Larkhall – Milngavie Railway Project Evaluation Study Final Report
9. REVIEW OF PASSENGER FORECASTS
9.1. This chapter compares actual passenger demand at the four stations on the line that were re-opened in 2005 against the forecasts as set out in the original Business Case. Where there are differences, possible reasons are discussed, for example by reviewing the methodology used to forecast demand.
9.2. Actual demand at the stations on the Larkhall – Milngavie line that benefited from service frequency improvements in 2005 was also reviewed. Demand forecasts for these stations were not explicitly documented in the original Business Case. Actual demand was therefore assessed against a set of control stations not impacted by the project to determine whether the improvements generated new demand above and beyond underlying growth in Scottish rail travel (which in effect is a proxy for the counterfactual i.e. what would have happened had the project not been implemented).
Actual Vs Forecast Demand – New Stations
9.3. This section compares the actual demand data against the forecast demand data from 2006/07 (the first full year in which the stations were open) to 2012/13 (the last year for which demand data is available) for the four newly opened stations in 2005:
- Larkhall;
- Merryton;
- Chatelherault; and
- Kelvindale.
Methodology
9.4. The actual demand data is based on the entries and exits total from the Office for Rail Regulation's (ORR) estimates of station usage[23],[24].
9.5. The forecast demand data was taken from section 7.3.17 of the Modelling Report for the Larkhall/Milngavie Rail Project prepared by Strathclyde Passenger Transport (SPT) in November 2000 (hereafter referred to as the 'SPT Modelling Report'). The SPT Modelling Report only provided disaggregated daily demand forecasts for the four stations in 2001, the date the stations were originally due to open ( Table 11).
Source: Table 7.3.17, SPT Modelling Report
9.6. In the SPT Modelling Report, the 2001 demand was grown by applying forecast planning data (i.e. population, number of households, employment etc) for each of the 599 Strathclyde Integrated Transport Model (SITM) model zones. This provided the demand for the two forecast years (2016 and 2031) at an aggregate level (i.e. total rail boardings) for the entire SPT network which covers all Strathclyde suburban rail services ( Table 12).
Source: Figure 3, SPT Modelling Report
9.7. For this evaluation, the demand in Table 12 was used to derive an average annual rail growth rate between 2001 and 2016 (0.7% per annum) which then was applied to the 2001 demand at the four stations to calculate the forecast daily demand for each year from 2006/07 to 2012/13. This was converted to annual demand by applying an annualisation factor of 283[25] (less than 365 as typically users do not use the network everyday e.g. most rail commuters do not travel at the weekend).
9.8. All annual demand figures presented below are rounded to the nearest 100 passengers.
Larkhall
9.9. Table 13 shows the actual versus forecast demand at Larkhall station from 2006/07 to 2012/13 and is represented graphically in Figure 19. These show that for every year from 2006/07 – 2012/13, actual demand at Larkhall station has been greater than forecast demand. Even in the first full year of opening, the actual demand was over 50,000 (23%) higher than forecast. Despite demand falling in 2009/10 and 2010/11 (presumably due to the recession), demand in 2012/13 was 116,000 or over 50% higher than the forecast.
Figure 19 Actual vs Forecast Demand – Larkhall, 2006/07 to 2012/13
Merryton
9.10. The actual versus forecast demand at Merryton station from 2006/07 to 2012/13 is shown in Table 14 and is represented graphically in Figure 20. In contrast to Larkhall, the actual demand at Merryton has been consistently lower than the forecast. Even though the actual demand growth rate has been higher than the implied forecast growth rate (with the exception of 2010/11) by 2012/13 demand was still 63,000 (36%) lower than forecast.
Figure 20 Actual vs Forecast Demand – Merryton, 2006/07 to 2012/13
Chatelherault
9.11. Table 15 shows the actual versus forecast demand at Chatelherault station from 2006/07 to 2012/13. This is represented graphically in Figure 21. The 2006/07 actual demand for Chatelherault was less than half the forecast. However, demand at Chatelherault is now outperforming the forecast and in 2012/13 was 23,000 (58%) higher.
Figure 21 Actual vs Forecast Demand – Chatelherault, 2006/07 to 2012/13
Kelvindale
9.12. The actual versus forecast demand at Kelvindale station from 2006/07 to 2012/13 is shown in Table 16 . This is represented graphically in Figure 22. The 2006/07 actual demand for Kelvindale was more than double the forecast. Despite actual demand falling by 18% in 2010/11 (possibly due to the recessional impact), demand in 2012/13 was still 51,000 or 111% higher than the forecast.
Figure 22 Actual vs Forecast Demand – Kelvindale, 2006/07 to 2012/13
Summary – All new stations
9.13. Analysis of 2012/13 actual against forecast demand at the four new stations shows there are some large differences with actual demand being at least 50% higher than forecast at three stations (Larkhall, Chatelherault and Kelvindale). Table 17 shows the actual versus forecast demand at all four stations from 2006/07 to 2012/13. This is represented graphically in Figure 23. Whilst actual demand in 2006/07 was very close to forecast (1% difference), a noticeable difference is observed from 2008/09 onwards. Although the recession dampened demand in 2009/10 and 2011/12, the divergence between actual and forecast has continued in the last couple of years with actual demand 26% higher than forecast in 2012/13.
Table 17 Actual vs Forecast Demand – All New Stations, 2006/07 to 2012/13
Figure 23 Actual vs Forecast Demand – All New Stations, 2006/07 to 2012/13
9.14. Potential reasons for the differences between actual and forecast demand are discussed at the end of this chapter.
Actual Vs Forecast Demand – Service Frequency Improvements
9.15. This section assesses the actual demand at stations that benefited from a service frequency improvement from 2005 onwards as a result of the Larkhall – Milngavie project. These stations are divided into two sections:
- Hamilton – Glasgow section (Hamilton Central, Hamilton West, Blantyre, Newton, Cambuslang, Rutherglen, Dalmarnock and Bridgeton); and
- Milngavie – Glasgow section (Milngavie, Hillfoot, Bearsden, Westerton and Anniesland).
Methodology
9.16. The SPT Modelling Report does not provide any demand forecasts for the stations that benefited from service frequency improvements in 2005.
9.17. As an alternative to using demand forecasts, the actual demand is therefore compared against:
- average demand growth at the ten stations on the Cathcart Circle section[26] of the network linking Glasgow to Cathcart in south Glasgow. This section of the network was not impacted by the Larkhall – Milngavie project and demand from 2006/07 onwards is relatively stable. This section therefore acts as a good control and it is reasonable to assume growth at stations that did receive a service improvement would have been similar to the growth observed on the Cathcart section of the network had there been no service improvements since exogenous demand drivers (i.e. population, employment) are likely to have been similar for the two rail lines; and
- the implied overall SPT rail demand growth between 2001 and 2016 as per the Modelling Report (as discussed in section 9.6).
Hamilton – Glasgow stations
9.18. Using estimates of station usage data from the ORR, Table 18 and Table 19 show the actual demand and annual growth rate respectively at the eight stations on the Hamilton to Glasgow section of the network and at a total level. This is also illustrated graphically in Figure 24.
Source: ORR
| Station | 06/07 | 07/08 | 08/09 | 09/10 | 10/11 | 11/12 | 12/13 |
|---|---|---|---|---|---|---|---|
| Hamilton Central | 5.0% | 5.9% | 9.2% | -1.4% | 5.2% | 3.2% | 2.2% |
| Hamilton West | -3.1% | 7.4% | 19.7% | 2.3% | 5.6% | 5.9% | 4.6% |
| Blantyre | 12.6% | 7.2% | 12.2% | -0.1% | 6.6% | 7.5% | 6.3% |
| Newton | 9.0% | 4.8% | 10.7% | 3.6% | 9.1% | 7.2% | 1.5% |
| Cambuslang | 3.9% | 4.4% | 9.1% | -0.7% | 5.3% | 9.8% | 2.6% |
| Rutherglen | 10.0% | 6.0% | 15.9% | 3.3% | 8.4% | 12.3% | 9.3% |
| Dalmarnock | 5.2% | 0.2% | 14.6% | 10.2% | -0.4% | 3.3% | -73.0%[27] |
| Bridgeton | 19.2% | 7.9% | 26.6% | 0.8% | 3.9% | 19.6% | 26.1% |
| All stations | 6.4% | 6.0% | 14.0% | 1.1% | 6.2% | 8.6% | 5.5% |
Figure 24 Hamilton – Glasgow Stations Indexed Actual Demand, 2005/06 to 2012/13
9.19. The majority of the stations on the Hamilton – Glasgow section exhibited strong growth over the period, averaging at 58% total growth over the seven years (an average growth of 7% per annum). In particular:
- growth was notably strong in 2006/07 and 2007/08 perhaps in immediate response to the frequency improvements;
- in 2006/07, demand decreased at Hamilton West and growth at Hamilton Central was less than that seen at other stations on the branch; this may be in part due to abstraction as passengers that previously used these stations switched to use the three new stations at Larkhall, Chatelherault and Merryton (this is supported by the evidence from the User Survey as discussed in Chapter 5);
- the high growth observed in 2008/09 was most likely to be in response to the introduction of more stringent revenue protection regimes by ScotRail on the Glasgow suburban rail network – this increase is therefore not necessarily new demand but instead represents the more accurate recording of existing demand;
- these stations were resilient during the recession period (2009/10 and 2010/11) with continued growth at most stations; this is in contrast to the national trend where rail demand growth was weaker (total ScotRail demand increased 0.7% in 2009/10 and 1.8% in 2010/11[28]).
9.20. Table 20 shows the demand-weighted average[29] actual annual demand growth for the eight Hamilton – Glasgow stations along with the demand-weighted average actual growth for the ten control Cathcart Circle stations and the implied forecast growth taken from the SPT Modelling Report. This is illustrated graphically in Figure 25.
Figure 25 Hamilton – Glasgow Stations Indexed Growth Comparison, 2005/06 to 2012/13
9.21. Figure 25 shows that average growth across the eight Hamilton – Glasgow stations has consistently outperformed that across the ten control Cathcart Circle stations. If it assumed that exogenous growth factors were broadly the same across both areas, then it can be inferred with reasonable confidence that the difference is largely attributable to endogenous factors, including the service frequency improvements.
Milngavie – Glasgow stations
9.22. Table 21 and Table 22 show the actual demand and annual growth rate respectively for the five stations on the Milngavie to Glasgow section of the network and at a total level from 2005/06 to 2012/13. This is illustrated graphically in Figure 26.
9.23. Table 21 and Table 22 show the actual demand and annual growth rate respectively for the five stations on the Milngavie to Glasgow section of the network and at a total level from 2005/06 to 2012/13. This is illustrated graphically in Figure 26.
Figure 26 Milngavie – Glasgow Stations Indexed Actual Demand, 2005/06 to 2012/13
9.24. All stations on the Milngavie – Glasgow section exhibited strong growth over the period, averaging at 40% total growth over the seven years (an average growth of 5% per annum). Growth was particularly high at Hillfoot, Bearsden and Anniesland. In particular:
- growth was notably strong at Bearsden and Hillfoot in 2006/07 and 2007/08 perhaps in immediate response to the frequency improvements;
- whilst it would be reasonable to have expected demand at Anniesland fall in 2006/07 due to abstraction at Kelvindale, in fact demand continued to rise in 2006/07 and 2007/08;
- as per stations on the Hamilton – Glasgow section, the high growth observed in 2008/09 was most likely to be in response to the introduction of the more stringent revenue protection regimes introduced by ScotRail and therefore the increase is not necessarily new demand but represents instead the more accurate recording of existing demand;
- with the exception of Anniesland, all stations suffered a drop in demand in 2009/10 which is likely to be due to recession.
9.25. Table 23 shows the demand-weighted average actual annual demand growth for the five Milngavie – Glasgow stations, along with the demand-weighted average actual growth for the ten control Cathcart stations and the implied forecast growth taken from the SPT Modelling Report. This is illustrated graphically in Figure 27.
9.26. Figure 27 shows that average growth across the five Milngavie – Glasgow stations outperformed that across the ten control Cathcart stations initially from 2006/07 to 2008/09, which was perhaps in response to the service frequency improvements. However, in 2009/10 onwards, average growth across the ten control Cathcart stations then outperforms that across the five Milngavie – Glasgow stations owing to a demand decrease in 2009/10 across the Milngavie – Glasgow stations. This trend is in contrast to the Hamilton – Glasgow stations which did not suffer a downturn in demand during the recession.
Figure 27 Milngavie – Glasgow stations indexed growth comparison, 2005/06 to 2012/13
Review of Forecasting Methodology
9.27. Section 9.13 showed there were some significant differences between the demand forecasts and actual demand for the four new stations.
9.28. This section investigates why differences between the forecasts and actual demand may have materialised through a critique of:
- modelling technique;
- forecasting assumptions; and
- service specification.
Modelling Technique
i. Choice of Model
9.29. Demand forecasts for the Larkhall – Milngavie project were undertaken using SPT's Strathclyde Integrated Transport Model (SITM), a comprehensive four-stage transport model used for transport planning in conjunction with varying land use or roads and passenger transport network development scenarios.
9.30. A full modal split run of the SITM model was carried out for years 2001, 2016 and 2031 for both the base and scenario networks. Travel demand forecasts were prepared to cover a 30 year period from the likely year of opening, as required by the economic assessment and appraisal guidance at the time.
9.31. It is noted that the SPT Report did not include any discussion of the justification for adopting the four stage modelling approach, nor whether alternatives were considered e.g. a direct demand (trip rate) model. However, under Passenger Demand Forecasting Handbook (PDFH) guidance, a four-stage model is considered ideal for considering major rail changes in an urban context. It has the advantage of being able to incorporate complex interactions between different modes within the wider transport network and this is perhaps the reason why a four-stage model was considered most appropriate.
ii. Model Zoning
9.33. The zoning structure of the SITM was reviewed to determine whether this was likely to be the reason why the Merryton station demand was over forecast whereas the adjacent station Larkhall was under forecast. As shown in Figure 28, a disaggregated zoning structure was applied with Merryton station in a separate zone from Larkhall station. It is unlikely therefore that the zoning structure is the cause of the difference. Indeed section 6.1.6 of the SPT Modelling Report discusses how the original single zone representing Larkhall was split into four to create a finer zone structure to allow the three new stations to be modelled more accurately.
Figure 28 STIM Model Zoning – Larkhall
iii. Demand Abstraction
9.34. Demand for new stations is typically from three sources: generated demand (i.e. the trip would not previously have been made); mode switching (i.e. the trip would have been previously made by another mode such as car or bus); and abstraction of demand from existing stations (i.e. the trip would have been previously made from an existing station).
9.35. Abstraction of demand from existing stations is a key consideration when considering demand at new stations, particularly when the new stations are within the vicinity of existing stations as was the case for all four new stations on the Larkhall – Milngavie line[30]. However, whilst the SPT Modelling Report explicitly documented the extent of mode switching as a source of new station demand, there was no reference to demand abstraction from existing stations so it is not clear whether this was considered or not.
9.36. If demand abstraction from existing stations was not explicitly considered, then this may be one reason why the forecasts at the four new stations were typically lower than actual demand.
Forecasting Assumptions
i. Exogenous Growth
- planning data was sourced from the Central Scotland Transport Model[31]; and
- the economic growth rate selected was the mid-point of the Government's low and high forecasts.
9.38. However, as discussed in section 9.6, the implied forecast average annual growth rate between 2001 and 2016 is 0.7% which can presumably be attributed wholly to exogenous growth drivers (assuming that other endogenous drivers such as timetable changes, marketing and fares initiatives are small enough to be ignored).
9.39. To verify this implied exogenous growth rate, key actual economic drivers over the period 2001 to 2013 were obtained specifically for the Glasgow area ( Table 24).
9.40. These growth drivers have been applied to the STP Modelling Report 2001 demand forecasts (as per Table 12) in line with PDFH methodology to grow demand from 2001 to 2012/13:
- commuting demand is grown in line with the total full-time employment growth i.e. 4.8% with an elasticity[32] of 1.3[33];
- other demand (business and leisure) is grown in line with total GVA per capita growth (i.e. 41.4%) with an elasticity of 1.2[34] and total population growth (i.e. 4.4%) with an elasticity of 1.0[35]; and
- commuting and other demand are weighted 59:41 as per Scotland's average journey purpose split[36].
9.41. Under this growth scenario, the 2012/13 forecast demand would compare to actual as shown in Table 25. For Larkhall, Chatelherault and Kelvindale, the forecasts are now closer to the actual demand and at a total level, the actual is only 6% higher than forecast. The overestimate of demand at Merryton is however even more pronounced.
9.42. This high-level analysis indicates that the actual exogenous drivers have outperformed the exogenous forecasts applied in the SITM and may be one of the reasons why the actual total demand is typically higher (although it is noted that the actual exogenous drivers applied for this evaluation were at the Glasgow level whereas in the SITM they were at the local i.e. SITM zone level and therefore not necessarily the same). Indeed, this is referenced in 'Station Usage and Demand Forecasts for Newly Opened Railway Lines and Stations' report[37]; in the absence of explicit exogenous growth factors in the SPT Modelling Report, a 4% per annum growth rate was assumed in the Station Usage report.
9.43. Further evidence that exogenous drivers may have been higher than forecast is provided by comparison of land-use supply in the Larkhall area between 2005 and 2013 which shows a large amount of land being released for industrial and housing development since the new stations have opened ( Figure 29). As a result, population and employment growth may have been higher than the year 2000 forecasts and the catchment areas of the stations may now be larger than originally expected. However, whereas this may explain the higher than forecast growth at Larkhall and Chatelherault stations, it makes the overestimate of demand at Merryton even more surprising given a lot of the new developments appear to be within the vicinity of this station.
Figure 29 Larkhall Land Use Supply Comparison, 2013 and 2005
9.44. One of the reasons for the over forecast of demand at Merryton may be due to the station's location. It is not on the main Larkhall to Hamilton main road (A72) so it may actually it is quicker for some Merryton residents to drive to Chatelherault station which is directly off the main road and which has the advantage of a larger car park and is in a cheaper fare zone[38]. A further reason could be that whilst there has been new housing development in Merryton, the level is still lower than what was predicted in 2000, perhaps as a result of the recession.
ii. Park and Ride Schemes
9.45. Three out of the four new stations have Park and Ride facilities:
- Chatelherault – 100 spaces;
- Merryton – 80 spaces; and
- Larkhall – 200 spaces.
9.46. Park and Ride facilities are designed to provide easy access to city centres; people tend to make the majority of their journey by car before transferring to rail to avoid inner-city congestion and parking charges. They are therefore attractive to those living further away as well as local residents and, as a result, increase the catchment area of the station.
9.47. There was no evidence in the SPT Modelling Report that Park and Ride was part of the scheme specification during the appraisal process and consequently that the potential demand impacts of Park and Ride provision were factored into the assessment. Demand may have been underestimated at these stations during the appraisal which could explain the underestimate at Larkhall and Chatelherault but not the overestimate at Merryton.
9.48. However, research[39] suggests that, except in specific circumstances, there is no conclusive link between car parking provision and rail demand (as people generally switch from on-street parking). Any additional demand generated by the Park and Ride facilities may well therefore have been relatively small.
iii. Annualisation Factors
9.49. The SITM provided station demand on a daily basis. This was converted into annual demand by applying an annualisation factor.
9.50. The SPT Modelling Report did not detail the annualisation factors which were used in the demand forecasting. However, as per section 9.5, an annualisation factor of 283 is implied.
9.51. As set out in the Department for Transport's Transport Users Benefit Appraisal (TUBA)[40], recommended annualisation factors for Commute/Business and Leisure are 253 and 353. Weighting these factors as per the Scotland average journey purpose36 (70% Commute/Business and 30% Leisure) would result in an annualisation factor of 283 which is identical to that used in the SITM. The annualisation factor applied is therefore unlikely to be a key reason why there is a difference between the forecast and actual demand.
iv. Demand Ramp-up
9.52. Under PDFH guidance, it is standard practice to apply demand ramp-up factors to reflect that improvements take time to occur and for passengers to be aware of the service changes. For example, PDFH suggests a 70% factor be applied to the full single year demand estimate in year 1, 85% in year 2 and 95% in year 3 with all lags complete i.e. 100% of demand evident by the end of year 4.
9.53. It would have been appropriate to apply such factors to the demand forecasts prepared for the four new stations. However, it is not clear from the SPT Modelling Report whether this was done or not.
Service Specification
9.54. The forecasting in the appraisal made a number of assumptions about the service specifications which would be implemented when the scheme was constructed. These are defined in the SPT Modelling Report as:
- introduction of a half-hourly passenger service between Milngavie and Larkhall;
- extension of the Northern Suburban line to Anniesland to allow a full Northern service timetable to be introduced;
- removal of operational restrictions at Westerton to enable a 15 minute service frequency to be operated on the Milngavie section;
- improvement from 6 trains per hour to 8 trains per hour at Westerton and Anniesland, due in part to the extension of the Northern Suburban to Anniesland and the removal of operational restrictions mentioned above;
- an increase in frequency from 10 trains to 12 trains an hour between Hyndland and Partick;
- an increase in frequency from 4 trains to 6 trains an hour between Rutherglen and Partick on the Argyle Line in the off-peak; and
- a 15 minute service frequency at Blantyre and both Hamilton stations.
9.55. Comparison of this planned service specification and the actual service specification implemented shows that they are closely matched suggesting that there is little, if any, variation between what was forecast in 2000 and what materialised in reality. There have also been no significant changes to this service pattern since 2005. Therefore, variations in the service specification are unlikely to have been responsible for the differences between the forecast and actual demand.
9.56. It is difficult to single out any one key reason why there are differences between the forecast and actual demand for the four new stations. This is further complicated by the lack of documentation in the SPT Modelling Report to verify key modelling assumptions or to determine whether impacts such as demand abstraction and demand build-up factors were applied.
9.57. Exogenous factors however are the dominant drivers of rail demand and evidence suggests that actual economic and population growth have been higher than what was forecast in 2000, particularly in the strong period of growth between 2000 and 2007 before the recession. This difference is therefore likely to be one of the most important reasons why actual demand has been higher than forecast at three of the four stations.
9.58. Other more minor reasons for the difference may include not accounting for:
- Park and Ride facilities; and
- demand abstraction from existing stations.
It should also be noted that there is no 'right' way of modelling demand at new stations, and that every new station will have its own unique set of circumstances that may lend itself to a particular modelling approach or set of assumptions. Furthermore, influences on long-term rail demand are multi-faceted, including potentially volatile drivers such as economic growth. This makes it almost impossible to achieve an exact match between forecast and actual station demand.