ). Prior studies have established that sitting in the rear seat of a vehicle is safer than sitting in the front seat in the event of a crash, and that child safety seats perform better at reducing injuries compared to car seats ( 17 , 18 ). While some studies have separately examined the sitting behavior and restraint use among child passengers and have identified common risk factors for both these variables ( 19 ), some other studies have also found a significant relationship between the two variables ( 20 ). Using Fatality Analysis Reporting System (FARS) ( 21 ) data, a study determined that both child restraint use and rear seating were associated with statistically significant reductions in the likelihood of a child dying in a crash ( 22 ). Despite concerns around the sitting behavior and low restraint use among child passengers, few studies have assessed the child restraint use and child’s seating appropriateness simultaneously. However, if the interrelationship between a child’s restraint use and sitting position depends on unobservable characteristics of the driver and the child passenger, then analyzing these two variables separately may produce bias estimates ( 14 ). A recent study from Ghana investigated sitting behavior and restraint use among child passengers and clearly indicated the existence of an interrelationship between child passengers’ seating position and restraint use. The key factors simultaneously influencing child passenger’s sitting position and restraint use included vehicle type, driver’s gender, driver’s belt use, child’s age, and the presence of other child or adult passenger. The time of the day, and the day of the week also had an influence but only on the child’s sitting position. Female drivers were less likely to position child passengers in the front seat and were more likely to restrain them. However, the age of the driver and the gender of the child did not have a significant effect on either the child’s seating position or the restraint use ( 23 ).

Existing research indicates that various sociodemographic and vehicle characteristics are associated with the likelihood of using CRDs. Robinson et al. ( 24 ) found that both ethnicity and parents’ age were significantly related to knowledge of neonatal car seat location. Results of their study indicated that African American parents and parents aged less than 22 years were least likely to know the correct child restraint use and seating position ( 24 ). Miller et al. ( 25 ) reported that children were more likely to be found unrestrained if the driver was male, young, a drinking driver, and traveling at night. Conversely, a few other studies indicated that male drivers are more likely to properly restrain the child in a vehicle than their female counterparts ( 8 , 26 ). Agran et al. ( 27 ) analyzed the use of restraints for children aged under 9 years in fatal crashes and determined negative associations between child restraint use and the age of children, number of vehicle occupants, older and larger vehicles, traveling between 3:00 and 6:00 a.m., and traveling in rural areas. Driver restraint use was found to be the strongest predictor of child restraint use, and unrestrained and young (less than 18 years old) drivers were associated with lower restraint use of child passengers ( 27 ). In an interview-based study in Australia, Keay et al. ( 28 ) showed that the inappropriate and non-use of restraint among children aged between 2 and 5 years was more likely when the driver was less than 36 years old, had a family with three or more children, or was from a family where a language other than English was spoken at home.

A study from Michigan found that the highest number of children aged between 4 and 8 years riding completely unrestrained (44%) were those traveling in passenger vehicles. With unrestrained drivers, more than 90% of the children were riding completely unrestrained. Moreover, middle-aged drivers (30 to 59 years) had the highest rate of proper booster seat use ( 26 ). A similar study determined that restraint use was low for older children in both pickup trucks and passenger cars ( 29 ). The strongest factor identified by a study that influenced child restraint use was the age of the child, with younger children being restrained more often. The other factors leading to increased child restraint use included cases where the driver was white, the parent was the driver, and there were three or fewer occupants in the vehicle. Interestingly, no significant relationships were found between the parent’s age and the frequency of restraining the child ( 30 ). A study by Rangel et al. ( 31 ) determined that the lowest rate of proper restraint use was among 4 to 8 year old children. The likelihood of any type of restraints being used was 2.3 times higher for Caucasian children than African American children. The compliance with restraint laws among African American children remains significantly lower than the national average. A cross-sectional study in Iran showed the prevalence of child safety seat use was significantly associated with higher income among parents ( 32 ). In a study designed to investigate child safety seat knowledge in post-partum mothers, Spanier et al. ( 33 ) determined that the higher level of education the mother had attained, the more knowledge she had about child passenger safety. African American mothers and mothers from lower socioeconomic statuses performed worse on child passenger safety questions. However, this relationship seemed to be related to the education level in their study.

For most studies, driver restraint use was identified as one of the key predictors for child restraint use ( 33 , 34 ). Macy and Freed ( 35 ) determined that unrestrained driver and traveling in a passenger car were predictors of a lower likelihood of CRD use. Additionally, children riding with a younger driver had lower odds of booster seat use. However, the gender of child or driver was not associated with appropriate restraint use ( 35 ). Glassbrenner ( 36 ) showed that when a driver is restrained, child passengers are restrained 92% of the time compared with 72% of the time when the driver is not restrained ( 36 ). However, a confounding study showed that there were no significant relationships between booster seat use and parental characteristics including parent’s age, education level, and their self-reported seatbelt use ( 37 ).

To summarize, several of these studies have attempted to determine child restraint usage rates, and the safety benefits of CRDs and booster seats. Yet there remains a dearth of research that simultaneously analyzes the appropriate child restraint use and proper seating position of the child passenger in motor vehicles. To address this knowledge gap, this study develops bivariate probit models to estimate the two variables simultaneously using the data from direct observation surveys in Michigan in 2015 and 2018.

Methodology

Site Selection

Study locations were selected to obtain CRD and booster seat use data from a representative sample of target-aged (0 to 7 years) child passengers in the State of Michigan. To ensure the representativeness of the sample, these observations were to be diverse in geographic coverage, vehicle mix, and socioeconomic characteristics of the drivers. To ensure such representativeness while maintaining efficiency of data collection, sites were sampled from 30 counties representing greater than 78% of the target population (children aged 0 to 7). The specific observation sites were selected from a statewide sample of locations expected to yield high volumes of target-aged child passengers, including fast-food restaurants, daycare centers, shopping centers, and recreational sites (e.g., zoos, museums, parks, etc.). Ultimately, a total of 263 sites in 30 different counties in Michigan were selected for data collection. The candidate counties were also classified into four strata based on the past safety belt use rates and vehicle miles traveled (VMT) ( 14 ). Stratum 1 consists of the counties with the highest historical safety belt use rates while stratum 4 has exhibited the lowest belt use rate. Combining counties with similar use and/or misuse rates into strata reduces the within-stratum variability and allows for a reasonable number of observations within each stratum while ensuring desired levels of precision. Figure 1 shows the strata of the counties and the data collection locations. To allow for a direct comparison between the results of these surveys, the same sites were utilized each year, where feasible.

Figure 1.

Data collection locations in different counties in Michigan.

Observer Training

A training program including classroom training and field training specific to direct observation of CRD/booster seat use was conducted each year before the start of the data collection. During the classroom training, data collectors were provided with information to aid them in assessing the age of child passengers, including height/weight information and sample photographs. At the end of the training, observers were tested on their ability to assess the ages of child passengers based on a series of photographs. The classroom training session was followed by practice field data collection at a local recreational location. The purpose of the practice field data collection was to provide observers with an opportunity to gain field experience in assessing child passenger age and determining the type of child restraint use. Observers worked as a group at the start of the field training, followed by practice sessions where they were instructed to record the information needed to the best of their ability. Following the field training, their performance was monitored to ensure consistency among observers. This included comparing the number of target-aged children identified by each observer, as well as the type of restraint used by each observed child. In addition to these training exercises, each data collector received a training manual, as well as all necessary field supplies.

Data Collection

In the data collection, the observers positioned themselves on the roadside near the entrance or exit of each study location as vehicles would be either stopped or moving slowly, ensuring a clear line of sight into the interior of each vehicle. At the primary observation sites where traffic volumes were relatively low, data were also collected from vehicles on the adjacent street. During the weekdays, the survey data collection schedule was arranged such that observations could be conducted at a fast-food restaurant at the start of the day, followed by shopping center locations en route to a daycare center scheduled to be visited later the same day. Each daycare center was researched to determine the start and release times, and other locations (e.g., shopping centers, fast-food restaurants, recreation centers) were also researched to ensure they were still in operation. Weekend data collection was performed at all types of locations, excluding daycare centers.

For all vehicles identified to have a 0 to 7-year-old child passenger, the driver and all target-age child passengers were observed for restraint use and non-use. During the direct observation use surveys, several factors were assessed as part of the data collection including vehicle type (passenger vehicles, sport utility vehicles [SUVs], vans/minivans, or pickup trucks); driver restraint use, gender, age group, and ethnicity; and child restraint use, age, and seating position. Driver restraint use was categorized as belted, not belted, or unknown. The seven restraint categories for each child were: belted, not belted, unknown, rear-facing child safety seat, front-facing child safety seat, high-back booster, or backless booster. For driver’s race, data were categorized into three groups, namely white, black, and other, to have an adequate sample size in each category. The data were collected in the morning and afternoon to obtain the largest possible sample size of child passengers.

Socioeconomic and census information, such as median household income, education level, and population density were also collected for the observation sites to account for some unobserved characteristics of the population in the survey data. The income level and education level were aggregated at county level using the “2013–2017 American Community Survey 5-Year Estimates” available from the U.S. Census Bureau website. Population density was collected for different zip codes of the observation sites using 2010 decennial census data. However, an initial review of these data indicated that none of this information had a significant association with either of the dependent variables, probably because those were aggregated at the county or zip code level and were not specific to individual observations, so they were excluded from further analysis.

Statistical Methodology

This study analyzes the factors that are associated with proper child restraint use and proper seating position of the child passenger simultaneously. Because the dependent variables consist of binary indicator variables (i.e., the child passenger was properly restrained or not, and the seating position was appropriate or not), discrete outcome models are an appropriate analysis framework. Given concerns as to a potential correlation between these two variables of interest, bivariate probit models are well suited to the context of this study. The bivariate probit models were developed to account for the common unobserved factors that are simultaneously associated with the proper child restraint use and seating position of a child passenger. The bivariate probit regression is an extension of the univariate binary probit regression and is designed to model two binary dependent variables that may be simultaneously estimated ( 38 ). A bivariate probit model essentially discounts the correlation between the disturbances resulting in inefficiency in model estimations when the univariate probit models are separately developed ( 39 ). The general form of a bivariate probit model with N observations can be expressed as:

y_{i 1}^{*} = β_{1} X_{i 1} + ε_{i 1}, y_{i 1} = 1, if y_{i 1}^{*} > 0, 0 otherwise

(1)

y_{i 2}^{*} = β_{2} X_{i 2} + ε_{i 2}, y_{i 2} = 1, if y_{i 2}^{*} > 0, 0 otherwise

(2)

where $y_{i j}^{*}$ are latent dependent variables; j = 1 for child restraint use (1 = proper, 0 = improper), j = 2 for child passenger’s seating position in the vehicle (1 = proper, 0 = improper), $β_{1}$ and $β_{2}$ are the vectors of estimable parameters; $X_{i 1}$ and $X_{i 2}$ are the vectors of explanatory variables, and $i = 1, 2, \dots, N$ . The disturbance or error terms, $ε_{i 1}$ and $ε_{i 2}$ , are assumed to follow a bivariate normal distribution;

[ε_{i 1}, ε_{i 2}] ~ N_{2} [0, 0, 1, 1, ρ], - 1 < ρ < 1

(3)

where the correlation coefficient, rho (ρ), is the cross-equation error term indicating the common unobserved correlated factors ( 40 ). In the current study, ρ estimates the correlation between correct child restraint use and correct seating position of a child passenger after the effects of the independent variables in the model are accounted for. In the model results, the sign of the parameter estimates ( $β$ ), exhibits the effect of each variable on both child restraint use and seating position of the child. A positive parameter implies that the factor would increase the likelihood of proper restraint use or proper seating position of a child passenger and vice versa.

More recently, random-effects models have become popular for their capability of accounting for spatial effects and heterogeneity across observations ( 41 ). If this issue is not taken into account and the effects of observable variables are held the same across all observations, predictions might be erroneous resulting from the biased estimated parameters ( 41 ). To address the issues with non-random sampling and unobserved heterogeneity in the data, random effect (intercept) was included in the bivariate probit models, wherein the intercept term was allowed to vary across observations. The random parameter, $β_{ijk}$ for ith observation, jth dependent variable, and kth coefficient can be generally defined as;

β_{ijk} = β_{jk} + ε_{ijk}

(4)

where $β_{jk}$ is the mean of the coefficient and $ε_{ijk}$ is a normally distributed disturbance or error term with a mean of zero and variance σ² ( 40 ).

Analysis and Results

Data Summary

The statewide survey on CRD use was performed between June and August of 2015 and 2018. During these observation periods, a total of 12,567 observations of 0 to 7-year-old child passengers were obtained. Data were screened for missing values in different categories and a total of 10,137 complete observations were utilized for further analysis. Descriptive statistics of the CRD use survey are provided in Table 1.

Table 1.

Descriptive Statistics (N = 10,137)

Factor	Frequency	Percentage	Factor		Percentage
Child restraint use			Vehicle type
Proper use	7,756	76.5	Passenger car	4,529	44.7
Improper use	2,381	23.5	Sport utility vehicle	3,395	33.5
Child seating position			Van or minivan	1,702	16.8
Proper	9,802	96.7	Pickup truck	511	5.0
Improper	335	3.3	Location type
Child age			Shopping	4,245	41.9
Under 2 years	1,598	15.8	Fast food	3,056	30.1
2 to 3 years	3,471	34.2	Recreation	2,119	20.9
4 to 7 years	5,068	50.0	Daycare	717	7.1
Number of children			Geographic location
One child	6,609	65.2	Stratum 1	2,439	24.0
Multiple children	3,528	34.8	Stratum 2	3,282	32.4
Driver belt use			Stratum 3	2,328	23.0
Driver belted	9,913	97.8	Stratum 4	2,088	20.6
Driver not belted	224	2.2	Weather
Driver gender			Clear	8,823	87.0
Male	3,581	35.3	Light fog/rain	1,314	13.0
Female	6,556	64.7	Day of the week
Driver age			Weekend	1,793	17.7
Under 30	2,344	23.1	Weekday	8,344	82.3
30 to 60	7,295	72.0	Time of the day
Over 60	498	4.9	Morning	5,238	51.7
Driver race			Afternoon	4,899	48.3
White	8,718	86.0
African American	923	9.1
Asian/Hispanic/others	496	4.9

For the purpose of this study, the “appropriate” child restraint use was defined based on the current Michigan law enacted in 2008 ( 42 ). Thus, the proper restraint for children aged between 0 to 2 years was to be restrained in a rear-facing child safety seat. Children aged more than 2 years and less than 4 years were defined to be correctly restrained if they were placed in either a rear-facing or a forward-facing child safety seat. Children aged between 4 and 7 years should be placed in a child safety seat or booster seat (high-back or backless) to be considered properly restrained.

Similarly, the proper seating position was defined as when the child was seated in the middle or back rows of the vehicle and occupying their own seating position (i.e., not on another occupant’s lap, regardless of restraint use status), and conversely, when the child was seated at the front seat or someone’s lap, it was considered an improper seating position.

The summary statistics of the data reveal that:

The proper child restraint use rates were 91.4% and 61.6% for children under 4 years and 4 to 7 years, respectively, indicating a much lower rate of restraint use for children aged 4 years or older.

Few children were improperly seated (in the front seat), approximately 0.4% and 6.2% for children under 4 years and 4 to 7 years, respectively.

In approximately one-third of the observations, there were multiple child passengers in the vehicle.

Overall, more than 97% of drivers were found to be properly belted. Middle-aged (30–59 years) drivers account for almost three-quarters of the sample, followed by the young drivers (16–29 years), and less than 5% of the drivers were old (60+ years). The majority of drivers observed were white.

Almost two-thirds of the drivers were female.

Among the different vehicle types, passenger cars were most commonly observed, followed by SUVs and van/minivans, and lastly pickup trucks.

With regard to the location types, shopping centers and fast-food restaurants were most commonly observed. Moreover, these locations have comparable representations from all four strata.

Lastly, the majority of the observations were obtained during clear weather, and during weekdays.

Model Results

For this study, all observations where at least one child passenger was present along with the driver in a motor vehicle were analyzed. A total of 10,137 such observations were analyzed after removing all unknown data. As the likelihood of proper child restraint use (age-appropriate) is not independent of the likelihood of proper seating position (rear/middle seat) of the child passenger, it is important to understand the correlation between these two variables. To analyze the factors that simultaneously affect both proper child restraint use and proper seating position of the child passenger accounting for the unobserved heterogeneity in the data, a bivariate probit model with random effects (intercept) was developed. The model included the information related to the child passengers and drivers, along with vehicle-, and site-related factors. The dependent variables in all models are coded as a binary indicator implying proper and improper child restraint use and child passenger’s seating position. The regression analysis in this study was carried out using NLOGIT6.

Table 2 displays the results of the final bivariate probit model for both the dependent variables. The independent variables in all models included child passenger’s age, number of child passengers in the vehicle, driver’s age, gender, race, vehicle type, stratum, weather, time of the day, and day of the week, each statistically significant at a 90% confidence level for predicting either child restraint use or child’s seating position, or both. Along with the socioeconomic factors as stated above, the types of location did not have any significant association with either of the dependent variables and were removed from the models.

Table 2.

Results of Bivariate Probit Models for Proper Child Restraint Use and Proper Seating of Child Passengers (N = 10,137)

Parameter	Proper child restraint use				Proper child seating
Parameter	Estimate	Standard error	z-Value	p-Value	Estimate	Standard error	z-Value	p-Value
Intercept	0.417	0.124	3.35	0.001	2.865	0.235	12.19	<0.001
Child age = up to 2 years	Baseline				Baseline
Child age = 2 to 3 years	−0.773	0.067	−19.12	<0.001	−0.384	0.159	−1.16	0.246
Child age = 4 to 7 years	−1.366	0.050	−17.45	<0.001	−1.352	0.141	−9.62	<0.001
Number of children in the vehicle = 1	Baseline				Baseline
Number of children in the vehicle > 1	0.037	0.038	0.99	0.324	0.615	0.063	9.69	<0.001
Driver belt use = Not properly belted	Baseline				Baseline
Driver belt use = Properly belted	0.688	0.104	6.59	<0.001	0.342	0.151	2.26	0.024
Driver age = 16 to 29 years	Baseline				Baseline
Driver age = 30 to 59 years	0.079	0.043	1.85	0.064	0.040	0.065	0.61	0.542
Driver age = 60+ years	−0.039	0.087	−0.45	0.656	0.223	0.121	1.84	0.066
Driver gender = Male	Baseline				Baseline
Driver gender = Female	−0.058	0.038	−1.53	0.127	−0.196	0.060	−3.25	0.001
Driver race = White	Baseline				Baseline
Driver race = Black	−0.392	0.059	−6.60	<0.001	−0.141	0.094	−1.51	0.132
Driver race = Other	−0.360	0.077	−4.66	<0.001	−0.099	0.109	−0.91	0.364
Vehicle type = Passenger car	Baseline				Baseline
Vehicle type = SUV	0.479	0.041	11.80	<0.001	−0.160	0.062	−2.57	0.010
Vehicle type = Van/minivan	0.876	0.056	15.74	<0.001	−0.162	0.080	−1.76	0.078
Vehicle type = Pickup truck	0.150	0.077	1.96	0.050	−1.029	0.087	−11.88	<0.001
Stratum 1	Baseline				Baseline
Stratum 2	−0.260	0.050	−7.18	<0.001	−0.366	0.078	−4.69	<0.001
Stratum 3	−0.237	0.054	−6.25	<0.001	−0.305	0.080	−3.81	0.005
Stratum 4	−0.268	0.054	−4.95	<0.001	−0.196	0.083	−2.37	0.018
Weekday	Baseline				Baseline
Weekend	0.183	0.046	3.95	<0.001	0.083	0.064	1.30	0.194
Weather = Clear	Baseline				Baseline
Weather = Light rain or fog	0.504	0.055	9.20	<0.001	0.298	0.084	3.54	<0.001
Time of the day = Morning	Baseline				Baseline
Time of the day = Afternoon	0.155	0.036	4.28	0.001	0.017	0.053	0.31	0.753
Rho (ρ)	0.8782	0.004	21.25	<0.001
Log-likelihood	−4453.66

Note: SUV = sport utility vehicle.

In addition to the model estimated parameters (β) for each independent variable, the marginal effects for each variable are also of interest. The marginal effects provide a means of examining the magnitude of the effect of each independent variable on proper child restraint use or proper seating position of the child passenger as presented in Table 3.

Table 3.

Marginal Effects of the Predictors (N = 10,137)

Parameter	Proper child restraint use		Proper child seating
Parameter	Estimate	p-Value	Estimate	p-Value
Child age = 2 to 3 years	−0.207	<0.001	−0.031	0.246
Child age = 4 to 7 years	−0.395	<0.001	−0.078	<0.001
Number of children in the vehicle > 1	0.011	0.324	0.033	<0.001
Driver belt use = Properly belted	0.211	<0.001	0.028	0.024
Driver age = 30 to 59 years	0.023	0.064	0.003	0.542
Driver age = 60+ years	−0.011	0.656	0.012	0.066
Driver gender = Female	−0.017	0.127	−0.012	0.001
Driver race = Black	−0.118	<0.001	−0.010	0.132
Driver race = Other	−0.108	<0.001	−0.007	0.364
Vehicle type = SUV	0.135	<0.001	−0.011	0.010
Vehicle type = Van/minivan	0.229	<0.001	−0.011	0.078
Vehicle type = Pickup truck	0.042	0.050	−0.125	<0.001
Stratum 2	−0.075	<0.001	−0.026	<0.001
Stratum 3	−0.070	<0.001	−0.022	0.005
Stratum 4	−0.079	<0.001	−0.014	0.018
Weekend	0.052	<0.001	0.005	0.194
Weather = Light rain or fog	0.137	<0.001	0.016	<0.001
Time of the day = Afternoon	0.045	0.001	0.001	0.753

Note: SUV = sport utility vehicle.

Child Restraint Use

The results of this analysis yielded several interesting findings. As can be seen from Table 2, the likelihood of proper child restraint use significantly reduced as the child’s age increased, especially when the child was aged between 4 and 7 years. This agrees with findings from other studies that determined a higher likelihood of proper restraint use for younger children compared with those aged 4 years or older ( 20 ). Although having more than one child passenger in the vehicle was not statistically significant for child restraint use, the positive parameter indicates that it could be given some consideration without disregarding the statistical significance. Similar to previous studies ( 19 , 23 ), the child was more likely to be properly restrained if the driver of the vehicle was properly belted. While middle-aged (30 to 59 years) drivers were found to be more likely to properly restrain the child passenger compared with the young drivers (16 to 29 years), old drivers were not found to have a significant association with the proper child restraint use. If majority of these elderly drivers traveling with child passengers were the grandparents of the child passengers, then a plausible explanation for this finding could be that the grandparent is typically not the child’s primary care-giver and might not have adequate information on the use of CRD/booster seat ( 5 ). This corresponds with some previous studies that allude to a lower rate of correct installations and restraint use for the child passengers by their grandparents ( 43 ).

The results also showed that, in proper child restraint use, there is no significant difference between male and female drivers, a finding that is consistent with previous studies ( 8 , 44 ). The white drivers were most likely to use the child restraint properly compared with African American drivers or drivers of other races. Drivers of passenger vehicles were the least likely to restrain their children properly, which has also been shown in other studies ( 35 ). Conversely, drivers of vans/minivans were the most likely to properly restrain their child passengers. Drivers in stratum 1, where historically driver’s seat belt use is the highest in the state, were most likely to properly restrain the child passenger. On the other hand, drivers in stratum 4, which comprises highly populated locations, were least likely to use the child restraint properly. Furthermore, the likelihood of proper child restraint use increased in light fog or rain compared with clear weather conditions. Finally, the use of proper restraint for child passengers was higher over the weekends and in the afternoons.

Child’s Seating Position

For child’s proper seating position in being seated in the middle or rear seats, the age of the child is negatively associated with the proper seating position, implying that the higher the age of the child passenger, the lower the likelihood of the child being properly seated. Having more than one child passenger in the vehicle also increased the likelihood that the child passenger would be properly seated. Moreover, similar to proper child restraint use, a properly belted driver significantly increased the likelihood of the child being properly positioned in the vehicle. With regard to driver age, older drivers (60+ years) particularly seemed to have a positive association with the child passenger being properly seated. These findings are reflective of the broader research literature ( 14 , 45 ) and may be attributed to greater risk aversion or more cautious driving habits among the oldest and most experienced group of drivers, in general. Furthermore, male drivers were more likely to properly position a child passenger than were their female counterparts. Similar to child restraint use, white drivers were most likely to place the child passenger correctly compared with drivers of other races. The proper seating of a child passenger was the least in pickup trucks and was the most likely in passenger cars. Additionally, the likelihood of positioning the child passenger in the rear or middle seats was the highest in stratum 1, followed by strata 4, 3, and 2, respectively. The proper seating positioning increased during light fog or rain compared with clear weather. The proper placing of a child passenger was greater during the weekends and in the afternoons.

Discussion and Recommendations

As shown in Table 2, the results displayed comparable parameter estimates for most of the independent variables in child restraint use and child seating position models. The correlation parameter (ρ = 0.8782, p-value < 0.001) is large and statistically significant, indicating the presence of common unobserved factors that may affect the proper use of child restraint and seating position of the child passenger. The correlation coefficient is positive, indicating that the use of proper child restraint is more likely to have the correct seating position of the child passenger as well. This finding is important and consistent with previous literature that identified a correlation between child passenger’s restraint use and seating position ( 23 ).

Additionally, model results for both dependent variables (child restraint use and seating position) indicate that the correct restraint use and seating position significantly decreases with the increase in child passenger’s age. This could be explained by the difference in the perceived safety of child passengers by parents as they feel a strong need to properly secure infants in CRDs, and become less concerned when the child grows older. Driver’s belt use was one of the most important variables that influenced child passenger’s restraint use and seating position, indicating belted drivers to be more associated with proper child restraint use and seating position. This result is expected as belted drivers are more risk-averse and already aware of the importance of good safety practices.

The results also indicate that female drivers were less likely to correctly position the child passenger. Although this factor was not statistically significant for child restraint use, the negative parameter suggests that it might be given some consideration, while not disregarding the statistical results. Although some previous studies indicates similar findings, he reason behind this finding cannot be directly explained from the results and it warrants further investigation.

The likelihood for a child passenger to be properly restrained and positioned was significantly lower on the weekdays than on weekends. Similarly, children were more likely to be properly restrained in the afternoons than in the mornings. While the reasons for these findings cannot be definitively explained, and require further investigation, a previous study associated less traffic in the morning with unrestrained child passengers ( 46 ).

The marginal effects for each independent variable are also comparable between the two dependent variables, although the extent or magnitude of these effects vary between the child restraint use and seating position.

To ensure proper CRD and booster seat use, parents must be provided with education and training on child restraint periodically throughout their child’s growth and development, particularly when a new CRD is utilized or upgrade/modification to the current CRD becomes necessary. Parents should also be encouraged to follow the current NHTSA CRD transitioning guidelines, which advise keeping children in each restraint type, including rear-facing, forward-facing, and booster seats, for as long as possible before moving them up to the next type of restraint ( 47 ). Although the sample of unbelted drivers was too small to draw meaningful conclusions, prior CRD surveys in Michigan have found that the most significant driver-related determinant of CRD or booster seat use among child passengers was driver belt use. CRD/booster use has historically been significantly lower when the driver was not belted appropriately as also seen in this study. Unbelted drivers present the greatest area of opportunity and should be the focus of future education and outreach programs aimed at informing the public of the importance of appropriate child restraint use. Similar programs have demonstrated effectiveness in enhancing safety belt use among Michigan drivers. Several educational/training opportunities are available to parents from different locations including hospitals, daycare facilities, healthcare providers, and public health officials. Also, NHTSA-certified inspectors are often available at fire stations and police stations at many locations throughout the State of Michigan. Additionally, mandatory legislation and stricter enforcement working synergistically with promotional campaigns of child restraint use could have a promising impact. The media can assist with education and advocacy for child passenger restraint in motor vehicles. Increasing the required age for forward-facing CRDs to booster seat use in child passenger restraint laws could also be an effective way to increase restraint use among older children.

Study Limitations

This study has some important limitations that can be looked further into future research. The data on CRD use and the seating position of the child passenger along with other variables of interest were collected by trained observers. However, errors are to be expected with respect to the data collection process, particularly for variables that require careful judgment such as age or race. While all data collectors underwent extensive training for repeatability and reliability, there may be some error that was necessarily introduced as a consequence of the data collection method. Additionally, data collection was performed only during the daytime to obtain the largest possible sample size of child passengers and to ensure maximum visibility by the observers. These are some areas where naturalistic driving data may provide some advantages given the availability of cameras both within and outside the motor vehicle. Direct observation studies also help to evaluate the rate of overall restraint use for a population, but do not shed light on individual patterns of use. Follow-up research can be carried out that uses alternative approaches, such as interventional or case-control studies, supplemental surveys, or focus groups, which would allow for an exploration of parents’ or drivers’ socioeconomic status, travel patterns, and driving behavior. Moreover, the authors also acknowledge that some of the apparent associations in the results are confounding and could be caused by small sample size or biased sampling. For example, the sample size for the older drivers was small (less than 5%) and showed no significant association with proper child restraint use. However, the older drivers demonstrated a greater likelihood of positioning the child passenger correctly than did the young drivers. The analysis should be extended further to examine these correlations, particularly for attributes with a small sample size to provide more definitive answers. Furthermore, the modeling technique also limits the capability of accounting for the correlations between the independent variables and focuses on the associations of the independent variables with the dependent variables. Therefore, some of the apparent findings, such as the lower likelihood of proper positioning of child passengers in pickup trucks, could also be influenced by having only one row of seats in the vehicle. Similarly, the apparent associations of the driver characteristics (e.g., gender or race) might also be reflective of various other sociodemographic factors that are not captured in the model, a limitation the readers should be mindful of.

Summary and Conclusion

This study simultaneously analyzes the factors that are associated with the child restraint use and seating position of child passengers of motor vehicles. For this study, the analysis was carried out on data that were collected from 263 sites across 30 counties in Michigan in 2015 and 2018 as part of direct observation surveys. The categories for the child restraint use included in this analysis were: properly restrained indicating age-appropriate child restraint use versus not properly restrained which includes using some restraint but not child safety device. Similarly, the other dependent variable was the proper seating position, which is essentially the rear or middle seats of the vehicle, versus improper seating position implying the front seats. Additional data considered in the study included child passenger age, number of child passengers in the vehicle, driver belt use, driver age, gender, and race, vehicle type, stratum, weather, time of the day, and day of the week.

In this study, more than 10,000 observations were analyzed where at least one child passenger was present in a vehicle. Preliminary analysis revealed that the proper restraint use rates were 91.4% and 61.6% for children under 4 years and 4 to 7 years, respectively. Few children, approximately 0.4% and 6.2% of children under 4 years and 4 to 7 years, respectively, were improperly seated in the front seat.

A bivariate probit model with random effects was developed that simultaneously estimated both the use of child restraint and the seating position of the child passenger in a vehicle, accounting for the unobserved heterogeneity in the data. The use of the bivariate model accounts for the correlation between the two dependent variables and provides increased efficiency compared with the separate univariate probit models. The correlation parameter (ρ) of the final bivariate model was found to be large and highly significant implying the presence of common unobserved factors affecting the use of child restraint as well as the seating position of the child passenger. The positive correlation parameter also indicated that the proper use of child restraint is more likely to have the child passenger correctly seated in the vehicle and vice versa.

Factors including child’s age, the number of child passenger(s) in a vehicle, driver’s belt use, driver’s age, gender, and race, vehicle type, strata, weather, time of the day, and day of the week were all found to be significantly associated with either child restraint use or child’s seating position, or both. In general, the effects of different factors associated with the two dependent variables were mostly similar. Location type along with the aggregated socioeconomic factors including population density, income, and education level did not have any significant association with proper child restraint use and child’s proper seating position.

The results reveal that the use of proper child restraint and proper child seating position are less likely with the increase in age of the child passenger. When more than one child is present in the vehicle, the likelihood of proper restraint use and seating of a child passenger is greater, and this is particularly significant for the appropriate seating position of the child. Driver belt use was found to be one of the key predictors indicating a higher likelihood of proper child restraint use and proper placement of the child passenger, if the driver is belted. Young drivers are, in general, found to be least responsible for using proper child restraint or placing the child in the correct seats. While male drivers were more likely to use the child restraint and place the child correctly, white drivers were found to be most likely to do the same. Proper child restraint use was least likely in passenger vehicles, but it was more common to have the child correctly seated in passenger cars. Furthermore, the proper use of child restraint and seating position of the child passenger is most likely in the stratum where the seatbelt use is historically the highest in the State of Michigan. Lastly, for both dependent variables, traveling in light rain or fog, on weekends, and in the afternoons increased the likelihood of proper use of child restraint as well as proper placement of the child passenger in a motor vehicle.

Overall, the results of this study support the previous research findings and provide more evidence that the interaction of various demographic, vehicle, and site-related characteristics are significantly associated with the use of child restraint and the seating position of the child passenger in motor vehicles. More importantly, the study contributes to the limited body of knowledge on the correlation between child restraint use and child passenger’s seating position. The findings from this study may help guide the development and implementation of intervention programs and policies by state and federal agencies to improve occupant safety practices. A particularly interesting and useful feature of the current study is the simultaneous analysis of the factors associated with child restraint use and seating position of child passengers. In particular, the correlation between the child passenger restraint use and seating in motor vehicles is explored, along with accounting for unobserved heterogeneity in the data.

Footnotes

Acknowledgements

The study utilized the data collected for a project supported and funded by the Michigan Office of Highway Safety Planning (OHSP).

Author Contributions

The authors confirm contribution to the paper as follows: study conception and design: M. Chakraborty; data collection: M. Chakraborty; analysis and interpretation of results: M. Chakraborty, Md. S. Mahmud; draft manuscript preparation: M. Chakraborty, Md. S. Mahmud, T. Gates. 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.

Funding

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

ORCID iDs

Meghna Chakraborty

Md Shakir Mahmud

Timothy J. Gates

The opinions expressed in this paper are those of the authors and not those of the Michigan Office of Highway Safety Planning (OHSP). OHSP disclaims any liability, of any kind, or for any reason, that might otherwise arise out of any use of this publication or the information or data provided in the publication.

References

Motor Vehicle Safety. Child Passenger Safety: Get the Facts. Centers for Disease Control and Prevention. https://www.cdc.gov/motorvehiclesafety/child_passenger_safety/cps-factsheet.html. Accessed June 18, 2020.

WISQARS^TM Injury Data. WISQARS^TM– Web-Based Injury Statistics Query and Reporting System. Centers for Disease Control and Prevention, Injury Center. https://www.cdc.gov/injury/wisqars/index.html. Accessed June 18, 2020.

National Highway Traffic Safety Administration. 2019 Children Traffic Safety Fact Sheet. Publication DOT HS 812 271. U.S. Department of Transportation, Washington, D.C., 2021.

MTCF. Michigan Traffic Crash Facts. https://www.michigantrafficcrashfacts.org/. Accessed June 30, 2019.

Gates

T. J.

Kay

J. J.

Crane

Brattin

2018 Direct Observation Survey of Child Restraint and Booster Seat Use and Misuse. Office of Highway Safety Planning, Dimondale, MI, 2018.

Chang

B. A.

Child Passenger Safety: Potential Impact of the Washington State Booster Seat Law on Childcare Centers. Injury Prevention, Vol. 8, No. 4, 2002, pp. 284–288. https://doi.org/10.1136/ip.8.4.284.

Aita-Levy

Henderson

Factors Affecting Booster Seat Use. Clinical Pediatrics, Vol. 55, No. 12, 2016, pp. 1132–1137. https://doi.org/10.1177/0009922815615824.

Tavakoli

Davoodi

S. R.

Azimmohseni

Factors Affecting Use and Nonuse of Child Safety Car Seats in Gorgan, Iran. Traffic Injury Prevention, Vol. 20, No. 6, 2019, pp. 661–666. https://doi.org/10.1080/15389588.2019.1634264.

Child Passenger Safety. Michigan State Police. https://www.michigan.gov/msp/divisions/ohsp/child-passenger-safety Accessed June 20, 2020.

10.

Arbogast

K. B.

Durbin

D. R.

Cornejo

R. A.

Kallan

M. J.

Winston

F. K.

An Evaluation of the Effectiveness of Forward Facing Child Restraint Systems. Accident Analysis & Prevention, Vol. 36, No. 4, 2004, pp. 585–589. https://doi.org/10.1016/S0001-4575(03)00065-4.

11.

National Highway Traffic Safety Administration. Fatality Reduction by Safety Belts for Front-Seat Occupants of Cars and Light Trucks. Annals of Emergency Medicine, Vol. 37, No. 6, 2001, pp. 728–729. https://doi.org/10.1067/mem.2001.115539.

12.

Arbogast

K. B.

Jermakian

J. S.

Kallan

M. J.

Durbin

D. R.

Effectiveness of Belt Positioning Booster Seats: An Updated Assessment. Pediatrics, Vol. 124, No. 5, 2009, pp. 1281–1286. https://doi.org/10.1542/peds.2009-0908.

13.

49 CFR § 571.208 - Standard No. 208; Occupant Crash Protection. Legal Information Institute, Cornell Law School. https://www.law.cornell.edu/cfr/text/49/571.208. Accessed November 2, 2021.

14.

Chakraborty

Singh

Savolainen

P. T.

Gates

T. J.

Examining Correlation and Trends in Seatbelt Use Among Occupants of the Same Vehicle Using a Bivariate Probit Model. Transportation Research Record: Journal of the Transportation Research Board, 2021. 2675: 288-298.

15.

Chakraborty

Gates

T. J.

Chakraborty

Causal Analysis and Classification of Traffic Crash Injury Severity Using Machine Learning Algorithms. 8th Road Safety & Simulation International Conference, Athens, Greece, 2022.

16.

Chakraborty

Stapleton

S. Y.

Ghamami

Gates

T. J.

Safety Effectiveness of All-Electronic Toll Collection Systems. Advances in Transportation Studies, Vol. 2, 2020, pp. 127–142.

17.

Winston

F. K.

Chen

I. G.

Elliott

M. R.

Arbogast

K. B.

Durbin

D. R.

Recent Trends in Child Restraint Practices in the United States. Pediatrics, Vol. 113, No. 5, 2004, pp. e458–e464. https://doi.org/10.1542/peds.113.5.e458.

18.

Doyle

J. J.

Levitt

S. D.

Evaluating the Effectiveness of Child Safety Seats and Seat Belts in Protecting Children From Injury. Economic Inquiry, Vol. 48, No. 3, 2008, pp. 521–536. https://doi.org/10.1111/j.1465-7295.2008.00170.x.

19.

Pan

Jiang

Bilston

Brown

Shen

Restraint Use and Seating Position Among Child Car Passengers: An Observational Study in Shanghai. Accident Analysis & Prevention, Vol. 43, No. 6, 2011, pp. 2195–2199. https://doi.org/10.1016/j.aap.2011.06.014.

20.

Lennon

Siskind

Haworth

Rear Seat Safer: Seating Position, Restraint Use and Injuries in Children in Traffic Crashes in Victoria, Australia. Accident Analysis & Prevention, Vol. 40, No. 2, 2008, pp. 829–834. https://doi.org/10.1016/j.aap.2007.09.024.

21.

NHTSA. Fatality Analysis Reporting System (FARS). https://www.nhtsa.gov/research-data/fatality-analysis-reporting-system-fars. Accessed November 4, 2019.

22.

Glass

R. J.

Segui-Gomez

Graham

J. D.

Child Passenger Safety: Decisions About Seating Location, Airbag Exposure, and Restraint Use. Risk Analysis, Vol. 20, No. 4, 2000, pp. 521–528. https://doi.org/10.1111/0272-4332.204049.

23.

Aidoo

E. N.

Ackaah

Appiah

S. K.

Appiah

E. K.

Addae

Alhassan

A Bivariate Probit Analysis of Child Passenger’s Sitting Behaviour and Restraint Use in Motor Vehicle. Accident Analysis & Prevention, Vol. 129, 2019, pp. 225–229. https://doi.org/10.1016/j.aap.2019.05.022.

24.

Robinson

D. C.

Reitan

R. E.

Jones

G. N.

Gist

R. S.

Knowledge of Neonatal Car Seat Location in a Predominantly Disadvantaged Prenatal Population. Clinical Pediatrics, Vol. 41, No. 7, 2002, pp. 455–459. https://doi.org/10.1177/000992280204100701.

25.

Miller

T. R.

Spicer

R. S.

Lestina

D. C.

Who Is Driving When Unrestrained Children and Teenagers Are Hurt?

Accident Analysis & Prevention, Vol. 30, No. 6, 1998, pp. 839–849. https://doi.org/10.1016/S0001-4575(98)00037-2.

26.

Eby

D. W.

Bingham

C. R.

Vivoda

J. M.

Ragunathan

Use of Booster Seats by Michigan Children 4–8 Years of Age. Accident Analysis & Prevention, Vol. 37, No. 6, 2005, pp. 1153–1161. https://doi.org/10.1016/j.aap.2005.06.013.

27.

Agran

P. F.

Anderson

C. L.

Winn

D. G.

Factors Associated With Restraint Use of Children in Fatal Crashes. Pediatrics, Vol. 102, No. 3, 1998, p. e39. https://doi.org/10.1542/peds.102.3.e39.

28.

Keay

Hunter

Brown

Bilston

L. E.

Simpson

J. M.

Stevenson

Ivers

R. Q.

Child Restraint Use in Low Socio-Economic Areas of Urban Sydney During Transition to New Legislation. Accident Analysis & Prevention, Vol. 50, 2013, pp. 984–991. https://doi.org/10.1016/j.aap.2012.08.002.

29.

Eby

D. W.

Kostyniuk

L. P.

Vivoda

J. M.

Restraint Use Patterns for Older Child Passengers in Michigan. Accident Analysis & Prevention, Vol. 33, No. 2, 2001, pp. 235–242. https://doi.org/10.1016/S0001-4575(00)00038-5.

30.

Meulenberg

P. S.

Parental Factors Affecting Child Restraint Device Use in Children Aged 3–5 Years. Doctoral dissertation. Grand Valley State University, Allendale, MI, 1998.

31.

Rangel

S. J.

Martin

C. A.

Brown

R. L.

Garcia

V. F.

Falcone

R. A.

Alarming Trends in the Improper Use of Motor Vehicle Restraints in Children: Implications for Public Policy and the Development of Race-Based Strategies for Improving Compliance. Journal of Pediatric Surgery, Vol. 43, No. 1, 2008, pp. 200–207. https://doi.org/10.1016/j.jpedsurg.2007.09.045.

32.

Moradi

Khanjani

Nabipour

A. R.

Determinants of Child Safety Seat Use Among Parents in an International Safe Community, Tehran, Iran. Traffic Injury Prevention, Vol. 20, No. 8, 2019, pp. 844–848. https://doi.org/10.1080/15389588.2019.1658872.

33.

Spanier

A. J.

Mercante

Barkemeyer

B. M.

Child Safety Seat Knowledge Among Postpartum Mothers in an Urban Setting. Southern Medical Journal, Vol. 95, No. 9, 2002, pp. 1017–1022.

34.

Ebel

B. E.

Koepsell

T. D.

Bennett

E. E.

Rivara

F. P.

Too Small for a Seatbelt: Predictors of Booster Seat Use by Child Passengers. Pediatrics, Vol. 111, No. 4, 2003, pp. e323–e327. https://doi.org/10.1542/peds.111.4.e323.

35.

Macy

M. L.

Freed

G. L.

Child Passenger Safety Practices in the U.S. American Journal of Preventive Medicine, Vol. 43, No. 3, 2012, pp. 272–281. https://doi.org/10.1016/j.amepre.2012.05.023.

36.

Glassbrenner

The Use of Child Restraints in 2002. National Highway Traffic Safety Administration, U.S. Department of Transportation, Washington, D.C., 2003.

37.

Bracchitta

K. M.

Factors Influencing Parental Use of Booster Seats for Their Children. Journal of Clinical Psychology in Medical Settings, Vol. 13, No. 3, 2006, pp. 270–281. https://doi.org/10.1007/s10880-006-9023-x.

38.

Gkritza

Modeling Motorcycle Helmet Use in Iowa: Evidence From Six Roadside Observational Surveys. Accident Analysis & Prevention, Vol. 41, No. 3, 2009, pp. 479–484. https://doi.org/10.1016/j.aap.2009.01.009.

39.

Greene

W. H.

Nlogit Version 4.0 Reference Guide. Econometric Software Inc., Plainview, NY, 2007.

40.

Lee

Mao

Abdel-Aty

Use of Bivariate Random-Parameter Probit Model to Analyze the Injury Severity of Highway Traffic Crashes Involving School-Age Children. Transportation Research Record: Journal of the Transportation Research Board, 2021. 2675: 530–537.

41.

Mannering

F. L.

Shankar

Bhat

C. R.

Unobserved Heterogeneity and the Statistical Analysis of Highway Accident Data. Analytic Methods in Accident Research, Vol. 11, 2016, pp. 1–16. https://doi.org/10.1016/j.amar.2016.04.001.

42.

Michigan Legislature. Section 257.710e. http://www.legislature.mi.gov/(S(g4jgxgx4dya4zv2kcwtnrzoi))/mileg.aspx?page=GetObject&objectname=mcl-257-710e. Accessed June 30, 2020.

43.

O’Neil

Bull

M. J.

Slaven

J. E.

Talty

J. L.

Grandparents and Child Passenger Safety. Accident Analysis & Prevention, Vol. 49, 2012, pp. 354–359. https://doi.org/10.1016/j.aap.2012.02.011.

44.

National Highway Traffic Safety Administration. Traffic Safety Facts: Additional Analysis of National Child Restraint Use Special Study: Characteristics of Those not Restrained. Publication DOT HS 812 477. U.S. Department of Transportation, Washington, D.C., 2018.

45.

Vivoda

J. M.

Eby

D. W.

Kostyniuk

L. P.

Differences in Safety Belt Use by Race. Accident Analysis & Prevention, Vol. 36, No. 6, 2004, pp. 1105–1109. https://doi.org/10.1016/j.aap.2003.04.001.

46.

Sam

E. F.

Don’t Learn Safety by Accident: A Survey of Child Safety Restraint Usage Among Drivers in Dansoman, Accra. Journal of Transport & Health, Vol. 2, No. 2, 2015, pp. 160–165. https://doi.org/10.1016/j.jth.2014.08.003.

47.

National Highway Traffic Safety Administration (NHTSA). NHTSA Releases New Child Seat Guidelines. https://one.nhtsa.gov/portal/site/NHTSA/menuitem.554fad9f184c9fb0cc7ee21056b67789/?vgnextoid=47818846139ce210VgnVCM10000066ca7898RCRD&vgnextchannel=c9f64dc9e66d5210VgnVCM100000656b7798RCRD&vgnextfmt=default. Accessed October 29, 2021.