Scholarly article on topic 'Automatic Preferences in the Choice of Transport Mode'

Automatic Preferences in the Choice of Transport Mode Academic research paper on "Economics and business"

CC BY-NC-ND
0
0
Share paper
OECD Field of science
Keywords
{"Automatic Preferences" / "Choice Model" / Psychology / "Revealed Preferences" / "Stated Preferences"}

Abstract of research paper on Economics and business, author of scientific article — J.E. Córdoba

Abstract There is a growing consensus currently among transportation researchers on the importance of psychological aspects in the choice of transport mode, but there has been difficulty in determining the strength of this aspect on the mode decision. This paper provides a new basic framework for automatic preferences (AP) data and automatic choice (AC) methods when applied subconscious choices on modes of transport and the associations of positive qualities and modal choices. The framework also includes the observable attributes of travel time, waiting time, number of transshipment, costs, and socioeconomic characteristics of the users, all of which may have an influence in the mode choice. The most relevant results obtained were the automatic association between the car and positive constructs in the car users, and a large difficulty in obtaining an association that is a positive construct to the bus. The bus users also can associate more easily to this mode with positive constructs than the car users.

Academic research paper on topic "Automatic Preferences in the Choice of Transport Mode"

Available online at www.sciencedirect.com

ScienceDirect

Procedia - Social and Behavioral Sciences 160 (2014) 381 - 389

11th Transport engineering conference (CIT 2014)

Automatic preferences in the choice of transport mode

J.E. Cordoba3*

"Department of Civil Engineering, National University of Colombia, Medellin, Colombia

Abstract

There is a growing consensus currently among transportation researchers on the importance of psychological aspects in the choice of transport mode, but there has been difficulty in determining the strength of this aspect on the mode decision. This paper provides a new basic framework for automatic preferences (AP) data and automatic choice (AC) methods when applied subconscious choices on modes of transport and the associations of positive qualities and modal choices. The framework also includes the observable attributes of travel time, waiting time, number of transshipment, costs, and socioeconomic characteristics of the users, all of which may have an influence in the mode choice. The most relevant results obtained were the automatic association between the car and positive constructs in the car users, and a large difficulty in obtaining an association that is a positive construct to the bus. The bus users also can associate more easily to this mode with positive constructs than the car users.

© 2014 The Authors. Published by Elsevier Ltd.This is an open access article under the CC BY-NC-ND license

(http://creativecommons.Org/licenses/by-nc-nd/3.0/).

Peer-review under responsibility of CIT 2014.

Keywords: Automatic Preferences; Choice Model; Psychology; Revealed Preferences; Stated Preferences

1. Introduction

The objective of this paper on automatic preferences (AP) data and automatic choice methods, analysis and applications is to demonstrate the benefits of developing a formal structure from which one can estimate a discrete choice model that can be applied quickly, accurately, and economically. Data collection can be minimalized but only when it is tested to be sufficient. The predictive results of discrete choice models applied to transport can be

* Corresponding author. Tel.: 574-425-5189; fax: 574-425-5150. E-mail address: jecordob@unal.edu.co

1877-0428 © 2014 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license

(http://creativecommons.org/licenses/by-nc-nd/3.0/).

Peer-review under responsibility of CIT 2014.

doi:10.1016/j.sbspro.2014.12.150

improved a can even reduce the complexity of the models, and can lower high costs in primary data collection. However, this can only result with the understanding and consideration of the psychological processes that occur in humans and have an influencing on decision making. Preferences to everyday activities tend to be automatic. Previously transportation data collection techniques utilized the revealed preferences (RP) and stated preference (SP) techniques. This research proposes a new technique for data collection and a new method for modal choice, referred to as automatic preferences (AP). Automatic preferences are the automatic association that exists in the subconscious of people between an object and their positive constructs, and that generates pro-choice behavior of that object. Automatic preferences (AP) has an advantage because the process is at the level of the subconscious, participants become less vulnerable to possible manipulation of the information revealed or stated consciously. This advantage is not present in the techniques (RP) and (SP) respectively. The strength of the automatic association that exists in a user's memory between transport modes and their positive qualities, and the relationship to modal choice, is evaluated. It should be easier and faster to determine the association between the positive constructs and modes because the two are subconsciously linked. Most people find it easier to make associations with positive statements and it is likely that car users will state a preference for positive phrases or constructs for the car mode, while bus users will state a preference for positive phrases or constructs for the bus mode.

2. Theoretical framework

To predict the behavior of users of a transport system, techniques traditionally have been based on observation of actual behavior of individuals, known as revealed preference (RP) (Ortúzar, 2000). However, RP has problems if used in the analysis of options that are not present or when using a variable of perception as comfort, reliability, and safety. There also are stated preference (SP) techniques utilized to obtain answers from the individuals in hypothetical situations. This technique does not have the disadvantage of (RP), but it is not certain that individuals really act as hypothetically stated. It is important to note that a model is the simplified representation of reality with a mathematical framework because it takes the most representative variables of a system and evaluates their impacts on the system by testing several alternatives. The revealed preference (RP) and stated preference (SP) techniques were applied to the collected data to estimate the model. In this section, revealed preference (RP), stated preference (SP) and the new automatic preference (AP) are presented.

2.1. Revealed preference (RP)

RP data show what is happening at the same time in different places, which is the critical characteristic of the origin-destination (O-D) surveys (Ortúzar, 2000). RP data typically are describing the world as it is now (current market equilibrium), possessing inherent relationships between attributes; having only existing alternatives as observables, embodying market and personal constraints on the decision maker, having high reliability and face validity, and yielding one observation per respondent at each observation point (Louviere, Hensher & Swait, 2000). According to Louviere, Hensher and Swait (2000) RP data contain information about current market equilibrium for the behavior of interest, and can be used to forecast short-term departures from current equilibrium. In contrast, SP data are especially rich in attribute tradeoff information, but may be affected by the degree of "contextual realism" that is established for respondents. Therefore SP data are more useful for forecasting changes in behavior and are presented here.

2.2. Stated preference (SP)

According to Ortúzar (2000), stated preference techniques are a set of methodologies that are based on judgments (data) declared by individuals about how they would act against different hypothetical scenarios that are presented to them. These techniques have origins in mathematical psychology (Luce & Tukey, 1964) and experimental designs are used to construct hypothetical alternatives. SP data typically are describing hypothetical or virtual decision contexts (flexibility), controlling relationships between attributes (which permits mapping of utility functions with technologies different from existing ones), may be including existing, proposed, or generic choice alternatives that cannot easily (in some cases, cannot at all) represent changes in market and personal constrains effectively, tending

to be reliable when respondents understand, committed to, and can respond to tasks, and yielding multiple observations per respondent at each observation point (usually), (Louviere, Hensher & Swait, 2000). A key role for SP data in combined SP-RP analyses also lies in data enrichment; that is, providing more robust parameter estimates for particular RP-based choice models, which should increase confidence in predictions as analysis stretches attribute spaces and the choice set of policy interest.

There are advantages of SP over RP regarding the cost and time required for analysis. While the RP need additional information to surveys (e.g., measuring costs and travel times for each individual), SP surveys are completely autonomous in relationship to the proposed scenarios in which the variables for modeling are defined. However, the SP also has disadvantages when compared to the RP. The most important disadvantage is that there can be large differences between what individuals declare they would do in a given situation and what they actually might do.

Information obtained through an SP survey can distort and create what is known as bias (Bonsall, 1984 or Bradley & Kroes, 1990). The best known types are: affirmation bias, rationalization bias, political bias and no restriction bias. In affirmation bias, the individual respondent may, consciously or subconsciously, be tempted to express preferences that he believes the interviewer wants to receive. In rationalization bias, the respondent can provide artificial responses in an attempt to rationalize their habitual behavior (Bradley & Kroes, 1990). In political bias, the respondent may respond deliberately biased as a way to influence the decisions or policies that he believes to be followed on the basis of the results of the survey. In no restriction bias, the respondent may respond unreal, if not taken into account practical restrictions on their behavior (Ortuzar, 2000). The above biases (of SP survey) are a consequence for those who seek to avoid the automatic preferences (AP), as well as the decrease in cost and time required for analysis to be completed.

2.3. Theory of automatic association (AA)

According to Kahneman (2012) two systems dubs agents handle all thinking: busy System (1) carries out fast thinking, while sluggish System (2) handles slow thinking. Fast thinking is intuitive as it engages the automatic mental activities of perception and memory. Slow thinking is deliberate and effortful. When System 1 presents a plausible story, System 2 will often pass it through uncritically. One may believe a rational choice has been made, when in fact it was not. This paper looks mostly at System 1 and explores how it relates and interacts with automatic preference.

Psychologists Keith Stanovich and Richard West originated System 1 and System 2 to describe the brain's two -system thought process. Kahneman personifies System 1 and System 2 as agents. He believes most people self-identify with System 2, which concentrates, reasons, performs complex mental tasks, makes choices, and is in charge of self-control. Surprisingly, most beliefs and choices actually begin with the automatic impressions and beliefs of System 1 (Kahneman, 2012).

System 1 is automatic and always active, sorting through feelings and memories to make suggestions to System 2, which produces the decision. Usually, this process serves one well. However, System 1 tends to have biases, and relies on the most readily available answers, which can cause judgment errors System 2 cannot detect. System 2 is too slow and effortful to sort through every decision; resulting in the two systems end up being compromised. Kahneman's working premise is that it is easier to recognize other people's mistakes rather than one's own. This is why he believes personifying Systems 1 and 2 helps illustrate how the mind works.

Kahneman investigates associative activation. System 1 makes associations, rapidly linking one idea with another in plausible ways that instruct the body and mind. If one hears the word "eat" and someone asks her to fill in the

missing letters in S__P, the answer is more likely to spell "soup" than "soap." This is called priming. "Eat" primes

the brain to think about food (Kahneman, 2012). Similarly it can show that there is an automatic association between the car and positive constructs in the car users, and it is difficult to obtain an association with a positive construct to the bus. Automatic association test are presented below.

2.4. Automatic preference (AP)

This research proposes a new technique for data collection and a new method for modal choice. The new technique, automatic preferences (AP), is the automatic association that exists in the subconscious of people between an object and their positive constructs, and that generates pro-choice behavior of that object. Automatic preferences (AP) have the advantage as the process operates within the participants' subconscious and their responses are less vulnerable to possible manipulation of the information revealed or stated consciously, as compared to the techniques (RP) and (SP) respectively.

2.5. Automatic association test (AAT)

This test consists of questions evaluating factors of perception, each of which is answered by "yes" or "no". Modal choice was performed between car and bus. These factors are described below.

Quickly read each row and place an X in the "Yes" only if the mode is "Car" and if the "word" is a "positive" word. Otherwise, place an X on "No". Try to respond as fast as possible (automatically / intuitively). If you make a mistake you can erase and correct. But remember to do it quickly; do not delay too much in thinking each response. (Time :_:_)

Mode Yes No Word

Car Comfort

Bus Successful

Car Discomfort

Car Insecurity

Bus Failure

Car Unreliability

Bus Insecurity

Car Successful

Bus Comfort

Car Failure

Bus Unreliability

Bus Discomfort

Car Insecurity

Bus Reliability

Car Reliability

Bus Insecurity

Quickly read each row and place an X in the "Yes" only if the mode is "Bus" and in the "word" is a "positive" word. Otherwise, place an X on "No". (Time :_:_)

Mode Yes No Word

Car Comfort

Bus Successful

Car Discomfort

Car Insecurity

Bus Failure

Car Unreliability

Bus Insecurity

Car Successful

Bus Comfort

Car Failure

Bus Unreliability

Bus Discomfort

Car Insecurity

Bus Reliability

Car Reliability

Bus Insecurity

We also carried out a survey of socioeconomic characteristics and attributes of the car and bus modes. Descriptive analyses are presented below.

3. Descriptive analysis

This research is about building a new basic framework for automatic preferences (AP) data and automatic choice (AC) methods. The research was conducted at the University of Cagliari (Italy) with undergraduate and graduate students of the Faculty of Engineering. We applied a test of automatic association, to determine the degree of association to the subconscious level between positive constructs and the car and bus modes, and its relationship to the modal choice, and inquired about the socio-economic aspects, mode choice and attributes of the modes: costs and travel time, transfer, and waiting time, in order to compare the results of the traditional model with the model of automatic preferences only. We surveyed students who made daily trips to get to the university on which mode (car or bus) they had used earlier that day. The vast majority of respondents used either one or the other of the modes for going to university. In the sample population was 41.9% female and 58.1% male. The 38.7% is undergraduate and graduate 61.3%. The ages of the participants varied between 19 and 36 years. 74.2% have access to or own a car and all participants have access to the bus. The automatic association test was applied to each participant by measuring the time spent for the association with the bus- positive word and car- positive word.

4. Hypothesis and analysis

Hypothesis: If the time delayed automatic association of positive words with the bus is greater than the time elapsed on the automatic association of positive words with the car, then the automatic preference is for the car, and therefore the choice will be car mode. For example:

AATB: Automatic Association Time, positive word and Bus AATC: Automatic Association Time, positive word and Car APC: Automatic Preference for the Car APB: Automatic Preference for the Bus NAP: No Automatic Preference

Note: If there are more than two modes, an order could be established according to the automatic association time of each mode.

If, AATB > AATC ^ APC AATB < AATC ^ APB AATB = AATC ^ NAP

Analysis of results: We used SPSS for the entire modeling and descriptive analysis. The main results are: For AATC and AATB we have:

Table 1. Descriptive statistics of Automatic Association Times (seconds) for bus and car

Variable Number Minimum Maximum Average Stand. Dev.

AATB 31 11.08 75.66 20.7342 11.19872

AATC 31 10.75 35.93 17.2984 4.69236

Valid number 31

Table 1 shows the descriptive statistics of Automatic Association Times (seconds) for bus and car. In general terms the AATB outweigh the AATC. The AATB vary between 11.08 and 75.66 seconds, while the AATC ranges from 10.75 to 35.93, and the standard deviation is less than the AATB. There were no AATB equal to AATC. That is, there was an automatic preference for a particular mode.

Table 2 shows the descriptive statistics of Automatic Preference for the Car (APC), Automatic Preference for the Bus (APB) and the Revealed preferences (RP), (modal choice made by the participants, CHOICE) are presented:

Table 2. Descriptive statistics of Automatic Association Times for the Car (AATC), Automatic Association Times for the Bus (AATB) and the modal choice made by the participants (CHOICE)

Variable Number Minimum Maximum Average Stand. Dev.

AATb 31 0.00 1.00 0.8065 0.40161

AATc 31 0.00 1.00 0.1935 0.40161

CHOICE 31 0.00 1.00 0.7419 0.44480

Valid number 31

Tables 3 show the frequency of the APB (19.4%) and APC (80.6%). Table 4 shows the frequency of the CHOICE; 25.8% for the bus and 74.2% for the Car. The most important outcome is the great similarity in the results of Automatic Preference for the Car (APC), with the modal choice made by the participants (CHOICE), 80.6% and 74.2% respectively. Moreover, the standard deviation is almost equal, 0.4 and 0.44 respectively (table 2). This situation is applied with a nonparametric test to test the above hypothesis "If time delayed automatic association of positive words with the bus is greater than the time elapsed on the automatic association of positive words with the car, then the automatic preference is for the car, and therefore your choice will be car mode." In the next section the test is performed.

Table 3. Automatic Preference for the Bus (APB) (1=Bus), and for the Car (APC) (0=Car)

Frequency Percent Valid Percent Cumulative

percentage

Valid 0.00 25 80.6 80.6 80.6

1.00 6 19.4 19.4 100.0

Total 31 100.0 100.0

Table 4. Modal choice made by the participants (CHOICE)

Frequency Percent Valid Percent Cumulative

percentage

Valid 0.00 8 25.8 25.8 25.8

1.00 23 74.2 74.2 100.0

Total 31 100.0 100.0

4.1 Statistical test to show the relationship between the responses arising from automatic preferences(AP) and responses of modal choice made by participants (CHOICE)

To statistically test this demonstration, the nonparametric Wilcoxon test is applied on the responses arising from automatic preferences (AP) and responses of modal choice made by participants (CHOICE). The result should test the hypothesis and indicate that the answers/responses are consistent. The results are in the Table 5.

Table 5. Test Statistics of contrast (b)

Significance_CHOICE - APC

Z -0.816(a) Asymptotic Significance (bilateral) 0.414 Exact Significance (bilateral) 0.688 Exact Significance (unilateral) 0.344 Probability point_0.234_

(a) Based on the positive range.

(b) Rank test with sign of Wilcoxon

The unilateral exact significance value 0.344 is greater than 0.05 (Table 5), indicating that the responses of the Automatic Preference for the Car (APC) and modal choice made by participants (CHOICE) are equal. Testing the hypothesis of AATB > AATC ^ APC and APC = CHOICE. Therefore the modal choice is: Car = 80.6% and Bus = 19.4%. Table 6 shows the ranges between CHOICE and APC.

Table 6. Ranks modal choice made by participants (CHOICE) and Automatic Preference for the Car (APC)

Number_Average rank_Rank Sum

CHOICE - APC Negative ranks 4(a) 3.50 14.00 Positive ranks 2(b) 3.50 7.00 Draw 25(c) _Total_31_

(a) CHOICE < APC

(b) CHOICE > APC

(c) CHOICE = APC

In addition to having tested the hypothesis, multinomial regression was performed to find the significance of automatic preference in the choice modeling and forecasting that the actual model works.

4.2 Multinomial regression

Tables 7, 8 and 9 show the summary of data processing and information model fit.

Table 7. Summary of case processing

Number Marginal rate

CHOICE 0.00 8 25.8%

1.00 23 74.2%

Valid 31 100.0%

Lost 0

Total 31

Sub-population 2

Table 8. Information model fit

-2 log Degree of

Model_likelihood Chi-square_freedom_Significance

Only the intersection 11.096

End 5.315 5.782 1 0.016

Table 9. Contrasts of the likelihood ratio

-2 log likelihood Degree of Effect reduced model Chi-square freedom Significance Intersection 5.994 0.680 1 0.410 APC_11.096_5.782_1_0.016

The chi-square statistic is the difference in -2 log likelihood between the final model and the reduced model. The reduced model is formed by omitting an effect from the final model. The null hypothesis is that all parameters of that effect are 0. Table 10 shows that Automatic Preference for the Car (APC) is a significant variable, whose value is 0.022 less than 0.05. The reference category was car running the model.

Table 10. Parameter estimates

CHOICE (a)

B Standard

Parameter error

Degree

of Signi-

freedom ficance

Confidence-at 95% for-Lower

_interval -Exp

0.00 Intersection 0.693 0.866 0.641 1

APC -2.351 1.024 5.278 1

0.423 0.022

0.095 0.013

(a) The reference category is: 1.00

The model predicts the modal choice of car with 80.6% and 19.4% of the bus (table 11), with equal automatic preferences, AATB > AATC ^ APC and APC = CHOICE. Therefore the modal choice is: Car = 80.6% and Bus = 19.4%.

Table 11. Classification and prognosis

Observed Predicted

0.00 1.00 Percent Correct

0.00 4 4 50.0%

1.00 2 21 91.3%

Overall percent 19.4% 80.6% 80.6%

5. Conclusions

A discrete choice model in transport can be estimated using automatic preference data and the automatic choice method. The results indicate that we can predict the modal choice with a high probability and the modeling requires

very little data, fewer variables, lower costs, and less time spent on data collection and modeling, also, testing the hypothesis of AATB > AATC ^ APC and APC = CHOICE.

This research creates a new basic framework for automatic preferences (AP) data and automatic choice (AC) methods that is based on an association in the user's memory at the subconscious level between modes of transport and their positive qualities, and its relationship to modal choice. The research also incorporates the observable attributes of alternatives such as travel time, waiting time, number of transshipment, costs, and socioeconomic characteristics of the users which may influence their choices. The most relevant results obtained were the automatic association (AA) between the car and positive constructs of the car users, and a large difficulty in obtaining an association that is a positive construct to the bus. The bus users also can associate more easily with this mode with positive constructs than the car users. Finally, this automatic association (AA) can affect the process of rational choice when posed with discrete choice models of perfect rationality, and can alter the responses of the respondents when applied to the stated preference techniques.

Part of the error term in the random utility theory could be explained by the fact that the modeler cannot observe the automatic preferences and that the decisions are made subconsciously. There is a part of the modal choice process that is not dependent on the attributes of the alternative, or either of the socioeconomic characteristics, but is directly dependent on the subject.

References

Bonsall. P.W. (1984) Transfer price data: its definition, collection and use. En E. Ampt, A.J. Richardson y W. Brog (eds.), New Survey Methods in Transport. VNU Science Press, Utrecht.

Bradley, M.A., & Kroes, E.P. (1990) Forecasting issues in stated preference survey research. Proceedings 3rd international Conference on Survey Methods in Transportation. Washington, D.C., EE.UU.

Kahneman, D. (2012) Thinking, Fast and Slow (1st ed.). California: Garamond Press.

Louviere, J.J., Hensher, D.A., & Swait, J.D. (2000) Stated Choice Methods. Analysis and Applications (1st ed.). New York: Cambridge University Press.

Luce, R.D., & Tukey, J.W. (1964) Simultaneous conjoint measurement: a new type of fundamental measurement. Journal of Mathematical Psychology I(1), 1-27.

Ortüzar, J. de D. (2000) Modelos de Demandas de transporte (1st ed.). Chile: Alfaomega.