Developing a methodology for ex post evaluation of the wider impact of the restoration of rail services to previously disconnected or isolated regions on employment and property prices and accessibility to jobs and essential services.

As well as improving access to jobs and essential services, the re-establishment of rail links to larger, more remote areas may produce wider economic and social impacts both within the region and beyond. Recent developments have highlighted the need for improved ex post evaluation of such impacts, particularly in formerly disconnected or isolated regions. The main thrust of this research was through investigating ex post situations both spatially and temporally to determine cause-effect relationships. This required developing a methodological approach which would match those objectives and adapted pre-existing methods to develop a methodology for appraisal particularly relevant to remote, rural or disconnected regions. Using three case studies - the Robin Hood line (1998), the Stirling-Alloa line (2008), and Borders Rail (2015) as representing different stages of recent rail investment in previously disconnected regions, and applying mainly secondary data sources, a counterfactual was developed which allowed a meaningful comparison between areas subject to treatment i.e. rail intervention, and those not treated i.e. either unaffected or minimally affected by the intervention and to establish any differences between findings in urban studies. Treatment groups were based on distance thresholds where the control group was selected from remaining locations in the region. There appeared to be some benefit in application of clustering and propensity matching to effect a more balanced comparison between similar locations in the treatment and control groups. An important consideration was the accessibility characteristic which conventionally has been distance to the nearest rail station. However, two additional measures were utilised here: a distance to station ratio (which measured the percentage improvement in distance to station following the rail intervention) and a job accessibility index which assessed the improvement in access to jobs based on skills matching and the cost of commuting. The job accessibility index was developed to take into account the limitations in travel in more remote communities where services are less frequent and commuting distances often greater than in the urban situation. The cost of travel was recognised as a key factor affecting accessibility and generalised cost allowed the cost of commuting to be calculated using local values of speed and cost of transport. Job accessibility was based on comparing the percentage skills share at each location, matched to actual jobs at all neighbouring destination locations. The job accessibility index allowed a measure of accessibility based on the original job market, but could also be used to assess the effect on accessibility of a slump in employment by considering the current job market. Without job skills matching, job accessibility appeared to be overestimated as the seemingly high attraction of job opportunities may not always synchronise with the skills set in that location. It yielded good results when used as an accessibility characteristic in the hedonic models, being a more complex measure than distance from rail station as it encompassed the whole regional employment picture relative to each location. Previous research had suggested some correspondence between rail access improvements and increased property price and employment levels. Four different approaches were examined here to assess causality: a descriptive comparison approach, a DID (difference-in-difference) model, a fixed effects hedonic model and a GWR (Geographically Weighted Regression) model. These incorporated other factors such as changes in local and property characteristics over the period spanning each intervention. The descriptive approach looked at individual variables in isolation pre- and post- intervention broken into treatment and control to assess any impacts but ignored the combined effect of other explanatory factors. The output indicated a discernible effect of treatment in some cases, and was useful in corroborating variables to carry forward to the model. For property impacts, the difference-in-difference approach produced contrasting findings for the case study regions. For job impacts, there was a positive effect on employment density of being closer to a rail station and of improvement in job accessibility, but for Borders Rail this was not statistically significant which may be due to the limited amount of data available at this stage. The fixed effects model showed that for property impacts the distance to rail station and distance ratio and improvement in job accessibility were all significant factors. A modified spatial-temporal version of Geographically Weighted Regression estimated local parameters through time by examining changes in coefficients for two separate years spanning the intervention for each case study. The property model showed variation across each region in the negative relationship between price and distance to the nearest station for both the established case study regions. For the jobs model, the relationship between employment density and distance to the nearest station showed that the distance from the rail network was critical in terms of the job market. The findings suggest some causality linking rail investment to house price changes and employment density, dependent on the scale of the rail intervention and the regional context. Improvements in rail transport infrastructure could produce economic benefits affected by the proximity to new stations and relate to the effect on property prices. Although improved job accessibility allowing increased commuting, spatial, temporal and economic barriers may still prevent more economically vulnerable neighbourhoods within each region from receiving the full benefit of the intervention. In conclusion, there are implications for practice in terms of making a case for new rail infrastructure, application in a WebTag style appraisal or evaluation, and new information on spatial patterns of employment and property prices. In addition, consideration is given to expansion of the methodology to other types of transport intervention as well as application in an urban context.