Calculating GVA and accessibility impacts to assess possible Sheffield City Region Investment Fund schemes

Transcription

1 in collaboration with Calculating GVA and accessibility impacts to assess possible Sheffield City Region Investment Fund schemes Report to South Yorkshire Passenger Transport Executive DRAFT Minor revisions 21 September 2012

2 Document control Project: c69.2 (FLUTE) File FLUTE GVA access v doc. Version Date Main author Other author(s) Reviewer(s) 1 07/09/12 David Simmonds 2 08/09/12 David Simmonds 3 (minor revisions only) 21/09/12 David Simmonds Paul Hanson (AECOM), Andy Dobson, Emma Revill This report has been prepared by David Simmonds Consultancy Limited, as subcontractors to AECOM, for SYPTE. Any opinions expressed in the report are not necessarily those of SYPTE. Copyright David Simmonds Consultancy Limited Suite 14, Millers Yard, Mill Lane Cambridge CB2 1RQ England Tel dsc@davidsimmonds.com ` ii

3 SUMMARY This report has been prepared by David Simmonds Consultancy and AECOM for South Yorkshire Passenger Transport Executive. It considers how the GVA and accessibility criteria set for the Sheffield City Region Investment Fund can best be implemented, making use of the work already commissioned for the Sheffield City Region Integrated Model (SCRIM), which combines the existing strategic transport model SYSTM+ with a new tool, FLUTE, for forecasting economic and land-use change. We have recommended a conventional and practical approach to the calculation of GVA impacts, based primarily on changes in employment multiplied by earnings per job, but also taking account of productivity effects arising both from transport cost savings and from agglomeration effects. The intention in adopting this approach is to maximise consistency of assessment across different schemes and different types of scheme. The analysis required can be split into three stages: forecasting the employment changes; calculating the GVA impacts resulting from those employment changes; taking account of the agglomeration and transport cost effects. Employment changes and transport cost changes are standard outputs of the SCRIM model now under development. These outputs will be used in the agglomeration calculations, and are also required for the accessibility calculations. The accessibility calculations proposed use a standard formula closely related to that already used in FLUTE. All of the schemes which need to be assessed will need to be quantified in ways that can be input either to FLUTE, to SYSTM+ or to a combination of the two. For schemes that can be directly represented in model inputs, this will be very simple for example, a new section of road between two existing junctions will be quantified essentially by its type (single/dual carriageway, number of lanes), free-flow speed and length. At the other extreme, for schemes that cannot be explicitly input to SCRIM, quantification will be more complex. In some of these more complex cases an initial estimate of the direct employment impact will need to be made outside SCRIM. SCRIM will then model the indirect impacts (for example, through impacts on population location and on property development); doing this will help to ensure that different types of schemes are assessed consistently and fairly. A number of potential enhancements to FLUTE were identified in the April 2012 proposal, but were excluded from the first stage of work on FLUTE in order to accelerate the work programme. The report considers which of these would be essential or helpful to the assessment of proposals for SCRIM investments. Whilst our report is written very much in terms of employment gains, cost savings, and additional productivity, it must of course be kept in mind that some of these impacts may iii

4 turn out to be negative if a scheme is forecast to have unintended side-effects for example by causing an increase in congestion on critical links of the transport networks. ` iv

5 Contents 1 Introduction Objectives Report structure Terminology Schemes and criteria Schemes to be assessed Criteria for assessment GVA and how it can be calculated What is Gross Value Added (GVA)? How is it estimated? Proposed general method of calculation for scheme assessment GVA per job Additionality, deadweight, etc Accessibility calculations: design Introduction Design requirements Design recommendations Calculations by scheme type Introduction Transport schemes Regeneration schemes Energy schemes Environmental schemes Education schemes GVA impact calculations (i): employment changes GVA impact calculations (ii): agglomeration effects GVA impact calculations (iii): transport cost savings GVA impact calculations (iv): combined results Accessibility calculations Timing, discounting etc Introduction Model years and forecast period Multi-year outputs and discounting: GVA Accessibility Conclusions Calculating the GVA and accessibility measures Required additions to SCRIM Possible additions to SCRIM, required for full assessment of particular schemes 33 References v

6 Appendix A Accessibility calculations A.1 Introduction A.2 Averaging generalised costs over modes A.3 Calculating accessibility A.4 Finding district or other averages Appendix B Outline of SCRIM and FLUTE B.1 Introduction ` vi

7 LIST OF TABLES Table 2.1 Types of scheme: transport... 3 Table 2.2 Types of scheme: regeneration... 3 Table 2.3 Types of scheme: energy... 4 Table 2.4 Types of scheme: environmental... 4 Table 2.5 Types of scheme: education... 4 Table 3.1 FLUTE Employment Sectors... 8 Table 4.1 FLUTE Socio Economic Levels Table 5.1 Ability to consider schemes of type: transport Table 5.2 Ability to consider schemes of type: regeneration Table 5.3 Ability to consider schemes of type: energy Table 5.4 Ability to consider schemes of type: environmental Table 5.5 Ability to consider schemes of type: education LIST OF FIGURES Figure B-1 Overall structure of the model Figure B-2 Time-marching process Figure B-3 Actors and markets Figure B-4 Main model components Figure B-5 FLUTE Modelled Areas: Fully Modelled and Buffer vii

8 ABBREVIATIONS Abbreviation APPI BRT DELTA DETR DfT DSC FLUTE GDP GVA LUTI MBC NIC NTEM R&D SAF SCR SCRIF SCRIM SYITS SYPTE SYSTM+ TEMPRO TRRL WEBs WebTAG Meaning Assembly of Planning Policy Information bus rapid transit Land-use/economic modelling package developed by DSC Department of Environment, Transport and the Regions Department for Transport David Simmonds Consultancy Limited Forecasting the Interaction between Land Use, Transport and the Economy application of DELTA under development for SYPTE, Barnsley and Doncaster Councils gross domestic product gross value added land-use/transport interaction Metropolitan Borough Council National Insurance Contributions National Trip End Model (DfT) research and development Single Assessment Framework Sheffield City Region Sheffield City Region Investment Fund Sheffield City Region Integrated Model (= FLUTE+SYSTM+ integrated to form a LUTI model) South Yorkshire Intelligent Transport System South Yorkshire Passenger Transport Executive South Yorkshire Strategic Transport Model developed by AECOM for SYPTE Trip End Model PROgram (DfT) Transport and Road Research Laboratory Wider Economic Benefits web-based Transport Appraisal Guidance (DfT) ` viii

9 1 INTRODUCTION 1.1 Objectives This Report has been prepared for SYPTE by David Simmonds Consultancy Ltd (DSC) working as subconsultants to AECOM within the contract for the development and operation of FLUTE The SYPTE brief and instructions specify that the objectives of the Report are to produce a position statement of the method to model and appraise in an equal, consistent and transparent way the impact of housing, regeneration, environmental, energy and transport schemes based on criteria set by the primary objective and the two secondary minima of the Sheffield City Region Investment Fund (SCRIF) Single Assessment Framework (SAF) The report is intended as a key input to workshops being held during September 2012 to determine the best way of calculating Gross Value Added (GVA) for each scheme type Material supplied to SYPTE by Doncaster MBC and by KPMG has been taken into account in preparing this document. 1.2 Report structure The report is structured as follows: Chapter 2 summarises the types of schemes which have to be considered, and the agreed criteria against which they have to be assessed; Chapter 3 discusses how GVA impacts may be estimated in general, and in broad terms how we propose that they should be calculated in the present case; Chapter 4 discusses how the accessibility criteria could be implemented, and how we propose that they should be operationalised; Chapter 5 reviews whether and how the currently commissioned modelling will provide the information needed to calculate the GVA estimates and accessibility effects, for each type of scheme, and explains our suggestions for dealing with the types of schemes where the presently commissioned modelling is insufficient or inappropriate; Chapter 6 considers the time dimension whether the assessment should be based on analysis for single target year or a sequence of years, whether discounting is required, etc; and Chapter 7 draws some conclusions about the use of FLUTE in providing the information required for SAF. 1

10 1.2.2 Appendix A sets out the mathematics of the proposed accessibility calculations A.4.2 provides an overview of the FLUTE model The report does not set out to review the specified criteria, though it does offer some alternatives as to how the specified criteria could be implemented. 1.3 Terminology We have tried to use the word assessment to refer to the process of calculating and applying the chosen criteria to avoid any confusion with WebTAG etc which might result from using the word appraisal. ` 2

11 2 SCHEMES AND CRITERIA 2.1 Schemes to be assessed The types of schemes that may need to be assessed have been discussed in various meetings and exchanges over recent weeks. They can be grouped into five headings: transport; regeneration energy; environment; and education The possible types under each heading are listed, with the examples previously raised, in the following tables. Table 2.1 Types of scheme: transport Type Examples Highway infrastructure Rail Tram Bus (including BRT) Active modes (this would need to be on a grand scale) New roads, widening, junction improvements etc. New lines, station improvements, rolling stock, changes in time table, electrification New lines, rolling stock, changes in time table Bus priority, vehicle investment, smart ticketing Traffic management SYITS type investment 1 Airports Canals and waterways Large scale infrastructure investment eg segregated routes Terminal improvement, car parking, enhancing airport capacity Investment in mooring sites Source: file Scheme Types_ fq DSC AECOM v2.docx, sent with brief for this report Table 2.2 Types of scheme: regeneration Type Housing Examples New development, refurbishment 1 South Yorkshire Intelligent Transport System (SYITS) is a Local Transport Plan and European Objective 1 funded project to provide a centrally controlled traffic management and information system for South Yorkshire (source: 3

12 Type Examples Commercial development New development, refurbishment Site remediation and preparation Site demolished and prepared for redevelopment Investment in the rural economy New development, refurbishment IT infrastructure including Broadband Homes and businesses given access to high-speed internet Urban design Improvement to commercial/residential space Source: file Scheme Types_ fq DSC AECOM v2.docx, sent with brief for this report Table 2.3 Types of scheme: energy Type Examples Energy generation (e.g. sub stations) Energy generation (biofuel stations, wind farms) Efficiency (e.g. lighting, insulation) Energy distribution (e.g. Combined Heat and Power network) New substation / upgrade Production of alternative fuel Replacement of bulbs to low-energy Energy efficient process Source: file Scheme Types_ fq DSC AECOM v2.docx, sent with brief for this report Table 2.4 Types of scheme: environmental Type Examples Flood defence Land decontamination Green infrastructure (parks and open spaces) Reservoirs Improved drainage Removal of waste Creation of green space Secured water supply Source: file Scheme Types_ fq DSC AECOM v2.docx, sent with brief for this report Table 2.5 Types of scheme: education Type Examples R&D facilities New R&D facilities; reduce a skill gap in growth sector Source: file Scheme Types_ fq DSC AECOM v2.docx, sent with brief for this report 2.2 Criteria for assessment We are informed by the brief 2 for this Report that the primary objective for the SAF has been agreed as follows: Increase in GVA in SCR per pound of whole life cost spent from the fund The agreed two secondary minima are: 2 from Dr Qaddir, SYPTE, Tue, 21 Aug :57:38 ` 4

13 Average connectivity/accessibility to employment for the people in most need of support improves by at least the average for the whole of SCR. The average % increase in employment connectivity/accessibility for each district across SCR is no less than 50% of the average % improvement for the whole of the SCR The definition of GVA, the ways in which the increase in GVA resulting from a proposed investment can (in theory) be estimated, and recommended approach for use in SAF, are set out in Chapter 3. The equivalent discussion for the accessibility measures are presented in Chapter 4. The practical details of how these calculations can be implemented using the commissioned models (or additional work, where necessary) are described in Chapter 5. 5

14 3 GVA AND HOW IT CAN BE CALCULATED 3.1 What is Gross Value Added (GVA)? GVA measures the contribution to the economy of each individual producer, industry or sector in the United Kingdom 3 - or any part of it It may be helpful to note the relationship between GVA and GDP, as GDP is the more widely-used measure of national economic performance. The relationship can be defined as: GVA (at current basic prices; available by industry only) plus taxes on products (available at whole economy level only) less subsidies on products (available at whole economy level only) equals GDP (at current market prices; available at whole economy level only) Or, in summary: GVA + taxes on products - subsidies on products = GDP 3.2 How is it estimated? GVA can in general be estimated using either production GVA(P) or income GVA(I) approaches (for more formal definitions see ONS, 2010, p53) The 'production' approach to estimating GVA measures the value of the outputs produced (goods or services) less the value of the inputs consumed in producing those outputs The income approach to estimating GVA measures the incomes earned by individuals (gross wages and salaries) and corporations (for example, profits) in the production of outputs (goods or services) GDP can also be estimated using the expenditure approach: total expenditure on final consumption of goods and services (ie including exports, but excluding intermediate consumption where goods and services are consumed by firms in order to produce other goods and services), less imports Official estimates of GDP use all three approaches, which are then balanced in order to arrive at a single figure. 3 ` 6

15 3.3 Proposed general method of calculation for scheme assessment We propose that for assessment we should use a common approach across all schemes as far as possible We propose that this should be based in effect the income method, calculated primarily as employment multiplied by average annual earnings, ie taking earnings as the best practical proxy for GVA. As KPMG pointed out in Appendix A of their note on implementing the GVA and accessibility criteria, this is the most common approach for impact assessment. The advantage of this is that it links directly to FLUTE, where employment (by sector and zone for each forecast year) is a standard output This means that as far as possible the GVA measure would be based on changes in jobs * earnings per job, adjusted for changes in productivity The changes in jobs can be disaggregated by sector (activity, in FLUTE terms) provided that the disaggregation fits with the FLUTE classification of employment activities, and provided that we have sufficient information to estimate different values of earnings per job by sector It should be kept in mind that FLUTE can both relocate jobs from adjoining regions into the SCR, or vice versa, and it can forecast higher (or lower) employment growth overall, and these impacts may vary by sector. The impacts of a scheme being assessed are forecast as a stream of impacts over time, not just as a single year value (this is discussed further in Chapter 6). The changes in jobs going into the GVA calculation will therefore be a set of numbers, not just a single value. FLUTE also forecasts changes in total population (eg because greater job opportunities in the SCR will typically result in fewer people leaving the region and more moving in) note that as a result, a 1% increase in GVA arising from a scheme will not necessarily represent a 1% increase in GVA per capita Changes in productivity per job will stem from agglomeration effects (using the DfT methodology 5 ); changes in transport costs accruing directly to SCR businesses (ie the costs, including staff time, of business travel and goods deliveries) The agglomeration effects will need to be applied as changes in earnings per job, but for all jobs, not just for the gains or losses resulting from the scheme being assessed Changes in transport costs could also be converted into values per job. However, this seems an unnecessary complication when values per job are only being 4 FLUTE does not identify individuals and therefore it is impossible to say precisely which jobs are forecast to be filled by people who, in the absence of the scheme, would have lived outside the SCR. It is however possible to estimate how many jobs have been filled by members of households that have moved to the SCR, or decided not to move away from the SCR, as a result of the scheme. There are of course flows of workers into and out of the SCR, so not all jobs in the SCR are filled by SCR residents. 5 see WebTAG Unit

16 considered in order to arrive at regional totals. We therefore propose simply to take the net saving in such costs accruing to SCR businesses and to add that to the total GVA gain resulting from changes in jobs and in agglomeration. This will be done using the cost savings per trip (for business travel and for goods movement) multiplied by the numbers of trips in the reference case. 3.4 Employment categories and GVA per job FLUTE will output employment by sector, for the following sectors: FLUTE Activity Description 41 Adult Education 42 Agriculture, Forestry and Fishing- Non Manual 43 Agriculture, Forestry and Fishing - Manual 44 Construction 45 Distribution 46 Factory Workshop Services 47 Health 48 Higher Education 49 Hotel and Accommodation 50 Manufacturing: Production and Administration Non Manual 51 Manufacturing: Production and Administration Manual 52 Mining Administration and Production Non Manual 53 Mining Administration and Production Manual 54 Office based services 55 Other businesses: non office based 56 Other businesses: office based 57 Primary and Secondary Education 58 Recreation and Sports 59 Restaurants and Bars 60 Retail 61 Transport 62 Utilities Table 3.1 FLUTE Employment Sectors There is a question of whether the GVA/job values used should take account of existing spatial differentials within the SCR. These are apparent in the data which DfT supplies for use in Wider Impact calculations (though these are described as GDP/job, not GVA/job). Using DfT values by local authority would mean that GVA gains could result from moving jobs across district boundaries within the ` 8

17 SCR, eg moving one manufacturing job from Sheffield to Barnsley in 2021 would add about 3000 to SCR GVA (all other things being unchanged) This kind of differential assumes that there are fixed differences between the industries in each local authority area, and could give rise to implausible differences (eg that new jobs in a new factory unit would be more productive at one end of the unit than the other, if the unit happens to straddle a boundary). We propose that an SCR average of GVA per job to avoid this. This value will of course change over time in line with the overall economic scenario adopted If a scheme involves investment in a project which would attract a very specific category of industry, and if it is clear that it will be successful in this, then it would be appropriate to treat those jobs separately, ie apply an industry- (and if necessary location-) specific GVA/job to the jobs directly generated by the project subtract those jobs from the impacts considered in FLUTE, and use the standard GVA/job for the remaining impacts on jobs (eg changes in jobs induced by population movements). 3.5 Additionality, deadweight, etc Official advice on the appraisal of government interventions (primarily the Green Book - HM Treasury, 2011) is much exercised with the issue of additionality, ie with establishing that an investment will generate benefits greater than those which would have occurred if the amount invested by the public sector had been left in the hands (or pockets) of tax payers, whether individual or corporate. Whilst the Green Book advice does not apply in full to SCRIF (not least because SCRIF is seeking specifically to benefit the SCR, not the whole economy), it should nevertheless be considered Page 54 of the Green Book specifies that: The net benefit of an intervention equals the gross benefits less the benefits that would have occurred in the absence of intervention (the deadweight ) less the negative impacts elsewhere (including displacement of activity), plus multiplier effects. For the present purposes we should note that the modelling approach is comparative to assess an intervention, we must run a reference case without the intervention and then a do-something test including the intervention. The differences between the two tests are the impact of the intervention; there is no need to make any further adjustment for deadweight ; the model and the investment criteria fund are both spatial in nature given the criterion of maximising GVA within the SCR, displacement effects outside the SCR are irrelevant. Displacement effects within the SCR (eg jobs relocating from Sheffield to Doncaster, or vice versa) will be forecast and will automatically be included in the GVA calculations; multiplier effects due to income effects are included in the modelling and do not require further analysis. 9

18 3.5.3 The comparison between model outputs does not have to be limited to comparing tests that differ only in including/excluding a single intervention; the modelling approach can readily be used to consider packages of interventions, and it may well be desirable to do those if there is an expectation that different schemes will interact with one with another (whether this is a concern about possible conflict, or a hope for complementarity). At the same time, it is highly desirable to test the components of packages in order to understand what they are each contributing to the overall results; if it is impractical (too time-consuming) to run the model system in full for each component, it may well be straightforward to devise partial model runs which will be sufficient to identify the relative contributions of each package component. Quick partial runs can also be carried out to to vary the assumptions under which the impacts are carried out, eg by suppressing induced development effects. Our experience is that additional model running to test the components of packages and of responses, and the associated reporting, is generally very valuable and has a high value-for-money in clarifying the discussion of the results Leakage is another issue. Some aspects of this can be tackled (but may require further work on the modelling) for example it should be possible to exclude travel time and cost savings for businesses/shippers outside the SCR by counting savings only for trips produced within the SCR. However, some (possibly many) of the GVA gains generated within the SCR may be captured in the form of increased rents for commercial property; for many properties, these increased rents will form increased income for investors, many of whom will be outside the SCR (and some of whom will be outside the UK). This does not invalidate the calculation of the GVA increases it is just a reminder that the additional value added in the SCR may not remain within the SCR The GVA calculations are based on where people work, not on where they live. This is consistent with standard uses of GVA as an indicator. Additional calculations could be done to trace the consequences for household incomes by place of residence, using the commuting data in SCRIM, but this would be going beyond the criteria specified. ` 10

19 4 ACCESSIBILITY CALCULATIONS: DESIGN 4.1 Introduction As a reminder, the agreed accessibility criteria are that: average connectivity/accessibility to employment for the people in most need of support improves by at least the average for the whole of SCR; and the average % increase in employment connectivity/accessibility for each district across SCR is no less than 50% of the average % improvement for the whole of the SCR There is no single definition of accessibility. The following section considers what is required from the accessibility measure in this context. From this we develop an appropriate definition and specify how it can be calculated. 4.2 Design requirements The criteria imply that we are only concerned with accessibility of SCR residents to employment. This Chapter therefore focuses on accessibility to employment, even when quoteing more general research. Other aspects of accessibility, such as residents accessibility to services or businesses accessibility to labour, to suppliers and to markets do not need to be explicitly assessed, though it should be kept in mind that they will be taken into account both within FLUTE itself (eg access to labour is one of the influences on employment location) and hence, through the employment forecasts, in the GVA calculations; and in calculation of agglomeration effects (these involve further accessibility calculations see section 5.8) The criteria also imply that it must be possible to calculate average changes in accessibility for districts and for the SCR as a whole; it must be possible (and meaningful) to express those changes as percentage changes, with a positive percentage change representing an improvement; we need to agree a category of people in most need of support ; and we must be able to make meaningful comparisons between the average change in accessibility to employment for this group (across the SCR) and the average change in accessibility to employment for the population as a whole (again, across the SCR). 11

20 4.2.3 The point about percentages means that for percentage changes to make sense and have the usual meaning - accessibility needs to be calculated as a continuous measure in which zero means no accessibility to employment. Then, in comparing two situations where there is at least some accessibility to employment, we can interpret a 100% improvement as twice as good, and so on (though in practice we would expect to be dealing with very much smaller changes). This is a normal and intuitive use of such a measure, but not all accessibility measures actually meet this requirement many of those used in FLUTE, for example, are cost-based measures where higher (more positive) values indicate worse rather than better accessibility In assembling data to calculate a measure of accessibility, it is generally necessary to identify the travel purpose to which the measure relates the group of people whose accessibility is being measured the modes that should be taken into account the times of day or days of the week that need to be considered In the present case, we need to consider one purpose travelling to and from work two different groups - people in most need of support and the working (or potentially working) population in general; in principle, all the modes of transport which might be used to travel to and from work, though we discuss the treatment of car travel further below; the times of day at which workers may need to travel to and from work. SYSTM+ distinguishes times of day; the interface to FLUTE, includes a process to average costs by purpose across time periods; we therefore do not need to consider this explicitly except possibly in the case of needing to assess a proposal which would significantly change either the hours (or days of the week) when people are expected to work, or the hours/days when transport facilities/services are available to carry them.) There is a wide range of possible accessibility measures which we have reviewed on various occasions, in particular for the former DETR (DSC, 1998). In that work, developing ideas drawn from an earlier review by Jones (1981), we identified six requirements for accessibility measures in policy and scheme assessment. These are that the chosen accessibility measure should take account of the location and characteristics of the individual or type of person should take account of the location and characteristics of opportunities for relevant activities should take account of the connecting transport systems should improve or remain constant if the number of opportunities for an activity increases anywhere ` 12

21 should improve or remain constant if the relevant travel opportunities by any mode become quicker or cheaper, or a new mode is added should be unaffected if changes are made to irrelevant travel opportunities (eg if the individual is not able to use the mode that changes) We identified five types of accessibility measurement (again extending earlier work by Jones): contour measures; Hansen measures; revealed value measures; time-space measures; and utility-based measures Contour measures consider either the number of opportunities (jobs, in this case) within a particular threshold or contour, eg the number of jobs that can be reached within 45 minutes travel from the location whose accessibility is being measured; or the time (or cost, or generalised cost) from the location whose accessibility is being measured within which there is a chosen number of opportunities, eg 10,000 jobs The critical feature of contour measures is therefore that they involve a threshold number, whether in terms of the chosen transport characteristic or in terms of the number of opportunities. Anything beyond that threshold (eg the number of jobs that are 46 minutes away, or the time taken to reach the next 1,000 jobs) is irrelevant and does not affect the result. These measures therefore do not satisfy the fourth of our criteria and it is difficult in practice to justify particular threshold values Hansen measures (named after Hansen, 1959) are those where in measuring accessibility for each zone i, the measure of accessibility is the sum over all the possible work zones j of the number of job opportunities in work zone j, weighted by a value in the range 0 to 1 which measures the relative acceptability or value of job opportunities in zone j for residents of zone i; the weighting value being one which gradually decreases as the generalised cost of commuting between i and j increases The weighting values must have the value 1 if generalised cost is zero (ie if the job opportunities at j are perfectly accessible to residents at i, without any in convenience or expenditure of time or money) This is essentially the type of measure proposed in the KPMG paper Revealed value measures are in effect proxies for accessibility the argument in some situations being that since accessibility, although widely discussed and used, cannot be observed, it is better to use a related, observable value such as rent (or 13

22 more likely, to use observed changes in rent as a proxy for changes in accessibility). In the present work, we are by definition forecasting future changes for which we cannot have observed data on any variables, so the argument for using observable data does not apply. One could perhaps make a case for using forecast rent impacts rather than forecast accessibility impacts as the means of considering the spatial impact of proposed investments, but that is out of scope Time-space measures are those which consider whether people can get to particular opportunities by a particular time of day. This is usually applied to specific opportunities such as reaching a secondary school by 09h00 or a hospital out-patient department by 10h00. They are often used in considering the social impacts of changes in public transport services. Since they require detailed public transport data (timetables rather than just speeds and average frequencies) they are only really suitable for short-term analysis, and hence irrelevant here NB in this exercise we don t necessarily have to use behavioural measures it would be reasonable to include a normative element which could say, for example, people shouldn t commute more than two hours each way, and accordingly exclude any work opportunities which are more than two hours one-way travel from the residents whose accessibility to employment is being measured Utility-based measures are generally measures of the expected utility, or disutility, or generalised cost, of the travel that individuals are expected to make from each zone, given the job opportunities available and the characteristics of the modes of transport for reaching those opportunities. Changes in such utility measures resulting from transport and/or land-use change are in effect measures of changes in consumer surplus, and can be used in appraisal. Depending on the exact formula used, the utility-based measure can be a transformation of the Hansen measure, converting it from a measure of opportunities weighted by a function of generalised cost to one in generalised costs weighted by opportunities. These measures are widely used in DELTA, and hence in FLUTE, but they have the disadvantage that a value of zero does not necessarily have any special significance, or if it does, it means perfect accessibility (all job opportunities can be reached without incurring any travel time, cost or inconvenience). This is not a problem if the way the measures are used always relies on absolute differences (as it does in DELTA), but it means that the utility-based measures do not meet our requirements for a measure where percentage changes are intuitively meaningful This leaves us in the position that only the Hansen-type measure is suitable for the present purpose. 4.3 Design recommendations We therefore recommend use of a Hansen-type measure (similar to that suggested in the KPMG paper), with a multi-modal measure of the generalised cost of travel By adopting a measure which is measured in terms of jobs weighted by their accessibility to the zone being considered, we have a measure where the values will always be zero or positive; a value of zero (if it occurred) would mean that no jobs could be reached at all from the residential zone in question; ` 14

23 higher values are better; percentage changes are meaningful eg if an intervention increased jobs in every work zone by 10%, the accessibility measure would increase by 10% for every residential zone The advantages of a Hansen-type calculation where the value or weight on a job decreases gradually with generalised cost, rather than a contour-type calculation where jobs are given a weight of 0 or 1, are that we don t have to justify a specific threshold or contour; we don t have the risk of unstable results due to tiny transport changes bringing a large employment zone just inside the threshold The weighting of generalised costs over modes will use a logsum formula, in order to comply with the Jones conditions listed earlier. This is a standard function in transport modelling and analysis, with a good basis in theory about individual behaviour. This will use data from SYSTM+ together with coefficients for alternative-specific constants, representing preferences which are not described by the generalised costs of travel; sensitivity to utility differences (lambda or spread parameters) The coefficients can by default be taken from those used in FLUTE, though further refinement would be possible FLUTE will output employment and working-age population by zone disaggregated into the following socio-economic levels (SELs, defined as groups of occupation levels): FLUTE SEL Contains 2001 Census of Population SOCs: 1 1 Managers and Senior Officials 2 Professional Occupations 2 3 Associate Professional and Technical Occupations 3 5 Skilled trade Occupations 4 Administrative and Secretarial occupations 6 Personal Service occupations 7 Sales and Customer Service Occupations 8 Process, Plant and Machine Operatives 4 9 Elementary Occupations Table 4.1 FLUTE Socio Economic Levels Again, changing the inputs to SYSTM+ and/or FLUTE itself may affect both SCR totals and their distribution across zones and socio-economic levels We suggest that 15

24 the most needy category should be taken as the least-qualified socioeconomic level), and that accessibility for this category should be calculated considering walk/cycle and public transport only ie excluding car travel as a means of getting to and from work the general accessibility measure should be based on the working population as a whole, and should consider all modes of transport The first suggestion is not asserting that people in the elementary occupations category (SEL 4) are necessarily needy, still less that they are necessarily the most needy. The suggestion is made on the basis that in so far as the most needy are people who might work and for whom accessibility to employment is relevant, low-skill and unskilled jobs are likely to be the most relevant; the distribution of residents working or likely to work in such jobs is then an appropriate basis for averaging accessibility scores over zones District and SCR averages will be weighted averages of the accessibility values by zone, the weights being the relevant number of working-age residents (ie workers and potential workers) in each zone. (Note that when considering the longer-term impacts of a proposal, the impacts may modify the distribution of residents as well as the generalised costs of commuting and the location of employment. This is good it means that if the impacts will include residents moving closer to work opportunities, that will appear as an accessibility improvement.) An alternative approach to measuring accessibility for the most needy category would be to consider as most needy only working-age persons (in total, or just the lowest SEL) living in areas of significant multiple deprivation, according to the Government Index of Multiple Deprivation (IMD). These could be defined as, for example, the 10% most deprived areas in the SCR. The criterion related to accessibility for the most needy would then in effect be that accessibility to employment for residents in those areas should improve by at least the average improvement The advantages of this approach would be that this criterion would bring SCRIF more into line with other initiatives which focus interventions on the basis of IMD scores; it is likely to be more sensitive to the spatial distribution of effects, because it will only look at a sample of zones There are also disadvantages, the first two related to the second of the above advantages: it might be excessively sensitive (ie highly likely to reject otherwise reasonable schemes); depending where the most deprived areas are, it might well be biased in favour of particular local authorities; it also assumes that the areas of deprivation are constant over time, unless a lot of much more detailed modelling is undertaken to forecast changes in all ` 16

25 the variables that go into the IMD. Whilst deprivation is highly persistent it is not constant We suggest that the merits of this alternative (compared to the basic proposal of defining the lowest socio-economic level as representing most needy ) should be discussed further. 17

26 5 CALCULATIONS BY SCHEME TYPE 5.1 Introduction This Chapter takes each of the scheme types and considers whether the currently commissioned models will provide appropriate inputs for calculating the specified GVA and accessibility effects. Then if they will do so, it outlines how the models will do so, ie it outlines the logical connections which will take the inputs describing a particular intervention and how these will produce the models outputs which become the inputs to the GVA calculations; if not, it sets out what it is proposed should be done instead, both in the short term (using the models as commissioned) and in the slightly longer term (with possibility of model enhancements and extensions). 5.2 Transport schemes SCRIM is (or will be) a full land-use/transport interaction (LUTI) model and as such it is intended that the land-use/economic component, FLUTE, will forecast how the quantity and location of employment, by sector, will be affected by transport interventions, provided that these are within the scope of the transport model itself (SYSTM+) The following table identifies which of the transport scheme types are within the scope of SYSTM+. Note that we are assuming that any changes will be significant, eg changes in timetable is taken as meaning a significant increase in speed and/or frequency, rather than detailed changes in the exact departure times of existing services. Table 5.1 Ability to consider schemes of type: transport Type Examples Within scope of SYSTM+? Highways Rail Tram New roads, widening, junction improvements etc. New lines, station improvements, rolling stock, changes in time table, electrification New lines, rolling stock, changes in time table Yes Yes for passenger operations (infrastructure and rolling stock changes will need to be converted into changes in frequency, speed or comfort) Yes (infrastructure and rolling stock changes will need to be converted into changes in frequency, speed or comfort) ` 18

27 Type Examples Within scope of SYSTM+? Bus (including BRT) Active modes (this would need to be on a grand scale) Traffic management Airports Canals and waterways Bus priority, vehicle investment, smart ticketing Large scale infrastructure investment eg segregated routes SYITS type investment Terminal improvement, car parking, enhancing airport capacity Investment in mooring sites Yes (infrastructure and rolling stock changes will need to be converted into changes in frequency, speed or comfort) Yes if time of travel is affected. Market research on perceptions of quality, safety etc may be required (or may be derived from previous work elsewhere) Yes if the effects of SYITS-type investment or similar can be converted into increases in road speeds and/or road capacities No: air travel/transport, and the possible impacts of improvements on economic interactions between SY and the rest of the world, are outside the scope of the commissioned model. Airport car parking is not considered in SYSTM+ See below. No: currently no consideration of waterborne transport in SYSTM+. A 'port' zone could be developed to reflect increased freight activity; this is not within the specified scenario. See below, and also Source: own assessment of model capabilities The points about converting infrastructure changes etc into changes in speed, frequency etc are standard aspects of transport modelling process For all the cases where the possible investment is within the scope of the transport model, then SYSTM+ will calculate the generalised costs (for all modelled modes and journeys) in the with-scheme situation, for the approximate year of opening; it will also calculate the traffic volume by zone in the with-scheme situation; the generalised costs will feed into FLUTE, which will typically forecast increases in employment, over time, in the zones where generalised costs improve, or where they improve most. This will typically be a mixture of redistribution from zones where there is little or no improvement to those with most improvement, and net gains if the accessibility of the area as a whole improves There may be positive feedback effects (over time) resulting from the influence of the industries directly affected by the improvement on suppliers or customers, even if these are not directly affected; 19

28 the influence of employment change on population if more people are attracted to the area by the improved work opportunities, additional service employment will typically follow; additional employment floorspace being provided in response to increasing demand (if planning policies allow this extra development) There may be negative feedback effects (over time) resulting from additional congestion, whether this is caused by the initial employment response or the feedback effects listed above. Households are also affected by the environmental effects of traffic, so a road improvement can actually have negative effects if the nuisance value of the additional traffic exceeds the positive value of the improved accessibility For these within-systm+ types of schemes, it will be straightforward to calculate changes in GVA/job using the DfT-recommended approach to estimating agglomeration effects (this will require the DELTA WEBs software to be implemented these are one of the enhancements already mentioned in discussion) Airports, waterways and rail freight. SYSTM+ is designed to look at the major components of the transport system for movement of people and goods within and around the city region, and as such does not consider either air or water transport, or rail freight. To assess the GVA implications of investment in facilities for these, it would be most efficient to carry out an additional analysis, outside SCRIM, to assess the impacts on jobs in transport-using sectors across the SCR eg to what extent (if any) would manufacturing employment in the SCR grow as a result of the investments (for passenger air travel, this would generally involve forecasting first what additional air services would be offered by airlines as a result of the investment) jobs at the airport, dock or whatever possibly including (for example) additional jobs in aerospace industries which would need the airport location; as ever, it would be necessary to avoid double-counting and to identify the net impacts to the SCR. For example, airport jobs within the SCR gained at the expense of other UK airports would be additional; if expansion of the SCR airports resulted in loss of employment at SCR railway stations, only the net gain could be counted) FLUTE would then be able to forecast the distribution of additional jobs due to extra growth in transport-using sectors (the first of the above); to forecast impacts on numbers and distribution of residents, and any multiplier effects on employment; in combination with SYSTM+, to calculate the changes in accessibility due to any of the categories of additional jobs mentioned, taking account of ` 20

29 potential changes in congestion due to other generated traffic, eg movements of goods, staff, passengers and others to and from the airport If it is going to be important to assess proposals for SCRIM to invest in transport schemes aimed at improving links with the rest of the UK, it would be appropriate to proceed with the development of the top-level national spatial-economic model as originally proposed Note that all of the schemes which can be specified as inputs to SYSTM+ will, at a minimum, require SYSTM+ to be run in the year in which the scheme opens. The most thorough analysis would be done by running the full SCRIM system from that point forward (ie FLUTE in one-year steps and SYSTM+ in the transport model years); this will capture both any change over time in the value of the scheme (for example, due to changing values of time; also, public transport schemes may become less valuable over time because growing car ownership makes them less relevant to users, or they may become more valuable because growing road congestion makes them more relevant); any feedback effects on congestion, and hence on travel time savings, due to induced land-use effects If circumstances make it necessary, a quicker but less thorough analysis could be done omitting the later SYSTM+ runs In the majority of cases below, the scheme would be specified as an input to FLUTE rather than to SYSTM+. The need to run SYSTM+ in the do-something case would be mainly to look at effects on congestion (ie more or less congestion compared to the reference case). Usual practice would be to run a FLUTE-only test first (keeping the travel costs the same as in the reference case) and then to decide whether a full test interaction with SYSTM+ is necessary. There is of course an element of risk in this, in that involves estimating whether the transport model would produce significant changes in congestion, rather than running the transport model to find out. 5.3 Regeneration schemes These are by definition land-use schemes, so they are likely to be modelled by changing inputs to FLUTE rather than SYSTM+. The following table summarises where/how this is possible. Table 5.2 Ability to consider schemes of type: regeneration Type Examples Within scope of FLUTE? Housing Commercial development New development, refurbishment New development, refurbishment Yes via inputs for quantity and quality of housing see below Yes via inputs for quantity and quality of any or all of the commercial floorspace types represented in FLUTE (retail, office, industrial, warehousing, hotels & other accommodation, leisure) 21

30 Type Examples Within scope of FLUTE? see below Site remediation and preparation Investment in the rural economy IT infrastructure including Broadband Urban design Site demolished and prepared for redevelopment New development, refurbishment Homes and businesses given access to high-speed internet Improvement to commercial/residential space Yes via inputs for permissible development see below Yes if the proposed development or refurbishment is of housing/commercial floorspace (as above). Otherwise no see Indirectly, yes see Yes see Source: own assessment of model capabilities As the table shows, many of the possible interventions in this category can be represented directly by changing the planning policy or exogenous development inputs to FLUTE (planning policy inputs being those which specify what is allowed to be built (if there is demand for it), and exogenous development inputs those which specify that something is definitely going to be built). These inputs are specified by floorspace type, and work through to employment as follows: increases in the amount of development permitted will, if they are taken up (ie if there is demand from developers to build the higher quantities now permitted), result in greater amounts of housing or employment floorspace (of different types) in particular zones; additional exogenous development inputs will definitely increase the supply of the chosen type(s) of development 6 ; the quality of floorspace (new or existing) can also be modified; the changed supply and/or quality of floorspace will influence the location of households and jobs, and potentially the total number of jobs by sector in the SCR (changes in the number or distribution of households may influence jobs both by changing the supply of labour (and hence changing the relative advantages of different locations for firms) and by generating demands for services (public and private)) There may be positive feedback effects (over time) resulting from (in particular) the influence of employment change on population, or of population change on employment; and 6 It should be kept in mind that the model is forecasting change through time and that (generally) there will always be development taking place and being occupied; development/regeneration interventions representing possibly SCRIF investments will be additional to these. This is why it is important that assessment should be based on comparing the with intervention results against the without intervention results, not on looking at the change over time in any one model run. ` 22

31 there may be negative effects (over time) resulting from increased congestion It will also be possible, if required, to model the effect of investment which reduces the cost of particular kinds of development in particular parts of the City Region (ie a subsidy to capital expenditure). This would be done by reducing the cost of development, instead of adding the development itself or changing the quality of development Other investments in the rural economy: if this involved development or refurbishment of space used as offices, factories, etc, that is located in a rural setting rather than an urban one, then it can be modelled in the same way as any other development or refurbishment. If it involves floorspace types that are not represented in the model, eg agricultural buildings, then it will be necessary to carry out a pre-model of the expected effects, in terms of either cost savings or employment gains, and to introduce those (if significant) into the model to examine the further consequences. (In the following paragraphs and sections, this approach is applied to several other types of schemes that cannot be directly introduced into the model.) Note that apart from the agriculture etc sectors (see Table 3.1) the model does not have separate urban and rural economies IT infrastucture: IT provision does not directly enter into any part of SCRIM. Some prior analysis would need to be carried out. We would suggest that if the effect was expected to be diffused across the SCR and to represent only a very small change in business costs, it would be sufficient simply to estimate that cost saving (ie the GVA impact would be the expected increase in business profits resulting from that cost saving); if it was expected to have a localised and larger effect, that effect could be introduced into the model as a cost reduction for appropriate sectors in the relevant zones, or possibly as an increase in the attraction of those zones for which they would be willing to pay a higher rent the sectors in question would then respond to the cost or attraction signal (and their responses would affect other sectors, both through competition for the available floorspace in those zones and through all the other model linkages); if the investment was aimed not just at creating an advantage for businesses in general, but at creating a quite exceptional level of services which would attract or generate activities not currently located in the SCR, then it would be necessary to carry out an ad hoc analysis of the direct effects (ie how many new-sector jobs would be generated, and what would be the associated GVA) and then to input these effects to SCRIM, using SCRIM to examine the impacts on the rest of the urban economy, its transport system etc (in much the same way as for major air or water transport changes see section 5.2 above) The pre-model analysis would need to allow for the take-up of new possibilities, eg any estimate of willingness-to-pay would need to be an average for a modelled sector, not just the willingness to pay of the households or firms which would actually adopt the improvement. 23

32 5.3.8 Urban design: the consequences of improvements in the quality of the public realm due to better urban design or other changes can be introduced into FLUTE either as quality improvements or as generalised cost changes. Initial analysis is required to quantify the effects of the proposals under consideration and to identify which groups of people or businesses they will affect. This can be done at various levels of detail, ranging from a limited exercise based entirely on previous research and professional judgements up to a major market research type study typically using stated preference methods to establish the value which individuals would put on different changes. We have considerable experience in such exercises, starting from work for DfT in reported in Whitehead et al (2006) Once the inputs have been specified, FLUTE (or the full SCRIM) is run; the direct impact of the changes will depend on the value of the scheme to each person it affects, and on the numbers of people affected by it. Indirect impacts will follow from all the linkages in the model for example, an increase in demand to locate in a particular zone may result in greater or more rapid development in that zone, but the ensuing increase in residents and/or jobs may generate more congestion, which will suppress demand and limit further growth. 5.4 Energy schemes Table 5.3 Ability to consider schemes of type: energy Type Examples Within scope of FLUTE? Energy generation (e.g. sub stations) Energy generation (biofuel stations, wind farms) Efficiency (e.g. lighting, insulation) Energy distribution (e.g. Combined Heat and Power network) New substation / upgrade Production of alternative fuel Replacement of bulbs to low-energy Energy efficient process Source: own assessment of model capabilities Indirectly, yes see below Indirectly, yes see below Indirectly, yes see below Indirectly, yes see below Several of these possibilities could be indirectly represented after some initial analysis to quantify their impacts in terms of cost reductions and/or a greater willingness-to-pay rents in particular locations. As for broadband schemes (see above), those quantified effects could then be introduced into the model and the impacts on location and other variables would follow That kind of analysis could be applied to the impacts of providing alternative forms of energy generation, although it is not obvious that this would have any significant effect in terms of reducing energy costs, or that such reductions would be particularly localised For energy efficiency measures, in particular, the issue of take-up would need to be considered in the pre-model analysis. We know from other research we have read ` 24

33 or heard that the take-up of energy-saving measures among households is often very low even when the pay-back period would be very short. 5.5 Environmental schemes Table 5.4 Ability to consider schemes of type: environmental Type Examples Within scope of FLUTE? Flood defence Improved drainage Partly see below Land decontamination Green infrastructure (parks and open spaces) Reservoirs Removal of waste See Site remediation in Table 5.2 Creation of green space Secured water supply, regulation of river flows, flood defence Source: own assessment of model capabilities Indirectly, yes see below Indirectly, yes see below All of these would require some pre-modelling analysis to estimate which types of households or businesses, in which zones, would be affected, and what value they would put upon the improvement. This would then enter into the model in the same way as for IT infrastructure, urban quality or energy schemes, as discussed in the preceding section Green infrastructure would be treated in the same way as urban quality and could make use of previous research in that area Reservoirs (if open, ie if they are artificial lakes rather than enclosed tanks) and any new waterways (whether for water supply, navigation or both) also have a value as landscape features which affects household preferences, property values etc and can be represented in the model. 5.6 Education schemes Table 5.5 Ability to consider schemes of type: education Type Examples Within scope of FLUTE? R&D facilities New R&D facilities see below [education/training to] reduce a skill gap in growth sector [Note that this was shown as an example of R&D facilities in the earlier material, but seems better treated as a separate type of scheme] Source: own assessment of model capabilities see below Research and development facilities cannot be directly represented in the model. It would be necessary to carry out a pre-model analysis of the expected effects within the SCR, which would probably be some combination of additional employment in the R&D facilities themselves; 25

34 an increase in employment as the results of research and development were exploited by firms/establishments within the SCR (obviously, the question of whether the intellectual property generated by the R&D activities would be taken up within the region is a major issue for the pre-model analysis) this might be a step increase in employment or a change in the underlying growth rate for a particular sector or sectors 7 ; alternatively, the forecast results of exploiting the R&D might be increases in productivity for a particular sector or sectors In the last case the GVA impact of exploiting the R&D could be calculated simply as the increase in GVA per job multiplied by the number of jobs in question Education and training to reduce skills shortages in growth sectors: the DELTA package, and hence FLUTE, has been developed on the basis that it should be able to match an exogenously given employment scenario, at least as a base case around which other forecast may pivot. It is therefore specified so that jobs are always filled there is no concept of unfilled jobs within the model design, and as a result it is difficult to represent a situation of skills shortages or an intervention which reduces skills shortages. For related reasons, labour costs do not influence the forecast growth rates in the base version of DELTA We have developed a variant version of DELTA which represents wages and wage changes, and in which changes in these can act as an influence on growth. This is more time-consuming to set up and run, and was therefore excluded from the FLUTE Stage 1 plans. Its main objective was to consider how transport improvements could improve labour supply to particular areas and hence, in some cases, release constraints on growth, but this could also be used to look at the impacts of changes in labour supply by socio-economic level. This would work at the level of the FLUTE socio-economic categorisation set out in Table 4.1. As shown there, these categories are based on broad occupation groups defined by skill levels. There is already a facility in DELTA to move workers and potential workers, living in selected zones, between these levels The version of DELTA with more detail of the labour market could therefore be used to explore the consequences of broad improvements in the availability of skilled workers on wages and hence on costs and growth rates. If however the effects of skill changes would be to allow employers to fill what are currently unfillable vacancies, and/or the shortage skills are much more precisely defined (eg the region has too many Java programmers and not enough Python programmers), then as with a number of other interventions an initial pre-model analysis would be needed to assess the employment effects, with SCRIM being used to assess secondary effects if these promise (or risk) being significant. 7 The employment impact would need to be expressed as a change for one or more of the modelled sectors; alternatively, some additional work could be undertaken to disaggregate the sectors so as to distinguish those that would be affected by the R&D activities. 8 This facility was introduced into DELTA as part of the modelling for the South and West Yorkshire Multi-Modal Study, circa 2001, though it was used there to define changes in the supply of labour as part of an aspirational economic/demographic scenario for South Yorkshire, not to define a policy intervention to be appraised (see A.4.2, section B.1for the distinction between scenario and policy elements).. ` 26

35 5.6.6 In any event there would need to be a critical pre-model analysis to consider who the proposed investment would convert into changes in the supply of skilled workers, since no form of investment can directly create skills it can only go into institutions and activities that seek to develop skills. 5.7 GVA impact calculations (i): employment changes FLUTE will output employment by zone and year for each of the sectors shown in Table 3.1. The modelled impact of a scheme on employment can therefore be found by comparing the with-scheme and reference case tests The model will not forecast employment that may be generated by the investment scheme itself, ie temporary employment in delivering the investment, eg in construction continuing employment in operation of the investment, eg in operating a new tram line, or a new biofuel power station Conventionally these are not included in the appraisal of public expenditure, on the basis that these are direct effects of expenditure which would arise however the money was spent whether by the public or the private sector (Green Book, p53). For SAF, this argument should perhaps be reviewed, on two grounds: SCRIF is aimed at maximising economic impacts within the SCR, so it is appropriate to ask where employment will be generated in delivering the investment (this is of course an important political question, as illustrated recently in the case of the Crossrail rolling stock); SCRIF is specifically an investment fund, and continuing employment in operating new facilities will presumably be paid for out of the revenue raised in operation or from other sources: as such it may be reasonable to include operational employment within the SCR as part of the benefit, provided that care is taken to allow for displacement effects (eg a requirement for tram drivers may be matched by lower demand for bus drivers) If employment in delivering and operating the investments is included in the GVA calculations, some attention will have to be given to the question of which firms, located where, should be assumed to win contracts to build or supply new facilities, and to ensuring that these are consistent (not necessarily identical) across scheme types. 5.8 GVA impact calculations (ii): agglomeration effects The DfT methodology involves calculating measures of employment density, which are in effect measures of accessibility to jobs; using the changes in employment density resulting from schemes to estimate changes in productivity for a number of broad sectors. 27

36 5.8.2 This will require the DELTA WEBs programs to be implemented these are one of the optional extras already mentioned. This is a relatively minor extension; all of the data required provided either available in FLUTE already or can be taken from the dataset which DfT supplies for calculation of wider impacts. (The data might need to be modified if the economic scenario(s) used for SAF work is/are significantly different from those assumed by DfT in WebTAG and NTEM.) 5.9 GVA impact calculations (iii): transport cost savings SYSTM+ has a distinct representation of travel during the course of business and for movements by HGV and LGV vehicles. The first two categories relate to businesses, and while the majority of use of light goods vehicles is for business, some interpretation would be required to exclude the minority of use for personal travel. The main limitation is that these types of trips are represented on an origindestination basis rather than a production-attraction basis 9, and an interpretation will therefore be required to attribute benefits for trips that cross the SCR boundary. Options include taking half the benefits for such trips (assuming that half such trips are produced within SCR) GVA impact calculations (iv): combined results A mechanism will be needed to combine the employment effects (multiplied by GVA/job), the agglomeration effects (which apply to all jobs, not just the net gain in jobs) and the cost changes, into an overall GVA impact for the SCR. How best this should be done depends in part on the number of years outputs to be processed, which is considered in the next chapter if it needs to be done for 20 or 30 years per test, it will be more important to automate the process than if the analysis is just done for a single year Accessibility calculations This again is readily implemented using data which will be available in the standard operation of the Stage 1 FLUTE model: the numbers of job opportunities will come directly from FLUTE but in any cases where the job changes aren t represented in FLUTE (eg investment in rural industries that would be insignificant in FLUTE) then these can be input directly to the accessibility calculations (though if they re not significant in FLUTE, they probably won t be significant in accessibility); the generalised costs of commuting will come from the output file which SYSTM+ supplies to FLUTE; the numbers of residents (in total or in the lowest SEL) will come directly from FLUTE. Note that these are used in weighting to find average 9 Representation of trips on an origin-destination basis means simply counting trips from A to B. Representation on a production-attraction basis identifies which of these trips from A to B are produced by households or businesses located at A, travelling from A to some attraction at B, and those produced by households or businesses located at B, returning from some attraction at A to their home or base at B. ` 28

37 accessibilities over zones, not in the calculation of accessibilities for any one zone The recommended accessibility calculations are closely related to those already implemented in FLUTE (ie in DELTA). We recommend that the process to extract the averaged accessibility values needed to test whether the accessibility criteria are passed should be implemented as a variant of, or add-on to, the standard DELTA calculations depending on the exact form chosen, this might not require any new accessibility calculations per se, only the calculation of averages over zones by district and for the SCR as a whole. 29

38 6 TIMING, DISCOUNTING ETC 6.1 Introduction Up to this point we have mentioned that SCRIM is a dynamic system, modelling change over time, but we have avoided the question of how we use results from different years. This Chapter deals with the time-related questions 6.2 Model years and forecast period FLUTE runs in one-year steps from its 2007 base year for as many years as required: it is currently set up to run to A modest amount of work to prepare additional scenario and policy inputs would be needed to extend the FLUTE forecasts further into the future SYSTM+ is currently implemented to run at approximately five-year intervals up to 2026; hence transport interventions can only be introduced (or removed) in those years, and generalised costs will only be updated in those years. The latter means that the transport components of accessibility and agglomeration calculations will only be updated in those transport model years, though the land-use inputs are updated every year A modest amount of additional work would be required to run SYSTM+ in additional years, whether they were inserted between the years listed above or added to extend the transport forecasting further into the future. If the sequence of transport model years is changed, the reference case forecast(s) need to be rerun or extended to include those years, before running the do something tests including those years. 6.3 Multi-year outputs and discounting: GVA We recommend that the GVA impacts should be calculated year by year and discounted to a base year; the required investment from SCRIF (the whole-life cost to SCRIF mentioned in the GVA criterion) should also be discounted to find its base year value; the GVA gained per invested will then be the ratio of these two base year values. ` 30

39 6.3.2 For initial sifting of schemes it would be attractive to simplify this, eg to looking at a single specific year or the single year 10 years after opening, though it might be tricky to find a way of doing this that wasn t biased towards certain kinds of projects and against others, especially if the schemes to be sifted involve different timing of investment Any calculation to find the base year value of a stream of benefits (or costs) raises the usual inter-related questions of how many years? what discount rate? There are standard answers to these in standard WebTAG/Treasury guidance, but SAF may not be constrained to these. 6.4 Accessibility It would be possible to calculate the accessibility changes and to check whether the specified criteria are satisfied for every year starting from the opening year of the scheme. That however would seem to be over-elaborate. We suggest just checking for the transport model year closest to 10 years after project opening. Other possibilities would be (for example) checking approximately 5 and 10 years after opening; checking in the final modelled year (horizon year) for all schemes (ie look at the impacts in the the longest (practical) term) We leave this for debate/decision at the forthcoming workshops. 31

40 7 CONCLUSIONS 7.1 Calculating the GVA and accessibility measures We have recommended a conventional and practical approach to the calculation of GVA impacts, based primarily on changes in employment, but also taken account of productivity effects arising both from transport cost savings and from agglomeration effects The analysis required can be split into three stages forecasting the employment changes; calculating the GVA impacts resulting from those employment changes; taking account of the agglomeration and transport cost effects Employment and transport cost changes will be used in the agglomeration calculations, and are also required for the accessibility calculations. The fact that the impacts of a scheme are found by comparing two forecasts, with and without the scheme, means that the issues of identifying additionality are very largely eliminated Employment changes and transport cost changes are standard outputs of the SCRIM model now under development. By making full use of standard outputs, we minimise the risk of inconsistency between schemes or types of schemes All of the schemes which need to be assessed will need to be quantified in ways that can be input to SCRIM. For some schemes that are directly represented in SCRIM inputs, this will be very simple for example, a new section of road between two existing junctions will be quantified essentially by its type (single/dual carriageway, number of lanes), free-flow speed and length. At the other extreme, for schemes that cannot be explicitly input to SCRIM, it will be more complex, and in some of these cases an initial estimate of the direct employment impact will need to be made outside SCRIM. SCRIM will then model the indirect impacts (for example, through impacts on population location and on property development); doing this will help to ensure that different types of schemes are assessed consistently and fairly Whilst our report is written very much in terms of employment gains, cost savings, and additional productivity, it must of course be kept in mind that some of these impacts may turn out to be negative if a scheme is forecast to have unintended side-effects for example by causing an increase in congestion on critical links of the transport networks. ` 32

41 7.2 Required additions The following additions to the currently commission FLUTE functionality are required in order to implement the approach recommended here: the Wider Economic Benefit calculations, in order to estimate agglomeration effects (see section 5.8) (these were item 1 in the list of potential enhancements included in the April 2012 Task Proposal); a process to calculate GVA impacts for SCR from FLUTE-forecast employment changes (see section 5.7), to extract the appropriate transport cost savings (see 5.9), and to combine these with the agglomeration results to produce the overall GVA impact (5.10); the accessibility calculations (as a variant on the DELTA zonal accessibilities program AC12 see section 5.11 including the necessary averaging over zones). Note that coefficients used in calculating the accessibility criteria may differ from those used in the main FLUTE calculations, and therefore separate runs of AC12 are likely to be needed for assessment purposes The Wider Economic Benefit calculations are existing DELTA software which can readily be set up for use with FLUTE. The other calculations will involve reading and processing large quantities of data from DELTA files, and we would therefore recommend that they be implemented as extensions of the DELTA software rather than as ad hoc utilities. 7.3 Possible additions to SCRIM, required for full assessment of particular schemes Two other items from the earlier list of possible enhancements would be relevant in particular cases: the national spatial-economic model to test impact of schemes to improve links between SCR and the rest of the UK (5.3.2 this was item 6 in the April 2012 list); wages-based model of the labour market (5.6.5 item 3) The national level of modelling would require a moderate amount of additional work, the wages-based model would be more substantial The dwellings-based version of the housing/household location model (item 4 in the proposal) would be an appropriate enhancement if detailed analysis of housing market impacts was required. Given the specified criteria for SAF, this seems unlikely Item 2 on the April 2012 list was explicit modelling of redevelopment and intensification. In our opinion this would be a significant enhancement to the model in any case, since it makes it easier to deal with the more distant years of the forecast for which local authorities are very reluctant to quantify or even to comment on their planning assumptions. The importance of this for SAF will depend what is decided about the timescale over which GVA impacts and accessibility outcomes need to be assessed if this is such that the period beyond 33

42 current local planning frameworks is important, the redevelopment/intensification enhancement would be very relevant. If the focus is on the much shorter term, it will be much less significant The final potential enhancement in the April list (item 5) was improved modelling of car parking, particularly taking account of parking as a land-use and as an element of transport supply that is modified by land-use decisions. If any proposed investment schemes would involve development or redevelopment in areas where car parking supply is or could become a constraint, this linkage should be considered, and if a range of alternative investments have to be compared it would be sensible, for consistency, to implement the linkage in the model rather than ad hoc. ` 34

43 REFERENCES David Simmonds Consultancy (1998): Accessibility as a criterion for project and policy appraisal. Final report to DETR. DSC in collaboration with University of Leeds Institute for Transport Studies, MVA Consultancy and Oxford Brookes University. Hansen, W G (1959): How accessibility shapes land use. Institute of Planners, vol 25, pp Journal of the American HM Treasury (2011): Appraisal and Evaluation in Central Government (The Green Book). TSO, London. Jones, S R (1981): Accessibility measures a literature review. TRRL report LR 967. TRRL, Crowthorne. Office for National Statistics (2010): Measuring the economic impact of an intervention or investment. Paper Two: Existing sources and methods. December Whitehead, T B, D C Simmonds and J Preston (2006): The effect of urban quality improvements on economic activity. Journal of Environmental Management, vol 80 (1), pp

44 APPENDIX A ACCESSIBILITY CALCULATIONS A.1 Introduction A.1.1 A.1.2 This Appendix sets out the calculations required for the proposed accessibility measures. These involve two steps: averaging generalised costs over modes in order to get an overall measure of the difficulty of getting from i to j calculating the accessibility itself, ie weighting job opportunities in each zone j by a function of the difficulty of getting to them. As mentioned in the main text, these calculations are already implemented as intermediate stages in the FLUTE accessibility calculations. A.2 Averaging generalised costs over modes A.2.1 A.2.2 A.2.3 A.2.4 This will use the conventional logsum formula to average over modes in each transport model year: g o Tij where 1 = ln exp Mo λ Tij m Mo o (. g Tij Tijm ) λ o g is the generalised cost in year T of travelling from i to j and carownership level o, to be found by averaging over Tij modes; o g is the generalised cost of travel from i to j by mode m and carownership level o, inclusive of alternative-specific constants, for the Tijm most recent transport model year T; and Mo is the value of the modal choice coefficient for zone pair (i,j) and carownership level o, again in the most recent transport model year λ Tij T. SYSTM+ supplies FLUTE with generalised costs by purpose these differ due to differing values of time and distributions of travel over times of day. Since in the assessment we are only concerned with accessibility to employment, the only purpose considered will be that for commuting (home-based work trips, in the transport model). The superscripts p for purpose which usually appears on any measure of generalised cost in the FLUTE specification has therefore been omitted here. Whilst the values of the modal choice coefficient can in principle vary by zone pair (i,j), for urban studies they are usually kept constant. We have suggested that calculations for the most needy should be done excluding car travel; that can be implemented either ` 36

45 by adding an extremely large constant to the generalised costs for car travel, which effectively makes car irrelevant, or by omitting car from the summation over modes m. A.3 Calculating accessibility A.3.1 The Hansen accessibility measure is found as ( λ ) A = W.exp. g o Do po ti tj t Tij j where po g is the generalised cost in the most recent transport model year T of Tij travelling from i to j for purpose p and car-ownership level o, found by the previous equation; a W tj is the relevant number of jobs at j in year t the suggestion is that this is either all jobs or jobs for workers of socio-economic level 4; Dao λ T is the distribution coefficient for this type of accessibility (reflecting preference for nearer opportunities, ie those which can be reached at lower generalised cost) and for car-ownership level o; and o A ti is the resulting measure of accessibility, for persons of car-ownership level o living in zone i at time t. A.4 Finding district or other averages A.4.1 The accessibility measure defined above cannot meaningfully be summed over zones, but it can be averaged if appropriately weighted. Our initial suggestion (for discussion) is that it should be averaged by the number of working-age persons, in total for the general measure and in SEL 4 for the most needy measure. The calculation would therefore be of the form A t( LA) = o i ( LA) o i ( LA) o A. P ti P o ti o ti A.4.2 where A is the average accessibility across local authority LA at time t t( LA) o P is the relevant population of working age living in zone i at time t. ti Our suggestion is that the most needy calculation would be done only for noncar-owners, in which case the calculation would be done only for o = households with no car, and the summation of o would be irrelevant. 37

46 APPENDIX B OUTLINE OF SCRIM AND FLUTE B.1 Introduction to the models B.1.1 B.1.2 This Appendix provides a brief description of the scope and main functions of SCRIM, and in particular of the FLUTE model, implemented in DELTA, which provides its land-use/economic functionality. Figure B-1 provides an overview of the complete SCRIM model. Any one run of the model produces a forecast of land-use and transport within the modelled area, given base year data from which the model starts top-down inputs defining the demographic and employment scenarios in the reference case these are defined for areas or for the whole model bottom-up inputs representing the planning and transport interventions to this particular test this are specifically defined by zone and floorspace type, or by link, etc, depending on the nature of the intervention. The user has to define when these interventions will occur, as well as what they are and where they will happen. Figure B-1 Overall structure of the model B.1.3 Note that the overall scenario (ie the rates of employment and population/household change) may be modified by the effects of the planning and transport interventions. ` 38

47 B.1.4 B.1.5 B.1.6 The impact of a proposed intervention is found by comparing two tests, one with and one without that intervention. There are no random elements to the model, and usually only a very low level of noise resulting from rounding errors in the calculations. The model works through the same sequence of submodels irrespective of the intervention being tested. It can be used to look at the impact of a single intervention or at complex packages of interventions. The model starts at a base year and then forecasts through time in single-year steps, taking the output from one year as the inputs to the next. For each year, the model calculates the changes in floorspace by land use type (ie residential, retail, office and industrial), as well as the changes in the activities that use that floorspace. This calculation generates information on the change year-by-year for each zone. The model starts at a base year and then forecasts through time in single-year steps, taking the output from one year as the inputs to the next. Figure B-2 shows part of the sequence for a typical model. Each test therefore has to be run in order, from the base year to the end of the forecasting period. Figure B-2 Time-marching process B.1.7 B.1.8 The FLUTE model sets out to represent the interactions between three different groups of actors : basically households and household members firms in general developers, as a special category of firms. (Each group of actors is disaggregated within the model, for example into different types of households per composition, age and socio-economic status.) These groups of actors interact with each other, and with the transport actors, through different markets, as illustrated in Figure B-3. 39

48 Figure B-3 Actors and markets B.1.9 The workings of the model are shown more specifically in Figure B-4, which identifies the main variables passing between different parts of the model. It also starts to show the spatial levels of the model: the components represented in the central urban block work at zone level the economic and migration components on either side work at area level. Figure B-4 Main model components B.1.10 The extent of the FLUTE model zone and area systems is illustrated in Figure B-5. The core of the model is the Fully Modelled Area, which has been defined to be rather larger than the SCR (so as to allow for location effects around the SCR boundary). Beyond that there is a large Buffer Area, where processes are modelled ` 40

49 in less detail (in particular, development processes are implicit rather than explicit) and zones are larger. 41

50 Figure B-5 FLUTE Modelled Areas: Fully Modelled and Buffer ` 42