Effects of Taichung bus policy on ridership according to structural change analysis

Abstract

Since 2001, the Taichung City Government has launched several policies to stimulate public transit ridership. Based on this successful experience of reforming the urban bus system, the aims of this study are to analyze the structural change in bus ridership time series over a period of more than a decade. The major objectives are to determine effective policies and explore critical factors. According to monthly ridership from 2001 to 2014, five regimes of bus ridership and three effective bus policies were defined. Important elements of successful policies include complete bus network, high frequency of buses and mileage subsidy. These policies have brought high growth to once very low bus ridership in Taichung City. This type of research process helped to provide an understanding of actual outcomes of past policies, as well as to obtain key factors, which can be used to continue to formulate or adjust public transportation policies. The results of this study may serve not only as a reference for formulating bus strategy in Taichung City, but also for providing direction in the development of public bus service in similar-sized cities.

Appendices

Appendix 1

In the present study, the following procedure was used to eliminate seasonal factors, with raw series data points $y_{t}^{\prime }$ and time sequence, t = 1–164.

• 1. Five data points were used to obtain central moving averages, the values of which were determined by $MA_{t} = \mathop \sum \nolimits_{k = - 2}^{k = 2} y_{t + k}^{\prime } /5$, t = 3–162 is the time sequence of series data, with a total of 160 data points.

• 2. The relative moving average ratio for each data point was calculated, $R_{t} = y_{t}^{{\prime }} /MA$, t = 3–162.

• 3. The mean ratio for each month, $\bar{R}_{i} = \mathop \sum \nolimits_{m = 1}^{N} R_{t = i + 12m} /N$, was calculated with i = 1–12 denoting months. $N = \left[ {160/12} \right] - 1 = 12$ is the number of ratio data point at each i.

• 4. Monthly seasonal factors $F_{i} = \bar{R}_{i} /\frac{{\mathop \sum \nolimits_{i = 1}^{i = 12} \bar{R}_{i} }}{12}$, were calculated with i = 1–12.

• 5. Deseasonalized series data were produced, with each data point represented as $y_{t} = y_{t}^{\prime } /F_{{i = \bmod \left( {t/12} \right)}}$, leading to $F_{0} = F_{12}$, t = 1–164.

Figure 3 presents monthly ridership data following the elimination of seasonal factors. In this figure, the dips in ridership in February and August are more gradual than in the raw data. In this study, this series were used to analyze potential structural changes.

Fig. 3

Taichung City deseasonalized monthly bus ridership

Appendix 2

See Table 5.