Attitude-Behavior Changes in a Before-After Mode Choice Situation

Mary Lynn Tischer, Urban Planning Division, Federal Highway Administration, U.S. Department of Transportation
Carl G. Shea, Urban Planning Division, Federal Highway Administration, U.S. Department of Transportation
ABSTRACT - A non-equivalent control group design was utilized in the evaluation of the effects of a high-occupancy reserved lane on the mode perceptions and mode choice behaviors of downtown commuters. Significant behavior and perception change was observed in both the treatment and control groups. The standardized change score analysis suggested that the transportation modification had some effect on perceptions although no effect on behavior was observed.
[ to cite ]:
Mary Lynn Tischer and Carl G. Shea (1978) ,"Attitude-Behavior Changes in a Before-After Mode Choice Situation", in NA - Advances in Consumer Research Volume 05, eds. Kent Hunt, Ann Abor, MI : Association for Consumer Research, Pages: 456-464.

Advances in Consumer Research Volume 5, 1978      Pages 456-464

ATTITUDE-BEHAVIOR CHANGES IN A BEFORE-AFTER MODE CHOICE SITUATION

Mary Lynn Tischer, Urban Planning Division, Federal Highway Administration, U.S. Department of Transportation

Carl G. Shea, Urban Planning Division, Federal Highway Administration, U.S. Department of Transportation

ABSTRACT -

A non-equivalent control group design was utilized in the evaluation of the effects of a high-occupancy reserved lane on the mode perceptions and mode choice behaviors of downtown commuters. Significant behavior and perception change was observed in both the treatment and control groups. The standardized change score analysis suggested that the transportation modification had some effect on perceptions although no effect on behavior was observed.

INTRODUCTION

A major objective of current transportation policy is the attainment of a more balanced transportation system through greater use of ridesharing modes. Programmatic efforts to facilitate the use of mass transit and carpooling must be based, however, on knowledge of the causal components of individual decisionmaking and an awareness of the level of transportation modifications necessary to affect behavior. Interviewing the same individuals both before and after transportation modifications have been effected can provide the necessary information regarding the relationships among system characteristics, individual perceptions, preferences and social situations and transportation behavior.

Panel data have previously been utilized in transportation research to evaluate the impacts of major changes in the transportation-related environment. Watson and Holland (1977) interviewed a Singapore sample twice to obtain behavioral information on the effects of an urban auto restraint policy. Skinner (1975) used repeated measurements of travel behavior to assess the effects of increased gasoline prices during the 1975-76 energy crisis. Longitudinal data have also provided information on the reliability of models in forecasting response to new modes (Parody: 1977; McFadden, et al.; 1977) and on the relationship between attitudinal structures and behavior in response to transportation innovations (Goss and Shuldiner, 1977; Gilbert et al., 1975). Little information is available, however: on the effects of transportation system management actions (TSM) on the perceptions, preferences and behaviors of individuals.

The relationship of the system to the individual and his social environment is represented in Figure 1. While the figure does not detail a theory of behavior, it does provide a framework for the analysis of potential effects of alterations in the transportation system. It is assumed that changes in the system affect the service provided by the modes which, when communicated to individuals, alters their perceptions. Similarly, it is suggested that changes in personality variables, for example an increase in the value of a clean environment, will change the perceptions of the instrumentation necessary to reach the goal and lead to the development of new behaviors.

The objective of this paper is to present preliminary findings on the effects of a reserved lane for buses and carpools implemented on the Santa Monica, California, freeway on the mode perceptions and mode choice behaviors for downtown Los Angeles commuters. Changes in individuals residing in this corridor are compared to those observed for a sample of commuters residing along comparable freeways which also feed into the central business district.

THE DESIGN AND METHODOLOGY

The introduction of a control group is a necessary feature when one is interested in the effects of a transportation modification; otherwise, changes in behavior could also be attributed to extraneous factors operating in the environment. The effects of the modification, or "treatment," are distinguished from general system activities through the comparison of two similar groups, only one of which was exposed to the transportation system changes. In the analysis of social phenomena where similarity of individuals cannot be obtained through random assignment into treatment conditions, the experimental design can be approximated with a "non-equivalent control group design," (Campbell and Stanley, 1963). The similarity of control and treatment groups was obtained, in this case, by sampling within census tracts matched by socio-demographic and travel-related characteristics, (Hartgen and Howe, 1976).

FIGURE 1

INFLUENCES UPON MODE CHOICE BEHAVIOR

There has been considerable debate as to the statistical technique appropriate for analyzing changes in the nonequivalent control group design. Selection of treatment group based on group differences can result in under-identification of the change model (Linn and Werts, 1977). However, if it is reasonable to assume that any group differences which might confound the treatment with the dependent variable are stationary over time, change score analysis can be applied (Kenny, 1975).

To perform the analysis of treatment effects, the variables are first standardized to correct for possible changes in variance. The time 1 treatment and control group scores for the variable of interest are combined and standardized; the time 2 scores are similarly adjusted. Differences between treatment and control groups are then evaluated with a correlational technique in which the treatment variable is dummy coded and the standardized pre-and-post test scores are correlated with the treatment variable. A significant difference between the pre-test-treatment correlation and the post test-treatment correlation is indicative of a treatment effect. Significance is determined with the t test for dependent correlations (McNemar, 1969). The t test for related samples is used to evaluate the significance of changes within the treatment and control groups (Bruning and Kintz, 1965).

THE SAMPLE AND DATA

Census tracts within two miles of the Santa Monica Freeway in Los Angeles, California with a high incidence of Central Business District (CBD) workers were identified. These census tracts were matched by socio-demographic characteristics with tracts surrounding alternative freeways which also fed into downtown Los Angeles. Within the census tracts, households were randomly located through reverse telephone directories. Only individuals who worked within the business district were eligible for the home interview. Where a household contained multiple workers, the individual taking the lesser used mode was selected. A supplemental sample of carpoolers observed entering the CBD and bus users located at bus stops was included to ensure representation of these modal groups.

Interviews with 996 individuals were obtained between September 1, 1975 and March 5, 1976; 512 lived in the Santa Monica corridor and 484 were in the control areas. From January - May 1977, attempts were made to recon-tact the 996 individuals. Interviews were completed for 502 of the Phase I respondents.

Individuals were asked how frequently each month they drove to work alone, drove another person, rode with another and took a bus. Perceptions for each of 3 modes (sharing a ride, driving alone, and taking a bus) were collected for 19 attributes using a seven point semantic differential format. The perceptions were coded such that seven represents the most positive aspect of each attribute. The attribute labels are presented in Table 1.

TABLE 1

ATTRIBUTE LABELS

The same questions were asked at the two interviews. However, at time 2, individuals were asked to recollect their behavior and perceptions from the period of March through August 1976.

THE TREATMENT

From March 15 to August 8, 1976, the median lanes on the Santa Monica Freeway were reserved for the exclusive use of buses and automobiles carrying three or more persons. No barriers separated the reserved lanes from general traffic, thereby permitting entry or exit throughout the 12.5 miles of the freeway. Preferential freeway access for vehicles carrying two or more persons had been implemented on 12 of 30 ramps prior to the Phase I interviewing although metering rates for the non-bypass ramp lanes were drastically altered at the start of the diamond lane project. Four new express bus routes were established and four existing bus routes took advantage of the lane. Three new routes were developed to serve the three newly-opened park-and-ride lots and a cross-town feeder service was initiated. Prior to the project, 18 express bus trips were offered every weekday morning, while on the first weekday in August, 74 express buses were in operation. Bus headways were also improved, a policy maximum of 15 minute headways was established during peak hours. Bus prices were increased resulting in an average rise of 204 per trip (Billheimer, et al., 1977). On August 11, 1976, the preferential lanes on the Santa Monica Freeway were terminated by court order; improvements in the bus system remained in effect. The freeways in the control areas were not modified during this period, al= though bus service was improved throughout the Los Angeles region and bus fares were increased in the Southern California Rapid Transit District.

SAMPLE MORTALITY, CONTROL AND TREATMENT GROUP EQUIVALENCE

Only 50% of the Phase I respondents were reinterviewed at time 2. Thirteen percent of the initial sample refused a second interview; 37% moved and could not be located. Differences between the initial sample and those retained for the second phase suggest some biases exist in the final sample. The individuals who responded to both questionnaires were dissimilar from the original sample in that they were more likely to be older, to have a higher income, to live in a single-family dwelling, and less likely to be black or to have children under six years of age. Additionally, fewer individuals were retained in the control sample (43%) than in the treatment group (57%)- The effects of the sample mortality cannot be determined although it is likely that the individuals who were lost from the sample would be more responsive to ridesharing incentives than the older and wealthier travelers.

The individuals who were retained for analysis in the treatment group differed in several respects from the control group. The control group contained a significantly larger proportion of whites, Hispanics, and sales workers than did the treatment group. Individuals from large households and dwellers in single-family residences were also overrepresented in the control group. The treatment group contained a larger proportion of blacks and clerical workers. The number and age of children, marital status and the number of licenses and vehicles in the household were not significantly different.

RESULTS: CHANGES OVER TIME

Behavior

Individuals were assigned to a mode use category based on use of a mode for 60% or more of the work trips. The three categories of mode use were single-occupant auto (SOA), carpool (CP), and bus (BUS). Where an individual used two modes at about the same rate, he is defined as a dual mode user an individual who uses three modes for a significant amount of traveling is designated as a mixed mode user. The proportion of trips made in each mode for the user groups at time 1 is presented in Tables 2 and 3- Over 90% of the treatment and control samples used one mode for most of their trips to work. The remainder divided their trips almost equally between two modes.

At time 1, 38% of the treatment group and 58% of the control group used one mode for 10% of their work trips. Almost a majority of the total sample used at least one other way of traveling during a monthly period. At time 2, a larger proportion of the sample had clearly defined an exclusive mode to work; 52% of the treatment group and 61% of the control group used the same mode every day. A slight increase was also found in the number of people who alternated between two or more modes. In the treatment group, 7% initially used dual modes whereas at time 2, 10% combined modes; in the control group in the percentages are 4 and 6, respectively.

TABLE 2

PROPORTION OF TRIPS MADE BY MODE USER GROUPS IN THE TREATMENT AREA AT TIME ONE

TABLE 3

PROPORTION OF TRIPS MADE BY MODE USER GROUPS IN THE CONTROL AREAS AT TIME ONE

Changes in the choice of mode to work can be observed in Tables 4 and 5. In the treatment group, 72% of the travelers used the same mode for a majority of their trips at both time periods. Most of the change in the treatment group is due to mode users combining their time one mode with another at time two, or the dual mode users at time one selecting one of those two modes for greater use at time two. Only 13% switched to a mode rarely used at time one. Switching from single-occupant autos was less pronounced than defections from carpools and buses; 19% of those who drove alone at time one switched to carpools, buses or combined modes while 36% of the carpoolers and 23% of the bus users altered their mode choice behaviors. In the control group, 80% of the respondents used the same mode at both points in time and 13% changed modes radically. Eleven percent of the time one single-occupant auto drivers altered their mode choice behavior compared with 31% of the carpoolers and 17% of the bus riders.

The small number of dual and mixed mode users necessitated their assignment to a modal group on the basis of the mode used most often in traveling to work. The average mode frequency was then analyzed to determine the significance of changes for time 1 mode-use groups (Tables 6 and 7). In the treatment group, single-occupant auto drivers significantly reduced their use of the mode at time 2 while increasing their use of both buses and carpools. Bus and carpool users significantly decreased the average frequency of use for their respective modes. Carpoolers more frequently drove alone and took buses at time 2; bus users increased their use of the carpool (p = .056). The same pattern was observed in the control group; mode use groups from time 1 decreased their average use of their respective modes at time 2. Single-occupant auto, bus and carpool users increased their use of the two alternative modes.

Mode choice behavior is based on trial and evaluation of alternative modes for much of the sample; at time 1, 38% of the treatment group and 58% of the control group used a second mode at least once during a monthly period. The single-occupant auto drivers' experience with ridesharing modes was surprisingly high; 75% of the treatment and 735 of the control auto users had taken a bus or a carpool at least once a month. While mode choice behavior changed over time in both treatment and control groups, only 14% of the time 1 single-occupant auto users took a ridesharing mode for more than half of their trips at time 2. This fact combined with the significant level of defections from carpools and buses reaffirms what is already known: ridesharing is not a particularly satisfactory alternative to driving alone for the vast majority of the traveling public.

Contrary to expectations, the mode distribution at time 2 was approximately the same as at time 1. No significant net decrease in driving alone was observed in either the treatment or control groups as a whole although a significant reduction in the frequency of driving alone was obtained for the time 1 single-occupant auto users. This suggests that some travelers who drive alone can be convinced to rideshare. However, the significant defections from the time 1 ridesharing groups indicate that retaining the carpoolers and bus users to produce a net increase in ridesharing is problematic.

Perceptions

The modal attributes which changed between measurement points are presented in Table 8. The single-occupant automobile attributes which were perceived to change, all changed in the negative direction for both the treatment group and the control group. For the treatment group as a whole, bus travel time, crowding and waiting in traffic improved over time. Carpool attributes of crowding, package space and ease of use changed negatively. Carpool travel time and waiting in traffic significantly improved and were subsequently more positive than those attributes for the alternative modes. Bus safety and cost were perceived more negatively at time 2 in the control group, although waiting in traffic significantly improved over time for the carpool mode.

The changes in perceptions for the total treatment and control groups indicate that some attributes had changed for all mode use groups. In the treatment group, changes are those expected: the attributes of driving alone were worse at time 2, while bus and car-pool attribute perceptions of travel time and waiting in traffic improved over time. In the control group, perceptions of the bus and single-occupant automobile were lower at time 2 than at time 1. When the sample was partitioned into subsets in terms of time 1 mode use, some changes in variables which were significant for the total sample were found to be insignificant within any one group (Tables 9 and 10). Furthermore, perceptions of other attributes were found to have significantly altered over time for certain user groups. Within the treatment group, only single-occupant automobile users perceived travel time in that mode to be significantly worse at time 2. No other significant differences were observed in single-occupant auto perceptions of the mode. Bus travel time, bus crowding, bus exposure to weather, and bus waiting in traffic significantly improved although they were still not perceived as strongly positive (# 4.1). The wait in traffic for the carpool was perceived more positively at time 2 than at time one. However, the carpool attributes of package space and crowding were perceived to have worsened by single-occupant automobile users in the treatment group. Time one carpoolers believed that st time 2, carpool travel time has significantly improved. Bus travel time and waiting in traffic improved but were still slightly negative (< 3.8) while extra time for the bus became marginally positive. Driving alone was considered to be more costly at time 2 than at time 1.

TABLE 4

CHANGES IN MODE CHOICE BY TIME ONE USER, GROUPS IN THE TREATMENT AREA, TIME ONE MODE CHOICE

TABLE 5

CHANGES IN MODE CHOICE BY TIME ONE USER, GROUPS IN THE CONTROL AREAS, TIME ONE MODE CHOICE

For treatment group bus users, the cost of bus use improved while the reliability of the bus decreased over time. Bus users believed the carpool was less easy to use at time 2 but the wait in traffic and parking cost of the mode had improved. No significant changes were observed in the bus users' perceptions of driving alone.

Within the control group, single-occupant automobile users believed the ease of use, the ease to the destination, personal safety and waiting in traffic had significantly worsened for driving alone. Attributes for the carpool and bus modes also significantly decreased over time. The carpoolers believed that the carpool mode had increased in reliability; their perceptions of bus safety and cost were more negative at time 2. Control group bus riders believed the bus was less comfortable at time 2 and more costly than at time 1 and that carpool convenience had decreased between the pre-test and post-test.

TABLE 6

CHANGE IN PROPORTIONAL USE OF MODES TIME 1-TIME 2: TREATMENT GROUP

TABLE 7

CHANGE IN PROPORTIONAL USE OF MODES TIME 1-TIME 2: CONTROL GROUP

TABLE 8

SIGNIFICANT ATTRIBUTE CHANGES FROM TIME 1 TO TIME 2 (ALL MODE USERS)

TABLE 9

PRE-TEST TO POST-TEST PERCEPTION CHANGES: TREATMENT GROUP

TABLE 10

PRE-TEST TO POST-TEST PERCEPTION CHANGES: CONTROL GROUP

For both the treatment and control groups, changes in single-occupant automobile and bus group perceptions corresponded with behavior. The decrease in use of the carpool by time 1 carpoolers was not indicated by the aggregate perceptions.

Changes in Attribute Importance

Perceptions of mode attributes may change and thereby induce behavior change; it is also possible that the salience of attributes changes over time and affects behavior. The value that individuals use in determining mode choice may be altered by changes in circumstances or reference groups, because basic motives have changed, because attribute importance is rationalized to correspond with behavior or because an attribute which was "taken for granted" or outside the range of possibility, is altered or becomes available. To evaluate changes in the importance of an attribute in determining behavior, the unstandardized regression coefficients from simple regressions were analyzed (Table 11). Simple, rather than multiple, regressions were used because of known multicollinearity in the data and entry into the multiple regression depends on residual unexplained variance rather than total unexplained variance. A t test was applied to test for significant differences over time.

Within the treatment group, the attributes of arriving on time, vehicle safety and extra time became more important over time in the determination to take a car-pool. The increased salience of safety and extra time could be due to the high accident rate on the Santa Monica Freeway and the requirement of three vehicle occupants to use the reserved lane. The increased importance of arriving on time in the decision to use either a single-occupant auto or a carpool may also be due to negative impacts of the reserved lane. Where arriving on time at time 1 had been a matter of personal responsibility in both automotive modes, at time 2 the unpredictable nature of the congestion on the freeway determined whether one arrived on time.

Ease of use in driving alone changed significantly in importance. The media representation of carpools and buses may have served as a negative referent by making that aspect of driving alone even more advantageous or it may be due to unexpected difficulties in use of the mode at time 2. Safety factors were more influential in the determination to take a bus at time 2 than at time 1. Again, the increase in traffic accidents may have highlighted this attribute. Alterations in the importance of parking cost also occurred; the fare increase may have focused individual attention on the lack of a parking cost as a way of rationalizing the cost savings usually attributable to the bus mode. Extra time was found to become more important while crowding decreased in importance. These changes may have resulted from the improvements in bus service which contrasted with what had been available.

In the control group, vehicle safety became more important and personal safety became less important in the determination to drive alone. The reported accident rates on the Santa Monica Freeway may have highlighted an attribute which had little meaning at time 1 and replaced the more traditional perception of the auto as a guarantor of personal safety. Changes in the importance of waiting in traffic, relaxing and exposure to weather in the decision to take a bus may have resulted from the improved bus service on the freeways. The causal structure with respect to the carpool mode changed significantly in the control group. Eight of the 19 attributes became more important over time.

Media coverage of events which occurred between measurements may have aroused interest in the mode and provided information where attitudes were not strongly held. The interview at time 1 may also have stimulated reassessment of the shared ride. Furthermore, trial of the mode would inform individuals about the characteristics of the mode, and if discrepant with expectations would generate reevaluation of the carpool and the value of its attributes.

Importance of the attributes can change through alterations in the external environment and through changes in individual values, perceptions or behaviors. Individuals may be indifferent to certain attributes at time 1 because they were associated with all 3 modes or were considered necessary or neutral qualities of a particular mode. Where the modes were later differentiated with respect to the attribute or where the attribute was found to significantly change, the individual may reassess its importance. Comparisons between the modes engendered through communications or personal experiences would serve to emphasize the attribute levels associated with usage of a particular mode. The changed reference structure may induce reevaluation of the importance of the characteristics of the modes. Furthermore, changing the importance of attributes may also represent dissonance reduction where behavior and perceptions were not congruent at time 1.

The Effect of the Treatment on Perceptions and Behavior

In order to assess the effects of the reserved lane on mode user groups, subjects in both the treatment and control groups were divided into three categories, on the basis of which mode-to-work (single-occupant auto, carpool, or bus) was most frequently used prior to the modifications in the Santa Monica Corridor. Changes in beliefs and behavior were then analyzed separately for the three groups.

A significant difference between the beliefs in the treatment group vis-a-vis the control group was observed in 15 of the 157 comparisons. These results are summarized in Table 12.

For ten attributes, differences between the control and treatment groups involve significant changes over time in the perceptions of at least one of the groups. The remaining five attributes, while not significantly different from time 1 to time 2 within the groups, display a shift in opposing directions for the treatment and control groups. The difference in the pattern of change suggests some treatment effect. The treatment would be expected to affect bus and carpool perceptions by making them more positive and to reduce the positive perceptions of driving alone. Ten of the 15 perceptual changes are in the expected direction for the treatment group.

For the time 1 single-occupant auto users, the treatment produced a significant change in the perception of the wait in traffic required for carpoolers. The attribute changed from slightly negative to positive for the treatment group while remaining negative in the control group. For the treatment group, carpool waiting in traffic was more positive at time 2 than perceptions of the wait in traffic necessitated by driving alone or riding in a bus. Most of the remaining perceptions of the attributes of the carpool were less positive than the comparable beliefs about single-occupant autos. No other modifications of carpooling beliefs were significantly different in the treatment group in comparison with the control group.

The single-occupant automobile treatment group perceived that the travel time of the mode had significantly increased from time 1 to time 2, while the control group did not alter their perceptions of this attribute. Both the perceptions of the carpool wait in traffic and the increased travel time incurred in driving alone can be attributed to the changes on the Santa Monica Freeway. However, a significant difference between the treatment and control groups with respect to single-occupant automobile personal safety and ease to destination was unexpected. There was no significant change in the treatment group but the control group perceptions of personal safety and ease to the destination decreased significantly. The actual cause of the change in perceptions for the control group is unclear and illustrates one problem with the non-equivalent control group design. Any event which affected those living in the control areas but not the treatment zone could result in a confounding of the treatment effect. The treatment group did reduce its frequency of single auto use significantly; however, the control group also reduced its single-occupant auto use by about the same amount. Thus, it is unclear whether the drop in the treatment group was caused by the policy change and the decrease in the control group by some other factor; or whether both decreases are a product of some other cause.

TABLE 11

CHANGES IN IMPORTANCE OF ATTRIBUTES IN EXPLAINING BEHAVIOR

TABLE 12

TREATMENT EFFECTS UPON PERCEPTIONS OF ATTRIBUTES

For single-occupant automobile users, three bus beliefs were found to differ significantly between the treatment and control groups. Bus crowding, waiting for the bus and exposure to weather were significantly more positive at time 2 than at time 1 for the treatment group. Transportation system modifications which decreased bus headways and increased frequency of bus service in the treatment zones would, therefore, seem to have had the desired effect upon this group. However, changes in the behavior of the treatment group were not significantly different from changes in the control group. Bus use increased for time 1 single-occupant auto users in both samples.

There was a significant treatment effect on the perceptions of the carpool users; the direction of change was, however, contrary to expectation. The treatment group showed a decrease in the ease of use of the car-pool mode while the control group found the mode to be improving in its ease of use. This may be a reflection of the reported difficulty in crossing traffic lanes to get to the diamond lane or confusion over the new policy. Perception of crowding in carpools also was less favorable in the treatment group at time 2 and may be due to the need to have at least 3 people in the car-pool in order to use the reserved lane. Travel time during rush hour in carpools moved in the desired direction; the treatment group exhibited an increasingly positive perception for this attribute.

The only significant treatment effect on the carpoolers' perceptions of driving alone was a relative increase in the time spent waiting in traffic for the treatment group, while the control group perceived a decrease in waiting time for the single-occupant auto. This is probably attributable to the Santa Monica ramp metering system and the use of ramp bypasses by the carpoolers.

The carpoolers in the treatment group also perceived a significant improvement in bus travel time while the control group perceptions were slightly worse at time 2 with respect to the attribute of the bus. The mean bus use of the carpoolers increased equally in each group.

Bus users in the treatment group considered the carpool to be less easy to use at time 2 than at time 1, and, thereby differed significantly from the control group. Again, this may be attributable to the observed difficulties in accessing the reserved lane. A treatment effect was also observed in the extra time required in driving alone. The treatment group changed positively with respect to this attribute while the control group changed to a less positive perception of single-occupant automobile extra time. Some confusion may have existed with respect to the definition of the attribute; individuals may have changed their parking locations or may be including the extra time required after leaving their car at a park-and-ride lot.

A treatment effect was observed for bus parking cost among the bus users. The more positive belief observed at the post-test for the treatment group is perhaps a reflection of reduced costs for those using a park-and-ride lot. Finally, no significant effect was observed in the behavior of bus users.

No significant treatment effect upon behavior was observed. While significant changes were found in the treatment zone in use of all three modes, behavior change was occurring at the same level in the control group. Several explanations can be offered. First, it is likely that exposure to media and to acquaintances who used the Santa Monica Freeway reserved lane heightened interest in the ridesharing modes and the transportation system in general for the control group. Individuals who were potential ridesharers may have become more aware of their alternatives and availed themselves of the opportunity to try a carpool or bus. Secondly, the definition of treatment as residence in the Santa Monica corridor may be too broad. It is likely that the reserved lane was perceived as an incentive only for those who could use the freeway for the entire 12.5 miles. Third, the reserved lane on the Santa Monica Freeway was controversial, and consistent threats to its continuation undoubtedly mitigated against the stabilization of attitudes and behaviors.

The changes that were observed in the panel of downtown commuters thus appear to be attributable to a dynamic ridesharing segment of the population for whom increased awareness and availability of the ridesharing mode facilitated the use of the alternatives to driving alone to work.

Conclusions

A significant change in mode choice behavior was observed in both the treatment and control groups over time. In each group, however, only 13% changed to modes with which they had had little experience at time one. In the treatment group, single-occupant auto, carpool and bus users at time 1 decreased their use of their respective modes at time 2. Single-occupant auto users significantly increased their use of the bus and car-pool while carpoolers increased their use of the bus and single-occupant auto. Bus users significantly increased their use of the carpool. The same pattern is found in the control group; time 1 mode groups decreased their use of those modes at time 2. Both carpoolers and single-occupant auto users increased their use of the bus at time 2; carpoolers and bus users also increased the use of the single-occupant auto. Both time 1 bus and single-occupant auto users increased their carpool usage.

It is likely that in the process of choosing a mode, individuals try at least one of the alternatives. A considerable number of individuals used two modes about equally in traveling to work. However, those who alternated between two modes at time one had generally selected between them by time two. Furthermore, a majority of the sample used a second way of traveling to work at least once at time one. It appears that most of the Los Angeles downtown commuters are not captive to a mode and that most have experienced at least one of the ridesharing alternatives. The significant defections from the bus and carpool modes suggest a dynamic ridesharing population. Continual promotional and information campaigns are necessary to continue to attract commuters and to thereby replace those who have switched.

Contrary to expectations, the treatment group did not display a large increase in the use of the carpooling mode. Travel counts on the Santa Monica Freeway revealed a 65% increase in carpools and a 300% increase in bus use (Billheimer, et al., 1977). This travel data is derived with respect to all the riders of the freeway, while the survey sample included only downtown workers. It is possible that downtown commuters were already ridesharing in greater proportions than the non-CBD workers or that many who had relocated had significantly altered their travel behavior. It is also likely that the survey sample was not representative of the total population. In the random survey of households the number of downtown workers and their mode to work were obtained; 56% of the total household workers drove alone to work, 20% used a carpool and 24% were bus users. This mode distribution is approximately equal to the mode distribution of the time 1 treatment group that was retained for analysis. It is possible, however, that the background characteristics of the mode users in the sample are significantly different from the population and~ therefore, explain the lack of change observed.

The reserved lane on the Santa Monica Freeway appears to have had relatively little effect on the behavior of the downtown commuter sample. Behaviors in the control group were not significantly different than those in the treatment group. However, one of the problems in the analysis is the inability to separate the increased use of the bus associated with express buses using a reserved lane from the mere availability of express buses. Increased bus service in all areas combined with additional express routes on control freeways confounded the analysis. The only conclusion which is indicated is that the reserved lane implemented on the Santa Monica Freeway for five months did not increase carpooling behavior among the downtown commuters. Whether this lack of carpooling behavior can be attributed to the controversial nature of the reserved lane and the continuous discussion of its termination, cannot be determined with the data available.

The importance of modal attributes, as measured by their influence upon behavior, changed over time in both the treatment and the control groups. System modifications, communication, and personal experience with the modes are presumed to have caused the reassessment of attribute salience. Reactions to alterations in service as well as the enhanced comparative evaluation of the attribute levels between modes contributed to the change in the importance of attributes in the mode choice decision.

Perceptions of the modes were also altered in both the treatment and control groups and in general corresponded with the behavior patterns. Significant differences in perceptions were observed in the treatment and control groups. Single-occupant automobile treatment perceptions were found to be more positive than single-occupant automobile control group perceptions of the bus and carpool modes and although the attributes of carpool and bus remained generally less positive than those of the single-occupant auto mode, the magnitude of difference was decreased. It is possible that had all the incentives to multi-passenger vehicle use remained in effect, further behavior changes in the direction of a more balanced mode split might have been attained.

The relationship between behaviors and perceptions of modal attributes can only be determined with individual-level analysis. Further research is required regarding the effects of perceptual consistency and behavior change, changed availability and mode choice, communication flows and mode trial, as well as the determination of the modal attributes which appear to influence behavior.

REFERENCES

John W. Billheimer, Robert Bullemer and Carolyn Fratessa, (SYSTAN, Inc.) "The Santa Monica Diamond Lanes: An Evaluation," preliminary report to the U.S. Department of Transportation. Transportation Systems Center, Contract No. DOT-TSC-1084, (Washington, D.C.: 1977).

James L. Bruning and B. L. Kintz, Computational Handbook of Statistics, (Glenview, IL: Scott, Foresman, 1968) p. 13-15.

Donald T. Campbell and Julian C. Stanley, Experimental and Quasi-Experimental Designs for Research, (Chicago: Rand McNally, 1963).

Gorman Gilbert and James F. Forester, "Public Response to the Chapel Hill Bus System," Department of City and Regional Planning, University of Carolina (Chapel Hill: 1975 ).

William Goss and Paul Shuldiner, "University of Massachusetts Bus Research and Demonstration Grant" awarded 1972 Mass-DMG-6 by UMTA Research and Development, Final Report expected 1977.

David T. Hartgen and Stephen M. Howe, "Irondequoit-Wayne Expressway: Before-After Study, Design Report," Preliminary Research Report 101 (Albany, New York: Planning Research Unit, New York Department of Transportation, June 1976).

David A. Kenny, "A Quasi-experimental Approach to Assessing Treatment Effects in the Non-Equivalent Control Group Design," Psychological Bulletin, 82:3, (May 1975) P. 345-362.

Robert L. Linn and Charles E. Wefts, "Analysis Implications of the Choice of A Structural Model in the Nonequivalent Control Group Design," Psychological Bulletin, 84:2 (March 1977) P. 229-234.

Daniel McFadden, Antti Talvitie, Stephen Cosslett, Ibrahim Hasan, Michael Johnson, Fred A. Reid, and Kenneth Train, "Demand Model Estimation and Validation," Institute of Transportation Studies, University of California, Berkeley, California (June 1977).

Quinn McNemar, Psychological Statistics, Fourth Edition, (New York: John Wiley and Sons, 1969), p. 158.

Thomas Parody, "An Analysis of Disaggregate Mode Choice Models in Predictions" paper presented at annual meeting of Transportation Research Board, (1977).

Louise Skinner, "The Effect of Energy Restraints on Travel Patterns," Urban Planning Division, Federal Highway Administration (Washington, D.C.: 1975).

Peter L. Watson and Edward P. Holland, "Road Pricing in Singapore: Impacts of the Area License Scheme," draft report, Urban Projects Department, World Bank (1977).

----------------------------------------