Identifying Hotspots of Transport Disadvantage and Car Dependency in Rural Ireland

12 This paper explores the concepts of car dependency and transport disadvantage and the correlation 13 between them in rural Ireland as a means of highlighting incidences of possible forced car ownership 14 with the use of Geographical Information Systems (GIS). Societal and cultural challenges associated 15 with the prevalence of the private car as the primary or in some cases the only form of mobility for people 16 living in rural areas are examined resulting in potential cases of forced car ownership (FCO) (Curl et al, 17 2018). Those defined as being forced to own a car are those who may find themselves in circumstances 18 with low transport accessibility and low income, which is intensified by the need to economically 19 participate in society for financial gain (Mattioli, 2014; Taylor et al., 2009; Currie and Delbosc, 2009; 20 Currie and Senbergs, 2007). This paper examines the existing gap between the necessity of transport 21 and the provision of reliable public transport in rural Ireland, which is frequently attributed as a major 22 determinant of FCO in the literature. While it is acknowledged that forced car ownership similarly exists 23 in urban areas under the same or similar conditions, this paper focuses exclusively on the incidence of 24 FCO in rural areas due to potentially higher levels of car dependency. The main objective of this paper 25 is to identify hotspots or areas that are susceptible to increasing rates of FCO and transport 26 disadvantage. Using the information gained from identifying the locations of these hotspots, transport 27 planners and policymakers can tailor interventions to improve sustainable mobility in these areas and 28 address equity concerns.

determinant of FCO in the literature. While it is acknowledged that forced car ownership similarly exists 23 in urban areas under the same or similar conditions, this paper focuses exclusively on the incidence of 24 FCO in rural areas due to potentially higher levels of car dependency. The main objective of this paper 25 is to identify hotspots or areas that are susceptible to increasing rates of FCO and transport 26 disadvantage. Using the information gained from identifying the locations of these hotspots, transport 27 planners and policymakers can tailor interventions to improve sustainable mobility in these areas and 28 address equity concerns. 29

INTRODUCTION 30
Three out of four journeys outside Dublin were made by car in 2016 (Department of Transport, Tourism 31 and Sport (DTTAS), 2017) and levels of car dependency tend to be even more exacerbated when there 32 is a need to travel over longer distances within rural areas (Currie, and Senbergs, 2007). However, 33 Distance to basic public services (e.g. hospitals, schools) was found to be related to 'forced' car ownership of low-income households in rural China.
Mattioli (2017) While Germany presents a higher incidence of FCO in rural and suburban areas (following the spatial trends of other continental countries like Australia), findings from the UK reveal FCO is also similarly present in urban areas due to the poor quality and high prices of public transport services. Walks (2018) Automobile dependence was found to be positively associated with the burden of automobile loan levels among low-income households in seven of the largest Canadian metropolitan areas.
Chevallier et al.
Low-income households in car dependent areas on the outskirts of Paris and Dijon (France) tend to reduce their trips to become less vulnerable to car-related economic stress (CRES) and avoid residential relocation Curl, et al. (2018) At the individual and aggregate levels, the relationship between financial difficulties and car ownership has weakened, indicating a more complex and dynamic relationship between financial circumstances and car ownership than conventional wisdom would indicate.
Transport subsidies such as concessionary fares for targeted populations, such as older people and disabled do little to address the widespread issues transport poverty.
Rock et al.
Results from the study survey pointed to considerable problems in suburban areas of Dublin that are disproportionately and unfairly impacting on particular population groups, including those that are not traditionally seen as disadvantaged.

Currie and Delbosc (2013)
The vulnerability of low-income households, living in the urban fringe is a major policy concern with regard to their inability to afford potential increases in fuel prices. Ahern and Hine (2012) Focus group discussions demonstrated that men find it more difficult to move from car use and car ownership to public transport and community transport use. Older women, while still experiencing difficulties in travelling, seemed to adjust to life without a car more easily than older men who were more likely to have driven themselves. Delbosc and Currie (2012) Voluntary and involuntary one-car households were more likely to be low-income and contain unemployed people than households running 2+ cars. Involuntary one-car households were still heavily reliant on car travel which resulted in greater problems with access, lower participation and social support and lower well-being.
Lucas (2012) Transport-related exclusion can be identified as a universal and operational concept, although it is differentially experienced within and between nations and by different social groups in different social and geographical contexts.
Challenges to providing accessibility and connectivity in rural communities include: understanding basic technological requirements in rural areas, considering trust and reliability issues with the crowd-sourced information provided by passengers during their journeys, and understanding an anticipating passenger behaviour change in response to technological innovations.
Currie et al.
FCO households make less trips (12.9%), travel shorter distances (-7%) and slightly shorter time (-6.8%) than average 2+ car households in Outer Melbourne. This propensity to travel less might be illustrative of financial pressures and a desire to reduce the costs of travel compared to other income groups in similar circumstances.

Preston and Rajé (2007)
Accessibility planning should not be limited to analysing social exclusion. In particular, charging mechanisms targeted should also be examined as they provide funding streams to promote personalised travel marketing and transport services that may more effectively deal with exclusion. Currie and Senbergs (2007) Results have shown that low-income households with high car ownership make 12.9% fewer trips, travel 7% shorter distances and have 6.8% shorter travel times than the average of 2+ car households in outer Melbourne (Australia).

Njenga and Davis (2003)
Transport is necessary in achieving a wide range of objectives including economic growth, personal welfare, governance and empowerment as well as security. However, the effectiveness of the sector in delivering these objectives is limited by an absence of policy links to other sectors to which it plays an important role.
In recent decades rural areas of Ireland have undergone a relatively dramatic demographic shift, which 141 has led to many young and educated people either moving to urban areas in Ireland in search of higher 142 paid employment opportunities in regional cities such as Dublin, Cork, Galway, Limerick and Waterford, 143 or have chosen to emigrate from Ireland. As a consequence of this, many rural areas have experienced 144 rapid depopulation, with the average age profile in such areas also increasing at a similar rate. The 145 extent of this shift in population from rural to urban areas in Ireland is clearly illustrated in Figure 1 and 146 2. Figure 1 and Figure 2  this has ultimately led to the closure and relocation of such services due to inadequate levels demand 164 to financially sustain the services. However, the Rural Transport Programme (RTP) is a true exception 165 to this trend, as it was introduced to address the mobility needs of the rural population in areas 166 experiencing a lack or in some case a total absence of public transport services. poor access to public transport services that only operate on one day per week from a 'hinterland 185 catchment area' to a market town and suggests that there must be a multi-faceted solution that must be 186 tailored to the needs of each specific area, with local community support. This paper presents a method 187 can be applied to such as solution by initially detecting the worse hits areas of transport disadvantage 188 risk and inaccessibility. 189 190 Thus, this paper seeks to offer a method of identifying areas of the country that are currently not being 191 serviced by the RTP and that are exhibiting signs of transport disadvantage risk and deprivation. It is 192 understood by the authors that research highlighting hotspots of FCO in Ireland has not been conducted 193 to date, therefore, this paper offers a novel approach that could aid transport planners in identifying 194 areas in need of service provision under the RTP and provide an evidence base for strategic investment 195 in public transport. 196 2015; Preston and Rajé, 2007;McDonagh, 2006), by adding an innovative tool to complement the 201 evaluation of areas in most in need of public transport. This methodology utilises a variety of GIS 202 analyses to enable a more objective verification of transport needs. In that sense, a fourfold analysis 203 was developed, which is presented in the following section. 204 The research presented in this paper was conducted as part of a fourfold methodology, which is 205 presented in Figure 3. Each part of this methodology will then be delineated in the subsequent sections Ireland. 225 The model that is proposed applies a Kernel Density (KD) function to estimate the availability of the 226 National Public Transport Access Nodes (NaPTAN, 2017). In total there are 19,630 nodes including bus 227 stops, rail stations, taxi ranks, and ferry ports, which were used in this research. It is important to mention 228 that the transport nodes introduced by the Rural Transport Programme were not included in this dataset 229 at this stage. Rather the focus for this step was to examine the existing level of mainstream public 230 transport. 231 This approach allows converting a point-based dataset into an area-based measure that can be 232 aggregated at the ED level to allow further comparisons with other socio-economic indicators (e.g. HP 233 deprivation index). Moreover, this method (represented in Figure 4) also considers the mutual influence 234 of nodes placed at neighbouring EDs, minimising thus the modifiable areal unit problem (MAUP) 3 235 (Openshaw, 1984). 244 Where: 245 ! " is the influence score generated by the transport nodes around grid cell " 246 # is the number of transport nodes within the threshold distance from grid cell " (i.e. only if +,-. < < +,-. /01 ) 247 $%&' ()* is the threshold distance (also called as search radius) that is further discussed later 248 $%&' + is the distance between the grid cell " and the transport node ,.

250
The KD analysis generates a raster grid in which every cell receives a value representing the density of 251 transport nodes considering a given search radius of 10 km and a distance-decay effect calculated by 252 the equation specified above. The average distance travelled to work in Ireland based upon the census 253 is 14 km and this informed the 10 km distance. The authors do note that this does not take into account 254 of the possibility of "park and ride" or "kiss and ride" possibilities, however, the data utilised in this study 255 was not conducive to multi-mode trips. The Irish census data only takes into account the main mode of 256 transport used and for the longest distance. This is a limitation of the work and when interpreting the 257 results this should be considered. The 10km distance is examined further in the results section in Table  258 3, where a sensitivity analysis is conducted examining 5km and 15km distances compared to the model 259

fit. 260
This search radius is needed to determine if a particular demand is covered (Radke and Mu, 2000) and The indicator of public transport density by ED is thus applied as a proxy for transport disadvantage risk 273 at a local level. In that sense, a region with a lower density of public transport nodes can be considered 274 as more at risk to transport disadvantage. It is important to remark that this proxy has been tailored to 275 the Irish context by applying a model that is compatible with the level of spatial data that is currently 276 available and is able to capture the socioeconomic and demographic characteristics described earlier. 277 Nonetheless, other accessibility indicators (e.g. cumulative opportunities, logsum benefit, two-step 278 floating catchment area, etc) may be also appropriate to better proxy transport disadvantage in other 279 contexts where further spatially disaggregated data is consistently available. 280 Finally, in order to further evaluate the presence of clustering patterns of ED's at transport disadvantage 281 risk, a 'Hot Spot Analysis' (Getis and Ord, 2010) was then undertaken by means of ArcGIS. This analysis applies the Gi*-statistics methods  to identify local "pockets" where spatial 283 autocorrelations are more likely to occur. In other words, these hotspots highlight areas with high 284 concentration of homogeneous conditions of poor transport accessibility. 285

Deprived areas mapping 286
For the purpose of this research, the deprivation values from the Pobal HP index (2012) were applied 287 to each ED in a shapefile extracted from the CSO database (2017). The HP Index is widely recognised 288 as an accurate proxy for deprivation in Ireland, which measures the relative affluence and/or 289 disadvantage of a particular area (Pobal, 2017). This measure of deprivation varies from a value of -35 290 (most disadvantaged) to +35 (most affluent) and it is based on a number of factors including age 291 dependency rate, level of education number of persons per room, unemployment rate, number of lone 292 parents, and professional classes (Pobal, 2012). Similar to the previous indicator, a hotspot analysis 293 was also carried out to identify clustering patterns of deprived ED's by means of ArcGIS. Following the 294 categorisation proposed by Haase and Pratschke (2017), all areas below the threshold of -10 in the HP 295 index were considered as deprived areas. 296 Finally, statistical correlations between transport disadvantage risk and deprivation were performed 297 through (i) a Spearman's correlation analysis, and (ii) linear regression between these two variables 298 with the aid of SPSS software. Since several hot and cold spots of Transport Disadvantage and HP 299 Deprivation were found at a higher level of aggregation, the final evaluation of correlation between these 300 two variables was conducted at county level. Moreover, as EDs from the main regional cities (e.g. Cork, 301 Limerick, and Galway), and from the whole county of Dublin presented extremely low levels of transport 302 disadvantage, due to the high density of public transport in urban centres, they had to be excluded 303 following a process of outlier removal. 304

Forced car ownership 305
In addition to the previous layers, another map was then plotted in order to assess potential of FCO. an absolute HP index score below -10 can be considered as socially disadvantaged. 315 section 3.1, areas with low access to public transport have been selected. Any ED with less 317 than at least one transport node in its average public transport density was considered to be in 318 transport disadvantaged. 319 3. High share of single car ownership: Finally, the third symptom considered that may indicate 320 potential FCO is the high share of single car ownership. Even though the Irish Census also 321 includes indicators accounting for multiple car ownership (2, 3, 4 or more), it is reasonable to 322 assume that the most socially deprived households would not be able to afford more than one 323 car even if experiencing FCO. Therefore, the intersection of high shares of single car ownership, 3.4 Rural Transport Programme Analysis 332 In order to further investigate the impacts of the RTP on the existing levels of access to the public 333 transport network, the stops/stations of fixed routes serviced by this programme were incorporated in 334 the transport disadvantaged risk analysis. The RTP stop nodes were sourced from the National 335 Transport Authority of Ireland (NTA) and then mapped in ArcGIS in order to conduct an analysis of the 336 distribution of these stops in the road network and to determine the accessibility of these stops by means 337 of a buffer/ catchment analysis. 338 In this way, spatial and statistical comparisons could then be conducted by contrasting the transport 339 disadvantage risk indicator with and without the RTP. To do so, the same methodology applied to 340 estimate transport disadvantage risk described in Section 3.1 was applied now also considering the 341 public transport nodes introduced by the RTP. The comparison of the percentage increase in availability 342 terms. 344 Moreover, due to the concentration of high levels of transport disadvantage risk in West region of Ireland, 345 further statistical analysis was conducted to explore the accessibility of jobs in this region. This analysis 346 examined data generated from the National Transport Authority (NTA) Regional Transport Modelling 347 System, specifically average journey time data for employment/ commuting trips in the West Regional 348 Model, which is a four-stage transport model. In this model trip times from and to each Electoral District 349 (ED) in the country are estimated as a result of computing a generalised cost function, which is 350 composed of the following components: 351    area of the country is covered by their services. As a result of the analysis conducted with GIS, it was 414 possible to substantiate that 109 rural settlements (48,375 people) were located in areas not covered 415 by the RTP, and in 100 out of these 109 settlements there were no transport nodes available within a 416 10 km radius. The calculated average of the deprivation index at ED level for these settlements was 417 found to be -8.1, with 54 out of these 109 settlements considered as deprived or very deprived (i.e. less 418 than -10) on the HP index. Since the settlement pattern of the rural population is dispersed and only a 419 minority of live in rural settlements, it is accepted that these numbers are only a measurable part of a 420 much larger problem. 421 As a means of conducting a comparison of the coverage of public transport services in Ireland with and 422 without the inclusion of the Rural Transport Programme; levels of transport disadvantage risk were 423 visually represented in the maps presented in Figure 6, based on the density of transport nodes. The 424 composite indicator of transport disadvantage risk utilised here was determined by the same criterion 425 used to detect potential FCO spots in Figure 6. By visually representing this indicator of transport 426 disadvantage risk, it was possible to determine the impact on possible incidences of forced car 427 ownership with and without the existence of the RTP. Figure 8 The study area for this analysis was the West region of Ireland, as the majority of counties in this region 456 showed strong and statistically significant regression coefficients, thus showing a strong correlation 457 between the transport disadvantage risk and deprivation variables in Table 3. As set out in the 458 methodology in Section 3.4, journey times between EDs in the west regional model of the NTA regional 459 modelling system were utilised in conjunction with employment figures taken from the Census of 460 population to determine the number of jobs accessible in 30 and 45 minutes by public transport and 461 private car. This data was then used to analyse the relationship between jobs accessible and transport 462 disadvantage risk values explored in Figure 6 by means of a Spearman Correlation test, shown in Table  463 4. 464 Furthermore, the results of the correlation tests, which are presented in Table 4, showed that there is a 465 statistically significant relationship between the number of jobs accessible and the transport 466 disadvantage risk measure, in the west region of Ireland, consisting of 743 ED's. This is supported by 467 the strong to moderate positive correlation coefficients and p-values being statistically significant at a 468 0.01 level, which are displayed in all cases in Table 4, both for car and bus journeys without and with 469 the inclusion of RTP. In other words, a positive correlation was found between the number of jobs 470 accessible and the transport disadvantage risk values in the west region, which suggests that these 471 variables influence each other. The correlation coefficients show that there was a statistically stronger 472 correlation for jobs accessible by car in 30 mins, while for bus the correlation coefficients were higher 473 for number of jobs accessible 45 mins. The results produced from this analysis ultimately provides 474 evidence to show that the number of jobs accessible in two different time bands is related to the degree 475 of transport disadvantage risk experienced in these electoral divisions. Overall, these results showed 476 that EDs with high a number of jobs accessible also had lower transport disadvantage risk score, 477 suggesting that the number of cumulative opportunities accessible is a key indicator in identifying 478 disadvantage and in this way, these measures are inextricably linked and influence each other. 479 areas of transport disadvantage. The characteristics found in rural Ireland are similar to those in many 484 other countries with large areas of rural populations. Therefore, the findings presented in this paper 485 maybe common to other similar areas across the world. One of the key aspects of the paper is the 486 identification of hotspots as a method to tailor sustainable mobility solutions to ease any future FCO or 487 transport disadvantage. This identification of these hotspots is one of the main contributions of this 488 paper. 489 Community-based and scheduled door-to-door style services will not always be viable for all rural 490 households, thus, this paper supports the view that community-based services are an appropriate 491 alternative to traditional high frequency and high capacity bus and rail services in rural areas. This 492 bespoke type of service caters for the distinctly dispersed, low-density travel requirements and demand 493 existent in such areas. For those whom require tailored transport services due to specific mobility 494 requirements such as disabled and elderly people, school children, young families, etc., community-495 based services may provide the most cost-effective solution to meet their mobility needs, when designed 496 appropriately. 497 While this study was focused on rural Ireland, it is acknowledged by the authors that the methodology 498 developed in this paper is not only exclusive to the context of Ireland, thus the same methodology could 499 be applied to rural settings in other countries. Moreover, this methodology is similarly appropriate for 500 analysis in an urban setting, for instance in suburban areas, as such areas also experience transport 501 disadvantage risk and difficulties accessing mainstream public transport services. In this way, the 502 approach conducted in this study can be easily transferable to other countries and urban locations that 503 experience comparable transport accessibility issues. 504 Even though GIS techniques have been widely applied in the literature to assess transport disadvantage 505 research is recommended to evaluate how FCO may respond to variations on these criteria in a 512 sensitivity analysis. Likewise, further studies and tailored surveys are also needed to explore 513 demographics factors (e.g. gender, education level, access to driving licenses, age of residents) of 514 households considered to be living under FCO. 515 or town specific characteristics on a more disaggregate, microscopic level in rural Ireland as a means 518 of determining the effectiveness of the RTP. 519 As presented throughout this paper, the majority of areas at transport disadvantaged risk in rural Ireland 520 are also deprived in socio-economic dimensions. As a result, this paper suggests a potential reinforcing 521 cycle between social deprivation and transport disadvantage, which appears to be exemplified by FCO, 522 particularly in remote areas where even programmes like the RTP are not proving to be beneficial to 523 everyone in the community. This study has highlighted the importance of demand responsive transport 524 solutions and vehicle borrowing schemes such as those under the RTP in Ireland as potential solutions 525 for tackling FCO and transport disadvantage. This is due to the fact that there is not sufficient travel 526 demand or political will to provide full mass transit or high capacity public transport solutions for 527 dispersed, low density patterns of settlement as seen many parts of rural Ireland. 528 Nonetheless, there is a need for further research to greater understand and assess the effectiveness of 529 other potential alternatives for rural Ireland such as car-sharing, carpooling and micro-mobility solutions. 530 These alternatives are not only useful in addressing inaccessibility issues associated with non-car 531 owning households and elderly and disabled people, but similarly they can enhance accessibility to 532 regional transport hubs and other public transport nodes to create a more integrated and sustainable 533 transport network that is open to everyone, consequently enabling the equal economic participation of 534 all people in society. In order to lift people who are structurally marginalised out of situations of transport 535 disadvantage and transport poverty, we must provide a built environment that is equitable and 536 welcoming to everyone, and a transport network that is inclusive, accessible and reliable is fundamental 537 in achieving this aim. In a broader extent, our findings also allude to the fact that promoting sustainable 538 car-shedding behaviour (Carroll et al., 2017), when combined with a proper access to the transport 539 system, acts not only as an environmentally friendly solution, but also a more socially inclusive transport 540 policy that should be considered nationally by policy and decision makers. 541