Elsevier

Accident Analysis & Prevention

Volume 49, November 2012, Pages 193-202
Accident Analysis & Prevention

A dynamic analysis of motorcycle ownership and usage: A panel data modeling approach

https://doi.org/10.1016/j.aap.2011.03.006Get rights and content

Abstract

This study aims to develop motorcycle ownership and usage models with consideration of the state dependence and heterogeneity effects based on a large-scale questionnaire panel survey on vehicle owners. To account for the independence among alternatives and heterogeneity among individuals, the modeling structure of motorcycle ownership adopts disaggregate choice models considering the multinomial, nested, and mixed logit formulations. Three types of panel data regression models – ordinary, fixed, and random effects – are developed and compared for motorcycle usage. The estimation results show that motorcycle ownership in the previous year does exercise a significantly positive effect on the number of motorcycles owned by households in the current year, suggesting that the state dependence effect does exist in motorcycle ownership decisions. In addition, the fixed effects model is the preferred specification for modeling motorcycle usage, indicating strong evidence for existence of heterogeneity. Among various management strategies evaluated under different scenarios, increasing gas prices and parking fees will lead to larger reductions in total kilometers traveled.

Highlights

► This study develops motorcycle ownership and usage models based on a large-scale, nationwide two-wave panel survey in Taiwan. ► Three types of discrete choice models − multinomial, nested, and mixed logit − are developed for motorcycle ownership. ► Three types of panel data regression models − ordinary, fixed, and random effects − are developed for motorcycle usage. ► The state dependence effect and heterogeneity do exist in motorcycle ownership and usage decisions, respectively. ► The management strategies of increasing gas prices and parking fees will lead to larger reductions in total motorcycle kilometers traveled and total motorcycle crashes.

Introduction

In many Asian countries such as China, Indonesia, Malaysia, Taiwan, Thailand, and Vietnam, motorcycles are a primary mode of urban transportation. Due to low purchase and running costs and convenient parking, demand for motorcycles has continuously risen in these countries. Traffic congestion, accidents, parking disorder, and air pollution are the inevitable consequences of high rates of motorcycle ownership and usage. Especially, in the context of accident analysis, because of limited protection design of motorcycles in comparison to cars, motorcycles are the most dangerous form of motorized transport, with injury rates eight times, and fatality rates 35 times that of car occupants (per vehicle mile traveled) (NHTSA, 2007, Ranney et al., 2010). In most developed countries, motorcycle fatalities typically comprise around 5–18% of overall traffic fatalities (Mohan, 2002, Koornstra et al., 2003, WHO, 2006). This proportion reflects the relatively low ownership and usage of motorcycles in many developed countries. However, the ownership and usage of motorcycles and other two-wheelers is generally relatively high in many developing countries. Reflecting this difference, the levels of motorcyclist fatalities as a proportion of those injured on the roads are typically higher in developing countries than in developed countries. For example, in India, 69% of the total number of motor vehicles are motorized two-wheelers (Mohan, 2002) and 27% of road deaths in India are among users of motorized two-wheelers, while this figure is between 70% and 90% in Thailand, and about 60% in Malaysia (Mohan, 2002, Umar, 2002, Suriyawongpaisal and Kanchanusut, 2003). In China, motorcycles accounted for 23.4% of all registered motor vehicles in 1987, increasing to 63.2% in 2001. Motorcyclist fatalities and injuries increased 5.5-fold and 9.3-fold between 1987 and 2001, respectively. A total of 7.5% of all traffic fatalities and 8.8% of all traffic injuries were sustained by motorcyclists, with the corresponding proportions increasing to 18.9% and 22.8% in 1987 and 2001, respectively. The changing proportions of both traffic fatalities and injuries sustained by motorcyclists were positively correlated with the change in the proportion of motorcycles among all motor vehicles (Zhang et al., 2004). In Indonesia, the population of motorcycles has reached 78.3% of total motor vehicles and 75% fatality victims of traffic accidents were motorcyclists (Indriastuti and Sulistio, 2010). In Taiwan, the motorcycle ownership rate reaches 645 per thousand people (the highest motorcycle ownership rate in the world). In 2009, motorcyclist fatalities in Taiwan accounted for 56.69% of total traffic deaths (MOTC, 2010). To compare the fatalities and injuries of car and motorcycle crashes further, Table 1, Table 2, respectively give the statistics of car and motorcycle ownership, usage and crash victims of 23 counties/cities in Taiwan. Noted from Table 1, there are a total of 5.7 million registered passenger cars with 502 fatalities and 71,564 injuries in 2009, while the number of registered motorcycles reaches 14.6 million with 889 fatalities and 129,200 injuries in Table 2. Even in terms of accident rates (i.e., number of victims per million kilometers traveled), motorcycle usage (in terms of total kilometers traveled) still exhibits higher rates of fatal and injured victims than car usage. On an average, there are 0.017 fatalities and 2.038 injuries per million kilometers traveled by motorcycles in comparison to 0.014 fatalities and 1.524 injuries per million kilometers traveled by cars. To further examine the relationship between motorcycle usage and numbers of victims of two severity levels, the Pearson correlation test is performed. Results show that correlation coefficients of motorcycle usage with fatalities and injuries are 0.586 (p-value = 0.003) and 0.886 (p-value < 0.0001), respectively, indicating a significant and positive correlation between motorcycle usage and crash victims. Compared with the relationship between motorcycle usage and crash victims, the correlation coefficients of car usage with fatalities and injuries are also significantly tested with slightly lower coefficients of 0.475 (p-value = 0.022) and 0.876 (p-value < 0.0001), respectively. Thus, effective management strategies for decreasing the ownership and usage of motorcycles and cars are urgently required.

In the past, substantial research has been devoted to the development of car ownership and usage models, which has contributed to enhance our understanding of choice behaviors (De Jong et al., 2004). These studies have been mostly conducted in the developed countries because people rely heavily on private cars for urban travel in these areas, and motorcycles play a secondary role as a mode of transportation. As a result, few studies have been done in the context of motorcycle ownership and use. In addition, previous works on ownership and usage of motor vehicles often used either cross-sectional or time-series data. Data sets that pool cross sections and time series (called panel or longitudinal data) have become increasingly common in transportation and other fields. Panel data analyses have many advantages, such as controlling for individual heterogeneity, less collinearity among the variables, more degrees of freedom, and more efficiency, over analyses only based on either cross-sectional or time-series data alone (Baltagi, 2005). Moreover, the panel data models allow for analyzing the repeated choices over time and can capture the state dependence effect by incorporating past choices.

The current literature lacks the analysis of motorcycle ownership and usage using panel data. To examine dynamic choice behaviors associated with motorcycle ownership and usage, this study conducted a large-scale, nationwide two-wave panel survey on owners of motorcycles in Taiwan. Thus, disaggregate models of motorcycle ownership and usage using panel data are proposed. The discrete choice model has been widely used as an appropriate methodology for examining vehicle ownership. Compared with the ordered choice model, the unordered discrete choice model derived from random utility theory provides a theoretical base for modeling the number of vehicles owned in the household (Bhat and Pulugurta, 1998). In addition to the use of a standard discrete choice model, i.e., the multinomial logit model (MNL), this study also adopts nested logit (NL) and mixed logit (MXL) models to accommodate the possible independence among alternatives and parameter heterogeneity among individuals. For modeling the usage of motorcycles, panel data regression models involving fixed and random effects approaches accounting for heterogeneity are developed and compared. These models can be used to identify factors influencing motorcycle ownership and usage, such as household structure, residential location, transportation system performance, driver's travel patterns, and vehicle characteristics. Our proposed panel data models provide reliable parameter estimates to evaluate the effects of management strategies in reducing motorcycle ownership and usage. Due to a significant relationship between motorcycle usage and crash victims, these strategies will also lead to reductions in traffic accidents by motorcycles.

The remainder of this paper is structured as follows. Section 2 provides a brief overview of previous literature. Section 3 presents the framework of the motorcycle ownership and usage models. Section 4 describes the data sources and estimation results of motorcycle ownership and usage models as well as scenarios and analyses of management strategies. Finally, this paper concludes with the research findings, and gives directions for further research.

Section snippets

Literature review

Studies on vehicle ownership often use either aggregate model (e.g., Jansson, 1989, Button et al., 1993) or disaggregate model (e.g., Train, 1980, Mannering, 1983). The aggregate models may suffer the shortcomings of aggregation bias and multicollinearity between explanatory variables (Potoglou and Kanaroglou, 2008). On the contrary, the disaggregate modeling approach overcomes the weaknesses of aggregate models by capturing individual choice behavior and explanatory variables at an individual

Motorcycle ownership model

Motorcycle ownership model examines the number of motorcycles (i.e., 0, 1,…) owned by households in each year. Households may alter or maintain the number of motorcycles next year. The existing methods for modeling this type of dependent variable typically used discrete choice model, ordered logit or probit model, and count data model (Karlaftis and Golias, 2002), but the unordered discrete choice model is preferred because it provides a theoretical framework on a basis of random utility theory

The data

The survey questionnaire contains three main components. The first part includes household characteristics, such as household location, age, and gender of household heads, family size and structure, household income, number of workers in the household, number of vehicles in the household, distance from home to nearest public transit stop, and purchases or sales of motorcycles. The second part includes principal driver/rider demographics, such as gender, age, occupation, educational level,

Discussion and conclusions

A nationwide panel survey of motorcycle owners was conducted to examine the dynamic choices of the number of motorcycles as well as usage. For modeling ownership, this study uses the MNL, NL, and MXL models to accommodate the possible existence of independence among alternatives and individual heterogeneity. Panel data regression models considering fixed and random effects were developed for the usage. To demonstrate the applicability of the ownership and usage models, the effects of various

Acknowledgements

The authors appreciate two anonymous reviewers for their insightful comments and suggestions, which helped improve the quality of this paper. This study was sponsored by the Institute of Transportation, Ministry of Transportation and Communications of the Republic of China, under contract MOTC-IOT-98-SDB004.

References (36)

  • R.C. Bhat et al.

    A comparison of two alternative behavioral choice mechanisms for household auto ownership decision

    Transportation Research Part B

    (1998)
  • T.S. Breusch et al.

    The Lagrange multiplier test and its applications to model specification in econometrics

    Review of Economic Studies

    (1980)
  • P. Burge et al.

    Modeling of motorcycle ownership and commuter usage: a UK study

    Transportation Research Record

    (2007)
  • K. Button et al.

    Modeling vehicle ownership and use in low income countries

    Journal of Transport Economics and Policy

    (1993)
  • G.C. De Jong et al.

    Comparison of car ownership models

    Transport Reviews

    (2004)
  • M. Hanly et al.

    Car ownership in Great Britain: a panel data analysis

    Transportation Research Record

    (2000)
  • J.A. Hausman

    Specification tests in econometrics

    Econometrica

    (1978)
  • J.A. Hausman et al.

    Specification tests for the multinomial logit model

    Econometrica

    (1984)
  • Cited by (0)

    View full text