The intention and willingness to pay moving violation citations among Taiwan motorcyclists

Accident Analysis and Prevention 49 (2012) 177–185

Contents lists available at ScienceDirect

Accident Analysis and Prevention journal homepage: www.elsevier.com/locate/aap

The intention and willingness to pay moving violation citations among Taiwan motorcyclists Rong-Chang Jou ∗ , Pei-Lung Wang Department of Civil Engineering, National Chi-Nan University, No. 1, University Road, Puli, NanTou Hsien 545, Taiwan

a r t i c l e

i n f o

Article history: Received 30 July 2010 Received in revised form 13 June 2011 Accepted 17 June 2011 Keywords: Violations Willingness to pay Spike model

a b s t r a c t This study uses three-level scenario design and Spike model constructs to investigate the risk premium Taiwan motorcycle operators are willing to pay for moving violations. The primary focus of the investigation is on four types of moving violations including speeding, running red lights, right turn on red violations, and drunk driving. In the model, these four types of violations influence the willingness to pay a risk premium. The results show that, in addition to increasing enforcement, raising the level of fines is one of the most effective methods to influence levels of compliance. Estimated results through the Spike model show that speeders will accept a risk premium of NT$740 (US$1 = NT$30), while motorcycle operators who run red lights will accept a risk premium of NT$1100, motorcycle operators who turn right on red will accept a risk premium of NT$367, and motorcyclists who drive drunk will accept a risk premium of NT$18,540. This indicates that current fines in Taiwan could be raised. © 2011 Elsevier Ltd. All rights reserved.

1. Introduction Taiwan has one of the world’s highest per capita rates of motorcycle ownership with 14,600,000 motorcycles for 23 million people. In contrast, the US has 7,130,000 motorcycles, Japan 12,930,000, Singapore 140,000, Hong Kong 50,000 and Korea 1,780,000. At the end of 2007, the number of motorcycles in Taiwan stood at 1,394,000, considerably more than any other country1 . At the end of 1991, the number of registered motor vehicles in Taiwan, including cars and motorcycles, totalled 16,310,000 vehicles, including 5,350,000 passenger cars and 1,095,000 motorcycles. By the end of 2009, the total number of registered vehicles had climbed to 21,370,000, including 6,760,000 passenger cars and 14,600,000 motorcycles. Over a period of ten years, the total number of vehicles increased by 5,070,000, including 1,410,000 cars and 3,650,000 motorcycles, for relative increases of 26%, 31%, and 33%, respectively. In addition, gas price rises in recent years have caused car drivers to switch to motorcycles. According to Ministry of Transportation statistics for 2008, following increases in gas prices, around 34% of short-distance car drivers switched to other modes of transportation, with 89% switching to motorcycles.

∗ Corresponding author. Tel.: +886 49 2910960x4944; fax: +886 49 2918679. E-mail address: [email protected] (R.-C. Jou). 1 The data presented in the first two paragraphs are quoted from International Road Federation (IRF) (2000–2010), World Road Statistics. R.O.C: Ministry of Transportation and Communications (MOTC), http://www.motc.gov.tw/mocwebGIP/ wSite/mp?mp=1. 0001-4575/$ – see front matter © 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.aap.2011.06.011

Undoubtedly, an increase of motorcycle usage will be accompanied by an increase of motorcycle traffic problems (i.e. moving violations, accidents, etc.). Although some research has been conducted on factors contributing to rule breaking by motorcycle operators (Broughton et al., 2009), little research has focused on analyzing motorcycle operators’ willingness to pay fines. Currently, Taiwan has no set formula for determining fines specifically for motorcycle operators, nor any reference data. Therefore current fines may be too high or too low. This study reports on an in-depth investigation of the risk premiums Taiwan motorcycle operators are willing to pay to break traffic rules. Aside from contributing to a better understanding of the factors influencing risk premiums for motorcycle operators, the results of this study will also provide a reference for the setting of fines. Becker (1968) researched rule breaking behavior and the use of government sanctions, assuming that violations are the result of individual acts of choice. That is, when the anticipated benefit of a violation outweighs its potential cost, the individual will violate the rule; otherwise he will obey the rule. In addition, the study pointed out that the expected cost to potential offenders usually depends on the offender’s perception of the expected penalty (probability multiplied by the potential fines). Therefore, there are two ways to deter violation2 : increase the probability of getting caught, or

2 Deterrence theory assumes drivers assess legal threats based on the perceived risk of punishment, including the perceived risk of being caught and the perceived certainty, severity, and swiftness of legal sanctions (Leal et al., 2009; Homel, 1988; Cameron et al., 1992; Watson, 2003; Vingilis, 1990).

178

R.-C. Jou, P.-L. Wang / Accident Analysis and Prevention 49 (2012) 177–185

increase the fine. The former is bound to increase costs for policing. The latter, while imposing a relatively modest social burden, will have different deterrent effects on drivers of different levels of relative wealth. For example, Polinsky and Shavell (1991) and Chu and Jiang (1993) showed that in societies with unequal wealth distribution, setting fines too high or too low will result in an ineffective social deterrent. Therefore, in setting a fine which will serve as a deterrent to the greatest number of people, finding the appropriate level of fines is very important. Analyzing the willingness to pay fines for violations will result in more realistic fines, thus increasing the likelihood that the potential offender will accept the corresponding fine. Generally speaking, rule breaking behavior is not a market good, and non-market goods cannot be traded through market mechanisms to establish their economic value. In recent years, cost-benefit analysis has been applied to policy implementation and the evaluation of decision-making. One widely used approach for assessing non-market goods is the Contingent Value Method (CVM), which has an advantage in assessing the value of nonmarket goods in the conversion of the value of goods, primarily through surveys or similar interview methods. CVM asks respondents to subjectively determine the dollar value of non-market goods, and determines what maximum sum the respondents would be willing to pay for a given good. The value of willingness to pay (WTP) for the individual is elicited from answers to hypothetical questions in the survey. Previous studies have used the logit or probit models to establish WTP as a research theme (Hanemann, 1984; Salvador, 2001). The main drawback of these models is that they can only deal with positive WTP. However, because of the high likelihood that respondents will respond with “zero” when asked what they would be willing to pay (i.e., individuals would rather not pay even if they commit the offence), many studies have adopted the Spike model as an alternative to avoid creating bias (i.e., overestimate WTP) in the model (Kristroöm, 1997; Saz-Salazar and Garcia-Menendez, 2001; Yoo and Kwak, 2002; Yoo et al., 2006; Bengochea-Morancho et al., 2005; Hu, 2006; Jou et al., 2011a; Jou et al., 2011b). These studies all showed the Spike model can effectively deal with a large number of zeros in the WTP survey data, and can accommodate other WTP factors. In the survey conducted for this study, 36% of respondents answered they would be WTP zero for speeding citations, 46% for running red lights, 41% for right turn on red, and 56% for drunk driving3 . Regardless of the type of violation, Zero WTP accounts for at least 35% of the responses4 . In summary, using the Spike model to handle traffic violation WTP risk premium problems can prevent overestimation of WTP and thus provide a more realistic result. This study is structured as follows. The model framework is presented in Section 2, followed by a description of the survey and data analysis in Section 3. Section 4 describes the model estimation results, along with different WTP for various violation behaviors. Conclusions are presented in the final section. 2. Model framework This study is based on random utility theory, combined with CVM, to establish a virtual market for the risk premium motorcycle operators WTP for traffic violations. Motorcycle operators are further classified into groups of traffic rule violation behavior by their WTP risk cost. But, in reality, the fines motorcycle operators are WTP for violations are not high, and the sample will return many

3 The high percentage of zero WTP for drunk driving may due to respondents either being unlikely to violate this regulation or to pay the fine. 4 Zero WTP includes two cases: (1) not willing to pay and (2) unlikely to commit the violation. The latter case is excluded in this study to avoid underestimating WTP.

zeros (as described above). Thus, using the Spike method of evaluation is more appropriate (Salvador, 2001; Bengochea-Morancho et al., 2005). 2.1. The Spike method First assume the individual random utility function V(Y,Z,Q) where Y is income, Z is the value of the asset being assessed (i.e. the cost to motorcycle operators for breaking a traffic rule), and Q is the set of socio-economic attributes and personal characteristics (i.e., age, gender, nature of the violation, etc.). Furthermore, assume the amount that individuals are willing to pay is a random variable; when the market provides an amount A, and WTP  A, the respondents will not accept the amount. Therefore, if WTP  A, the probability function can be expressed as Pr(no) = Pr(WTP ≤ A) = FWTP (A)

(1)

where FWTP (A) is a right-continuous non-decreasing function. Further integrating the expected value concept, WTP = 0 is a boundary, and the group expected E(WTP) can be obtained by

 E(WTP) =



∞

0

(1 − FWTP (A))dA −

(FWTP (A))dA

(2)

−∞

0

In the Spike model, suppose the distribution function of WTP is

 FWTP (A) =

0, P, GWTP (A),

A<0 A=0 A>0

(3)

where P is (0,1), while GWTP (A) is a continuous and increasing function, so that GWTP (0) = p and limA → ∞, GWTP (A) = 1. Thus, P is greater than zero, and manifests as a discontinuous spike in the graphic. P is also the probability of respondents whose WTP is zero. It can be obtained from model estimation results. Generally speaking, CVM survey methodology encompasses not only single bound dichotomous choice, but also double bound dichotomous choice, one-and-one-half bound dichotomous choice, and triple bound dichotomous choice (Hanemann et al., 1991; Cooper et al., 2002; Barreiro et al., 2005). This study uses the triple bound dichotomous choice (closed end), so the proposed amount for answers received at each level are only “yes” and “no”. The Spike model’s probability model corresponds to the Logit model. Therefore, when the respondents answered “yes” to pay amount (A), then the probability function derived from the Logit model is Pr(yes) = 1 − FWTP (A) =

eV1 1 1 = = eV1 + eV0 1 + eV0 −V1 1 + e−V

(4)

in which V is the utility difference after respondents are willing to pay the amount (A). Therefore, Eq. (4) is expressed in the form of Pr(yes) = 1 − FWTP (A) = {1 + exp(− ˛ + ˇA)} −1 . Since this study uses the Triple-bounded questionnaire form, there will basically be 2 × 2 × 2 = 8 types of answers. However, when the questionnaire is actually administered, if respondents are reluctant to continuously answer the final three questions, the respondents may actually have two possible WTP situations, one of which entails the respondent being unwilling to pay any price, in which case the response for the WTP price is NT$0 (see the last line of Eq. (5)). The other possibility is that the respondent has a certain willingness to pay, but the WTP amount is lower than the amount on the questionnaire (i.e., 0
R.-C. Jou, P.-L. Wang / Accident Analysis and Prevention 49 (2012) 177–185

adding these two situations, the total number of possible cases is nine, as in Eq. (5), below.

equation (11) as follows:

I YYY = 1{Yes − Yes − Yes}

ln L =

I YYN = 1{Yes–Yes–No} I YNY

{I YYY ln[1 − Fwtp (B3U2 ; )] + I YYN ln[Fwtp (B3U2 ; )

i=1

− Fwtp (B2U1 ; )] + I YNY ln[Fwtp (B2U1 ; ) − Fwtp (B3D2 ; )]

= 1{Yes–No–Yes}

I YNN = 1{Yes–No–No} I NYY

N 

179

= 1{No–Yes–Yes}

(5)

I NYN = 1{No–Yes–No}

+ I YNN ln[Fwtp (B3D2 ; ) − Fwtp (B1O ; )] + I NYY ln[Fwtp (B1O ; ) − Fwtp (B3U2 ; )] + I NYN ln[Fwtp (B3U2 ; ) − Fwtp (B2D1 ; )]

I NNY = 1{No–No–Yes}

+ I NNY ln[Fwtp (B2D1 ; ) − Fwtp (B3D2 ; )] + I NNNY ln[Fwtp (B3D2 ; )

I NNNY = 1{No–No–No–Yes}

− Fwtp (0; )] + I NNNN ln[Fwtp (0; )]}

I NNNN

Among these, the indicator Iijk shows the actual answer. i,j and k are, respectively, the first, second and third-level price responses. Suppose that the first level price is B1O . If the second-level recommended price is higher, it is B2U1 ; otherwise it is B2D1 . If the third-level recommended price is higher than the second-level price, it is B3U2 ; otherwise it is B3D2 . Therefore, probability of the nine individual cases in the preceding Eq. (5) can be obtained by solving derived Eq. (6) as follows: I YYY (B1O , B2U1 , B3U2 ) = Pr(WTP ≥ B1O , WTP ≥ B2U1 , WTP ≥ B3U2 ) = Pr(WTP ≥ B1O , WTP ≥ B2U1 |WTP ≥ B3U2 )Pr(WTP ≥ B3U2 )

(6)

in which the formula Pr(WTP ≥ B1O , WTP ≥ B2U1 |WTP ≥ B3U2 ) equals 1. Therefore, Eq. (6) can be further simplified as Eq. (7): I YYY (B1O , B2U1 , B3U2 ) = 1 × Pr(WTP ≥ B3U2 ) = Pr(WTP ≥ B3U2 )

Finally, derived from the inverse formula Pr(yes) = −1 1 − Fwtp (A) = {1 + exp(−˛ + ˇA)} of Eq. (4), we can obtain FWTP (A) = {1 + exp(˛ − ˇA)} −1 , then FWTP (A) is the Logit-type probability model. Therefore, Eq. (3), above and the distribution range FWTP (A) can be further expressed as Eq. (12): FWTP (A) =

⎧ ⎨ [1 + exp(˛ − ˇA)]−1 , A > 0 [1 + exp(˛)]−1 ,

(7)

I

(B1O , B2D1 , B3D2 , 0)

= Pr(WTP <

B1O , WTP

<

B2D1 , WTP

<

B3D2 ,

E(WTP) =



∞

(8)

Additionally, given a second case response of IYYN and a thirdlevel not WTP price of B3U2 , the individual WTP will fall between B2U1 ≤ WTP < B3U2 . Therefore the probability of IYYN can be expressed by Eq. (9) below: I YYN (B1O , B2U1 , B3U2 ) = Pr(B2U1 ≤ WTP < B3U2 ) = Pr(WTP < B3U2 ) − Pr(WTP < B2U1 ) = Fwtp (B3U2 ; ) − Fwtp (B2U1 ; )

(9)

Similarly, the sequence of six answers develops as follows: I YNY (B1O , B2U1 , B3D2 ) = Pr(B3D2 ≤ WTP < B2U1 ) = Fwtp (B2U1 ; ) − Fwtp (B3D2 ; ) I YNN (B1O , B2U1 , B3D2 ) = Pr(B1O ≤ WTP < B3D2 ) = Fwtp (B3D2 ; ) − Fwtp (B1O ; ) I NYY (B1O , B2D1 , B3U2 ) = Pr(B3U2 ≤ WTP < B1O ) = Fwtp (B1O ; ) − Fwtp (B3U2 ; ) I NYN (B1O , B2D1 , B3U2 ) = Pr(B2D1 ≤ WTP < B3U2 ) = Fwtp (B3U2 ; ) − Fwtp (B2D1 ; )

=

0

(1 − FWTP (A))dA −

(FWTP (A))dA −∞

1 ln[1 + exp(˛)] ˇ

(13)

In addition, because Spike is defined as the value P when A = 0 (see Eq. (3)), therefore we can obtain Eq. (14): Spike =

1 1 + exp(˛)

(14)

3. Data survey and analysis

WTP ≤ 0) = Pr(WTP < B1O , WTP < B2D1 , WTP < B3D2 |WTP ≤ 0) × Pr(WTP ≤ 0) = 1 × Pr{WTP ≤ 0) = Fwtp (0; )

(12)

in which ˛ is the marginal utility for environmental improvement, and ˇ is the marginal utility of income. In the case that a respondent’s WTP > A, the estimated willingness to pay is calculated according to Eq. (2), substituting the FWTP (A > 0) = [1 + exp(˛ − ˇA)]−1 fragment of Eq. (12), and can be expressed as Eq. (13):

0

Therefore, for the first response IYYY , the probability value is indicated in the above Eq. (7). Derived from the same concept, the final response INNNN is similar in concept to IYYY , of which the formula Pr(WTP < B1O , WTP < B2D1 , WTP < B3D2 |WTP ≤ 0) fragment is also equal to 1. Therefore, for INNNN the probability value can be derived by Eq. (8):

A=0 A<0

⎩ 0,



= 1 − FWTP (B3U2 ; )

NNNN

(11)

= 1{No–No–No–No}

(10)

I NNY (B1O , B2D1 , B3D2 ) = Pr(B3D2 ≤ WTP < B2D1 ) = Fwtp (B2D1 ; ) − Fwtp (B3D2 ; ) I NNNY (B1O , B2D1 , B3D2 , 0) = Pr(0 ≤ WTP < B3D2 ) = Fwtp (B3D2 ; ) − Fwtp (0; )

The Spike model uses the maximum likelihood method (MLE) for assessment. Therefore the probability can be expressed as

3.1. Questionnaire design and survey description The questionnaire designed for this research focuses on motorcyclists, with the sampling population taken from students (including graduate and part-time students) at National Chi Nan University in Nantou County, Taiwan. The main objective of this study is to propose a TBDC approach to estimate WTP for traffic violations, thus we focus exclusively on students. The questionnaire can be divided into three basic parts: (1) socio-economic and trip characteristics, (2) law-abidingness and familiarity with traffic rules, (3) assumed market scenarios. The first part can be further broken down into two parts: demographic information and trip characteristics. The demographic information gathered includes the respondent’s gender, age, level of education and monthly personal disposable income. Trip characteristics questions include, “On your last trip, did you have a passenger?”, “What was your relationship with the passenger?”, “How many days a week do you use your motorcycle?”, “How many times a day do you use your motorcycle?”, and “What was the origin, destination, purpose and total travel time of your last trip?”. The second part included the following items: “Awareness of traffic rules”, “Proportion of non-compliance”, “Number of

180

R.-C. Jou, P.-L. Wang / Accident Analysis and Prevention 49 (2012) 177–185

violations in the last ten trips”, “Who paid the fine?”, “Was the motorcycle operator speeding?”, “Excess speed above the speed limit (k/h)”. The first item was further defined as follows. Respondents were classified as “familiar” with a given rule if they knew the specific fines for a given violation, and “very familiar” if they were also had detailed knowledge of the provisions of the violation. For example, for speeding, respondents were “familiar” if they were aware of the different fines for different levels of speeding, and were “very aware” if they could associate different penalties with their related ordinances. The third part uses scenario design to simulate market trading prices. To more effectively elicit the real price respondents are willing to pay, this research uses a Triple-bound style questionnaire to set three levels of WTP prices (Fig. 1). Taking speeding as an example, the 1st level bid is the current fine (CF: NT$1200), and our question is “given the current fine, would you commit a speeding violation?” The current fine is used rather than an arbitrary fine because the starting bid provides the respondent with a better reference point and, as a result, elicits a more realistic and practical WTP. The use of three levels of WTP price items results in nine possible answer permutations: (1) Yes–Yes–Yes; (2) Yes–Yes–No; (3) Yes–No–Yes; (4) Yes–No–No; (5) No–Yes–yes; (6) No–Yes–No; (7) No–No–Yes; (8) No–No–No–Yes; (9) No–No–No–No. In the 8th case, respondents answered “no” three times consecutively, suggesting they are unwilling to pay the recommended lowest price on the third level, but further questioning indicates that the respondent will accept a price above zero. The respondent is informed that they can respond with “zero”. In this part, we specifically ask respondents their reason for the amount they are willing to pay to determine whether they are rational or not. Differentiating these respondents from those who truly would pay zero effectively reduces the likelihood of error when calculating the WTP price. Finally, if the WTP price truly is zero, the respondents were asked to explain their reasons, which basically fell into one of two categories5 : (1) not willing to pay and (2) unlikely to violate. The survey was conducted from 9 am to 6 pm over five days from October 19 to 23, 2009. It targeted students (including graduate and part-time students) at National Chi Nan University in Nantou County. Respondents were intercepted randomly in campus parking lots. Respondents were informed of the survey’s purpose, and, in collecting the data, all reasonable efforts6 were made to avoid possible bias. Over the five day period, a total of 320 students were interviewed, with a total of 300 valid questionnaires completed, for a completion rate of 94%. 3.2. Data analysis Data was analyzed according to the three main topics mentioned above. The socio-economic and trip characteristic instance distribution is given below. 3.3. Socio-economic and trip characteristics Of the questionnaire respondents, 224 (75%) were male, 47% were between the ages 18 and 20, and the average age was 21.12. In terms of monthly disposable income, 22% reported having between NT$5000 and NT$10,000, while 23% reported disposable income below NT$5000. The average disposable income was NT$7969. Sixty-five percent were undergraduates with 19%, 21% and 24%

5 Note that samples unlikely to violate were excluded to avoid underestimating WTP. 6 The key was to ensure that all respondents fully understood the questions listed in the questionnaire. The representativeness of the sample was ensured through sampling according to the distribution of gender, schools affiliation, and grade level.

in their second, third, and fourth year of university, respectively. Graduate students accounted for 19%. In the section related to trip characteristics, 47% of respondents had carried a passenger on their last trip. Further breaking down results by gender reveals that women were 20% more likely to carry passengers than men. 94% of passengers were between 18 and 25 years old. In terms of the relationship between the motorcycle operator and passenger, classmates, friends and boy/girlfriends accounted for 47%, 26% and 21% of passengers, respectively. In addition to being asked about their passenger’s ages and relationships, respondents were asked about their personal driving habits. For example, 56% reported driving every day in the past week, and 11% had driven five days out of the past seven. Forty-two percent reported driving for between 31 and 60 min a day, while 36% reported driving for less than 30 min a day. Thirty-six percent indicated that their most recent trip started from their place of residence, with 30% responding that their most recent trip began at school. Sixty-five percent indicated the destination of their most recent trip was school, with 53% reporting that the purpose of their most recent trip was to go to class, while 25% reported answers including “shopping”, “sport”, “studying”, “eating” and other purposes. Seventy-nine percent reported that their total travel time was less than 30 min, indicating that most trips were brief.

3.4. Law-abidingness and familiarity with traffic regulations The second part investigates the driving behavior of motorcyclists and their knowledge of traffic regulations, sorted by speeding, running red lights, right turn on red, and drunk driving. Thirty-five percent of respondents were familiar with the speed limit rules, and 33% were somewhat familiar, indicating a basic knowledge of the penalties for non-compliance. Thirty-six percent were familiar with the red light rules, but it is worth noting that as many as 32% were very familiar with these regulations, with most aware of the specific fine for running a red light. Thirty-five percent were familiar with the rules for turning right at a red light, and 26% were very familiar, but it should be noted that three respondents (1%) did not know this behavior could result in a fine. This may be because right on red rules may not be stringently enforced in the relatively rural sample locations, leading some residents to believe this is not a violation and therefore would not incur a fine. Finally, 34% were very familiar with the drunk driving rules, and 25% were somewhat familiar with them. When asked about their most recent trip, 4%, 37%, 56%, and 70% of respondents admitted to drunk driving, running red lights, right turn on red and speeding, respectively. The significantly lower frequency of drunk driving could possibly be attributed to respondents’ awareness of the safety risks associated with drunk driving. To further understand the motorcycle operators’ experience in violating traffic rules, they were asked about their past ten trips. Speeding was the most commonly reported violation, with 31% reporting having exceeded speed limits in their last ten trips, against 27% who reported that they had not. It is worth noting that proportion of all violations in the past ten trips was 16%, which indicates that motorcycle operators are more likely to speed than drivers of passenger cars. Fifty-four percent reported that they had not run a red light in the past ten trips, indicating that the respondents may consider this to be a riskier violation, though 37% reported having done so one to four times in the past ten trips. But the results were reversed for right on red, where 51% of all respondents reported engaging in this behavior one to five times in the past ten trips. Last was drunk driving, where 97% reported not having driven while intoxicated in the last ten trips. The final part was concerned with the respondents’ driving conditions and who had paid fines for moving violations.

R.-C. Jou, P.-L. Wang / Accident Analysis and Prevention 49 (2012) 177–185

181

1st level bid (current fine (CF))

Yes

No

2nd level bid (CF x 1.25)

2nd level bid (CF x 0.5)

Yes

No

Yes

No

3rd level bid (CF x 1.5)

3rd level bid (CF x 1.125)

3rd level bid (CF x .75)

3rd level bid (CF x 0.25)

Yes

No

Yes

No

Yes

No

Yes

No

Determine actual highest amount respondent is WTP (can be zero)

Highest amount respondent is WTP

Yes

No

Respondent is willing to pay a certain amount, but this amount is lower than the recommended lowest bid scenario

Reason for zero bid: 1.Not willing to pay 2.Would never violate

Fig. 1. Assumed market scenarios – triple-bounded structure.

Sixty-nine percent of motorcycle operators responded that they normally speed while driving their motorcycles, with 37% exceeding speed limits by 1–10 k/h, followed by 23% at 11–20 k/h, and an average of 11.25 k/h above the speed limit, which is just within the margin generally enforced by police. When asked who normally paid the fine, 52% responded that they pay their own fines, while the rest mainly have their family members pay their fines. The data for responses to the second part of the questionnaire are presented in Tables 1 and 2. 3.5. Assumed market scenarios Assumptions regarding market scenarios are sorted by speeding, running red lights, right turn on red and drunk driving. In general, speeding fines are the same as for passenger cars and are based on excess speeds of less than 20 k/h, for which

the current fine (CF) is NT$1200. The highest and lowest fines are NT$1800 (CFx1.5) and NT$300 (CFx0.25). A closer look at the responses shows that 5% of respondents were willing to pay all three consecutive prices, while 44% were not willing to pay any of three consecutive prices. Generally, as the fine increases, a lower percentage is willing pay. The current fine for running a red light is NT$1800, with the highest and lowest fines set at NT$2700 and NT$450. Taking the nine types of responses together, and keeping the fine at NT$1800, the percentage of respondents who would be willing to pay a NT$1800 fine for running a red light was 5%. However, when the fine is raised to the next level of NT$2050, the percentage drops to 2%. Overall, when the fines rise, the percentage of respondents willing to pay falls. The current fine for making a right turn at a red light is NT$600. In the assumed scenarios, NT$900 and NT$150 were the two extreme

182

R.-C. Jou, P.-L. Wang / Accident Analysis and Prevention 49 (2012) 177–185

Table 1 Familiarity with traffic regulations and law-abidingness (percentage in parenthesis). Traffic rules

Samples

Samples

Percentage of violations in last trip

Speeding

Very unfamiliar Unfamiliar Somewhat familiar Familiar Very familiar

0(0.00) 35(11.67) 98(32.67) 104(34.67) 63(21.00)

0% 1–49% 50% 51–99% 100%

91(30.33) 89(29.67) 40(13.33) 40(13.33) 40(13.33)

0 times Fewer than 4 times 5 Times More than 6 times 10 Times

81(27.00) 95(31.67) 32(10.67) 42(14.00) 50(16.67)

Running red lights

Very unfamiliar Unfamiliar Somewhat familiar Familiar Very familiar

0(0.00) 27(9.00) 69(23.00) 107(35.67) 97(32.33)

0% 1–49% 50% 51–99% 100%

190(63.33) 85(28.33) 13(4.33) 10(3.33) 2(0.67)

0 times Fewer than 4 times 5 Times More than 6 times 10 Times

163(54.33) 111(37.00) 16(5.33) 8(2.67) 2(0.67)

Right turn on red

Very unfamiliar Unfamiliar Somewhat familiar Familiar Very familiar

3(1.00) 41(13.67) 74(24.67) 105(35.00) 77(25.67)

0% 1–49% 50% 51–99% 100%

133(44.33) 103(34.33) 30(10.00) 23(7.67) 11(3.67)

0 times Fewer than 4 times 5 Times More than 6 times 10 Times

114(38.00) 119(39.67) 35(11.67) 18(6.00) 14(4.67)

Driving while intoxicated

Very unfamiliar Unfamiliar Somewhat familiar Familiar Very familiar

0(0.00) 54(18.00) 75(25.00) 69(23.00) 102(34.00)

0% 1–49% 50% 51–99% 100%

289(96.33) 10(3.33) 0(0.00) 0(0.00) 1(0.33)

0 times Fewer than 4 times 5 Times More than 6 times 10 Times

291(97.00) 9(3.00) 0(0.00) 0(0.00) 0(0.00)

fines. When fines are set at NT$600, 10% are willing to pay, but this drops to 5% when the fine rises to NT$675. Generally speaking, the indication is the same as before: as the fine rises, the percentage willing to pay falls. The current fine for drunk driving7 is NT$15,000, specifically for driving with a blood alcohol content of between 0.25 and 0.4 mg/L. Therefore, fine prices were adjusted accordingly, with the highest and lowest fines set at NT$22,500 and $3750. Looking at the nine response scenarios together, 6% of respondents were willing to pay a fine of NT$15,000, but that drops to 4% when the price rises to $16,875. But it is worth noting that, when the WTP price exceeds NT$22,500, 18% are willing to pay, which would be of interest to government agencies. Generally speaking, drunk driving fines perform the same as for the other violations: as the fine rises, the percentage willing to pay falls.

3.6. Acceptable risk premiums for traffic violations This study further investigated risk premiums for traffic violations under different classifications and different situations. The classifications are (1) gender, (2) personal income, (3) level of educational achievement, (4) understanding of the penalty level and (5) past experience with violations. These classifications are discussed in order below: The gender variable was taken from demographic data collected in part one of the questionnaire, and used to investigate differences in acceptable risk premiums for traffic violations by male and female drivers. For speeding violations, men were willing to pay an average price of NT$611.12, as opposed to NT$510.53 for women. For running red lights, the figures were NT$605.71 and NT$583.32, respectively; for right turn on red, NT$289.51 and NT$30.68; and for drunk driving, NT$8320.76 for men and NT$9334.36 for women. However, the gender-based difference for actual incidence of traffic violations was not significant (with significance set at 10%), which

7 Taiwan law provides for license suspension or disqualification for drunk driving in addition to a fine. Prior to filling in the questionnaire, we highlighted the relevant regulations to respondents to enhance their knowledge of the consequences of drunk driving.

Violations in past ten trips samples

Samples

Familiarity with fines

suggests that for the violations listed here, gender does not cause a difference in acceptable risk premium. Among the 300 university students surveyed, average monthly disposable income was about NT$7969, with 69% reporting monthly disposable income between NT$5001 and NT$10,000. A Bureau of Statistics survey of young people in 2001 reported that people aged 15–24 had average monthly disposable income of NT$4711, and the average rose to NT$7932 for those who were already employed. Taking this into account, we set NT$7932 as the dividing point for those with higher and lower disposable incomes. For speeding, those with high disposable incomes were willing to pay NT$643.56 on average, while those with low disposable income were willing to pay NT$538.24. For running red lights, the amounts were NT$684.96 and NT$531.94, respectively; for right turn on red, NT$370.12 and NT$230.09; and for drunk driving, NT$8176.67 and NT$8910.61. For these two groups, a significant difference was found in right turn on red violations (with significance set at 10%), and those with higher disposable income were willing to pay an NT$140.03 greater risk premium than those with lower disposable incomes. The survey results differentiated respondents as undergraduates, master’s degree students, and PhD candidates, and level of educational attainment was checked as a factor in acceptance of risk premiums for traffic violations. Among the sample, 241 respondents were undergraduates, while 59 were graduate students (including part-time graduate students). However, in the final test results, there was no significant difference for any type of violation (with significance set at 10%).

Table 2 Law-abidingness analysis (percentage in parenthesis). Items

Samples

I normally exceed speed limits

Yes No

208(69.33) 92(30.67)

Normal amount of speeding

0 k/h 1–10 k/h 11–20 k/h Over 21 k/h

92(30.67) 111(37.00) 70(23.33) 27(9.00)

Person who pays fines

Self Relatives

157(52.33) 143(47.67)

R.-C. Jou, P.-L. Wang / Accident Analysis and Prevention 49 (2012) 177–185

183

Table 3 Different levels of citation awareness and acceptable risk premium. Violation type

Item

t-Test

Familiar

Speeding Running red light Right on red Drunk driving *

Unfamiliar

Samples

WTP

Samples

WTP

167 204 182 171

536.11 531.40 260.75 7344.88

133 96 118 129

647.82 748.28 343.01 10218.02

−1.33 −1.90* −1.68* −1.91*

Confidence interval under 90%; t-value is significant at 1.65.

In addition to socio-economic factors, this study also explored familiarity with traffic violations and law-abidingness. In tests done for each type of violation, aside from speeding, the WTP of respondents who were familiar with traffic rules was significantly lower than those who were not familiar with traffic rules (with the significance level at 10%). That is, those less familiar with the regulations regarding running red lights, right turns on red and drunk driving were more willing to pay fines. Therefore, we suggest that the lessaware motorcycle operators should attend driver education classes (e.g. road safety workshops) to raise their awareness of fines and the related regulations, which may reduce their incidence of traffic violations (Table 3).

Table 4 Variable definition. Variable Socio-economic variables Traveler’s law-abidingness and adherence to traffic rules variables

Explanation Personal income (NT$10,000) Past history of violation

4. Model estimation results This study used a calibrated Spike model, arranged sequentially according to the four above-mentioned violations shown in Tables 5–8, respectively. Taiwan has one of the world’s highest per capita rates of motorcycle ownership, and this needs to be taken into consideration and illustrated for each model variable. Aside from the situation price variable, each variable in the model can be categorized as (1) socio-economic variables, (2) driver lawabidingness and adherence to traffic rules variables, (3) interaction variables. These three variables are explained in Table 4.

Familiarity with fines

Excess speeding Interaction variables

Male social achiever

4.1. Speeding When considering as a single scenario bid variable, the average WTP price is NT$740, and NT$693 when considering multiple variables, including personal income, ratio of past speeding violations, excess speed above the posted speed limit, and familiarity with speeding fines. Scenario bid and familiarity with speeding fines are negatively correlated, while personal income, past speeding violations and excess speed are positively correlated. As an example to illustrate the scenario bid variable, suppose that when the fine for speeding is under NT$500 the incidence of speeding will rise. But if the fines are raised to a given level, it will result in a drop in the level of speeding. Therefore as the scenario bid (fine) rises, people are less willing to pay, which matches the direction of this research’s expectations. The proportion of speeding violations in the past ten trips also correlates positively, indicating that if the motorcycle operator has violated speeding rules in the past, and the ratio is not low, the motorcycle operator will be more inclined to pay the fine which, perhaps because motorcycle operators feel that the fine amount is acceptable. Therefore, such violations would continue unhindered, and this fine would have little deterrent effect. The excess speed above the speed limit was primarily obtained from a follow-up question on the questionnaire item, “Do you normally speed?” Therefore, if the motorcycle operator normally speeds while driving, the higher their excess speed, the less effective fines are likely to be as deterrents. In other words, this type of motorcycle operator will be able to accept higher fines. Finally, awareness of speeding fines correlates negatively. In this research, awareness of fines is defined as whether the respondent clearly

Familiarity with fines

Committed violations on past trips Committed violations on past trips and self-paid fines

Number written by the respondent 1 = “No violations in past ten trips” 2 = “4 violations or fewer in past ten trips” 3 = “5 violations in past ten trips” 4 = “6 or more violations in past ten trips” 5 = “violations in all past ten trips” 1 = “very unfamiliar” 2 = “unfamiliar” 3 = “somewhat familiar” 4 = “familiar” 5 = “very familiar” Number (0–60 k/h) written by the respondent If level of familiarity is “familiar” or “very familiar”, then “1”; otherwise “0” Male social achiever (classified as “1”) is defined as a male graduate student who pays his own fines and has a high-level income (otherwise “0”) Past violation behavior = “1”; otherwise “0” If the respondent owns the vehicle and pays the fines himself, then “1”; otherwise “0”

knows the actual fine amount. If the respondent knows the true amount, they are said to be aware. Therefore, when the motorcycle operator knows the actual amount of the fines, the motorcycle operator will tend to speed less or, if he does speed, he will do so to a limited extent to avoid being fined. 4.2. Running red lights For running red lights, when considering “scenario bid” as a single variable, the average WTP is NT$1100. When other factors (i.e. awareness of fines, past violations ratio, and male social achiever status) are considered, the average WTP is NT$1264. Scenario bid is negatively correlated and when the scenario bid (fines) rise, the motorcycle operators WTP drops. Awareness of fines is negatively correlated, and the more aware motorcycle operators are of fines, the less willing they are to pay. When a motorcycle operator’s awareness of fines covers exact amounts for given speeds,

184

R.-C. Jou, P.-L. Wang / Accident Analysis and Prevention 49 (2012) 177–185 Table 7 Spike model for motorcycle right on red violations.

Table 5 Spike model for motorcycle speeding violations. Variable

Single variable −0.53(−4.39)** −1.34(−11.59)**

Constant Scenario bid Personal income (NT$10,000) Past speeding violations Excess speed (k/h) Awareness of speeding fines Spike(t-value) Log-likelihood Wald statistic (p-value) Average WTP Sample size * **

0.37(13.14)** −303.27 127.69(0.00) 740.88

Multiple variables −1.58(−3.08)** −1.47(−11.68)** 1.26(2.76)** 0.45(2.61)** 0.25(1.90)* −0.42(−1.76)* 0.36(8.46)** −286.72 67.91(0.00) 693.08 300

Variable

Spike (t-value) Log-likelihood Wald statistic (p-value) Average WTP Sample size *

Confidence interval under 90%; t-value is significant at 1.65. Confidence interval under 95%; t-value is significant at 1.96.

Single variable

**

Multiple variable

0.37(3.16)** −0.24(−10.08)**

Constant Scenario bid Personal income (NT$10,000) Awareness of fines Male social achiever Past right on red violations

0.41(14.47)** −288.27 94.63(0.00) 367.18

0.50(0.88) −0.27(−10.09)** 0.69(1.50) −0.26(−2.28)** 1.31(2.53)** 0.40(1.75)* 0.31(5.67)** −274.92 35.12(0.00) 441.29 300

Confidence interval under 90%; t-value is significant at 1.65. Confidence interval under 95%; t-value is significant at 1.96.

4.4. Drunk driving violations the motorcycle operator will be less likely to speed and, if he does speed, he will do so within a range unlikely to result in a citation and, thus, the motorcycle operator’s WTP will drop. The ratio of past violation behavior correlates positively and, when the ratio rises, so does WTP. Finally, male social achievers – defined primarily as males with high educational achievement (graduate school and above), high average income and the practice of paying their own fines – will have a certain understanding of risk perception, based on their higher income. Therefore, they are less sensitive to fines and will be more willing to pay fines for violations.

4.3. Right turn on red violations For right turns on red, when the scenario bid is treated as a single variable, the average WTP is NT$367, and $441 for multiple variables, and both single and multiple variables are all negatively correlated in that, the higher the fine, the less willing the motorcycle operator is to pay. Personal income was positively correlated, and the higher a motorcycle operator’s personal income, the more willing he was to pay. Although personal income was not very significant, it still had a certain impact on the model and is therefore included in the model. Awareness of fines correlated negatively; when motorcycle operators were more aware of fines, they were more reluctant to pay them. Male social achievers were more willing pay, and past violation behavior was also positively correlated. If motorcycle operators had turned right on red lights in the past, it indicated their acceptance of fines was higher, and that fines did not have the desired deterrent effect on them. Therefore, they are likely to continue making right turns on red and tend to pay the fine.

Table 6 Spike model for motorcycle right turn on red. Variable Constant Scenario bid Awareness of fines Past violations Male social achiever Spike (t-value) Log-likelihood Wald statistic (p-value) Average WTP Sample size * **

Single variable

Multiple variable

0.15(1.29) −0.70(−8.84)**

0.46(16.01)** −281.96 72.09(0.00) 1100 300

0.56(0.96) −0.72(−8.92)** −0.24(−2.05)** 0.32(1.67)* 0.73(1.65)* 0.40(5.87)** −276.89 24.32(0.00) 1264

Confidence interval under 90%; t-value is significant at 1.65. Confidence interval under 95%; t-value is significant at 1.96.

When only the price variable is factored in, the average WTP price is NT$18,540. The average WTP price rises to NT$21,435 after two more variables (“familiarity with the penalty” and “previous drunk driving and self-paid fines”) are included. The sign of scenario bid is negative indicating that motorcycle operators are less inclined to pay when the scenario bid (fine) rises. Meanwhile, consistent with the above violation models, the level of familiarity with the fines also has a negative influence. Finally, past experience with drunk driving and paying the fines oneself has a positive influence. That is, people who are belong to the driving category are more likely to pay the fine, suggesting that, due to previous experience with drunk driving, these motorcycle operators may be more accepting of higher fines. The higher WTP for drunk driving violations may due to the fact that motorcycle operators are very unlikely to commit this violation. As a result, they raise their bid to prevent others from driving drunk and consequently promote a safer driving environment for everyone. 5. Conclusions and suggestions This study uses the Spike method to build a WTP pricing model for traffic violations by motorcyclists, including speeding, running red lights, right turns at red lights, and driving while intoxicated. This section explains our results and conclusions. The difference between WTP and the current fine is examined because we assume the WTP is the amount that the respondents can afford. If the difference is significant, then the current fine is high enough to prevent from committing violations; otherwise, the student will not take the current fine seriously since he/she would be willing to pay more than the actual fine. As a result, he/she will be more likely to violate this traffic rule. Table 8 Spike model for motorcycle drunk driving violations. Variable Constant Scenario bid Awareness of fines Past drunk driving and self-paid fines Spike (t-value) Log-likelihood Wald statistic (p-value) Average WTP Sample size * **

Single variable

Multiple variables

0.25(2.13)** −0.31(−6.48)**

1.06(2.62)** −0.32(−6.47)** −0.36(−3.40)** 1.87(1.66)*

0.56(19.54)** −279.79 40.36(0.00) 18540.40

0.50(14.55)** −272.58 38.52(0.00) 21435.40 300

Confidence interval under 90%; t-value is significant at 1.65. Confidence interval under 95%; t-value is significant at 1.96.

R.-C. Jou, P.-L. Wang / Accident Analysis and Prevention 49 (2012) 177–185

When speeding violations are only considered in the context of the price factor, the average WTP is NT$740, which drops to NT$693 when other factors are included. Currently the actual fine for exceeding the speed limit by 20 k/h is NT$1200, a difference of about NT$460 from the study results. For university students who work part time, $460 is an easily affordable sum. Therefore, this study suggests that there is room for raising this fine, especially given that motorcycles are a bigger traffic problem than passenger cars in Taiwan. When the running of red lights is only considered in the context of the price factor, the average WTP is NT$1100, which grows to NT$1264 when other factors are included. Currently the actual fine for running a red light is NT$1800. At first glance, this seems like a considerable amount, but the difference from our study results is only NT$700. This is not a small sum for students, but could be considered acceptable by students who work part-time. Furthermore, incidents of motorcycles running red lights are considerably higher than for passenger cars. Therefore this study suggests that the fines for running a red light could be increased slightly, which may increase the deterrent effect. Currently, the fine for turning right at a red light is NT$600. When considered by price alone, the average WTP was NT$367, an insignificant gap of only NT$233. Furthermore, the nature of motorcycle driving leads motorcycle operators to frequently turn right on red. At small intersections this may not cause much of a problem. But at a major intersection it could have a serious impact on traffic. Given that the government does not intend to legalize right on red, this study recommends that fine for this violation be raised. When considered on price only, the willingness to pay for drunk driving is NT$18,540. Currently, the actual fine for driving with a blood alcohol level of between 0.25 and 0.4 mg is NT$15,000, a positive gap of NT$3540. This may possibly be attributed to motorcycle operators being very unlikely to actually commit this violation. Thus, they raise their bid to prevent others from driving drunk and consequently promote a safer driving environment for everyone. The government could raise the current fine to match the popular expectations expressed in the higher WTP and thus further reduce accidents due to drunk driving. Note that the estimation results above consider zero WTP only for respondents unwilling to pay and all results from spike models indicate that respondents previous violation experience are likely to have higher WTP. Finally, the findings and model estimation results may or may not be transferrable (e.g., to car drivers, other trip purposes or countries), which is always an issue of concern to transportation planners. Our main aim in this study is to propose a TBDC approach to estimate WTP for traffic violations, and transferability is not within our scope of study. In addition, this study does not include information about likelihood of detection which may affect motorcycle operators’ WTP. Nor does it include respondents other than students, and comparing results from this study to groups with

185

different socio-economic characteristics may provide interesting insights. References Barreiro, J., Sanchez, M., Viladrich-Grau, M., 2005. How much are people willing to pay for silence? A contingent valuation study. Applied Economics 37, 1233–1246. Becker, G.S., 1968. Crime and punishment: an economic approach. Journal of Political Economy 76 (2), 169–217. Bengochea-Morancho, A., Fuertes-Eugenio, A.M., Saz-Salazar, S.D., 2005. A comparison of empirical models used to infer the willingness to pay in contingent valuation. Empirical Economics 30, 235–244. Broughton, P.S., Fuller, R., Stradling, S., Gormley, M., Kinnear, N., O’Dolan, C., Hannigan, B., 2009. Conditions for speeding behaviour: a comparison of car drivers and powered two wheeled riders. Transportation Research Part F 12, 417–427. Cameron, M.A., Cavallo, A., Gilbert, 1992. Crash-based evaluation of the speed camera program in Victoria 1990–1991, Phase 1: General Effects, Phase 2: Effects of program mechanisms (Report No. 42). Monash University Accident Research Centre: Melbourne. Chu, C.Y., Jiang, N., 1993. Are fines more efficient than imprisonment? Journal of Public Economics 51, 115–124. Cooper, J.C., Hanemann, M., Signorello, G., 2002. One and one-half bound dichotomous choice contingent valuation. Review of Economics and Statistics 84, 742–750. Hanemann, W.H., 1984. Welfare evaluations in contingent valuation experiments with discrete responses. American Journal of Agricultural Economics 66 (3), 332–341. Hanemann, W.H., Loomis, J., Kanninen, B., 1991. Statistical efficiency of doublebounded dichotomous choice contingent valuation. American Journal of Agricultural Economics 73, 1255–1263. Homel, R., 1988. Policing and Punishing the Drinking Driver: A Study of Specific and General Deterrence. Springer-Verlag, New York. Hu, W., 2006. Use of spike models in measuring consumers’ willingness to pay for Non-GM oil. Journal of Agricultural Economics 38 (3), 525–538. International Road Federation (IRF) (2000–2010), World Road Statistics. R.O.C: Ministry of Transportation and Communications (MOTC). Jou, R.C., Wu, Y.C., Chen, K.H., 2011a. Analysis of environmental benefits of a motorcycle idling stop policy at urban intersection. Transportation 38 (No. 2). Jou, R. C., Jian, R. Y., Wu, Y.C., 2011, A study of passengers’ willingness to pay for business class seats of high speed rail in Taiwan, Transportmetrica, forthcoming paper. doi: 10.1080/18128602.2011.565816. Kristroöm, B., 1997. Spike models in contingent valuation models. American Journal of Agricultural Economics 79, 1013–1023. Leal, N., Watson, B., Armstrong, K., King, M., 2009. There’s no way in hell I would pull up”: Deterrent and other effects of vehicle impoundment laws for hooning. In: Australasian Road Safety Research, Policing and Education Conference, Sydney, New South Wales. Polinsky, A.M., Shavell, S., 1991. A note on optimal fines when wealth varies among individuals. American Economic Review 81, 618–621. Salvador, L., 2001. Restoring dockland areas for recreation purposes: A contingent valuation study. In: European Real Estate Society Conference. Saz-Salazar, S.D., Garcia-Menendez, L., 2001. Willingness to pay for environmental improvements in a large city. Environmental and Resource Economics 20, 103–112. Vingilis, E.R., 1990. A new look at deterrence. In: Wilson, R.J., Mann, R.E. (Eds.), Drinking and Driving: Advances in Research and Prevention. Guilford Press, New York. Watson, B.C., 2003. The Road Safety Implications of Unlicensed Driving: A Survey of Unlicensed Drivers. Australian Transport Safety Bureau, Canberra, Australia. Yoo, S.-H., Kwak, S.-J., 2002. Using a spike model to deal with zero response data from double bounded dichotomous choice contingent valuation surveys. Applied Economics Letters 9, 929–932. Yoo, S.-H., Shin, C.-O., Kwak, S.-J., 2006. Inconvenience cost of spam mail: a contingent valuation study. Applied Economics Letters 13, 933–936.