Exploring cyclists’ travel behaviour in different attitudinal market segments: case study of Quelimane, Mozambique

ABSTRACT

Distinguishing bicycle market segments is relevant to promote bicycle use, as customised policy strategies can be targeted to each submarket. This paper aims to adopt the attitudinal market segmentation approach to identify potential markets of bicycle commuters and their travel behaviour differences. A survey was conducted among 902 commuters in Quelimane, Mozambique. Factor analysis is used to identify key latent factors. Two-step cluster analysis is used to segment the bicycle commuting market into several submarkets. Then, Mann-Whitney U test is used to compare pairs of segments based on a set of travel behaviour factors such as cycling frequency, travel time and commuting mode. Three segments are obtained: Reluctant cyclists (S1), Livelihood cyclists (S2), and Demanding cyclists (S3). The study reveals that S1 cycle less frequently when compared to S2 and S3. S1 can be motivated to cycle more frequently by promoting educational campaigns to increase their perception towards cycling benefits. S1 and S3 commute long hours when compared to S2, thus, promoting distribution of land-use activities to reduce travel distance could motivate bicycle use. S1 frequently walk commute compared to S2 and S3, therefore policies to increase bicycle use were pointed to making bicycles more accessible to Reluctant cyclists.

KEYWORDS:

• Bicycle
• market segmentation
• attitude
• travel behaviour
• Quelimane

1. Introduction

In recent years, most cities in Sub-Saharan Africa (SSA) observed rapid growth at an average rate of 4.5% (Sietchiping et al., 2012), however, this growth is not always followed by the provision of transport infrastructure. Cities are becoming slow and expensive to get around as road infrastructure is disconnecting (Lall et al., 2017). This has an influence on travel capability, particularly for the low-income population who are usually located in the outskirts of the city (Mendiate et al., 2020). Due to their heavy reliance on expensive and inefficient private public transport, these households spend about 8% of their monthly income on transport costs (Kumar & Barrett, 2008). This decreases the ability to access basic services and limits active contribution to the city’s economy, reinforcing social segregation and widening the social gap (Cervero, 2013; Goswami & Lall, 2016). Consequently, it seems crucial to make daily trips cheaper and more affordable for urban residents, and for this reason, bicycles are seen as alternative modes of transport.

Cycling is commonly recognized as an active, sustainable and effective transport mode for urban trips (Mendiate et al., 2022; Goel et al., 2022; Handy & Xing, 2011). Furthermore, cycling increases social cohesion by being affordable and providing inexpensive commuting (Gatersleben & Appleton, 2007). Moreover, in SSA cities, cycling allows fast access to decent employment and health facilities, contributing strongly to mobility over short or medium distances at substantially low costs (Howe & Davis, 2002; Mutiso & Behrens, 2011). Notwithstanding all the above cycling benefits, it has not been recognized as a formal mode of urban transport (Nkurunziza et al., 2012; Pochet & Cusset, 1999). For instance, in Maputo (Mozambique), the modal share of the bicycle was estimated at 1.2% in 2014, while the mode share for personal motorized vehicles was 12.0%, that for public transport was 42.1%, and that for walking at around 44.7% (JICA, 2014). A key challenge to better understanding of the low cycling trend is the lack of in-depth research into existing cycling behaviour, ensuring that decision makers can effectively target promotional initiatives and prioritise dedicated projects that meet their specific travel needs.

Moreover, to address the growing concerns of air pollution and traffic congestion in many African cities, researchers have begun to promote pro-cycling initiatives, which include traffic calming measures and traffic signals at main road intersections (Acheampong, 2017). Other studies have sought to improve road safety by adapting road networks in favour of cycling (Pochet & Cusset, 1999). For example, Sietchiping et al. (2012) suggest adding safe cycling lanes in an effort to reduce motorized traffic accidents. However, the existing literature offers incomplete information on differences in travel behaviour among cyclists in medium-sized SSA cities, and how they may be motivated to more cycling. A limited number of studies identify, for example, 2% of people in Bamako (Mali) cycle commute (Pochet & Cusset, 1999). In Johannesburg (South Africa), 68% of people cycle a short distance as cycling is used as a feeder to public transport (Moyo et al., 2018). Furthermore, in Dar es Salaam (Tanzania) 76.5% of people taking cycling action have bicycle available for use (Nkurunziza et al., 2012). These studies are often limited to large African cities and the identified travel behaviour cannot be exported to a medium-sized SSA city context for the following reasons: (i) people in these cities tend to cycle more frequently than those in larger cities (Acheampong & Siiba, 2018); (ii) in small African cities, people often cycle long distances because they have fewer transport mode options (Mendiate et al., 2020); (iii) in addition, they often give less importance to factors linked to the commercial bicycle use (Mendiate et al., 2022; Mutiso & Behrens, 2011), which are unique to these cities, contributing to increasing the experience of bicycle even for those who do not own this mode.

Based on those important issues, it can be said that little research has gone into bicycle use in a medium-sized SSA city. In particular, there have been very few attempts to identify distinct cycling behaviours based on different groups of cyclists. This study aims to address this gap by identifying existing cycling market segments to explore cyclists’ travel behaviour differences using attitude-based segmentation with a focus on the city of Quelimane, Mozambique. The rest of the paper is structured as follows. The second section is a literature review on cycling market segmentation. The third section talks about the methodology used. The fourth section presents and discusses the results. The last section presents the conclusion and provides remarks for further research.

2. Literature review

The relevance of market segmentation when investigating travel behaviour has been widely discussed by the academia (Badoe & Miller, 1998; Dill & McNeil, 2016; Elgar et al., 2004; Gatersleben & Appleton, 2007; Nkurunziza et al., 2012). Market segmentation presupposes that in given population there exists variety of subgroups with relatively homogenous characteristics that can easily be targeted to promotional strategies (Anable, 2005). In travel demand studies, segmentation is often based on two approaches. The first approach is based on socio-demographic variables (Bergström & Magnusson, 2003; Jensen, 1999; Kroesen & Handy, 2014), and the second approach is based on attitudinal variables (Anable, 2005; Beirao & Cabral, 2007; Nkurunziza et al., 2012).

Important socio-demographic segmentation approach in travel behaviour studies is often based on age, gender, household size, employment type, vehicle ownership and trip purpose (Elgar et al., 2004; Hunecke et al., 2007; Jensen, 1999). Bergström and Magnusson (2003) divided cyclists based on their stated choice of mode for their journey to work, in summer and winter. Four categories of cyclists were found which are winter cyclist, summer-only cyclist, infrequent cyclist and never cyclist. They found that all-year cyclists were more motivated by exercise, summer-only cyclists were negatively impacted by road and weather conditions, and the other two were mainly influenced by travel time. Kroesen and Handy (2014) defined four behavioural clusters based on different combination of levels of bicycle commuting and non-work cycling. These clusters are non-cyclists, non-work cyclists, all-around cyclists and commuter cyclists. They found that non-cyclists for both work and non-work purposes present the most stable travel behaviour. These studies have shown that this segmentation approach over simplifies the structure of obtained segments as they present some overlaps in terms of motivation and attitude, which is very relevant in travel behaviour studies (Anable, 2005; Li et al., 2013).

In regard to the segmentation based on attitudinal variables, Nkurunziza et al. (2012) uses the stage of behaviour change to segment commuters in Dar es Salaam (Tanzania). Six attitudinal segments were obtained which are Pre-contemplation, Contemplation, Prepared for action, Action, Maintenance and Relapse. They found that the effect of personal, social and physical factors on bicycle commuting have different effects on people depending on the stage of cycling behaviour. Li et al. (2013) used the attitudinal market segmentation approach to identify potential bicycle commuter market segments. Four attitudinal factors were used to identify six market segments. This study demonstrated that different submarkets have distinct attitudinal features and actual bicycle usage. Dill and McNeil (2016) used predefined cyclist typologies such as fearless, enthused and confident, interested but concerned, and no way, no how to find out its applicability to the whole nation and explored motivating and barrier factors. They found that the four types of cyclist typologies show heterogeneous comfort patterns. These findings have demonstrated that attitude-based market segmentation has a merit in identifying the main factors influencing cycling behaviour per each cluster, for this reason it was used in this study.

While few travel market segmentation studies exist in African cities and in particular for cycling, the existing studies often focus on large cities (Bechstein, 2010; Nkurunziza et al., 2012), and thus the identified segments cannot be transferred to a medium-sized SSA city as they tell little in explaining who are the cyclists in these urban contexts due to likely cycling perception differences influenced by different urban settings. For instance, studies in medium SSA cities like Kisumu (Kenya), Mzuzu (Malawi) found important association between cycling frequency and its economic benefits (Mutiso & Behrens, 2011; Moyo, 2013). From these studies, it is shown that individuals cycle more if bicycles can contribute to their livelihood or reduced travel costs. Moreover, Nkurunziza et al. (2012) when estimating cycling use models in Dar es Salaam (Tanzania) revealed that long cycling distances lower the likelihood of being in Maintenance cycling stage. This behaviour is quite different from permanent cyclists in medium SSA like bicycle taxi operators, who are motivated to cycle longer distances in order to increase their daily revenues (Howe & Davis, 2002). These studies reveal that the relationship between attitude and behaviour may vary depending on the urban context, and the cycling promotional strategies should be targeted to the audience. Therefore, a systematic segmentation approach to identify homogeneous cycling market segments is still lacking in empirical research, especially in a medium-sized SSA city setting. This would help to uncover the different cycling segments and their travel behaviour differences in a medium-sized African city, which often receives little attention in scientific research.

3. Methodology

3.1. Case study description

This study is based on data from commuter population of Quelimane, a medium-sized city in Mozambique. From the national census of 2017, the total population is about 246, 915 inhabitants (INE, 2019) and the density of 92,61 inhabitants per hectare. The average annual temperature is about 24.7°C (INAM, 2019). Administratively, the city is divided into 51 neighbourhoods and 3 urban zones. The inner city (Z1) aggregates 7 neighbourhoods and has 54.05% of services. The city-periphery (Z2) is composed of 26 neighbourhoods and 28.82% of services. The suburban area (Z3) aggregates 18 neighbourhoods and 17.11% of the total services (see Figure 1). Generally, cycling in Mozambique has a low modal share, particularly in larger cities (1.2%) like Maputo -the capital city of Mozambique- and Quelimane has the highest share of cycling among the main Mozambican cities.

Figure 1. Case study location.

From this study, it is found that the current modal split is composed of 40% walking, 35% cycling, 8% motorcycles, and 17% cars (Mendiate et al., 2020). The increased cycling modal share could be due to the flat topography of the city, low motorized traffic particularly in the city periphery, and average short commuting distance for the majority population, where more than 61% of inhabitants are located within 2–3 km from the inner city. This has geared to an increase in cycling and other sustainable modes of transport, such as walking. The city is also negatively remarkable for having a poor road network, especially in the city periphery and sub-urban area, and for not having urban public transport. To fill the gap, bicycle-taxi services are used for intra-urban trips. This service is provided by around 2355 operators, most coming from the city periphery (Quelimane, 2017). This has a great contribution to the overall cycling modal share. Different from conventional public transport, bicycle-taxi service as means of public transport is door-to-door public transport, flexible, has no fixed time, stops and routes. However, during day time, they are mainly located near large markets, commercial areas, and workplaces, whereas during the evening time, they are mainly gathered near schools and universities.

3.2. Data gathering

The survey was conducted on 902 individuals in Quelimane (64% male, with a mean age of 25.73 years, SD = 11.36). This has a confidence level of more than 95% and a margin of error of 5%. Using a hybrid survey approach, data were collected through an online survey and face-to-face interviews. For the online survey, 1284 cards containing individual access codes to the questionnaire were randomly distributed in 4 shopping areas, and 3 university campuses. The face-to-face interview survey was administered to 535 individuals who did not have regular access to the Internet. This survey was administered in the main bicycle-taxi corridors and in larger formal and informal markets. It is believed that individuals with different socio-economic backgrounds, travel behaviours and cycling attitude may be found in these locations. This survey ran from July to December 2017. For additional survey methods used, Monzon et al. (2020).

The questionnaire was divided into three main sections:

1. The first section covered the travel behaviour characteristics of the respondents. Respondents were asked to indicate their cycling frequency, average cycling time and frequent mode of transport. Cycling frequency was coded as follows: 1 = once a month or less; 2 = twice a month; 3 = once a week; 4 = twice a week and 5 = daily. The average cycling time was coded as 1 = <10 min; 2 = 10 min-20 min; 3 = 20 min-30 min; 4 = 30 min-40 min; 5 = 40 min-50 min and 6 = >50 min. Regarding daily travel modes, responses were coded as follows: 1 = Walking, 2 = Bicycling, 3 = Combination of cycling with all other modes and 4 = Motorised modes (Motorcycle, three-wheeler and car).

2. The second section contained 24 statements measuring attitude toward cycling. The attitudinal statements were formulated to reflect the reality that commuters encounter in the urban context of SSA cities. These attitudinal statements were adapted from several previous studies (Iwińska et al., 2018; Mutiso & Behrens, 2011; Acheampong & Siiba, 2018). Respondents used a Likert scale ranging from 1 (strongly disagree), 4 (undecided) and 7 (strongly agree) to indicate the extent to which they agree or disagree with the 24 attitudinal statements in the questionnaire (see details in Table 1).

   Table 1. List of attitudinal statements used in the interview.

   Content	Mean score	SD
   Cycling long distances daily is tiring	4.40	2.10
   Frequent disrespect to speed limits discourages cycling	3.92	1.83
   Lack of street signalization discourages cycling	4.23	1.90
   Cycling on uneven pavement is uncomfortable	4.10	1.93
   Cycling is discouraged by the increasing rates of road accidents	2.71	1.68
   Mixed road traffic composition discourages cycling	3.74	1.74
   Cycling in the poorly lit street is uncomfortable	4.35	1.99
   Road encroachment caused by street vendors discourages cycling	3.87	1.78
   Not enough tree canopy shade along the streets discourages cycling	3.93	1.86
   The absence of a convenient place to shower and change clothes discourages cycling	3.16	1.81
   The absence of convenient bicycle parking discourages cycling	2.96	1.77
   Bicycling for bicycle-taxi services is convenient	2.26	2.02
   Bicycling to sell goods on the street is convenient	2.68	2.05
   Bicycling to carry loads in the street is convenient	2.15	1.86
   My available time encourages cycling	3.47	1.93
   The prevalence of bicycle theft discourages cycling	3.12	1.84
   Bicycling in the rain is uncomfortable	2.57	1.67
   Cycling in heat would make me sweat	2.80	1.76
   Bicycling would present a risk of accidents or fall	3.04	1.61
   Bicycling would be good for health promotion	3.94	1.88
   Bicycling is flexible	4.30	1.68
   Bicycling is comfortable	3.92	1.72
   Bicycling is cheaper if compared to other means of transport	4.07	1.84
   Bicycling is pleasant	4.89	1.73

3. The third section of the questionnaire covered the standard list of socio-demographic variables: age, gender, average monthly income, education level, employment status, household composition, vehicle ownership and place of residence.

Out of the 1284 cards delivered, only 521 responded to the questionnaire. Adding to the face-to-face interview survey, a total of 902 responses were obtained with a completion rate of 83.22%. The average completion time of the questionnaire was 26 minutes. Almost 69.31% of the respondents were male and 30.69% were female.

3.3. Cycling attitudinal market segments and daily travel behaviour differences

To identify the attitudinal bicycle market segments and understand their differences in travel behaviour, three-fold phases were followed. In the first phase, latent factors were extracted from a series of attitudinal statements using Factor Analysis. In the second phase, a Two-step cluster analysis was conducted to group cyclists into homogeneous attitudinal market segments, which allowed us to understand the differences in travel behaviour of people in each attitudinal market segment. In the third phase, the differences in travel behaviour were assessed after conducting a Mann-Whitney U test of pairs of attitudinal market segments.

3.3.1. Phase 1: factor analysis to explore commonalities between attitudinal statements

Factor analysis is conducted to determine the underlying factor structure that exists among the 24 attitudinal statements and to identify a set of latent factors. Principal component is selected as the extraction method and the rotation used is Promax, due to its ability to handle a large dataset (Field, 2013). The robustness of the result is tested using Kaiser–Meyer–Olkin (KMO) and Bartlett’s Test of Sphericity. The obtained factor scores provide inputs to the cluster analysis. SPSS version 23 was used for the statistics (Table 2).

Table 2. Factor loading and factor group influencing bicycling markets.

Latent factors and correspondent attitudinal statements	1	2	3	4	5
Cycling long distances daily is tiring	0.57
Frequent disrespect to speed limits discourages cycling	0.64
Lack of street signalization discourages cycling	0.77
Cycling on uneven pavement is uncomfortable	0.81
Mixed road traffic composition discourages cycling	0.65
Cycling in the poorly lit street is uncomfortable	0.76
Road encroachment caused by street vendors discourages cycling	0.61
Not enough tree canopy shade along the streets discourages cycling	0.62
Cycling is discouraged by the increasing rates of road accidents		0.55
The prevalence of bicycle theft discourages cycling		0.52
Bicycling in the rain is uncomfortable		0.76
Cycling in heat would make me sweat		0.84
Bicycling would present a risk of accidents or fall		0.62
Bicycling for bicycle-taxi services is convenient			0.89
Bicycling to sell goods on the street is convenient			0.73
Bicycling to carry loads in the street is convenient			0.90
Bicycling is flexible				0.70
Bicycling is comfortable				0.62
Bicycling is cheaper if compared to other means of transport				0.66
Bicycling is pleasant				0.57
The absence of a convenient place to shower and change clothes discourages cycling					0.75
The absence of convenient bicycle parking discourages cycling					0.79
% of Initial Eigenvalues variance	5.70	2.73	1.87	1.50	1.36
% of initial Eigenvalues cumulative (n=54.89)	23.75	11.39	7.79	6.27	5.67

3.3.2. Phase 2: attitudinal cycling market segments

Homogeneous segments of the cycling market are identified through cluster analysis. This served as the basis for decision-making in defining targeted policies to encourage people to cycle. Two-step clustering is selected as the clustering method for handling large data sets either categorical or continuous and for automatically determining the ideal number of clusters (Martínez et al., 2006). The score each latent factor received was considered as continuous variables. Akaike’s Information Criterion (AIC) was selected as a clustering criterion for being sensible to variables and providing the best possible solution. The average silhouette measure of cohesion and separation is 0.4, which is statistically acceptable. The ratio of sizes between the largest cluster and smallest cluster is 1.36. The arithmetic means of each latent factor point belonging to the cluster was extracted and ranged between −0.75 and 1.02 (Table 3).

Table 3. Cluster centres of cycling market segments.

Latent factors	Reluctant cyclists (36.7%)	Livelihood cyclists (27.1%)	Demanding cyclists (36.2%)
Need for cycling facilities	−0.73^2,3	−0.44^1,3	0.89^1,2
Need for the economy	−0.71^2,3	1.02^1,3	−0.16^1,2
Sensitivity to weather changes and safety	−0.75^2,3	0.82^1,3	0.05^1,2
Sensitivity to comfort and stress	−0.51^2,3	0.10^1,3	0.74^1,2
Perceived cycling benefits	−0.41^2,3	0.07^1,3	0.44^1,2
n	331	244	327

Note: Items in superscript indicate the attitudinal market segments the latent factor group is significantly different in comparison with the current attitudinal market segment. For example, a 1,3 written in the superscript of the first line under the Livelihood cyclists segment indicates that the effect of the latent factor need for cycling facilities is statistically different to Reluctant cyclists (1) and Demanding cyclists (3) (Analysis of variance Bonferroni post-doc test p<0.05). The meaning of the positive/negative value, they indicate the extent people in each attitudinal market segment agree or disagree with the influence of the latent factors on cycling. For example, the negative value for the latent factor sensitivity to weather changes and safety under the segment of Reluctant cyclists indicates that people in this segment strongly disagree that they are sensitive to weather changes and safety. Differently, the positive value for the latent factor need for the economy under the segment of Livelihood cyclists indicate that people in this segment strongly agree that the economic benefits of cycling influence the cycling behaviour.

Then, to test the stability within the clusters, one-way ANOVA post-hoc multiple comparisons, Bonferroni, were used. Bonferroni test was selected for performing many independent or dependent statistical tests at the same time. The five standardized latent factor scores were considered as dependent test variables, while the selected clusters solution was considered as an independent test variable with the mean differences significant at the 0.05 level. The result of this analysis is presented in Tables 3.

3.3.3. Phase 3: differences in cycling behaviour

After identifying the attitudinal market segments of cyclists, a non-parametric test (Mann-Whitney U test) was conducted to test the existing cycling behaviour differences between the attitudinal market segments. Mann-Witney U test was employed to compare two independent pairs of attitudinal market segments based on a set of travel behaviour factors such as cycling frequency, travel time and frequent travel mode. From the output of this analysis, those pairs of attitudinal market segments that present significant statistical differences at p-level <0.05, error bars were applied to explore the existing travel behaviour differences. Error bars are graphical representations of the variability of data and are used to indicate the error in a reported measurement. They often represent one standard error, at a particular confidence interval (e.g. a 95% interval) (Field, 2013). The result of this analysis is presented in Table 5 and Figure 2a-c.

Figure 2. Main travel behaviour differences between pairs of cycling market segments.

4. Results

4.1. Identifying commuters’ attitudes using factor analysis

In Table 2, a total of 5 latent factors are presented. Factor Group (FG) 1 comprises 8 attitudinal statements associated with the statement’s concerns about road infrastructure conditions and traffic. It shows respondent sensitivity to comfort and stress caused by poor street pavements (0.81), lack of street signalization (0.77), and lack of street lights for night riders (0.76). This FG is called ‘sensitivity to comfort and stress’. FG 2 comprises 5 attitudinal statements showing respondents’ sensitivity to weather changes and safety such as heat (0.84), rain (0.76) and lack of safety, risk of accident or fall (0.62). This FG is called ‘sensitivity to weather changes and safety’. FG 3 comprises 3 attitudinal statements that show the commercial potentiality of cycling such as carrying loads on the street (0.90), for taxi-services (0.89) and selling goods on the street (0.73). This FG was called ‘need for the economy’. FG 4 consists of 4 attitudinal statements reflecting the benefits of cycling, such as flexibility (0.70), comfort (0.62) and low travel costs if compared to motorized modes (0.66). This FG was called ‘perceived cycling benefits’. FG 5 consists of 2 attitudinal statements presenting respondents’ need for cycling facilities such as proper bicycle parking (0.79) and a place to shower and change clothes after cycling (0.75). This FG was called ‘Need for cycling facilities’.

The goodness of fit indexes, KMO = 0.86, chi-square (χ²) = 6998.85; degree of freedom (Df) = 276 and Sig = 0.00, which is statistically acceptable.

4.2. Cyclists’ attitudinal market segments and their socio-economic characteristics

From the identified 5 latent factors, 3 attitudinal market segments were obtained and presented in Table 3. These segments are Reluctant cyclists (S1), Livelihood cyclists (S2) and, Demanding cyclists (S3). In ANOVAs, the difference between the segments with regard the mean of the constitute latent factor was tested. Significant differences are indicated in Table 3 superscripts. Their socio-economic characteristics are presented in Table 4. The summary of the profile of each attitudinal cycling market segment is presented as follows:

Table 4. Socio-economic characteristics and travel behaviour profile of each attitudinal segment.

	Cycling market segments
Items	Reluctant cyclists (S1)	Livelihood cyclists (S2)	Demanding cyclists (S3)
Socio-economic factors
Gender
% Male	69.5	84.2	61.7
Age
% young (<35 years old)	88.5	78.4	82.8
% adult (>35-<55 years old)	9.1	14.9	14.7
% elderly (>55 years old)	2.4	6.6	2.5
Average monthly income (1000Mts ≈14USD)
% <3642Mts	66.8	56.8	30.7
% >3642Mts to <10000Mts	29.0	25.3	60.7
% >10000Mts to <25000Mts	2.7	10.0	8.0
% >25000Mts	1.5	7.9	0.6
Education level
% Other levels	5.7	11.6	1.5
% Primary	29.0	20.7	21.7
% Secondary	57.1	48.5	70.9
% University	8.2	19.1	5.5
Employment
% Unemployed	19.9	10.8	10.1
% Informal job	30.5	45.2	32.5
% Formal job	49.5	44.0	57.4
Household composition
% Single parents with children	59.1	46.7	64.0
% One-person household	5.3	4.2	1.2
% Couple without children	6.6	6.3	6.5
% Couple with children	29.1	42.9	28.3
Vehicle ownership
% Only bicycle	32.6	53.3	52.6
% Only motorcycle/car	8.2	9.8	2.4
% Bicycle and motorcycle/cars	10.0	15.6	18.7
% None (walking)	49.2	21.3	26.3
Place of residence
% inner city	8.8	15.8	12.9
% city periphery	78.2	66.8	77.8
% sub urban area	12.4	17.4	9.2

Table 5. Mann-Whitney U test for differences in travel behaviour patterns per pairs of cycling attitudinal market segment.

a) Cycling frequency			b) Travel time
	S1	S2		S1	S2
S2	0.000		S2	0.005
S3	0.000	0.670	S3	0.225	0.045
c) Commuting mode
	S1	S2
S2	0.000
S3	0.000	0.001

Reluctant cyclists: This segment represents 36.7% of the total sample and is the largest. In general, they seem to disagree on the influence of various barriers and motivators on cycling. They strongly disagree with being sensitive to weather changes and safety. They also show reluctance to cycle as they have a slight disagreement on perceived benefits of cycling. Regarding their socio-economic characteristics, this segment is composed of a slightly higher proportion of females. A total of 66.8% have a monthly income below 3.642Mts and 57.1% have a secondary education level. This attitudinal segment presents the highest share of unemployed individuals as 19.9%; however, 49.5% have a formal job. About 59.1% reside in households of single parents with children, 49.2% do not own a transport mode.

Livelihood Cyclists: This is the smallest segment, with a percentage of 27.1% of the sample. It is characterised by agreeing with the motivators for the use of the bicycle, namely by strongly agreeing that cycling may improve livelihoods as it enables commercial purposes such as the transportation of cargo on the street, bicycle taxi services and the sale of goods on the street. This segment is mostly composed of males (84.2%) and a relatively high share of individuals with low income (below 3.642Mts) and high income (above 25000Mts), individuals with low education level and high education level, showing that this segment is a mix of bicycle taxi operators and passengers. About 45.2% have informal jobs, 46.7% are in the household of single parents with children and 53.3% own only a bicycle.

Demanding Cyclists: This segment has a share of 36.2% of the total sample. Individuals in this segment are very demanding for agreeing with the cycling barriers, particularly they agree that the absence of a convenient place to shower and change clothes discourages cycling. They present a similar socio-economic characteristic as Reluctant cyclists in regard to gender since 61.7% are male, the age where 82.8% are young, education level where 70.9% are in secondary education level, employment status where 57.4% have a formal job, and household characteristics where 64.0% are in households of single parents with children. About 60.7% present an income between 3642–1000Mts and 52.6% own only a bicycle.

In short, Reluctant cyclists disagree on most cycling barriers and motivators, particular they strongly disagree to be sensitive to weather and safety and slightly disagree on perceived cycling benefits. Livelihood cyclists agree on the most cycling motivators, particularly for bicycle enabling economic purposes. Demanding cyclists agree with most cycling barriers, in particular, that the lack of cycling facilities influences cycling.

4.3. Travel behaviour

4.3.1. Travel behaviour differences per attitudinal market segments

For each cycling market segment differences in cycling behaviour are assessed based on cycling frequency, daily travel time and frequent commuting mode. Regarding cycling frequency, it is depicted in Table 5a that statistically significant differences are observed between Reluctant cyclists and Livelihood cyclists (S1-S2; p = 0.000), as well as between Reluctant cyclists and Demanding cyclists (S1-S3; p = 0.000). Moreover, significant statistical differences in regard to travel time (Table 5b) are observed between Reluctant cyclists and Livelihood cyclists (S1-S2; p = 0.005) and Livelihood cyclists and Demanding cyclists (S2-S3; p = 0.045). Individuals in all cycling market segments present significant statistical differences in regard to daily commuting mode (Table 5c); Reluctant cyclists and Livelihood cyclists (S1-S2; p = 0.000); Reluctant cyclists and Demanding cyclists (S1-S3; p = 0.000) and Livelihood cyclists and Demanding cyclists (S2-S3; p = 0.001).

4.3.2. Comparing pairs of attitudinal market segments with significant travel behaviour differences

Figure 2 presents the comparison of travel behaviour between individuals in different cycling market segments, which results from the Mann-Whitney U test.

Figure 2a compares cycling frequency between S1-S2 and between S1-S3. It’s found that S1 do not cycle frequently (Mean = 3.38; 95% CI = 3.26–3.52; SD = 1.20) in comparison to S2 (Mean = 3.82; 95% CI = 3.65–3.99; SD = 1.34). A similar finding is observed when comparing S1-S3 (Mean = 3.98; 95% CI = 3.86–4.09).

Figure 2b indicates that when looking for differences in travel time is found that S2 cycle shorter time (Mean = 2.70; 95% CI = 2.50–2.90; SD = 1.57) when compared to S1 (Mean = 2.95; 95% CI = 2.77–3.14; SD = 1.69) who cycle longer time. However, comparing the travel time between S2-S3, it is found that S3 (Mean = 2.71; 95% CI = 2.55–2.85; SD = 1.40) cycle relatively long distances in comparison to S2.

Figure 2c compares the daily transport modes used by people in different attitudinal market segments. It is observed that S1 mostly walk for their daily activities (Mean = 1.66; 95% CI = 1.57–1.76; SD = 0.86) compared to S2 (Mean = 2.00; 95% CI = 1.90–2.10; SD = 0.80) and S3 (Mean = 1.81; 95% CI = 1.74–1.88; SD = 0.64). Different travel behaviour is depicted when comparing S2 and S3, which shows that although both attitudinal market segments use cycling as the main commuting mode, a significant number of S2 considers motorcycles and cars as their main commuting mode.

5. Discussion and policy implications

This study contributes to identifying travel behaviour differences among clustered attitudinal market segments in Quelimane, a medium-sized Mozambican city. The segments were identified through Two-step cluster analysis based on a set of attitudinal variables. The travel behaviour differences were identified through analysis of several cycling behaviour factors such as cycling frequency, travel time and frequent commuting mode.

From Figure 2a, a significant travel behaviour difference is observed between S1-S2 and S1-S3. In general, Reluctant cyclists (S1) cycle less frequently when compared to Livelihood cyclists (S2) and Demanding cyclists (S3). Based on Table 3, Reluctant cyclists have a negative attitude towards cycling, and they disagree on the influence of various motivators on cycling. This is as expected, based on Muñoz et al. (2013) people not experiencing cycling are less likely to have a clear judgment on the cycling benefits. This is particularly evident for women such as those in S1 (Table 4). This is consistent with Acheampong (2016) who mentions that in Ghana women do not perceive cycling as the most comfortable, safe, easy, convenient and flexible mode of transportation. This can probably explain the cycling behaviour differences between S1 with S2 and S3. These finds suggest the need for educational campaigns targeted to non frequent cyclists such as woman to perceive the benefits of cycling, which could encourage a more frequent bicycle use. Similar finds have been reported in Pochet and Cusset (1999) in an urban context with bicycles having enormous prominence. They point out that women constitute a group worth working with for awareness campaigns as they will create a chain reaction within their family and to the entire community. Doing so might influence the next generation of confident cyclists who can view cycling as a beneficial transport mode.

From Figure 2c-d, it is found that people in S1 and S3 cycle relatively longer time compared to those in S2. Based on Table 4, S1 and S3 are composed of individuals having formal jobs often located in the inner city while residing in the city periphery and the suburban areas. Based on Lall et al. (2017), the spatial mismatch between jobs and residence places influences long travel time. This reinforces the segments’ attitude towards cycling as S1 has a negative attitude towards cycling, while S3 often complains about the lack of cycling facilities such as a place to shower after long cycling time, despite presenting a generally positive attitude towards cycling. These results confirm previous studies like Heinen et al. (2010) who found cycling being preferable for short distances since it is very physically demanding. Moreover, Muñoz et al. (2013) cite that the lack of cycling facilities such as a place to shower cannot be overcome by the cycling experience. This finding suggests the need of reducing daily travel time to main destinations and provision of cycling facilities such as places to shower as initiatives to encourage more cycling. Therefore, to remove these cycling barriers from individuals in S1 and S3, policies should focus on improving land use distribution by allocating more formal jobs in the city periphery where the larger workforce resides. In addition, authorities must ensure providing proper places to shower at formal workplaces.

Regarding frequent travel mode, Figure 2c indicates that S1 frequently commutes by walking compared to S2 and S3. Based on Tables 3–4, people in S1 have a negative attitude toward cycling and have the highest share of low-income individuals and most do not own any transport mode. According to Fernández-Heredia et al. (2014), not owning a bicycle decreases the odds of cycling. While people in S2 and S3 have similarities in bicycle use, those in S2 slightly use more motorcycles/cars than those in S3 (Figure 2c). This could be explained based on Table 3, where the S2 presents the highest share of high-income individuals which obviously tends to increase motorcycle/car ownership and use. This is in line with Biernat et al. (2018), who indicate that the higher the income, the lower the bicycle ownership and use. This finding indicates the need to increase bicycle ownership and use among people in S1 and high-income people from S2 who show a positive attitude toward bicycle taxi (Table 3). According to UNEP (2012), where it is shown that due to the high costs of bicycle acquisition and maintenance in SSA cities, it is crucial promoting bicycle acquisition schemes and lowering taxes, to anticipate increasing bicycle ownership, particularly among low-income families as those in S1. Moreover, according to Howe and Davis (2002) efforts to attract high-income individuals as those in S2 should include improving bicycle taxi service quality, as they often present inexperienced driving and poor appearance and personal hygiene.

6. Conclusion

This study identifies the main travel behaviour differences in Quelimane, a Medium sized city in Mozambique through the use of an attitudinal market segmentation approach. The result of this study revealed that the attitudinal segmentation approach is useful in identifying attitudes and travel behaviour differences that drive each cycling market and the policies and strategies that would be the most effective in targeting the different market segments. In this study, three attitudinal market segments were identified: Reluctant cyclists, Livelihood cyclists and Demanding cyclists. When exploring their main travel behaviour differences, it was observed that they exhibit different travel behaviour characteristics and thus different barriers and motivations for increased cycling.

This study brings three main contributions. First, this study confirms previous studies that people who have a negative attitude towards cycling, for example, women are less likely to cycle frequently in SSA cities. This is particularly evident in Quelimane, where most women are relegated to household activities and limited to cycling rarely to the market or farms. Second, the study highlights inequalities of cycling access to formal work in SSA cities as people cycling daily for formal jobs often do it for long travel time as most reside in the periphery far from formal jobs often concentrated in the inner city. Third, for travel behaviour research, this study adds that differently from many previous studies, in an urban setting where people have few transport alternatives like medium sized SSA cities, bicycle taxi service is also positively perceived even by high-income individuals.

The results presented in this study are an effort to study differences in travel behaviour among cyclists through the use of attitude-based segmentation in a medium-sized SSA. The study provides a basis for policy-makers to plan and develop cycling promotion strategies tailored to the urban context of medium-sized cities in SSA and targeted to each segment of the cycling market. Given that the quality of the road network in these cities changes spatially and temporally, it is recommended that further studies should analyse how cyclists’ behaviour is influenced by these changes. Furthermore, it is also crucial to understand the influence of socio-economic factors on cyclists’ behaviour in the context of SSA cities, particularly for those who use the bicycle as a means of public transport.

Acknowledgments

The Ministry of Science and Technology, Higher and Technical Vocational Education of the Republic of Mozambique founded this research. The authors would like to thank the support received from the TRANSyT-UPM, the University of Rwanda’s College of Science and Technology and Quelimane city council staff.

Disclosure statement

No potential conflict of interest was reported by the author(s).