The willingness-to-walk to access public transit in Indonesia

In order to support the implementation of public transportation, an understanding of people’s interest in walking is essential. The farther people’s walking distance means public transportation will have a larger catchment area. This study aims to get parameters that can be used to improve walking distance in Balikpapan (Indonesia). Data collection in this study was carried out through stated preference surveys to 800 respondents divided into 400 motorcycle users and 400 car users. The data were analysed using the Binary Logistics Regression Analysis method, and the result shows that the respondent’s walking distance was 300 meters with a probability of 47% (motorcycle) and 59% (car). The result shows that two aspects significantly influence the willingness to walk, comfort and aesthetics. So, the strategies that are prepared to increase willingness to walk can be started by providing street furniture to support the comfort aspect and planting trees and plants to add aesthetic value.


Introduction
Walking is an option for continuing the journey after public transportation [1].Walking is an essential mode of travel to reach the nearest bus stop from the place of origin [2].Therefore, bus stops should be within walking distance acceptable to the community.This makes the willingness to walk necessary for transport planners and related agencies in determining the position of bus stops along urban routes.There is a relationship between walking distance and the use of public transportation [3].The distance of willingness to walk can determine the user base of public transport and the catchment area [4].Thus, the distance, which is the measure of the catchment area, is often used to make predictions about transit passengers, the basis for regulatory arrangements, as well as a reference for financial planning for the concept of Transit Oriented Development (TOD) in a city [5].
The relationship between walking distance and public transport users makes pedestrian paths an important element in a city [6].There is a need to increase the characteristics that support the accessibility of pedestrian paths because sidewalks that give the impression of being walkable will increase walking motivation [7].Various factors can potentially influence the characteristics and variability of people's walking distance, one of which is the built environment in the local area [8].The same thing was expressed by previous study, which confirmed that one aspect of the built environment that can increase walking activity is the walkability of the sidewalk paths that are passed [7].The friendliness of the pedestrian path to pedestrians can be assessed from its wider dimensions and higher quality and the provision of supporting facilities such as chairs and bicycle paths.Increasing walkability in pedestrian corridors leading to public transportation stops can increase the distance people are willing to walk [9].Someone tends to walk farther on a more walkable corridor [10].On average, people in Kupang Indonesia are willing to walk 115 meters [11].On the other hand, the walking distance accepted by pedestrians on Penang Island, Malaysia, is 400 to 500 meters [12].This indicates that the willingness to walk in each city is different [2].Therefore, even though the standard walking distance to the bus stop has been regulated in the service quality indicators issued by the World Bank (1987), this value does not distinguish the conditions of the city served, even though the characteristics of each city are very different [13].Therefore, an analysis was carried out regarding the willingness to walk of the community towards the public transport in Balikpapan.This analysis aims to determine the extent of walking distance people accept and identify the factors that most influence this distance.The resulting parameters can be a reference for the development of bus stop locations and the basis for developing strategies to increase people's willingness to walk.
In this study, the analysis was carried out using the binary logistic regression method to know how significant the probability of willingness to walk is for the people of Balikpapan City.Through this, it can also be known what factors influence people's walking decisions, especially private vehicle users.The analytical method used in this study is quite different from the previous research [11] [13], which used the Probability Density Function (PDF) and Cumulative Distribution Function (CDF) methods.This is because this research is expected to produce a probability percentage of the willingness of motorbike and car users for each variation of walking distance.Also, the study's target respondents in Balikpapan City were private vehicle users.This is different from the case studies in Kupang (Indonesia) and Penang Island (Malaysia), which determined pedestrians as respondents.This difference is because by conducting this research, it is hoped that it can simultaneously increase public knowledge about the existence of the Public Transport, which will later be able to participate in growing the people's interest in walking.Along with the increasing distance of willingness to walk, the catchment area will also be more significant and increase transport users' potential.

Method
In general, this study was divided into two stages: data collection and analysis.At the data collection stage, interviews were carried out with people active around public transport corridors who use private vehicles (cars and motorcycles) for mandatory activity.The survey technique used was random sampling with the stated preference interview technique.Several alternative situations of imaginary pedestrian facilities will be used, which will be answered by the respondent.The survey was conducted online and offline.Because this study focuses on people who use private vehicles, the number of respondents is divided into two classes, namely motorcycle and car users.The number of respondents used in this study was 800, consisting of 400 motorcycle users and 400 private car users.This aims to obtain a balanced comparison of the analysis results regarding the number of respondents.

Questioner Design
The questionnaire used was divided into two types, namely the characteristics of the respondents and part of the willingness to walk.The characteristics of the respondents used in this study include age, gender, occupation, the purpose of travel, direction of travel, number of trips, reasons for using private vehicles, number of family members, number of motorcycle and car ownership, the income of respondents, household income, frequency of use public transportation, distance to access the stop point of public transport, and time to the public transport stop point.The second part of the questionnaire is about the willingness to walk, which contains questions about the pedestrian facility scenario.The pedestrian facility scenario is a combination of several facilities that support walkable, including distance (X1), access (X2), convenience (X3), aesthetics (X4), and safety and comfort (X5) with details shown in Table 1.
Table 1 shows that each facility has several variations.The several facilities in Table 1 are then combined into six scenarios that will be given to the respondents so that each respondent will provide six responses to the pedestrian facilities offered.The various scenarios used in this study are shown in Table 2.

Binary Logit Model
Use the Binary Logit Model to obtain the probability of willingness to walk and the factors influencing it.Binary Logit Model Analysis is used because the variables used only produce outputs, namely willing (1) and not willing (0).Logistic regression is usually used to predict the probability of a phenomenon or event.

Characteristic of Respondents
Respondent characteristic data obtained includes respondents' answers to the questions asked.Questionnaire results data on motorbike users and car users are shown in Table 3. Table 3 informs that respondents who use motorbikes are dominated by people aged 25 years (34.5%), while car driver respondents are dominated by the age group over 26 years (40.5%).Nonetheless, the purpose of travel for motorcyclists and cars is dominated by work (46.5% and 52.5%).Regarding vehicle ownership, respondents who use motorbikes own at least two motorbikes (46%) with more than four family members (37.75%).On the other hand, most car drivers only own one car, with a percentage of 87.75%, and the number of family members is four.
For private vehicles, the motorcycle category prefers it because it saves time (49%), while the car category, as much as 35%, chooses it for convenience.Almost all motorcycle user respondents did not own a car, so they only used motorbikes for daily mobility.This was quite different from car user respondents, who almost owned at least one motorbike but still used a car as their primary vehicle.There is a similarity in the frequency of use of transportation for motorbikes and cars, where almost all respondents did not use public transportation services at all in the past month (87.5% and 97%).
Table 3 also shows that almost half of the motorcycle user respondents (29.5%) have a place to live near the pick-up point (stops or highways).On the other hand, almost half of car user respondents (35.5%) had to walk for quite a long to access public transportation.This condition was one of the factors that made the percentage of car users' interest in using public transportation lower when compared to motorbikes.From a total of 17 question items tested for validity and reliability, there were 12 invalid items for motorcycle respondents, and for car respondents there were 14.The conditions mean that there were three invalid items in the MC category, namely age, gender, and frequency of use of public transportation.At the same time, frequency of public transport is invalid in the car user.Furthermore, the reliability test indicates that both motorcycle and car user respondent characteristic questionnaire has low reliability.This condition occurred because the question items such as age, occupation, and income per month, it is possible that the answers will change if the same questions are asked again.So, the questionnaire needs to be more consistent and more reliable.

Willingness to Walk
The willingness to walk questionnaire from this survey will produce 6 data for each respondent, so 400 respondents will produce 2,400 data.Interpretation of the binary logistic regression analysis results consists of several model tests, namely the overall model test, the goodness-of-fit test, and the coefficient of determination contained in Table 5.When the block number = 0 (the independent variable has not been included in the model), the -2 log-likelihood value for the motorcycle user is 3010.479and decreases to 2846.594 when the block number = 1 (the independent variable is included).The same applies to car user, the -2 log-likelihood value has decreased from 3326,625 to 3244,121.The decrease in the -2 Log Likelihood value indicates that the regression model is improving, so the hypothesized model fits the data, and H0 is accepted.
Furthermore, the table chi-square value for Degree of Freedom 4 (Sum of independent variables -1) for the 0.05 significance level is 9.488.The chi-square value on the Hosmer and Lemeshow's Test is low, namely 5.871 (MC) and 7.861 (LV).The significance value of MC (0.209) and LV (0.097) is more than 0.05, so the model is acceptable.
The Nagelkerke R Square MC and LV values obtained in the final survey were 0.165 and 0.045, respectively.So, it was concluded that the ability of the independent variables to explain the dependent variable was only 16.5% for motorbike users and 4.5% for car users.These values experienced a slight decrease compared to the Nagelkerke R Square values obtained in the preliminary survey, which yielded 18.7% for MC and 9.3% for LV.
Furthermore, the results of the Wald test showed that the variables of distance, access, comfort, and aesthetics in the MC and LV categories had a value greater than the chi-square table (1.960955), so they were declared to affect the model partially.Simultaneously, the results of the omnibus test showed that the chi-square value of motorbike and car users was more significant than the chi-square table (2.375643).This indicated that they had a significant effect after the independent variables were entered into the logistic regression model.
The results of the Odds ratio test on motorcycle and car users show that distance has the highest value, which means that a longer walking distance under certain conditions does not reduce the respondent's interest in walking at all.Furthermore, in the MC category, comfort also contributed to an increased willingness to walk, which was quite large by 0.732 times.Meanwhile, in the LV category, the aesthetic variable gave the second largest contribution after distance, namely being able to increase the willingness to walk by 0.689 times.
Then a probability calculation is also carried out based on the value of the coefficient B presented in Table 5.Then an example of calculating the probability for scenario 1 is carried out.
The calculation of the probability for the MC category for scenario 1 produces a value of 33%.It indicates that only that is the probability of the willingness to walk for the motorcycle community in scenario 1, which has a distance of 100 meters without being supported by wide access and not equipped with other supporting facilities.
A similar method was carried out for the following scenarios for each category of vehicle users, and the recapitulation was obtained as shown in Table 6.The results from Table 6.show that the highest probability for motorbike and car users is the same in scenario 5, with percentage values are 47% and 59%, respectively.Scenario 5 offers a longer walking distance but prioritizes aspects of comfort and aesthetics by providing various supporting facilities such as seats, shade, and bicycle paths, and is accompanied by good environmental management such as planting trees and plants and bollards.Based on scenario 5, the willingness to walk of respondents, both motorbike and car users is sequentially 300 m.

Conclusion
In conclusion, the willingness to walk of respondents who are prospective passengers of the Public Transportation and divided into motorbikes and cars is the same, namely 300 meters.Several factors influence the distance, but the results of binary logistic regression analysis show that two factors most influence willingness to walk, namely the comfort and aesthetics of a sidewalk.Thus, improvement strategies formulated to support walking comfort and foster community willingness to walk can be carried out by focusing on these two aspects.In terms of comfort, this can be done by providing flat pedestrian paths equipped with street furniture such as seats, canopies, and bicycle paths.To improve the aesthetic quality of space, this can be done by planting and arranging trees to create shady conditions, adding decorative vegetation, improving pedestrian path designs, and installing bollards.

Table 1 .
Variable for Pedestrian Facilities

Table 2 .
Scenarios for Pedestrian Facilities

Table 3 .
Characteristics of Respondents

Table 3 .
Characteristics of Respondents (Cont'd) Validity and Reliability of Questioner DataValidity and reliability tests were carried out on the respondent characteristic questionnaire at the end of the survey.The r table value for 400 respondents with a significance level of 5% is 0.098, and for a significance level of 1% is 0.128.The results of the validity and reliability tests for the motorcycle and car categories are presented in Table4.

Table 4 .
Validity and Reliability of Questioner Data

Table 5 .
Estimated Model