Scholarly article on topic 'The Long-term Effectiveness of a Reward Scheme in Changing Daily Travel Choices'

The Long-term Effectiveness of a Reward Scheme in Changing Daily Travel Choices Academic research paper on "Economics and business"

CC BY-NC-ND
0
0
Share paper
OECD Field of science
Keywords
{"Pricing policies" / "reward scheme" / "long-term effect" / "ML model"}

Abstract of research paper on Economics and business, author of scientific article — Elaheh Khademi, Harry Timmermans

Abstract Understanding behavioral response and attitudes of car users towards pricing policies and the public's acceptance of new pricing schemes are of the highest priority for generating a successful policy. The assumption underlying these policies is that travelers will adapt their current behavior when faced with pricing schemes. Behavioral psychology however claims the benefits of reward over punishment in terms of its effectiveness of influencing behavior. Although the arguments are compelling, the effects of reward schemes have not received much attention in the context of travel behavior. A fascinating reward scheme has been suggested in The Netherlands because of the massive public resistance against the implementation of road pricing in 2010. Based on Stated Intention (SI) data from the Dutch SpitsScoren reward scheme, we study the long-term effectiveness of the reward as an opposite of punishment or price. More specifically in this first analysis, we seek to understand travelers’ behavioral change over time as a function of socio-economic and situational variables using cross- sectional Mixed Logit (ML) models. Results support the results of former related studies. Driving off-peak is recognized as the most popular alternative. However, its utility decreases during the studied period. Results also show that by exploring and experiencing the different alternatives across time, travelers change from driving off-peak to the teleworking and changing route options. Socio-economic and situational variables strongly affect travelers’ decisions regarding alternatives chosen. However, their effectiveness changes over time.

Academic research paper on topic "The Long-term Effectiveness of a Reward Scheme in Changing Daily Travel Choices"

Available online at www.sciencedirect.com

ScienceDirect

Procedia - Social and Behavioral Sciences 111 (2014) 380 - 389

EWGT2013 - 16th Meeting of the EURO Working Group on Transportation

The Long-Term Effectiveness of a Reward Scheme in Changing

Daily Travel Choices

Elaheh Khademia*, Harry Timmermansb

a'bUrban Planning Group, Eindhoven University of Technology, P.O.Box 513, 5600 MB Eindhoven, the Netherlands

Abstract

Understanding behavioral response and attitudes of car users towards pricing policies and the public's acceptance of new pricing schemes are of the highest priority for generating a successful policy. The assumption underlying these policies is that travelers will adapt their current behavior when faced with pricing schemes. Behavioral psychology however claims the benefits of reward over punishment in terms of its effectiveness of influencing behavior. Although the arguments are compelling, the effects of reward schemes have not received much attention in the context of travel behavior. A fascinating reward scheme has been suggested in The Netherlands because of the massive public resistance against the implementation of road pricing in 2010. Based on Stated Intention (SI) data from the Dutch SpitsScoren reward scheme, we study the long-term effectiveness of the reward as an opposite of punishment or price. More specifically in this first analysis, we seek to understand travelers' behavioral change over time as a function of socio-economic and situational variables using cross-sectional Mixed Logit (ML) models. Results support the results of former related studies. Driving off-peak is recognized as the most popular alternative. However, its utility decreases during the studied period. Results also show that by exploring and experiencing the different alternatives across time, travelers change from driving off-peak to the teleworking and changing route options. Socio-economic and situational variables strongly affect travelers' decisions regarding alternatives chosen. However, their effectiveness changes over time.

© 2013 The Authors.PublishedbyElsevier Ltd.

Selection and/or peer-reviewunderresponsibilityofScientificCommittee

Keywords: Pricing policies; reward scheme; long-term effect; ML model.

1. Introduction

Pricing policies as a means of reducing congestion have received increasing attention in travel behavior research during the last decades (Nielsen, 2004; Arentze et al., 2004; Dissanayake and Kouli, 2007). The assumption underlying these policies is that travelers will adapt their current behavior when faced with pricing schemes.

Corresponding author. Tel.: +31-40-247-2934; fax: +31-40-243-8488. E-mail address: e.khademi@bwk.tue.nl.

1877-0428 © 2013 The Authors. Published by Elsevier Ltd.

Selection and/or peer-review under responsibility of Scientific Committee

doi:10.1016/j.sbspro.2014.01.071

Behavioral psychology however claims the benefits of reward over punishment in terms of its effectiveness of influencing behavior. Although the arguments are compelling, the effects of reward schemes have not received much attention in the context of travel behavior and there is not much international experience about the effects of reward schemes on individuals' travel behavior. Temporary free bus tickets as a reward scheme, to reduce car use, have been implemented in a few short-term studies, although their results are inconclusive (Fujii et al. 2001; Bamberg et al. 2002; Fujii and Kitamura 2003; Bamberg et al. 2003).

There are two fascinating implemented projects in The Netherlands, SpitsScoren and Spitsvrij, which are based on the opposite of congestion charging. The basic idea in both projects is to reward those travelers who are willing to avoid the morning and evening peak. GPS enabled smart phones provide information on travel alternatives, and enable motorists to keep track of their trips (and credits for not traveling). The phone should be on during trips, as it is part of the scheme to ensure that people are not travelling at peak hours in the specified corridors. The Spitsvrij project offers a smart phone travel app (but unlike SpitsScoren no actual phone) with a carpooling database (to facilitate people car sharing) and customized alternative travel options and also the installed OBU in the car. SpitsScoren or "profit from the peak" just covers one road (A15 highway corridor), whereas the whole area of Utrecht region is covered by the Spitsvrij project.

Before conducting these projects, a pilot experiment called "Spitsmijden" was designed that aimed to empirically explore the potential impacts of rewards on travel behaviour for avoiding rush-hour trips. The vicinity of The Hague in the west of The Netherlands was covered by the experiment and 340 participants were involved for 13 weeks. Participants could gain reward in the form of money or credits to keep a Smartphone, by changing their departure time for their work trips outside the morning rush-hour, switching to another travel mode, and teleworking. Various comprehensive research projects have been conducted based on this pilot experiment. For example Ben-Elia and Ettema (2009, 2010, 2011) identified the most important factors influencing travel behaviour in response to the reward stimuli. Authors concluded that the reward scheme is effective in the short-term. They also emphasized that choosing the response behaviour to avoid the peak trips is related to socio-economic characteristics and particular situations. In another study, Knockaert et al. (2012) estimated a number of discrete choice models that described commuters' behaviour with respect to departure time and transport mode choices. The results indicated that rewards can be used as an effective policy instrument and shifting departure time within the peak is likely a more important behavioral response. Tillema et al. (2010) compared two congestion management schemes: road pricing (a time differentiated kilometer charge) and peak avoidance reward (Spitsmijden), and their impact on changing commuter behavior based on two very different Dutch studies. They focused specifically on three points of comparison: effectiveness of both schemes in shifting trips during the peak hours, effectiveness of both measures on the alternative chosen, and the influence of price and reward levels and other exogenous respondent-related factors on these changes. Their results suggested that a reward scheme can be more effective than a pricing scheme and that both measures show the same influence regarding the alternatives chosen. Moreover, a higher impact was achieved by the 'shock' effect of introducing the measures. Gender, income and work flexibility are the most effective factors. Despite the effectiveness in decreasing peak-hour travel and the public acceptance of reward schemes, they believe that the long-term feasibility of this scheme is still in doubt.

Thus, these studies concluded that the reward schemes have been a useful solution in the short-term especially in specific local situations. Questions about the long-term feasibility and effectiveness of this scheme however remain. Does a reward scheme preserve behavioral changes in the long-term? To address this issue, this research investigates the impact of a reward scheme on travelers' behavior over time. More specifically, we seek to understand travelers' behavioral change over time as a function of socio-economic and situational variables. The analysis is based on the Stated Intention (SI) data from the Dutch SpitsScoren reward scheme. It should be emphasized from the outset that this data collected by a consultant was not collected with academic research in mind. Consequently, as we will see later, the interpretation of the results may be subject to confounding.

The paper is organized as follows. First, we describe the SpitsScoren project and the data in more detail. Next, we discuss the modeling approach and estimation results. The paper is completed with conclusions.

2. Data

The data used in our analysis were collected in one of the reward scheme projects in The Netherlands called SpitsScoren or "profit from the peak". The project started on October 26, 2009, and aims at reducing congestion on the Dutch A15 motorway corridor which connects the Rotterdam harbour with the Dutch and German hinterland. Because of the considerable success, the project was extended until December 21, 2012. Around two thousands regular users of the A15 motorway were identified by collecting plate license information to identify those vehicles that travelled at peak hours at least 5 times in four consecutive weeks. The drivers were then approached and invited to participate in the project. Similar to the other reward projects in The Netherlands, the basic idea was to pay participants not to drive on the mentioned corridor during morning (6-9 am) and afternoon (3-6 pm) rush-hours, thereby reducing the usual number of commuter trips in this corridor during peak hours. During the project, which thus lasted for 3 years, the reward scheme was changed several times. It started with 5 for avoiding the morning peak in the direction of the harbor From May 2011, participants could earn 1 .5 for avoiding the afternoon peak in addition to the morning reward; From August 2012 to the end of the project, the reward level decreased to 3 for the morning peak a nd increased to 3.50 for the afternoon peak.

The participants received a smart phone to provide information on travel alternatives, and to keep track of their trips. They were supposed to indicate their daily decision for the next day using a special application on the smart phone. The possible alternatives were: driving to work before or after peak hours; using mass transport; using slow mode; working from home; carpool appointment; using alternative route outside the corridor; using group transport; a special option which indicates they are on holidays and don't travel to work; and other options. Some additional services were also added to the phone to encourage and assist participants to choose the predefined alternatives and change their departure time and travel mode appropriately. GPS signals from smart phones and camera detection were used to verify participants stated intention (Palm et al. 2010, 2012).

The data is unique because of (i) the nature of the data - Stated Intention (SI) - collected in a real world project; (ii) the large number of participants, and (iii) the duration (2010-2012). Unfortunately, it lacks sufficient variation in reward levels in each year and also in general. In addition, due to strict privacy issues, it does not come with much background information related to the activity program of participants and their residence.

380 participants of the SpitsScoren project of whom we had socio-economic, situational, and background information were selected for our analysis. To answer the question about the long-term impact of the reward scheme, three similar periods of four consecutive weeks from 2010-2012 were used to analyze in this research. Table 1 presents the socio-economic variables and sample composition, while Table 2 shows the background and situational variables considered in this study. In addition to these variables, we also extracted the weather information for that area for different years to include in the analysis. It should be noted that the participants are men and as Table 1 shows they are car owners. Most of the sample is also highly educated and it is likely that they have more flexibility regarding their working hours and teleworking or working from home. The average age is 46, the youngest 26 and the oldest 66. According to Table 1, they are mostly married and half of them have children. However, there is no information about the age of the children in the household. According to Table 2, more than half of the sample made between 15-20 morning peak trips in the four consecutive weeks and have the possibility of teleworking.

Table 1. Socio-economic variables and sample composition

Description

Variable

Percentage

Aggregation

Marital Status (MS)

Having Children (HC)

MS=1 (Married)

MS=2 (Single)

HC=1(Yes)

HC=2(no)

MS and HC variables were used to make a new variable that reflects the household status (MC). This variable has 4 following categories: 1=single; 2=single parent; 3=married without children; 4=married with children

Number of cars

NC= 1 (one car)

NC can also be categorized to the two

in the household NC=2 (Two cars) 38.5% categories, respondents who have one car

(NC) NC=3 (More than two cars) 34.1% and others who have more than one in their household. In this case it can show the dependency of household to the car.

IN1 = <30000; 5.7%

IN2 = 30000-60000; 33.9% Aggregated to 4 categories, low, middle, and high plus the last group.

Income (IN) IN3 = 60000-90000; 18.2%

IN4 => 90 000; 7.8%

IN5 = I prefer not to answer 34.4%

1 = No schooling / education; 0.8%

2 = LBO / VBO / VMBO; 6.5%

Education 3 = MBO; 32.3% Aggregated to 4, low, middle, and high plus the last group.

(ED) 4 = HAVO / VWO; 9.9%

5 = HBO; 31.5%

6 = WO; 13.5%

1= Ag< 40 25.3%

Age (AG) 2= 40<=Ag<55 52.3%

3= Ag>=55 22.4%

'able 2. Situational variables

Description Variable Percentage Aggregation

Number of 6<= MPT<10=1 12.2%

morning peak trips 10<= MPT<15=2 31.5% Coded variables using effect

in four consecutive weeks (PT) 15<= MPT<20=3 56.5% coding

1 = Yes; 47.4%

Possibility of working at home (PH) 2 = Yes, but in practice it never happens; 7.3% Aggregated to 2 groups, who have

3 = yes but my activities will not allow it; 3.9% the possibility and who don't. Coded variables using effect

4 = No; 27.6% coding

5 = No, but in practice it is possible 13.8%

1 = every day same time start; 28.1%

2 = shift with fixed times; 7.3%

Flexibility of working hours (FH) 3 = can decide myself on start and end times; 14.3%

4 = can decide myself on start and end times but within certain time window 47.1%

5 = Other 3.1%

3. Modeling Approach and Results

In order to study the long-term effectiveness of the reward scheme and investigate the effects of socio-demographic variables, it was decided to start the analysis with cross-sectional models, estimated from annual data. This also gives better insight into the available data set before any further and more complicated dynamic analysis will be conducted. Thus, in this section the results of cross-sectional models are reported. There are three kinds of the data in the SpitsScoren project, Stated Intention (SI) of the participants, GPS traces and camera detection data. Because of privacy issues, we only have access to the SI data. In this paper, we only consider

participants' SI for the morning peak trips. As mentioned before, there are 11 predefined alternatives (including driving during the peak that can be interpreted as "no change" or base alternative) to avoid morning peak trips. Figure 1 shows the changes in participants' travel decisions regarding alternative chosen in three years. As can be seen from Figure 1, the most popular alternative for all three years is driving off-peak or changing the departure time of the trip, although its popularity decreases over time. This finding is in line with other related studies. Changing route and tele working or working from home are the second and third most popular options. The percentage of teleworking increases from 2010 to 2012.

Fig. 1. Changes in travelers' behaviour regarding alternative chosen over time

Considering the low percentage of some alternatives, the 11 predefined alternatives were aggregated into six discrete alternatives: (i) Driving off-peak or changing departure time; (ii) Changing route outside the corridor; (iii) teleworking or working from home; (iv) Switching to bicycle (the most common transport mode in The Netherlands); (v) Other (including carpooling, public transport, group transport, motorcycle and other), and (vi) Driving during the peak or no change (base alternative). Since each participant indicated his intention for working days in four consecutive weeks (up to 20 working days), the data was constructed as a panel. The number of working days can be different for each participant. Thus, the panel is not balanced in this data set.

The Mixed Logit (ML) formulation was used to estimate the cross-sectional models. Generally the ML model takes in to account random taste variation, substitution patterns, correlation in unobserved factors, and it also can accommodate panel effects. Equations 1, 2, and 3 show the general form of the ML model:

Uni (Xni + PnXni + £ni (1)

Pm = f Lnl(fi)f(fi\8)d0 (2) where

ImiP) =—]-a .+ex . (3)

Uni is the utility of alternative i for individual n and Pni is the probability of individual n choosing alternative i. The parameters of the utility function were estimated using Nlogit 4.0 in a stepwise manner. The Normal distribution was used for the random parameters in the utility function. Random alternative specific constants for "changing route" and "other" alternatives provided the best results. The base utility of "driving during the peak"

or "no change" was assumed to be 0. The numbers of Halton draws was set to 1000 to estimate the final model. The estimated utility functions for 2010 is summarised in Table 3.

The goodness-of-fit of the model in terms of Rho-square is 0.31. The p-value for all random parameters is less than alpha equal to 0.05. Thus, the mean of each random parameter is statically different from zero. Parameter estimates for the estimated standard deviations of the random parameters show the existence of significant heterogeneity in these parameters. Income and number of morning peak trips in the four consecutive weeks play an important role in participants' decision regarding changing their current pattern or not. The probability of adaptation to the reward increases with lower income and with a lesser number of morning peak trips. For "driving off-peak" or "changing departure time", household status and flexibility of working hours show significant effects. Interestingly the probability of changing travel behaviour by "driving off-peak" is less for single and single parent groups and higher for married people. This adaptation option is more chosen by travelers have more flexible working hours. Education levels affect the participants' choice about "changing route" but the second level that reflects the middle educated participants is not significantly different from zero. It should be noted that heterogeneity in the sample is captured in the random constant for this alternative. Possibility of teleworking influences their decision regarding working from home to avoid morning peak trips. Participants' age shows a significant effect on changing the mode of travel from car to bicycle. The effect is lowest for the youngest age group and highest for the middle age group. However, weather type does not play a role in choosing bicycle as the estimated parameter is not significant. For the last option that represents the aggregation of other mentioned options, the significant heterogeneity in the sample captures by the random alternative specific constant.

In order to compare the results for the different years and study the effectiveness of explanatory variables over time, the same model structure was estimated for the other two years. Table 3 also shows the estimated parameters for 2011 and 2012. Rho-square is around 0.30 and the p-value for all random parameters is less than 0.05. However, some of the non-random parameters like the first level of income, the first level of household status, the second level of education and the alternative specific constant for teleworking are not significant in 2011. It may be caused by the fact that participants had more holidays in the studied period in 2011, and consequently less regular travel patterns.

3.1. Comparison of the base utility of the different alternatives

Figure 2 shows the trend line of the base utility of the different alternatives to the base utility of "driving during the peak" (no change), which was assumed to be zero. It suggests that although the reward level decreases from 5 in 2010 to 3 in 2012, "driving off-peak" h as the highest base utility compared to the "no change" option. But according to the trend lines, this utility decreases 21.3% for "driving off-peak" and increases 38% for "changing route", 17.8% for "bicycling", and 28% for the "other" alternative from 2010 to 2011 with the same reward level. As reward level decreases from 2011 to 2012, the base utility of "driving off-peak", "changing route", and "bicycling" further decreases with 11.8%, 1.40%, and 38.6% respectively. Despite the reduction of reward level from 2011 to 2012, the utility of "other" alternative increases by 20%. The increase in utility of "teleworking" option is remarkable. The base utility of this alternative is negative in 2010 and positive in 2011 and 2012. It can be concluded that by exploring and experiencing the different options during the project, respondents shift from the "driving off-peak" alternative to "teleworking" and "changing route" options.

Table| 3. Estimated cross-sectional ML models for three years

Alt. Variable Description P P{lZl>z} St. dev. P{ IZI>z}

2010 2011 2012 2010 2011 2012 2010 2011 2012

No IN1 First level -0.427 -0.680 -0.735 0.000 0.111 0.000

Change IN2 Second level -0.236 -0.418 0.142 0.000 0.000 0.000

(Driving IN3 Third level 0.435 0.501 0.420 0.000 0.000 0.000

during the peak) PT1 First level -0.199 -0.220 -0.367 0.000 0.000 0.000

PT2 Second level -0.068 0.000 0.064 0.005 0.995 0.005

D0 Constant 1.053 0.829 0.731 0.000 0.000 0.000

MC1 Single -0.353 -0.010 0.143 0.000 0.705 0.000

MC2 Single parent -0.561 -0.523 -0.506 0.002 0.000 0.000

Driving Off-Peak MC3 Married without children 0.157 0.205 0.225 0.000 0.000 0.000

FH1 First level -0.453 -0.494 -0.595 0.000 0.000 0.000

FH2 Second level -0.181 -0.256 -0.522 0.000 0.000 0.000

FH3 Third level 0.133 0.247 0.290 0.00 0.000 0.000

FH4 Forth level 0.060 0.079 0.184 0.008 0.034 0.000

C0 Constant -2.726 -1.685 -1.709 0.000 0.000 0.000 3.192 2.833 2.924 0.000

Changing ED1 First level -2.106 -1.001 -1.024 0.000 0.053 0.016

Route ED2 Second level 0.139 0.003 -0.508 0.652 0.990 0.058

ED3 Third level 1.077 0.652 0.700 0.000 0.020 0.010

Tele H0 Constant -0.283 0.024 0.388 0.000 0.329 0.000

working PH First level 0.773 0.400 0.196 0.000 0.000 0.000

B0 Constant -1.238 -1.018 -1.411 0.000 0.000 0.000

Bicycling AG1 First level -0.825 -0.350 -0.308 0.000 0.000 0.000

AG2 Second level 0.540 0.324 0.181 0.000 0.000 0.000

WT First level 0.033 0.122 0.210 0.788 0.056 0.001

Other 00 Constant -4.022 -2.898 -2.323 0.000 0.000 0.000 3.843 3.169 2.346 0.000

Log-likelihood=-8909.940, p 2 =0.311

Log-likelihood=-9117.827, p2 =0.281

Log-likelihood=-9247.821, p 2 =0.275

3.2. Driving during the peak or do nothing

As the cross-section models show, income and number of morning peak trips in four consecutive weeks strongly affect participants' decision regarding changing their current travel behavior and shift to other alternatives. Figure 2 also illustrates the effect of different levels of number of current morning peak trips on "driving during the peak" or "no change" for different years. The significant and strong effect of this variable reflects the considerable effect of participants' current situation on their decisions. For all years, the utility of "no change" increases as the number of morning peak trips increases. It shows that participants, who have more current morning peak trips have a lower tendency to change, may be because they have less flexibility compared to the other travelers during the morning peak. The negative effect of this variable for the first group who make between 6 to10 peak trips, increases by 11% from 2010 to 2011, and 67% between 2011 and 2012. It means that

the tendency of changing current travel pattern increases over time and a decrease in reward level did not reduce the effect of this variable since the percentage of change increases 56% between two periods. For the second group with 10 to 15 peak trips, the utility is negative in 2010 and positive in 2012 for this alternative. It shows this group of travelers is not willing to change its current travel pattern over the studied period. The positive effect of the number of morning peak trips decreases 17.6% during the first period and increases 38% during the second period for the last group with 15 to 20 morning peak trips. It indicates that decreasing the reward level in the second period affects travelers' decisions in this group and results in a lower tendency to change the current travel pattern.

In the context of income, as expected, participants with higher income are not inclined to change their current travel pattern. In contrast, lower income group have higher tendency to adapt in response to reward.

Fig. 2. Trend line of the base utility of the different alternatives to the base utility of driving during the peak (no change) and effect of number of morning peak trips on driving during the peak (no change) alternative for different years

3.3. Driving off-peak (changing departure time)

Regarding effects of household status on "driving off-peak" alternative, all groups show the positive tendency to drive at off-peak hours. Married with and without children groups have a higher utility to use this option. It seems that being married reduces their burden since the partner may take on some of the responsibilities at home. The utility for the single group increases by 17% in the period with a higher reward level, and 6.7% in the lower one. That means that although the utility is increasing, the rate of increase is affected by decreasing the reward level. The utility of this alternative decreases for single parents by 38% in the first period, and 26.5% in the second period with the lower reward. In this case, this result indicates that single parent burden strongly decreases the benefits of the reward. The married travelers without children show decreasing utility which is more for the first period than the second with the lower gain, although the effect of the variable itself increases in both periods. Similarly, the utility of "driving off-peak" alternative for married travelers with children decreases in the first period more than in the second one.

Another variable for choosing this alternative is flexibility of working hours. As Table 2 shows, it has 5 categories. The differences between the groups are interesting as utility increases from highly inflexible in the first group to totally flexible in the third group. Utility, however, decreases from the third group to the fourth group, which again has less flexibility. There are different variations in the two different periods for different groups of travelers. Overall utility seems decreasing from the first period to the lower reward level period and this reduction is more remarkable for the second group of travelers since the utility in the second period decreases 29% more than the first one.

3.4. Changing route

Education plays an important role in participants' decision to change route. The utility of choosing this option increases as the education level increases and this effect is the same in the all years. It demonstrates that high-educated people have a high ability to search a new route through internet or other navigation devices. For the low educated travelers, the utility of this option increases 44% from 2010 to 2011 and decreases 2% between 2011 and 2012. The same pattern can be seen for the middle educated travelers but the increase is 35% in the first period and the reduction is 32% in the second period. For the high educated group, the utility of "changing route" increases by 37% and 3% in the first and second period respectively.

3.5. Bicycling

For Bicycling, age is an important variable. However, the young age group (26-40) has a lower tendency of switching from car to bicycle. Conversely, participants older than 40 have a higher utility and as a result are willing to shift from car to bicycle. But this tendency is lower for those who are older than 55 years of age. This result may come from the fact that the new generation in The Netherlands is more dependent on the car rather than bicycle that has been the common means of transport for many years. E-bike which has recently received increasing attention worldwide can be another reason for this finding. For the young travelers, the utility of bicycling increases by 34% in the first period, but decreases 26% in the second period. The utility for the middle age does not change that much in the first period. However, it decreases dramatically in the second period. For travelers older than 55, utility is reduced with 4% and 29% in the first and second period respectively.

Another interesting result for this alternative is the effect of weather conditions. This variable was aggregated into sunny and rainy in our analysis. Although the estimated parameters of this variable are not at the predefined significance level in 2010 and 2011, an increase in the effect of weather can be observed.

3.6. Teleworking or working from home alternative

As expected, the possibility of teleworking has a significant effect on participants' decision regarding this alternative. This effect is positive for participants' who have this possibility and is negative for the other group. The utility of choosing "teleworking" for the travelers with this possibility is reduced by 13.5% in the first period and increased by 38% in the second period. It should be noted that the effect of variable itself decreases in both periods. The utility of choosing this option for the other group increases considerably by 63% for the first period. In the second period, however, the utility of teleworking is positive for this group. This result reflects the recent interest in teleworking in The Netherlands. Although the reward level decreases in the second period, the utility of "teleworking" increases implying that the reward level does not seem affect the utility of this alternative for travelers of the second group. The effect of this variable itself for this group of travelers also increases in the first and second period.

4. Conclusion and Discussion

In this paper, we discussed long-term effectiveness of a reward scheme on travel behavior. More specifically, we used cross-sectional analysis to understand travelers' behavioral change over time as a function of socioeconomic and situational variables. The analysis was based on the SI data from the Dutch SpitsScoren reward scheme. Three similar periods of four consecutive weeks from 2010-2012 were used and 380 participants were selected for the analyses in this paper. Results of cross-sectional ML models support the results of previous studies. Results also indicate that by exploring and experiencing the different options over time, respondents changed their decision from "driving off-peak" to "teleworking" and "changing route". Income and number of morning peak trips for "driving during the peak", household status and flexibility of working hours for "driving off-peak", education for "changing route", possibility of teleworking for "working from home", and age for

"bicycling" showed strong effects for all three years. However, their effect changed over time. It should be noted that we could not consider the effects of reward levels directly in this cross-sectional analysis as the project lacked variation in each year. Also, because of a strict privacy clause, we did not have any data on travel time and distance that strongly affect travelers' decisions.

Acknowledgements

The research leading to these results has received funding from the European Research Council under the European Community's Seventh Framework Programme (FP7/2007-2013) / ERC grant agreement n° 230517 (U4IA project). The views and opinions expressed in this publication represent those of the authors only. The ERC and European Community are not liable for any use that may be made of the information in this publication.

We thank Goudappel Coffeng Consultants for making available the data.

References

Nielsen, O.A. (2004). Behavioral responses to road pricing schemes: Description of the Danish AKTA experiment. Journal of Intelligent Transportation Systems, 8(4), 233-251.

Arentze, T.A., Hofman, F., & Timmermans, H.J.P. (2004). Predicting multi-faceted activity travel adjustment strategies in response to

possible congestion pricing scenarios using an Internet-based stated adaptation experiment. Transport Policy, 11, 31-41.

Dissanayake, D., & Kouli, S. (2007). Intercity route choice stated preference model to investigate driver response to road pricing.

Proceedings of the 86th Annual Meeting of the Transportation Research Board, Washington, D.C.

Fujii, S., Garling, T., & Kitamura, R. (2001). Changes in drivers' perceptions and use of public transport during a freeway closure: Effects of temporary structural change on cooperation in a real-life social dilemma. Environment and Behaviour, 33, 796-808.

Bamberg, S., Ajzen, I., & Schmidt, P. (2002). Choice of travel mode in the theory of planned behavior: The roles of past behavior, habit, and reasoned action. Basic and Applied Social Psychology, 25, 175 -187.

Fujii, S., & Kitamura, R. (2003). What does a one-month free bus ticket do to habitual? An experimental analysis of habit and attitude change. Transportation, 30, 81-95

Bamberg, S., Rolle, D., & Weber, C. (2003). Does habitual car use not lead to more resistance to change of travel mode? Transportation, 30, 97-108.

Ben-Elia, E. & Ettema, D. (2009). Carrots versus sticks: Rewarding commuters for avoiding the rush-hour- a study of willingness to participate. Transport Policy, 16,68-76.

Ben-Elia, E. & Ettema, D. (2010). Commuters choice behavior with rewards for avoiding peak-hour driving. Proceedings of the 89th Annual meeting of the Transportation Research Board, Washington DC.

Ben-Elia, E. & Ettema, D. (2011). Rewarding rush-hour avoidance: a study of commuters' travel behavior. Transportation Research Part A Policy and Practice, 45 (7) 567-582.

Knockaert, J., Tseng, Y-Y., Verhoef, ET., a& Rouwendal, J. (2010). The Spijtsmijden experiment: a reward to battle congestion. Transport Policy, 24, 260-272.

Tillema, T., Ben-Elia, E., & Ettema, D. (2010). Road pricing vs. peak-avoidance rewards: A comparison of two Dutch studies. Proceedings of the 12th World Conference on Transportation Research, Lisbon, Portugal.

Palm, H., Kooistra, A., & Van Der Meulen, M. (2010). The behavioral impacts of rewarding scheme during two and a half years. European Transport Conference, Glasgow, Scotland.

Palm, H., Kooistra, A., & Van Der Meulen, M. (2012). The peak avoidance project in Rotterdam and its behavioral impacts. European Transport Conference.