1

Managed Lane Choices by Carpools Comprised of Family Members 1

Compared to Non-Family Members 2 3

Mark W. Burris, Ph.D, P.E. 4 Snead I Associate Professor 5 Zachry Department of Civil Engineering 6 Texas A&M University 7 CE/TTI Building Room 301G 8 3136 TAMU 9 College Station, TX 77843-3136 10 Corresponding author: mburris@tamu.edu 11 Ph: 979-845-9875 12 Fax: 979-845-6481 13 14 Mandeep Singh Pannu, EIT 15 Transportation Planning Analyst 16 Traffic Engineers, Inc. 17 8323 Southwest Freeway, Suite 200 18 Houston, TX 77074 19 mandeep@trafficengineers.com 20 Ph: 713-270-8145 21 22

23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39

40 Paper Submitted for Publication in Transportation Letters 41 November 2010 42 43 44

2

1 ABSTRACT 2 Carpools can be comprised of family members (fampools), non-family members (non-fampools) 3 or a combination of both. By analyzing the characteristics of fampools and non-fampools, we can 4 better understand both groups and how policies may impact each group differently. One area of 5 particular interest is the impact of managed lanes on the mode choice of fampools and non-6 fampools. Many managed lanes offer travelers a choice to use the lanes toll-free as a carpool or 7 use the lanes as a single occupant vehicle for a toll. This may influence some carpools to break 8 up, with one hypothesis being fampools would be more likely to stay together. 9 10 This research found that the majority of fampools and non-fampools are interested in using 11 managed lanes. The analysis of survey data proved that there are many similarities between the 12 fampools and non-fampools but very few differences. Fampools carpooled on a smaller 13 percentage of their weekly trips than non-fampools did. Fampools more often formed carpools 14 for dropping off kids at school or day care, non-fampools formed carpools more often for sharing 15 the vehicle expenses. The simulation study performed to analyze the impact of different travel 16 scenarios available on managed lanes on the mode choice of carpoolers showed that a very small 17 percentage of fampools and non-fampools chose to travel as a single occupant in managed lanes. 18 Both fampools and non-fampools were insensitive towards the toll cost. The analysis of the 19 descriptive statistics and the results from the simulation models confirmed that fampools and 20 non-fampools have similar commute characteristics, interest in managed lanes, and socio 21 economic characteristics – and had similar reactions to managed lanes. 22

23 KEY WORDS 24 Managed Lanes, Carpools, Fampools, mode choice 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46

3

INTRODUCTION 1 Carpooling declined during the 1980’s and 1990’s, even as the policies were in place to 2 encourage carpooling (Ferguson (1997); Pisarski (1996); Pisarski (2006); Poole and Balaker 3 (2008)). Overall, the carpool mode share has decreased from 20 percent of all commuters in 4 1980 to 12 percent in 2000 (Pisarski (2006)). The recent census data shows that the overall 5 carpooling mode share decreased to 11 percent in 2006 (Liberles (2009)). Carpoolers are 6 comprised of two very different groups of travelers: fampools and non-fampools. Additionally, 7 the majority of carpools these days are fampools (Pisarski (1996); Burris et al. (2009); Li et al. 8 (2007) ). Alan Pisarski, in his study of “Commuting in America – II” found, 9 10

Most carpooling today is not carpooling in the sense we knew it just a few years ago: 11 a voluntary arrangement among co-workers or neighbors. That is dying; most of the 12 surviving carpool activity consists of family members with parallel destinations and 13 timing. A term “Fampool” is used to describe the carpoolers who travel with other 14 household travelers. In major metropolitan areas fampoolers are one third to two 15 third of total carpoolers (Pisarski (1996)). 16 17

With the decreased carpool mode share and increased interest in using high occupancy/toll 18 (HOT1) lanes by transportation authorities in the United States, it became crucial to examine the 19 role of HOT lanes in encouraging or discouraging carpools (Parkany (1998)). Most research has 20 focused on how carpooling mode share changes with time, fuel costs, and socio-economic 21 characteristics like income, economic growth, availability of jobs, etc. A great deal of the 22 research done on carpool mode share has used the National Household Travel Survey (NHTS) 23 dataset which is limited to socio-economic and demographic variables (Pisarski (1996); Bard 24 (1997); Dohery et al. (2002); Teal (1987)). But none of these studies examined the behavior of 25 fampools and non-fampools in the vicinity of managed lane options which allow them to travel 26 as a SOV after paying a toll. It is important to understand the impact of toll and travel time 27 savings on mode choice of carpoolers as HOT lanes may negatively impact carpooling. 28

Managed lanes may greatly impact the different carpool groups and the impacts may be 29 considerably different by group. In the case of managed lanes2 (MLs), there is often the option to 30 pay for travel on the lanes as a single occupant vehicle (SOV) or travel toll-free as a HOV. It 31 would not be surprising if fampools would be more likely to stick together as carpools under this 32 scenario while non-fampools may more frequently take the opportunity afforded by the ML and 33 break up into (toll paying) SOVs. This research investigates the mode choice of fampools and 34 non-fampools in the presence of a ML option and identifies important factors influencing their 35 mode choice through descriptive statistics and discrete choice modeling of survey data. 36

The primary objective of this research is to examine the differences and similarities in the 37 characteristics and mode choice of these two types of carpoolers, including how fampools and 38 non-fampools react to different travel options on the managed lanes. This includes how 39 carpoolers from these two categories may switch their mode if given an option to travel alone on 40 a managed lane for a particular price. This information will be important to planners looking at 41

1 A high occupancy vehicle (HOV) lane is a lane that restricts usage to people who are driving vehicles with a certain number of occupants (often 2 or more). A high occupancy/toll (HOT) lane is a HOV lane that also allows people to use the lane for a toll if they have too few occupants to travel toll-free. 2 Managed lanes (MLs) is a generic term that encompasses all lanes that are managed (access controlled) by pricing, occupancy, etc. Therefore, HOT lanes are one form of MLs

4

adding MLs in their jurisdiction as MLs could have a significant, and negative, impact on 1 carpooling. If fampools were more likely to stay together and an area had a very high percentage 2 of fampools than the negative impact on carpooling of adding MLs would be less than in an area 3 predominated by non-fampools (more traditional carpools). 4

5 REVIEW OF CHARACTERISTICS OF FAMPOOLS AND NON-FAMPOOLS 6 A study analyzed traveler characteristics by mode choice in HOT corridors in Houston (Burris 7 and Figueroa (2006)). In this study, particular effort was spent examining the characteristics of 8 QuickRide travelers. In the QuickRide program, HOV3+ are allowed to travel for free and 9 HOV2 carpools are allowed to travel during peak periods for a toll of $2 while SOV’s are never 10 allowed to travel on the two HOT corridors. This research found in almost 60 percent of 11 carpools, the passengers were family members. QuickRide trips were more likely to be school 12 trips (11%), and in 76% of these trips, travelers were traveling with a child. 13

The issues related to carpooling were explored directly by examining reasons behind 14 people’s travel mode choice decision through a survey of travelers in Dallas-Fort Worth and 15 Houston, the two largest metropolitan areas in Texas (Li et al. (2007)). In the survey, 16 respondents rated different reasons for carpooling. Fampools rated “dropping off kids at school 17 or day care” and “enjoying travel with others” higher than other reasons. While, non-fampools 18 rated other factors including travel time and cost related factors higher on average. Comparing 19 the two groups, non-fampools rated factors including “access to HOV lanes”, “travel time 20 savings”, “sharing vehicle expenses”, “reliability of arrival time”, “splitting tolls on toll roads”, 21 “encouraged by program at work” and “preferred parking at work” significantly higher than 22 fampools. It also found that 75 percent of carpools were fampools. 23

The commute surveys from 1998 and 2003 in the San Francisco Bay Area estimated that 24 fampools make up one-third of carpools (Poole and Balaker (2008)). Similarly: 25

In California, fampooling increased from 49 percent in 1996 to 55 percent in 1999 26 (Southern California Association of Governments (1999)). 27

In Minneapolis-St. Paul, study found 67 percent of carpoolers were fampools (Poole and 28 Balaker (2008)). 29

In Houston, previous studies found that between 70 percent and 75 percent of carpools 30 were fampools (Burris and Figueroa (2006)). 31

In a nationwide estimate of all work commute carpools from 1990 to 2001, an increase of 32 fampools from 75.5 percent to 83 percent was found (McGuckin and Srinivasan (2008)). 33

Based on the NHTS data, the majority of carpool peak period trips were likely to be 34 fampools. The majority of highest occupancy trips during peak time were not work trips. The 35 highest occupancy trips were for social and family purposes, which also suggested the likelihood 36 of fampooling (Poole and Balaker (2008)). 37

38 SURVEY DATA COLLECTION 39 The survey responses from two Texas metropolitan areas (Houston and Dallas), which have 40 numerous HOV and HOT lanes, were used for this research. Data regarding traveler’s commute 41 characteristics, views towards managed lanes, stated preference questions, and socio economic 42 characteristics were collected. Many of the respondents already traveled on a roadway with a toll 43 or an HOV lane, while the remainder traveled on roadways where an HOV, HOT, or managed 44 lane might be added at a future date. The survey was advertised using newspaper articles, TV 45 news, push cards given at tollbooths and links on different websites. In both Houston and Dallas, 46

5

the majority of the respondents completed the survey online, but this data did not have many 1 respondents from lower income and minority populations. For this reason, a paper and laptop 2 based survey was also conducted at the community centers and department of public safety 3 offices in low income and minority neighborhoods. These surveys were identical to the ones 4 online. (See the project final report at https://ceprofs.civil.tamu.edu/mburris/Papers/0-5286-5 02_transmittal_032409.pdf for additional details regarding survey design and administration). 6 Unfortunately, the 4620 total respondents still did not reflect the ethnic and economic makeup of 7 the two cities, so a weighting process was undertaken. 8

The data was weighted to better represent the demographic features of the population. The 9 sampling weights (fixed) were developed using the income (four groups), ethnicity (four groups), 10 and toll-road usage (two groups). Due to the data collection process used for this survey, the 11 sampling design was not simple random sampling (SRS). In this scenario, the use of fixed 12 weights based on the assumption of SRS would imply lower standard error (SE). Thus, using 13 fixed weights may lead to some results from non-SRS surveys being found statistically 14 significant when in fact they are not. To address this issue, replicate weights were calculated 15 using fixed weights as the input (Burris et al. (2009)). All analyses in this paper use these 16 replicate weights. 17

The survey data from Houston and Dallas were compared to determine if they need to be 18 analyzed separately or can be analyzed as one group. Houston and Dallas respondent data were 19 analyzed based on their response to the stated preference questions. If the respondent selected a 20 managed lane option (SOV or HOV) in at least one of the four stated preference questions then 21 they were categorized as selecting a managed lanes option. The selection of a managed lanes 22 option was fairly consistent between the two cities in most strata (Table 1). There was no 23 significant difference in choosing MLs between the two cities, and therefore a model that 24 combined the data from these two cities was developed. 25

26 DESCRIPTIVE ANALYSIS OF CHARACTERISTICS OF FAMPOOLS AND NON-27 FAMPOOLS 28 With the decline of traditional carpooling and the increase in fampooling , there is an increasing 29 need for a better understanding of these two groups. This understanding canhelp identify reasons 30 for changes in the carpooling behavior and possible ways to encourage carpooling. This may be 31 even more critical in areas considering HOT lanes and MLs as the reaction to these new lane 32 choices may have considerably different impacts on fampools versus non-fampools. 33

For this purpose, carpoolers were divided into fampool and non-fampool groups. Fampools 34 included all carpools that had only family members – regardless of age. This is consistent with 35 HOV lane policy that considers all people, even those too young to drive, as members of the 36 carpool. Similarly, non-fampools were comprised of any set of multiple people in the same 37 vehicle who were not related, again regardless of age. 38

There were a few respondents who had both family members and non-family members in the 39 vehicle, termed mixed carpools. Due to the small sample size of mixed carpools, it was not 40 possible to keep the mixed carpools as an independent group for the descriptive statistics 41 analysis. The mixed carpools were further investigated to decide whether to merge them with the 42 group of fampools or non-fampools. The comparison of the mode choice of all three groups 43 revealed that mixed carpools showed different tendencies than did fampool and non-fampool 44 groups. Hence mixed carpools were excluded from all data analysis in this paper. The detailed 45

6

descriptive statistics were generated for the commute characteristics, interest in managed lanes, 1 and socio-economic characteristics for fampools and non-fampools (Table 2). 2

There were very few significant differences found between the two groups. Fampools 3 carpooled a on a smaller percentage of their weekly trips than non-fampools. Approximately 45 4 percent of fampools continued to their final destination after dropping off their passenger in 5 comparison to 23 percent of non-fampools. This makes sense as the majority of non-fampools 6 had the same destination. These were the only significant differences (p=0.10) between the two 7 groups. Many interesting similarities were found with fampools and non-fampools. Both 8 fampools and non-fampools were most likely to be on a commute trip. The amount of carpooling 9 increased for both fampools and non-fampools with increased travel distance. Fampools had an 10 average carpool formation time of 6.4 minutes in comparison to 6.2 minutes for non-fampools. 11 Overall, fampools and non-fampools had very similar characteristics. 12

In addition to examining the similarities and differences between the characteristics of 13 fampools and non-fampools, their reaction toward the use of managed lanes was analyzed. After 14 a brief description of MLs, survey respondents were asked if they would be interested in using 15 MLs. The majority of respondents in both groups indicated an interest in using managed lanes. 16 To explore further, many different factors for the interest or disinterest in using managed lanes 17 were also examined. The respondents ranked different factors on a scale from 1 to 5, with rank 1 18 as the least important and 5 as the most important. Both fampools and non-fampools stated 19 “travel time reliability” as the most important factor for their interest in using managed lanes. 20 “Able to travel faster than GPL” was the second most important reason for travelers from both 21 groups. When asked about the reason for their disinterest in using managed lanes, fampools rated 22 “other” the highest, followed by “not interested in paying a toll”. Because “other” was so 23 important, the text that respondents entered corresponding to “other” was examined. The vast 24 majority of the typed notes were anti-toll. Non-fampools were not interested in using managed 25 lanes because of the toll and also because they were not interested in driving alone. Note that 26 despite these small differences in rankings, the difference was not statistically significant and 27 there were no significant differences between the two groups found regarding their interest or 28 disinterest in using MLs. 29

The important reasons for the formation of the respondent’s current carpool were also 30 examined to find any differences between fampools and non-fampools. Fampools stated 31 “dropping off kids at school or day care” was more important to them than non-fampools. Non-32 fampools rated “sharing vehicle expenses” higher than fampools. These differences were 33 significant at a 5 percent level of confidence. Overall, fampools rated “relaxation while 34 traveling” as the most important while “Preferred parking at work” and “encouraged by program 35 at work” were the least important factors in the formation of their current carpool. Non-fampools 36 indicated “access to HOV lanes” was the most important reason while “dropping off kids at 37 school or day care” was the least important reason in the formation of their current carpool (see 38 Table 3). 39

For further analysis, fampools were split into two sub-groups: with children and adults only. 40 The three groups were examined to see if fampools with a child(ren) were different than 41 fampools of adults-only or non-fampools and if fampools of adults-only were most similar to 42 non-fampools. It was found that non-fampools carpooled more regularly in a week in comparison 43 to the two fampool groups. Within fampool subgroups, fampools of adults-only carpooled more 44 often than fampools with a child(ren). The difference was significant at a 5 percent level of 45 confidence. When comparing their average carpool formation times, fampools with a child(ren) 46

7

reported the highest time of 9.1 minutes in comparison to the 5.3 minutes for fampools with 1 adults-only and 6.2 minutes for non-fampools. However, this difference was not significant and 2 there were few other differences found. 3

The data were analyzed further to examine the characteristics of travelers from these three 4 groups by dividing them into sub-groups based on their current number of passengers. However, 5 none of the differences between the groups were found to be significant. This was partially due 6 to the similarities between the groups and partially due to lower sample sizes (and higher 7 standard error). 8

The descriptive statistics analysis proved that there were many similarities between the 9 fampools and non-fampools but very few differences. The differences found in the descriptive 10 statistics analysis did not provide sufficient information to conclude any specific reasons for the 11 mode switching by different groups of carpoolers. Overall, this information was only minimal 12 help in determining what variables should most likely be included in the mode choice model 13 (discussed below). 14

15 DISCRETE CHOICE MODELING 16 Another objective of this research was to estimate the mode choice of fampools and non-17 fampools on managed lanes. The reaction of carpoolers from both groups were examined for 18 different toll and travel time options available on managed lanes based on their survey responses 19 to ML stated preference questions. The mode choice models were developed for fampools and 20 non-fampools using Nlogit (NLogit (2009)). Different random parameter logit models were 21 developed to predict traveler’s responses to the GPL and ML options. Different variables were 22 used as dummy variables in the model built to predict mode use of the six options given in the 23 stated preference section of the survey. These options were: 24

25 SOV on the MLs 26 HOV2 on the MLs 27 HOV3+ on the MLs 28 SOV on the GPLs 29 HOV2 on the GPLs 30 HOV3+ on the GPLs 31 32 In the next step a market segmentation approach was used was to decide upon whether to use 33

a pooled logit model (a single model for both fampool and non-fampool respondents) or a 34 separate model for respondents from each group (Koppelman and Bhat (2006)). Market 35 segmentation approach can be used to determine whether the impact of the variables is different 36 among population groups. Models were estimated for the sample associated with each segment 37 and compared to the pooled model (all segments represented by a single model) to determine if 38 there are statistically significant and important differences among the market segments. Based on 39 the test statistic, it was found that two separate models better predicted the mode choice of 40 fampools and non-fampools. 41

42 Mode Choice Model for Fampools 43 As described above, a random parameter logit model was estimated to better understand the 44 factors influencing the mode choice of fampools. Various utility equations with different 45 variables were tested. The utility functions given in Table 4 were found to have best fit and 46

8

explanatory ability. All the variables were significant at a 5 percent confidence level. Driving 1 alone on the GPLs was the base mode. 2

The utility equations for the mode chosen by fampools revealed relatively few surprises. To 3 begin, the only positive alternative specific coefficient (ASC) was for HOV2 on the GPLs. Exact 4 interpretation of the coefficient was difficult since we could not be sure if the result was more 5 due to the fact that it was a fampool or the mode chosen. For example, fampools on a commute 6 trip were less likely to choose HOV2 on MLs. It was hard to be sure if that was because 7 fampools were not on commute trips or if they avoided HOV2 on MLs. In any case, the results 8 appeared reasonable and will be compared to the non-fampool results for additional clarity. The 9 overall model provided an acceptable adjusted rho square value of 0.24. The estimated VTTS 10 was $ 22.80 per hour for the fampools. 11

12 Mode Choice Model for Non-fampools 13 Similar to the development of the fampool mode choice model, a random parameter logit model 14 was developed for non-fampools. All the variables used in the non-fampool model were 15 significant at a 5 percent confidence level (Table 5) except for income in the final mode (HOV3+ 16 on the GPLs). The overall model provided an acceptable adjusted rho square value of 0.18. It 17 was interesting to note that time was not a significant variable in this model. It implied non-18 fampools were not very sensitive to travel time. Note that this is only valid for the range of tolls 19 displayed in the survey. However, this range covered all reasonable toll levels as most 20 respondents were shown tolls between $0.25 and $7, but a small percentage of respondents were 21 shown tolls up to $15. 22

As time was not a significant variable, according to the mode choice model in Table 5, any 23 changes in travel time did not impact the mode choice of non-fampools. To investigate this 24 unusual result the survey data was analyzed further. Data was analyzed based on the mode 25 choice by the non-fampools in the stated preference section of the survey and travel time savings 26 offered by the managed lanes for a specific toll. It was found that even with the large travel time 27 savings non-fampools were equally likely to choose between MLs and GPLs (Figure 1). It 28 confirmed that non-fampools were not sensitive to the time and were equally likely to choose 29 between MLs and GPL. The findings from the descriptive statistics analysis also confirm that 30 non-fampools were more sensitive to the toll cost in comparison to the fampools. 31

32 Simulation Results for the Fampools and Non-fampools 33 Next, to analyze the impact of travel time and tolls on the mode choice of fampools and non-34 fampools, many different managed lane travel scenarios were simulated. The mode choice 35 models described in the previous section were used to estimate the percentage of travelers 36 willing to use managed lanes under realistic toll and travel time savings scenarios. The main 37 focus was on to analyze any mode switching by carpool groups. The simulation included several 38 assumptions that resemble some typical ML scenarios found in the US: 39

40 The trip length was assumed to be 10 miles 41 The managed lane speed was assumed to be 65 mph 42 The speed on general purpose lanes3 (GPLs) was assumed to be 40 mph 43 The toll varied between $0.00 and $5.00 in increments of $0.50 44

3 General purpose lanes (GPLs) refer to the lanes that are open to all traffic without tolls or occupancy restrictions. They are the ‘typical’ freeway lanes, where the MLs are the unusual, managed, lanes.

9

Total travel time was the combined time of the actual travel time plus carpool formation 1 time 2

3 The time required to pick up and drop off the passenger was calculated from the reported 4

carpool formation time by the respondents in the survey. HOV2s reported an average carpool 5 formation time of 5.0 minutes and HOV3+s reported an average carpool formation time of 9.8 6 minutes. This time was added to the actual travel time to get the total travel time for the 7 simulation runs. 8

To analyze the mode switching behavior of fampols and non-fampools, three different 9 scenarios were simulated. In many ML situations higher occupancy vehicles receive some price 10 discount or free travel – so these scenarios include such price discount. In the first scenario, the 11 SOV toll was varied from the $0.50 to $5.00, the HOV2 toll was kept at half of the SOV toll, and 12 HOV3+ were allowed to travel for free. As the toll increased from $0.50 to $5.00 for SOVs (and 13 to $2.50 for HOV2s), the decrease in SOVs and HOV2s on MLs was compensated by an 14 increase in all GPL modes (Figure 2, 3). To summarize, with the increased toll, the majority of 15 travelers who switched chose SOV or HOV2 mode on GPLs. The percentage switching to 16 HOV2s on GPLs was slightly higher for fampools. While for non-fampools, SOVs on GPLs 17 observed the highest percentage switch. Overall, the carpool mode share decreased for both 18 fampools and non-fampools by 1.2 percent and 1.5 percent respectively. 19

In the second scenario the removal of the toll discount for HOV2s on managed lanes was 20 investigated. The SOV toll on the managed lanes was kept constant at $5 but the HOV2 toll on 21 the managed lanes varied from $0.00 to 4.50. HOV3+s were still allowed to use the managed 22 lanes for free. For both fampools and non-fampools, as the toll increased, a very sharp decrease 23 was observed for HOV2s on MLs (Figure 4, 5). The majority of travelers, who switched, chose 24 SOV and HOV2 modes on GPLs. For fampools, the highest percentage increase was observed in 25 HOV2 mode travelers on GPLs while the highest percentage of non-fampools switched to SOV 26 mode on GPLs. With the increased toll, there was a similar decrease observed in the overall 27 carpool mode share for both fampools and non-fampools of 3.5 percent and 3.3 percent 28 respectively. 29

Third scenario examined the removal of free travel for HOV3+s on managed lanes. The 30 HOV3 toll varied from $0.50 to $2.50 but the SOV toll was constant at $5 and the HOV2 toll 31 was also kept constant at $2.50. For both fampools and non-fampools, as the toll increased, the 32 decrease in HOV3 travelers on the MLs was compensated by the increase in all other modes 33 (Figure 6, 7). There was an overall decrease observed in the total number of HOVs of 0.9 percent 34 for the fampools and 1.3 percent for the non-fampools. 35

From the simulations, it was observed that both fampools and non-fampools switched their 36 modes to GPLs in response to the increased toll on the MLs. To examine the extent of this mode 37 switching, the elasticities were calculated. Elasticities were calculated for scenario 1 to test the 38 reasonableness of fampools and non-fampools switching to the SOV and HOV2 modes on ML 39 for every $0.50 increase in the toll value. For fampools, SOV elasticity varied between -0.04 and 40 -0.38 and for HOV2s elasticity varied between -0.01 and -0.13. For non-fampools, SOVs 41 elasticity varied between -0.00 and -0.35, and the HOV2s elasticity varied between -0.01 and -42 0.11. According to the literature these values seemed reasonable and demand elasticity was fairly 43 inelastic (Matas and Raymond, 2003). 44

45 46

10

Comparison between Fampools and Non-fampools 1 Both fampools and non-fampools were not very sensitive towards the cost of the toll. For both 2 fampools and non-fampools, as the toll increased, the decrease in the number of travelers on the 3 MLs was compensated by an increase of travelers in the GPLs. The majority of travelers 4 switched to SOV and HOV2 modes on the GPLs. The percentage of fampool and non-fampool 5 travelers switching to SOV and HOV2 modes on GPLs were too close to reach any concrete 6 conclusion regarding their mode switching behavior although fampools remained HOVs slightly 7 more often. 8

The results from the mode choice analysis agreed with the findings from the descriptive data 9 analysis. For fampools, the percentage of travelers who switched to HOV2 on GPLs was slightly 10 higher. These results confirmed the finding that “dropping off kids at school or day care” was 11 more important reason for fampools than non-fampools for the formation of their current 12 carpool, hence were more likely to stay with their current carpool. For non-fampools, the 13 percentage of travelers switching to SOV mode on GPLs was slightly higher. The descriptive 14 statistics analysis found that “sharing vehicle expenses” was more important to non-fampools 15 than fampools for the formation of their current carpools. This could be one reason for non-16 fampools switching to SOVs on GPL. However, this percentage shift was too small to validate it 17 as a conclusion. Overall, in the mode choice analysis, there were no contrasting differences 18 observed between fampools and non-fampools. 19 20 CONCLUSIONS 21 To better understand the mode choice of fampools and non-fampools, this research examined 22 survey data collected from travelers in Houston and Dallas, Texas. The majority of both 23 fampools and non-fampools were interested in using managed lanes. There were very few 24 significant differences found in the commute characteristics of fampools as compared to non-25 fampools. Fampools carpooled a smaller percentage of their weekly trips than non-fampools. 26 Fampools indicated that dropping off kids at school or day care was the most important reason 27 for their current carpool formation in comparison to the non-fampools. Sharing vehicle expenses 28 was more important to the non-fampools than the fampools. This was one of the few clues 29 behind any possible differences in how the two groups may change mode in response to MLs. 30 There were almost no significant differences found in the socio-economic characteristics and 31 reasons for interest or disinterest in using managed lanes of fampools and non-fampools. 32

To further investigate how these travelers may react to MLs, mode choice models were 33 developed for the fampool and non-fampool carpool groups. The value of travel time for the 34 fampools was estimated to be $ 22.80 per hour. Non-fampools were not sensitive to travel time. 35 Next a simulation study was done using these models to estimate the percentage of fampools and 36 non-fampools that would likely use managed lanes under different travel scenarios. A very small 37 percentage of fampools and non-fampools chose SOV mode on ML. Both fampools and non-38 fampools were insensitive to the tolls. With an increased toll, some travelers from both groups 39 switched to SOV and HOV2 modes on GPLs. However, this shift was very small and so similar 40 that nothing concrete could be concluded about differences in the mode switching behavior of 41 fampools versus non-fampools based on these findings. This also indicates that factors other than 42 price are influencing their choice to carpool – and the most important reasons for carpooling are 43 found in Table 3. 44

Although the survey sample size was large, the number of travelers in some categories, for 45 example HOV3+ non-fampool, was small. A considerable variation in the characteristics of the 46

11

travelers sometimes resulted in large standard deviations for the descriptive statistics of the 1 groups. This, combined with the fact that respondent answers from fampools and non-fampools 2 were similar meant little difference between the two groups. 3

4 ACKNOWLEDGEMENTS 5 The authors wish to thank the Texas Department of Transportation for funding the Texas 6 Transportation Institute (TTI) project entitled “The Role of Preferential Treatment for Carpools 7 in Managed Lane Facilities”, from which the data are used for this research. However, the 8 contents of this paper reflect the views of the authors, who are responsible for the facts and the 9 accuracy of the data presented herein. The contents do not necessarily reflect the official views 10 or polices of the Texas Department of Transportation. The authors would like to thank the TTI 11 research team led by Ginger Goodin as research supervisor for conducting the survey to collect 12 data and helping with the data analysis. Special thanks go to the TxDOT project director, Matt 13 MacGregor, P.E., for his leadership and guidance. We would especially like to mention Dr. 14 David Ungemah for programming the online survey and Maneesh Mahlawat for the initial data 15 cleaning. 16

17 18

REFERENCES 19 Bard, E. A. Transit and Carpool Commuting and Household Vehicle Trip Making. In 20 Transportation Research Record: Journal of the Transportation Research Board, No. 1598, 21 Transportation Research Board of the National Academies, Washington, D.C., 1997, PP. 25-31. 22

Burris, M. W., D. H. Ungemah, M. Mahlawat and M.S. Pannu. Investigating the Impact of Tolls 23 on HOVs Using MLs. In Transportation Research Record: Journal of the Transportation 24 Research Board, No. 2099, Transportation Research Board of the National Academies, 25 Washington, D.C., 2009, pp. 113-122. 26 27 Burris, M. and C. Figueroa. Analysis of Traveler Characteristics by Mode Choice in HOT 28 Corridors. Journal of the Transportation Research, Vol. 45, Issue 2, 2006, pp. 103-117. 29

Chen, T. D. Compendium of Student Papers: 2003 Undergraduate Transportation Engineering 30 Fellowship Program: Improving Houston Quickride Through an Investigation of Former Users. 31 Texas Transportation Institute, Texas A&M University, College Station, Texas, 2003. 32

Doherty, S.T., M. Gosselin, K. Burns and J. Andrey. Household Activity Rescheduling in 33 Response to Automobile Reduction Scenarios. In Transportation Research Record: Journal of 34 the Transportation Research Board, No. 1807, Transportation Research Board of the National 35 Academies, Washington, D.C., 2002, pp. 174-182. 36

Ferguson, E. The Rise and Fall of the American Carpool: 1970-1990. Transportation, Vol. 24, 37 1997, pp. 349-376. 38 39 Koppelman, F. S., Bhat, C. A Self Instructing Course in Mode Choice Modeling: Multinomial 40 and Nested Logit Models. University of Texas, Austin, Texas, 2006. 41

12

Li, J., P. Embry, S. P. Mattingly, K. F. Sadabadi, I. Rasmidatta, and M.W. Burris. Who Chooses 1 to Carpool and Why?. In Transportation Research Record: Journal of the Transportation 2 Research Board, No. 2021, Transportation Research Board of the National Academies, 3 Washington, D.C., 2007, pp. 110-117. 4

Liberles, J. Riding Along Together: Carpooling. http://www.carectomy.com/riding-along-5 together-carpooling/. Accessed September 17, 2009. 6 7 Matas, A. and J. Raymond. The Demand Elasticity on Tolled Motorways. Journal of 8 Transportation and Statistics, Vol. 6, Numbers 2/3, 2003, pp. 91-108. 9

McGuckin, N. and N. Srinivasan. The Journey-to-Work in the Context of Daily Travel. 10 http://ctpp.transportation.org/Future/resource_papers/Journey-to-Work.pdf/. Accessed November 11 10, 2008. 12

Nlogit. Modeling Individual Choice with NLOGIT: Model Estimation, 2009. 13 http://www.limdep.com/features/capabilities/nlogit/model_estimation_2.php/. Accessed July 9, 14 2009. 15

Parkany, E. Can HOT Lanes Encourage Carpooling? A Case Study of Carpooling Behavior on 16 the 91 Express Lanes. Institute of Transportation Studies, University of California, Irvine, 17 California, 1998. 18

Pisarski, A. E. Commuting in America II, the Second National Report on Commuting Patterns 19 and Trends. Urban Land Institute, Washington D. C., 1996. 20

Pisarski, A. E. Commuting in America III, the Third National Report on Commuting Patterns 21 and Trends. Transportation Research Board, Washinton D. C., 2006. 22

Poole, R. W. and T. Balaker. Virtual Exclusive Busways: Improving Urban Transit while 23 Relieving Congestion. Policy Study 337. Reason Foundation. 24 http://reason.org/files/9428e00f34f5a65fd35f5b6589f1a0dd.pdf/. Accessed July 10, 2008. 25

Southern California Association of Governments. State of the Commute Report 1999. Los 26 Angeles. http://www.scag.ca.gov/publications/pdf/SOC_1999.pdf/. Accessed February 05, 2009. 27

Teal, R. F. Carpooling: Who, How and Why. Transportation Research Part A, Vol. 21, Issue 3, 28 1987, pp. 203-214. 29

30 31 32 33 34

35

13

1 TABLES 2 3 TABLE 1 Respondents Choosing a Managed Lanes Option 4 Characteristic Percentage Choosing a Managed Lane Option Dallas (n=1,503) Houston (n=2,366) Proportion Std.

Error Proportion Std.

Error p-value

Household Income Less than $25,000 75.3% 7.9% 73.9% 2.7% 0.39 $25,000 - $50,000 64.2% 3.8% 67.7% 2.5% 0.30 $50,000 - $100,000 66.9% 9.2% 74.9% 4.3% 0.29 Greater than $100,000

75.6% 2.5% 76.5% 1.1% 0.38

Ethnicity

Caucasians 73.4% 3.1% 73.6% 2.7% 0.40 Afro-American 64.9% 6.9% 76.7% 2.4% 0.11 Hispanic 68.7% 7.7% 71.9% 5.6% 0.38 Others 58.2% 4.9% 68.3% 2.9% 0.08

Trip Purpose

Commute 70.3% 3.2% 72.4% 2.8% 0.35 Recreational 71.7% 6.8% 77.8% 3.2% 0.29 Work 67.1% 6.7% 70.6% 5.1% 0.37 School 59.1% 16.3% 82.4% 6.4% 0.16 Other 46.0% 25.6% 56.0% 13.9% 0.38

Mode

SOV 69.7% 3.6% 71.0% 2.5% 0.38 HOV-2 72.1% 5.0% 78.8% 5.4% 0.26 HOV-3+ 80.6% 16.4% 75.8% 8.5% 0.39 Transit 62.2% 6.3% 72.0% 5.8% 0.21 Motorcycle 65.3% 27.3% 64.0% 16.4% 0.40 Vanpool 68.9% 31.5% 78.3% 5.9% 0.38 Total 69.5% 3.0% 72.9% 1.9% 0.25 Note that the number of respondents (n) varied slightly from question to question as not all 5 respondents answered every question. 6 7 8 9 10 11 12 13

14

TABLE 2 Descriptive Statistics of Fampools and Non-fampools 1

Characteristics Fampools Non-fampools

Proportion Std Error Proportion

Std Error

P- Value

Interested in using ML

Yes 68.2% 4.9% 66.5% 18.9% 0.93

No 31.8% 4.9% 33.5% 18.9% 0.93

525 217

Trip Purpose Commute 42.7% 8.1% 57.4% 13.0% 0.34 Recreational 39.0% 8.9% 17.2% 11.3% 0.13 Work related 11.3% 6.0% 21.5% 10.3% 0.39 School 4.6% 4.7% 0.7% 1.0% 0.41 Other 2.5% 1.9% 3.2% 3.7% 0.87 Total (N) 561 221 Typical Trip Length Short (0-3 miles) 1.5% 1.6% 1.9% 2.7% 0.90 Medium (4-9 miles) 17.8% 6.6% 10.2% 9.7% 0.52 Long (10-20 miles) 48.5% 8.3% 40.3% 10.6% 0.54 Very Long (more than 21miles) 32.2% 7.4% 47.6% 14.4% 0.34 Average trip distance (miles) 20.2 3.1 23.4 3.8 0.51 Total (N) 532 214

Number of Trips per Week

1 or 2 17.4% 8.7% 7.3% 5.2% 0.32

From 3 to 5 29.2% 6.3% 33.0% 11.0% 0.77

From 6 to 9 7.6% 3.1% 14.9% 7.8% 0.38

10 25.8% 5.7% 28.9% 7.9% 0.75

more than 10 20.0% 5.7% 15.8% 11.3% 0.74 Average number of trips 7.8 1.4 8.0 1.3 0.94 Total (N) 545 218

15

TABLE 2 (Continued) Descriptive Statistics of Fampools and Non-fampools

Characteristics Fampools Non-fampools

Proportion Std Error Proportion

Std Error

P- Value

Carpool Trips per Week Percentage of trips carpooled 57.1% 7.1% 78.0% 9.0% 0.07** Total (N) 515 208 Pay Toll Yes 15.7% 3.7% 20.2% 7.2% 0.58 Total (N) 543 205 Carpool Formation Time Average Time (min) 6.4 2.7 6.2 1.2 0.95 Total (N) 414 170 Age From 16 to 24 years old 17.4% 10.2% 11.9% 10.5% 0.71 From 25 to 34 years old 33.3% 8.6% 30.7% 5.1% 0.79 From 35 to 44 years old 22.3% 7.9% 25.1% 8.1% 0.80 From 45 to 54 years old 14.4% 10.2% 22.8% 5.6% 0.47 From 55 to 64 years old 7.9% 7.6% 8.8% 7.7% 0.93 More than 65 years old 4.7% 4.7% 0.7% 0.5% 0.39 Total (N) 555 218 Gender Male 63.6% 10.2% 49.7% 6.3% 0.24 Female 36.4% 10.2% 50.3% 6.3% 0.24 Total (N) 559 213 Number of Vehicles One or None 19.8% 10.4% 28.5% 17.6% 0.67 Two 48.5% 9.9% 35.7% 10.0% 0.36 Three or more 31.7% 6.1% 35.8% 11.7% 0.76 Total (N) 495 201

16

TABLE 2 (Continued) Descriptive Statistics of Fampools and Non-fampools Characteristics Fampools Non-fampools

Proportion Std Error Proportion

Std Error

P- Value

Household Type Single adult 7.9% 5.6% 25.7% 13.5% 0.22 Unrelated adults (e.g., roommates) 6.1% 4.5% 12.2% 6.1%

0.42

Married without child 25.1% 11.2% 16.6% 9.6% 0.57 Married with child(ren) 47.6% 7.3% 29.2% 15.0% 0.27 Single parent 10.4% 6.7% 12.2% 9.5% 0.88 Other 2.9% 1.1% 4.0% 5.4% 0.85 Total (N) 544 214 Household Size One 4.4% 4.8% 12.3% 5.5% 0.28 Two 30.0% 13.4% 30.1% 7.7% 1.00 Three 27.6% 6.3% 30.3% 10.3% 0.82 Four 23.3% 4.6% 15.3% 7.0% 0.34 Five or more 14.6% 5.0% 12.0% 9.7% 0.81 Total (N) 517 208 Ethnicity Caucasian 39.8% 22.0% 48.4% 14.7% 0.74 Afro-American 15.8% 8.9% 15.6% 10.0% 0.99 Hispanic 31.3% 12.3% 18.8% 9.8% 0.43 Other 13.2% 10.0% 17.1% 5.7% 0.73 Total (N) 546 212 Income Less than $34,999 30.6% 9.3% 38.4% 9.7% 0.56 From $35,000 to $49,999 30.3% 11.6% 18.1% 7.0% 0.37 From $50,000 to $74,999 10.2% 5.8% 13.9% 6.6% 0.67 From $75,000 to $99,999 14.1% 5.0% 13.0% 5.6% 0.88 More than $100,000 14.8% 8.7% 16.6% 11.0% 0.90

Total (N) 516 204

17

TABLE 2 (Continued) Descriptive Statistics of Fampools and Non-fampools

Characteristics Fampools Non-fampools

Proportion Std Error Proportion

Std Error

P- Value

Road

Houston: Beltway 8 (only Houston toll road in list) 2.1% 0.9% 0.9% 0.8% 0.99 Houston: All other roads listed 61.7% 16.2% 67.3% 16.7% 0.87

Dallas: George Bush Turnpike and Dallas North Tollway (only Dallas toll roads in list) 3.9% 2.9% 2.4% 4.6% 0.96 Dallas: All other roads listed 26.9% 14.1% 24.5% 16.5% 0.87 No road selected 5.4% 5.5% 4.9% 5.8% 0.95 Total (N) 563 220 Occupation Administrative 17.5% 8.5% 18.1% 6.8% 0.96 Sales, service, manufacturing, student, and self-employed 22.4% 6.3% 26.4% 10.7%

0.75

Stay-home, unemployed, others 19.1% 4.6% 10.4% 7.1% 0.30 Total (N) 552 215 Education High school graduate or less 22.7% 7.8% 23.1% 8.6% 0.97 Some college/Vocational 31.5% 9.2% 34.4% 12.9% 0.85 College graduate 31.7% 5.9% 26.7% 8.7% 0.64 Postgraduate degree 14.1% 12.5% 15.8% 7.6% 0.91 Total (N) 552 214 Pay to Park at Destination Yes 86.8% 3.2% 75.1% 8.6% 0.20 No 13.2% 3.2% 24.9% 8.6% 0.20 Total (N) 562 219 After Passenger Drop Off Driver / Passenger have Same Destination 40.4% 14.4% 57.9% 6.6% 0.27 Continue to Final Destination 44.7% 10.3% 23.2% 5.9% 0.07** Pick up Additional Passengers 2.3% 2.6% 0.3% 0.4% 0.44 Perform Errands 11.3% 6.9% 18.2% 8.5% 0.53 Other 1.3% 1.5% 0.5% 1.1% 0.66 Total (N) 434 185

** significantly different at 10 percent level of confidence 1 These are the responses from the survey, often the driver of the fampool/non-fampool. So each 2 respondent basically represents 1 carpool. 3

18

TABLE 3 Important Factors in the Formation of Current Carpool by Fampools and Non-1 fampools 2 Factor Fampools Non-fampools

#Obs MeanStd Error #Obs Mean

Std Error

P-value

Drop off Kids at School/Day Care 487 2.9 0.4 189 1.9 0.2 0.03* Access to HOV Lanes 497 3.4 0.4 201 4.0 0.3 0.26 Relaxation while Traveling 80 3.7 0.5 32 3.8 0.8 0.91 Help Environment and Society 488 3.4 0.3 197 3.5 0.4 0.81 Travel Time Saving 497 3.1 0.2 201 3.5 0.3 0.26 Enjoy Travel with Others 507 3.5 0.3 193 3.5 0.3 0.98 Sharing Vehicle Expenses 503 3.0 0.2 211 3.8 0.2 0.01* Reliability of Arrival Time 462 3.0 0.3 180 3.5 0.2 0.15 Splitting Tolls on Toll Roads 63 2.4 0.6 30 3.1 0.5 0.35 Get Work done while Traveling 84 2.7 0.7 32 2.3 1.1 0.77 Carpool Partner Matching Program 494 2.3 0.2 192 2.3 0.2 0.90

Encouraged by Program at Work 487 2.2 0.3 197 2.3 0.3 0.91 Preferred Parking at Work 492 2.2 0.1 201 2.2 0.3 0.92 Other 69 3.2 0.8 31 3.1 1.4 0.92

* significantly different at 5 percent level of confidence 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22

19

TABLE 4 Random Parameter Logit Model for Fampools 1

Mode Variable Coefficient P-Value

All Travel Time (min) -0.04 0.00 Toll Cost ($) -0.11 0.01

SOV on the MLs Alternative Specific Coefficient -0.82 0.00 On a Commute Trip -0.92 0.00 Paying a Toll for Current Trips 1.09 0.00

HOV2 on the MLs

Alternative Specific Coefficient 0.36 0.00 On a Work Trip -0.54 0.01 Household Income between $35,000 and $50,000 -0.51 0.00 Education: Post Graduate Degree -0.55 0.00

HOV3+ on the MLs

Alternative Specific Coefficient -0.89 0.00 Household Income between $15,000 and $25,000 2.79 0.00 Traveling in Evening Peak 0.48 0.00

SOV on the GPLs Base Mode

HOV2 on the GPLs Alternative Specific Coefficient 0.27 0.01 Ethnicity: Hispanic -1.74 0.00 Married with No Child(ren) 0.73 0.00

HOV3+ on the GPLs

Alternative Specific Coefficient -1.04 0.00 Education: Post Graduate Degree -1.74 0.01

Standard Deviation Travel Time (min) 0.01 0.00 Toll Cost ($) 0.02 0.01

Summary

Number of Observations 1887 Log Likelihood -2496 Rho Square 0.17 Percent Estimated Correctly 24.4%

Utility Equation Models: 2 V 0.82 0.04 TT 0.11 Toll 0.92 TPCommute 1.09

PayTollV V 0.36 0.04 TT 0.11 Toll 0.54 TPWork 0.51

Income 0.55 Education V 0.89 0.04 TT 0.11 Toll 2.79 Income

0.48 EveningPeak V 0.04 TT 0.11 Toll V 0.27 0.04 TT 0.11 Toll 1.74 Hispanic 0.73

MarriedWithoutChildrenV V 1.04 0.04 TT 0.11 Toll 1.74

Education

20

TABLE 5 Random Parameter Logit Model for Non-fampools 1

Mode Variable Coefficient P-Value All Toll Cost ($) -0.11 0.00

SOV on the MLs Alternative Specific Coefficient -0.76 0.00 Number of Weekly Trips between 6 to 9 1.92 0.00 Number of Weekly Trips between 1 to 2 1.39 0.00

HOV2 on the MLs

Alternative Specific Coefficient 1.98 0.00 On a Work Trip -1.56 0.00 Household Type: Married with No Children -1.17 0.00 Trip Distance -0.05 0.00

HOV3+ on the MLs

Alternative Specific Coefficient -1.22 0.00 Household Income between $35,000 and $50,000 -1.99 0.00 Traveling in Evening Peak 1.41 0.00 On a Work Trip 1.32 0.00

SOV on the GPLs Base Mode

HOV2 on the GPLs Alternative Specific Coefficient -0.47 0.01 Household Type: Married with Children 0.62 0.01 Number of Weekly Trips between 3 to 5 0.49 0.02

HOV3+ on the GPLs

Alternative Specific Coefficient -0.75 0.00 Household Income between $35,000 and $50,000 -0.89 0.09

Standard Deviation Toll Cost ($) 0.03 0.01

Summary

Number of Observations 676 Log Likelihood -1083 Rho Square 0.16 Percent Estimated Correctly 23.4%

Utility Equation Models: 2 V 0.76 0.11 Toll 1.39 TripWeek 1.92 TripWeek V 1.98 0.11 Toll 1.56 TPWork 1.17 HHTypeMnoChild

0.05 TripDist V 1.22 0.11 Toll 1.99 Income 1.41 EveningPeak

1.32 TPWork V 0.11 Toll V 0.47 0.11 Toll 0.62 HHTypeMwithChild 0.49

TripWeek V V 0.75 0.11 Toll 0.89 Income 3

1 2 3 4 5 6 7 8 9

10 11 12 13 14

15 16 17 18

FIGUREFIGUREFIGUREFIGUREFIGUREFIGUREFIGUREFIGURE

FIGURE

ES E 1 Mode CE 2 Mode CE 3 Mode CE 4 Mode CE 5 Mode CE 6 Mode CE 7 Mode C

E 1 Mode C

Choice of NoChoice of FaChoice of NoChoice of FaChoice of NoChoice of FaChoice of No

Choice of No

on-fampoolsampools: Scon-fampoolsampools: Scon-fampoolsampools: Scon-fampools

on-fampools

s Based on Tenario 1. s: Scenario enario 2. s: Scenario enario 3. s: Scenario

s Based on T

Travel Tim

1.

2.

3.

Travel Tim

e Savings.

e Savings.

21

22

1 FIGURE 2 Mode Choice of Fampools: Scenario 1. 2 3

0%

10%

20%

30%

40%

50%

60%

0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0

Percentage

 in Each M

ode

Toll (SOV ML)

SOV‐ML HOV2‐ML HOV3+‐ML All GPLHOV2 Toll = 50% HOV3 Toll = $0

23

1

2 FIGURE 3 Mode Choice of Non-fampools: Scenario 1. 3 4 5

‐10%

0%

10%

20%

30%

40%

50%

60%

0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5

Percentage

 in Each M

ode

Toll (SOV ML)

SOV‐ML HOV2‐ML HOV3+‐ML All GPLHOV2 Toll = 50% HOV3 Toll = $0

24

1

2 FIGURE 4 Mode Choice of Fampools: Scenario 2. 3 4

0%

10%

20%

30%

40%

50%

60%

70%

0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0

Percentage

 in Each M

ode

Toll (HOV2 ML)

SOV‐ML HOV2‐ML HOV3+‐ML All GPL SOV Toll = $5 HOV3 Toll = $0

25

1 2

3 FIGURE 5 Mode Choice of Non-fampools: Scenario 2. 4 5

0%

10%

20%

30%

40%

50%

60%

0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5

Percentage

 in Each M

ode

Toll (HOV2 ML)

SOV‐ML HOV2‐ML HOV3+‐ML All GPLSOV Toll = $5 HOV3 Toll = $0

26

1 2 3

4 FIGURE 6 Mode Choice of Fampools: Scenario 3. 5 6

0%

10%

20%

30%

40%

50%

60%

70%

0.0 0.5 1.0 1.5 2.0 2.5 3.0

Percentage

 in Each M

ode

Toll (HOV3+ ML)

SOV‐ML HOV2‐ML HOV3+‐ML All GPLSOV Toll = $5HOV2 Toll = $2.5

27

1 2

3 FIGURE 7 Mode Choice of Non-fampools: Scenario 3. 4 5 6

0%

10%

20%

30%

40%

50%

60%

0 0.5 1 1.5 2 2.5 3

Percentage

 in Each M

ode

Toll (HOV3+ ML)

SOV‐ML HOV2‐ML HOV3+‐ML All GPL SOV Toll = $5HOV2 Toll = $2.5