Scottish Government

5 Economic impact

Evidence from the literature on the economic impact of ICT/broadband

Studies on the economic impact of ICT

5.1 There is a growing body of evidence in the literature that ICT does have significant positive impacts on productivity and economic output. For example:

• London Economics (2003) found that ICT capital deepening contributed an average of 0.8 percentage points annually to UK economic output from 1992 to 2000 (25% of the total output growth), and accounted for an average contribution of 0.76 percentage points annually to UK average labour productivity growth over that period.
• Rouvinen and Maliranta (2003) analysed a sample of 2,000 Finnish firms over the period 1998-2001, and found that, on average, the productivity of ICT-equipped labour is 8% to 18% higher than for other workers.
• Oulton and Srinivasan (2005) found that ICT capital accounted for 28% of the growth in output per hour in the UK market sector in 1990-2000 (with the proportion rising to 46% in the second half of that decade).
• Bloom et al (2005) estimate that a doubling of IT stock is associated with an increase in productivity of between 2% and 4%. They also report that US Multinational Enterprises ( MNEs) in the UK have 8% higher productivity than domestic UK firms, while other MNEs in the UK are only 5% more productive: over 80% of this productivity advantage for US-owned plants is attributed to better use of ICT. They suggest that superior management practices of US subsidiaries may explain their ability to extract higher returns from IT.
• DCITA (2005) analysed productivity growth in Australia's service sector in the period 1984/5 to 2001/2, and found that, after taking away the effect of increased capital spending per worker, technological factors (the ICT revolution in particular) accounted for between 59% and 78% of productivity growth in service industries.
• Farooqui (2005) analysed the impact of UK business spend on telecoms services, and found that telecoms use has a large positive and significant effect on firm output - explaining up to 7.5% of the productivity differences in manufacturing firms (after accounting for the effects of IT capital). There is also a strong association between IT investment and spend on telecoms - with increased spending on the latter compounding the effects of IT investment. Enabling manufacturing workforces with internet raises productivity by 2.9% for every 10% of the workforce enabled, and this is additional to the effects of increased IT investment.

5.2 A number of studies emphasise that there is a substantial lag effect at work, as businesses take time to recognise, design and implement complementary 'organisational investments' (in business process change) before they realise the full benefits of their ICT investments:

• In OECD (2004), Gretton et al report on a study into the impact of ICT on productivity in Australia. They found that the productivity response followed an inverted 'U' pattern as the duration of ICT use increased - with the productivity impact of computer take-up being largely complete after a period of adjustment of around 5 years.
• Brynjolfsson and Hitt (2003) analysed the effect of computerisation for 527 large US firms over the period 1987-1994, and found that the contributions associated with computerisation are up to five times greater when measured over long (5-7 year) periods than when measured over short (1 year difference) periods.

5.3 There are significant variations between sectors - with the nature of the business determining the extent to which it can benefit from ICT:

• London Economics (2003) estimated that the annual contribution ¹⁰ of ICT capital deepening to labour productivity growth in the period 1992-2000, varied from 0.0 percentage points (Mining and quarrying) to 1.24 (Financial intermediation) to 6.27 (Post and telecommunications - which should, perhaps, be regarded as a special case).
• Bank of England (2005) notes that business services, distribution and finance - which together account for c. one third of aggregate UK output - were responsible for almost two thirds of capital deepening in the 1990s. The growth of labour productivity of these sectors in the UK substantially outperformed that of the market sector as a whole, with particularly marked gains in the past couple of years.

Studies on the economic impact of broadband

5.4 While the above research is focused on the impact of ICT in general, some studies have sought to estimate the economic impact of broadband, in particular - at local, national and supra-national levels.

5.5 Given the relatively recent availability and take-up of affordable broadband services, these studies have not been able to analyse economy-wide historical impacts attributable to broadband - and it may yet be a couple of years before sufficiently long and robust series of data are available for the UK regarding business adoption of broadband, to allow such an analysis. The previous studies - and our study - have necessarily had to make 'educated guesses' as to what the economic impact of broadband may be.

5.6 The existing studies essentially take one or both of the following approaches:

• estimating consumer surplus and producer surplus - e.g. Criterion Economics (2001); PricewaterhouseCoopers et al (2004);
• estimating the impact on Gross Domestic Product - e.g. Gartner (2002); Criterion Economics (2003); CEBR (2003); ACIL Tasman (2004).

5.7 In SQW's view, the first of these approaches is problematic. It relies on being able to judge the price elasticity of broadband services ¹¹ at some point in the future when broadband penetration has saturated (bringing the market into equilibrium). Given the pace of change in broadband markets, and the difficulties of accurately determining price elasticities (now, yet alone many years' hence), we consider there to be particularly high error margins associated with this method. Furthermore, Scotland's economy is far from being a 'closed system': much of the 'consumer surplus' associated with broadband in Scotland will, in practice, contribute to the wealth of other nations ( e.g. through increased Scottish expenditure on music, videos, games etc. sold online by companies entirely outwith the country).

5.8 For the purposes of our research, therefore, we have chosen to focus on the impact on business productivity and Scotland's GDP - the most widely recognised measure of Scotland's wealth - rather than on the consumer/producer surplus.

5.9 Annex C summarises the methodologies adopted by four previous studies into the impact of broadband on GDP. The key features are captured in the table below.

Table 5 1 Key features of four previous studies into the GDP impact of broadband

Firm-level evidence on the impact of broadband

5.10 In support of the above economy-level analyses, there is ample evidence that businesses themselves consider broadband to be important to their operations.

5.11 For example, the DTI has compiled a large selection of broadband case studies (available at http://www.liveworkplaybroadband.org.uk/). In Annex D we have summarised examples from these case studies for each of the SIC groupings considered in our model, in order to provide qualitative context for our critical assumption that broadband helps improve business productivity.

5.12 Large scale surveys confirm that broadband is perceived to be an important technology for many businesses. For example:

• In an SQW/ NOP survey for DEFRA [ SQW (2005)], business broadband users across England were asked to rate the importance of broadband to their firm, on a scale of 1 to 10. The average 'scores' from both urban and rural businesses were remarkably high, at 7.8 and 7.3 (out of 10) respectively.
• In SEBS 2005, 82% of business broadband users agreed (or strongly agreed) that broadband provided them with more scope to make continuous improvements, as illustrated - by SIC grouping - in the chart below.

Figure 5 1 Broadband users' response to the statement "[Broadband] provides more scope to make continuous improvements to how we do business", by SIC grouping [source: SEBS 2005]

Business applications enabled by broadband

5.13 In this sub-section, we briefly consider the business applications enabled by each broadband generation, in order to inform our estimates of their respective incremental productivity impacts.

5.14 For the purpose of this analysis, we compare the 'entry level' bandwidths at each broadband generation. Our definitions in section 2 of this report considered only the downstream peak bandwidths; however, in practice, many business applications are dependent on the upstream bandwidth, as well as the downstream connectivity. We have therefore needed to assume a 'typical' upstream bandwidth associated with the entry level to each generation of broadband, informed by the current 1B market offerings, the emerging 2B services ( e.g. via cable modem, ADSL2+ and 8Mbit/s ADSL), and the mooted capabilities of 3B+ services ( e.g. via VDSL2 and DOCSIS 3.0). The 'representative' upstream and downstream bandwidths considered in this analysis are shown in the table below:

Table 5-2 Representative downstream and upstream bandwidths considered for our analysis of business applications [source: SQW]

5.15 In order to illustrate the improvements in response time associated with the upgrades between broadband generations, we have considered a set of example business applications at various file sizes as shown in Table 5 3 below.

Table 5 3 Example file sizes and business applications [source: SQW]

5.16 Figure 5 2 and Figure 5 3 below illustrate the improvements in response time ¹² associated with each generation upgrade, for each of the above example file sizes/applications, in the downstream and upstream directions respectively.

Figure 5 2 Improvement in download response times due to upgrade (log scale)

Figure 5 3 Improvement in upload response times due to upgrade (log scale)

5.17 These charts serve to illustrate some important points:

• For some high usage, modest file size applications there are negligible benefits from upgrading to 2B and 3B+ bandwidths. For example, the Google home page (c. 20kB) could be loaded in 0.03s with 2B as opposed to 0.3s with 1B, and a 50kB intranet page would load in 0.008s with 3B, as opposed to 0.08s with 2B: differences that are likely to be imperceptible to the end user.
  • If we assume a 40kB average (non-spam) email message size in 2004 [source: Verisign (2004)], and a 10% p.a. average annual growth rate in email message size [source: HP (2005)], then we could expect average email message sizes to be c. 100kB by 2015. Such messages would take 16s to download using dial-up, 1.6s on 1B, 0.2s on 2B and 0.02s on 3B: representing a substantial improvement for a dial-up to 1B upgrade, some improvement for a 1B to 2B upgrade, but an imperceptible difference between 2B and 3B+.
• For the dial-up to 1B upgrade, the download response time is perceptibly better ( i.e. a difference greater than 1s, say) for files of c. 7kB or more; the equivalent threshold file size is c. 70kB for the 1B to 2B upgrade, and c. 700kB for the 2B to 3B upgrade.
• Applications requiring data to be uploaded benefit from the upgrades at significantly smaller file sizes. For the dial-up to 1B upgrade, the upload response time is perceptibly better (with a difference greater than 1s) for files of 4kB or more, and the equivalent threshold file size is about 60kB for both the 1B to 2B upgrade and the 2B to 3B upgrade.

• When we consider much larger file sizes, we observe improvements in response times which make previously impossible or cumbersome applications feasible: accessing a 10 MB spreadsheet in a company's (off-site) KM system would take c. 16s with 2B, versus more than two and half minutes with 1B; a 200 MB backup file could be uploaded to a secure remote server in c. 50s with 3B, versus 53 minutes with 2B.

5.18 We now turn to consider streamed applications, such as voice and video communications, which are consumed in real-time and are therefore particularly intolerant to delay.

5.19 According to Cisco, an uncompressed (G.711) VOIP call can be supported by a symmetrical bandwidth in the order of c. 90kbit/s, whilst compression techniques ( e.g. using G.729, G.723 codecs) can reduce this requirement to c. 30kbit/s.

5.20 With 1B typically providing c. 250kbit/s upstream, we can see that VOIP calls can readily be supported by 1B connections. Whilst upgrades to 2B and 3B+ would allow a greater number of simultaneous VOIP calls to be supported by a single connection, an SME could obtain a similar outcome simply by using multiple 1B connections.

5.21 The situation is different, however, when we consider video. While this is a much more bandwidth-hungry application than voice, the advent of advanced video compression techniques - in particular, H.264 ( MPEG-4 Part 10) and VC-1 (Microsoft's competitor to H.264) - is leading to significant improvements in the quality of video that can be delivered over modest bandwidths.

5.22 The table below shows the resolution that can be achieved at different levels of VC-1 coding, for example, together with our estimate of which broadband generation is required to support each level: both for video streaming ( e.g. of a webcast, which is dependent on downstream bandwidth, but not upstream bandwidth at the end user's site) and for video conferencing (which is dependent on upstream bandwidth as well as downstream bandwidth).

Table 5 4 Representative resolutions for different levels of VC-1 coding [source: Microsoft]

5.23 Note that the above table indicates the maximum bit rates at each level of coding: i.e. the actual bit rates at each coding level will be below these rates. According to Apple, H.264 could typically provide 640x480 pixel resolution (24 frames per second) at c. 1-2Mbit/s, 1280x720 (24fps) at c. 5-6Mbit/s, and full high definition 1920x1080 (24fps) video at 7-8Mbit/s.

5.24 Given these indicative resolutions at various bandwidths, we anticipate that Scottish businesses will start to experiment with person-to-person video-conferencing via their PCs over 1B connections; it will gradually grow in popularity as a business application over 2B connections, and both person-to-person and multi-party video-conferencing will become mainstream once 3B+ services start to become widely available in urban areas (2008 onwards).

5.25 Video streaming ( i.e. in the downstream direction only) is already reasonably well supported over 1B connections, and we would expect the availability of 2B services to further enhance the quality of video delivery (such as company webcasts, product training videos etc.). Large corporates can potentially make substantial cost savings through video-based product training for their large and distributed salesforces, for example. However, we suspect that the economics of video-streamed training will be less compelling for SMEs - not least because the up-front costs of production are spread across a much smaller audience than in a corporate. We anticipate that video conferencing will be a more important, and widely used, application for Scottish businesses than video streaming, in the medium term.

5.26 Our overall assessment of the incremental benefits to businesses associated with each generation of broadband is summarised in the table below:

Table 5 5 Incremental business benefits associated with each broadband generation [source: SQW]

5.27 In summary, our assessment is that there are very considerable incremental benefits associated with businesses upgrading from dial-up to 1B broadband, which are likely to translate into very significant productivity improvements over time. We also believe that there will be substantial incremental benefits in upgrading from 1B to 2B and from 2B to 3B+ (some applications for which will not have not been invented yet). At this stage, we are of the view that the incremental productivity impacts of the 1B to 2B upgrade and of the 2B to 3B+ upgrade are unlikely to be as great as those of the move from dial-up to 1B (each providing in the order of a further 25% to 50% of the potential productivity gains from 1B).

Broadband's impact on the GVA of the telecoms supply chain

5.28 In addition to the productivity impacts for ICT users, we need to consider whether there is a significant impact on the GVA of the telecoms supply chain in Scotland, from provision of the various generations of broadband service.

5.29 Our conclusion is that this is likely to be a relatively minor source of economic impact, in Scotland's case, for the following reasons:

• There is little evidence to date that the telecoms sector GVA has been substantially increased as a result of 1B broadband. Competition between telcos/ ISPs appears to be ensuring that the value created by this technology is primarily accruing to the users rather than the telcos/ ISPs, with intense price competition bringing broadband prices down towards - or even below - dial-up prices.
  • For example, BT's 2005 Annual Report states that "The [ BT Retail] gross margin percentage decreased by 0.9 percentage points in the 2005 financial year after a decrease of 0.2 percentage points in the 2004 financial year. The decline primarily reflects the change in revenue mix from traditional business to lower margin new wave services."
  • As illustrated in the chart below: UK retail fixed telecoms revenues have flattened out, and the number of employees in telecommunications ( SIC 64.2) in Scotland has been flat/falling.

Figure 5 4 UK retail fixed telecoms revenues, and telecommunications employees in Scotland [sources: Ofcom (2005), Annual Business Inquiry]

• In contrast to Criterion (2003) - which analysed the impact of broadband for the US economy - we cannot assume that much if any value from telcos' investments in broadband in Scotland will be captured by equipment manufacturing plants located in Scotland.
• The development of the broadband market has led to the emergence of Next Generation Networks - and in particular, BT's 21 CN. In the absence of broadband, there would be no 21 CN, which would imply:
  • no £10 billion investment over five years (say £200 million p.a. in Scotland - though, as noted above, we expect that the bulk of this investment will be to the benefit of equipment manufacturers using plants located outwith Scotland);
  • but also no subsequent £1 billion p.a. operational cost savings, as a result of 21 CN (say £100 million p.a. in Scotland) - some of which is likely to be achieved through a reduction in staff costs.
• If there was no mass market broadband, then it would appear likely that competition would be less intense in the telecoms sector, and that IP telephony would not be posing such a threat to voice revenues (fixed and mobile). In the hypothetical case of a Scotland with no broadband, therefore, it is possible that current and future GVA in the telecoms sector would actually be higher than in the with-broadband scenario.
• While 2B appears unlikely to involve major incremental network investment, 3B could potentially involve significant revenues for the construction industry in Scotland (digging up roads, laying new duct, installing new cabinets etc.).
  • If BT eventually mirrors what Deutsche Telekom and SBC have already started in their respective territories, this could involve c. £1 billion p.a. additional investment across the UK (say £100 million p.a. in Scotland). If c. 50% of this investment goes to the Construction sector (employing people in Scotland) this would imply an increase in Construction turnover in the order of £50 million p.a., or c. £20 million p.a. increase in direct Construction GVA (assuming a c. 40% ratio of GVA to turnover for the Construction sector). Applying a (Type 2) multiplier of 1.9 for Construction gives an overall potential GVA impact in the order of £40 million p.a..
  • However, this does not take into account the wealth creation effects of alternative uses of a 3B network investment ( e.g. dividends to shareholders, including funds managed in Scotland). We therefore suspect that the overall net GVA impact for Scotland of this 3B construction boom (if it happens) is likely to be less than £40 million p.a.. As will be seen later in this section, this is small relative to the impacts of improved productivity for broadband users.

Modelling assumptions

5.30 Our economic impact model applies the projected business take-up rates to a set of assumptions regarding the impact of broadband on the productivity of broadband adopters, in order to generate a first order approximation of the effect on the overall Gross Value Added ( GVA) of Scotland's market sector.

5.31 The overall structure of the model is illustrated in section 2. In summary, our projections for cumulative annual broadband connections in each employment size band and SIC grouping are combined with assumptions regarding the average number of employees per site in each size band, to derive annual projections of the total number of employees in each SIC grouping at sites newly-connected to the various generations of broadband. We then apply 'S-curves' describing the productivity impact of the various generations of broadband by year after adoption, and an assumed baseline GVA per employee in each SIC grouping, in order to derive our estimate of the overall annual impact on Scotland's market sector GVA. Our key assumptions are summarised below.

Employment levels and supply chain

5.32 We have assumed no impact of broadband on employment levels, given the current low levels of unemployment in Scotland. For simplicity, we have assumed constant employment levels in every SIC grouping; increases in productivity per employee therefore feed through directly into increases in GVA.

5.33 Given the above discussion regarding the telecoms supply chain, we have also assumed no material incremental effect on the telecoms supply chain GVA resulting from broadband network investment.

Productivity impact

5.34 We have divided our market sector SIC groupings into high, medium and low impact categories, on the basis of the differential impacts of ICT observed in the literature:

Table 5 6 Assigned broadband impact categories of our SIC groupings [source: SQW]

5.35 A lag effect has been incorporated, such that broadband adopters do not see the full productivity benefits of broadband for some years after adoption, given that it takes time to recognise, design and implement the complementary 'organisational investments'. Informed by our literature review, we have assumed a lag ranging from four to six years. Broadband productivity effects are assumed to be largely complete after four years for High impact sectors, five years for Medium impact sectors, and six years for Low impact sectors. Diffusion (S) curves have been applied to these productivity shocks, rising from zero impact in the year of adoption.

5.36 For our Medium impact sectors, we have assumed that 1B broadband adopters eventually become 5% more productive (over five years) than they would have been if they had not adopted broadband - i.e. an average annual productivity shock ¹³ of c. 1%. This assumption ¹⁴ has been informed by our literature review, including the observation in Farooqui (2005) that telecoms use accounts for up to 7.5% of the productivity differences in manufacturing firms, after accounting for the effects of IT capital investment.

5.37 We have assumed that the annual impact of 1B for our High impact sector adopters is approximately twice that for the Medium impact sector adopters: c. 2%, leading to an eventual impact of 8% (over four years). The annual impact of 1B for our Low impact sector adopters has been assumed to be approximately a quarter that for the Medium impact sector adopters: c. 0.25%, leading to an eventual impact of 1.5% (over six years). These assumptions regarding the relative impacts for different sectors have, again, been informed by our literature review, including London Economics (2003), which found that ICT capital deepening accounted for c. 1.24 percentage points growth in labour productivity in Financial intermediation in the UK in the period 1992-2000, versus 0.55 for Manufacturing and 0.10 for Agriculture, forestry and fishing (taking the average of the low software and high software scenarios).

5.38 Guided by the discussion above re the business applications enabled by broadband, the incremental impact from the use of 2B (rather than 1B) has been assumed to be 33% of the incremental impact of the use of 1B (rather than dial-up). Similarly, the incremental impact from the use of 3B (rather than 2B) has been assumed to be 33% of the incremental impact of the use of 1B (rather than dial-up).

5.39 When these incremental impacts are combined, therefore, a firm in a High impact SIC grouping, for example, (such as a financial services company) is assumed - eventually - to be up to c. 13% more productive (8% + 2.7% + 2.7%) using 3B than it would have been if it had remained on dial-up. With our assumed 3B take-up profiles and productivity lag effects, this maximum productivity impact only starts to be achieved from 2011 onwards.

5.40 To put this maximum impact into context, Barnes and Haskel (2000) found a wide dispersion of productivity levels within UK (manufacturing) industry sectors: with the 90th percentile firm being 3.5 to 6 times more productive than the 10th percentile firm in most sectors. Productivity dispersion tends to be higher in service industries.

Baseline productivity

5.41 We have used a baseline GVA per employee in 2000 (at 2000 prices), derived by dividing Scottish GVA data per SIC grouping from ONS (2004) by the relevant employment data from the Annual Business Inquiry ( ABI).

5.42 This will tend to overstate the baseline GVA per employee in each SIC grouping, as the ABI does not include the self-employed (c. 11% of market sector employment in Scotland). However, this effect should be cancelled out in our overall GVA impact modelling approach, as the employee data from ABI is also used in our model to estimate the numbers of broadband-connected people in the market sector workforce (which will therefore be understated, as it excludes the self employed). While there are higher proportions of self-employment in certain sectors ( e.g. construction, business services), this should have no effect on our overall GVA impact calculations.

5.43 The FISIM (Financial Intermediation Services Indirectly Measured) adjustment in ONS (2004) has not been included in our baseline GVA per employee, as we are not able to distribute this adjustment across our SIC groupings. This could potentially mean that our total GVA projections for the market sector are overstated by 3% to 4% - but we consider this to be well within the margin of error for our projections.

5.44 The assumed baseline values of GVA per employee in Scotland are shown in the table below.

Table 5 7 Baseline GVA per employee in the year 2000 [sources: ONS (2004), ABI]

Economic impact projections

5.45 Our economic impact projection is, of course, based on a combination of assumptions around which there are inherent uncertainties - the incremental productivity impacts per broadband adopter at different broadband generations, the growth of 2B and 3B+ coverage and the rate of broadband take-up. Having tested the sensitivity of our model to these various assumptions, we estimate that the annual GVA of Scotland's market sector in 2015 will be in the order of £2 billion to £6 billion higher than it would have been otherwise (at 2000 prices).

5.46 Our central projection is that annual GVA of Scotland's market sector will be c. £3.4 billion higher in 2015 than it would have been otherwise (at 2000 prices) - up from an impact of £0.8 billion in 2005.

5.47 As shown in Table 5 8 and Figure 5 5 below, much of the overall impact (45%) is expected to accrue to the J&K SIC grouping ¹⁵ (financial services; real estate, renting and business activities). These are the industries that have adopted broadband most rapidly, and the sectors in which ICT has a disproportionately large impact, given the nature of their operations.

Table 5 8 Projected impact of broadband on Scotland's market sector GVA, by SIC grouping £ million (2000 prices) [source: SQW]

Figure 5 5 Projected impact of broadband on Scotland's market sector GVA, by SIC grouping £ million (2000 prices) [source: SQW] (note that SIC grouping AB is at the bottom of this chart, but not visible)

5.48 The bulk of the overall impact (77% in 2015) derives from the incremental benefits of 1B broadband vs dial-up (see Table 5 9 and Figure 5 6 below). This is due to three factors: the roll-out and take-up of 2B and 3B+ broadband happens later than that of 1B; the incremental productivity benefits of 2B vs 1B and of 3B+ vs 2B are each assumed to be lower than the incremental benefits of 1B vs dial-up; and there is a lag of four to six years before businesses upgrading to these bandwidths realise the full incremental productivity benefits associated with that upgrade.

Table 5 9 Projected impact of broadband on Scotland's market sector GVA, by broadband generation, £ million (2000 prices) [source: SQW]

Figure 5 6 Projected impact of broadband on Scotland's market sector GVA, by broadband generation, £ million (2000 prices) [source: SQW]

5.49 The above chart and table need to be interpreted with care: the 'incremental impact of 1B' in any year does not only relate to those businesses using 1B (as opposed to 2B or 3B+) connectivity in that year. To take an example: a business in SIC grouping C&D that adopted 1B broadband in 2003, upgraded to 2B in 2007, and upgraded to 3B+ in 2011 is assumed, by 2015, to be 8.1% more productive than it would have been if it had remained on dial-up: 5% from adopting at least 1B broadband (fully realised by 2015), plus 1.7% incremental benefit from adopting at least 2B broadband (fully realised by 2015), plus 1.4% incremental benefit from adopting 3B+ (four fifths of the way along an S-curve to the maximum 1.7% benefit for 3B+).

Cross-checks on our economic impact projection

5.50 In this sub-section we briefly provide some cross-checks on the above projection, in order to confirm that it appears reasonable.

5.51 Firstly, we note that our central projection of market sector GVA being £3.4 billion higher than it would be otherwise in 2015 would correspond to c. 4.6% of market sector GVA in that year (assuming 1.9% p.a. real growth, the average Scottish market sector growth rate for 1995-2004). Given the strategic importance of broadband for economic development, as perceived by Scotland's business organisations and policy makers, this implied proportion of 2015 market sector GVA does not appear to be outlandish.

5.52 Secondly, we have considered what share of total market sector GVA growth this would imply for broadband over the period.

• If we assumed that the future real growth of market sector GVA is constant at 1.9% p.a. (the average Scottish market sector growth rate for 1995-2004), then our projections would imply an average 20% share of annual GVA growth attributable to broadband (and the ICT investments and business process changes enabled by broadband) in the period 2001-2015.
• If we assumed a future 2.5% market sector GVA growth rate (the long term UK trend rate for real GDP growth), then our projections would imply an average 14% share of annual GVA growth attributable to broadband.

5.53 We consider this implied 14%-20% growth share to be credible, in the light of recent evidence [Oulton and Srinivasan (2005)] that ICT capital deepening accounted for 46% of the growth in productivity in the UK market sector in the second half of the 1990s. Assuming that this level of overall ICT impact will be sustained in the period 2001-2015, our projection would therefore imply that something in the order of a third to half of the productivity benefits associated with ICT would not be realised if broadband were not available.

5.54 This appears to be reasonable (or potentially conservative), given that ICT investments being made by businesses are increasingly reliant on sites having broadband connectivity - especially for multi-site firms ¹⁶, as illustrated in the table below.

Table 5 10 Indicative broadband dependence for various business ICT investments [source: SQW]

5.55 Thirdly, we have cross-checked our projections against those of the four previous studies summarised in Annex C. A ballpark indicator of what these previous studies would imply for a developed economy of c. 5 million people can be derived from a very crude pro rata calculation, on the basis of relative population sizes. This results in a wide range of implied annual GDP impacts in 2015 for a country the size of Scotland - from £0.3 billion to £15 billion, as shown in the chart below.

Figure 5 7 Comparison of SQW's projections for the impact of broadband in 2015 with extrapolated/pro rata projections from other studies (for a developed economy of 5 million people) [source: SQW]

5.56 Of the four studies, the CEBR study from 2003 is the most relevant comparator, as it is UK-focused. Applying our simple pro rata calculation on the basis of population would imply an annual GDP impact in 2015 of c. £1.8 billion for Scotland (under CEBR's 'inclusive' scenario, which they considered to be the more likely). However, we note that actual broadband take-up has been accelerated by no less than five years compared with CEBR's demand projections, which were themselves relatively bullish at the time: CEBR expected 10 million broadband connections in the UK by the end of 2010 (6.1 million by the end of 2005), whereas the latest estimates from Ofcom indicate that there were 11.1 million connections as of March 2006.

5.57 While our economic impact projections are higher, on a pro rata basis, than those of three of the four comparator studies from 2002 to 2004, we are comfortable that this is justified in the light of recent actual growth in broadband availability and take-up, and given the growing body of evidence in the literature regarding the importance of ICT-related innovation for productivity growth.

5.58 Finally, we note that there is some emerging evidence that broadband is indeed having a real and measurable impact on the economic growth of developed countries.

• In a US-wide study, Lehr et al (2006) report that between 1998 and 2002, communities in which mass-market broadband became available by December 1999 experienced more rapid growth in employment, in the number of businesses overall, and in the businesses in IT-intensive sectors. This finding is supported by matched sample regressions, which control for community level factors known to influence broadband availability and economic activity.
• There appears to be some positive correlation between broadband penetration in developed countries and their time-shifted GDP growth. For the G7 countries ¹⁷, we plotted the annual 2005 GDP growth rate against the annual growth in broadband penetration in the years 2002, 2003, 2004 and 2005. Interestingly, we found a negative correlation when considering the most recent broadband growth rates, but a positive correlation when considering the broadband growth rates for 2002 and 2003. In particular, there was a strong and significant positive correlation (R-squared of 0.84) ¹⁸ between 2005 GDP growth and the 2002 growth in broadband penetration, as shown in the chart below.
  • While this correlation needs to be treated with caution, as there are only seven observations, we found it striking that the observed R-squared was as high as 0.84 given the many other factors affecting GDP growth in these countries ¹⁹. By way of comparison we found, for example, that the positive correlation between these countries' 2005 GDP growth and their 2001 R&D expenditure as a proportion of GDP was much weaker - at an R-squared of 0.54.
  • The three year difference between 2002 (the year taken for the growth in broadband penetration) and 2005 (the year taken for GDP growth) would appear to provide some support for our assumption of a lag of four to six years before the full benefits of broadband are realised, as this corresponds to a peak annual impact after two to three years.

Figure 5 8 2005 GDP growth versus 2002 growth in broadband penetration, for the G7 countries [sources: SQW analysis, The Economist (for GDP data), OECD (for broadband data)]

Impact of the Broadband for Scotland intervention

5.59 We now turn to consider the economic impact of the Scottish Executive's Broadband for Scotland intervention.

5.60 In December 2002 (when the coverage of affordable broadband services stood at c. 40% of Scotland's population) the Scottish Executive announced that it would seek to increase the level of coverage to 70% by March 2004. Following an analysis of the intervention options, the Scottish Executive, Scottish Enterprise and Highlands and Islands Enterprise set about an intensive demand stimulation programme, under the brand Broadband for Scotland - including a major advertising campaign ( TV, radio and press), the launch of an impartial website ( www.broadbandforscotland.co.uk), and assistance to local broadband campaigns. These activities had the effect of raising awareness of broadband in Scotland, and accelerating the rate of registrations of interest in receiving ADSL service, under the 'trigger point' mechanism that was being run at that time by BT²⁰.

5.61 As at the end of March 2003 there were 94 Scottish exchanges on BT's trigger point list (with a median exchange size of c. 6,000 household lines). However, in November 2003 BT announced a major extension to the trigger point scheme - allocating trigger points to every exchange serving more than 300 lines. This brought the total number of Scottish exchanges on the trigger point list to 465 as at the end of November 2003 (with a median exchange size of c. 600 household lines).

5.62 The chart below illustrates that all but 101 of these 465 exchanges had reached their trigger levels by the time BT called a halt to the scheme in April 2004 (at which point they committed to enable every exchange that had been assigned a trigger level, by summer 2005).

Figure 5 9 Exchanges yet to reach their trigger levels in Scotland [sources: bt.com, SQW 2004]

5.63 The rate of ADSL registrations appears to have been particularly high in Scotland. The chart below compares the trends in 'registrations yet to go' on the trigger point lists of the UK nations and regions ²¹ in early 2004.

Figure 5 10 'Registrations yet to go' on the trigger point lists of UK nations and regions in early 2004 [sources: bt.com, SQW analysis, 2004]

5.64 In Figure 5 11 below, we present our estimate of the 'counterfactual' coverage of affordable broadband in Scotland ( i.e. what the coverage would have been in the absence of any intervention by the Scottish Executive and its agencies) in terms of the proportion of Scotland's households connected to an ADSL-enabled exchange (a reasonably close proxy to overall population coverage for 2003 onwards).

Figure 5 11 Actual vs counterfactual broadband coverage in Scotland [sources: SQW analysis, Samknows (for exchange activation dates)]

5.65 Our key assumptions in developing the above chart are as follows:

• We have assumed that the public sector demand stimulation intervention started in April 2003 (although, in fact, there were some initiatives pre-dating this - especially in the Highlands and Islands).
• For those exchanges on the trigger point list at the end of March 2003, our counterfactual projections apply their March 2003 average rates of registrations per day per line.
• For those exchanges on the trigger point list at the end of November 2003, our counterfactual projections apply the lowest regional average rate of registrations per week per exchange observed across the UK nations and regions over the period January 2004 to April 2004 (which was 2.4 registrations per exchange per week).
• The exchanges in the Executive's Supply Side Intervention have been included as scheduled in the 'actual' coverage line; for the counterfactual we have assumed that these exchanges would have been upgraded at the end of 2009 - i.e. towards the end of BT's 21 CN roll-out.
• We have assumed that the average lag between an exchange reaching its trigger level and being enabled was four months.
• We have assumed that BT would have let the trigger point mechanism run its course (and not committed to enabling all exchanges with trigger levels), in the absence of the interventions that substantially accelerated the rate at which exchanges were reaching their trigger points. It should be noted that it was, of course, the combined effect of interventions and campaigns throughout the UK - not just in Scotland - that accelerated demand registrations and ultimately led to BT's decision to enable all exchanges with trigger levels.

5.66 The implications of this analysis are that the Broadband for Scotland interventions accelerated affordable broadband coverage as follows:

• 70% coverage was brought forward by c. 3 months;
• 90% coverage was brought forward by c. 17 months;
• near 100% coverage was brought forward by c. 4 years.

5.67 As a cross-check on our counterfactual projection, we compare in the table below the actual and counterfactual coverage levels versus the affordable broadband coverage in other countries as at 3Q05, as reported in Ovum (2006). Our counterfactual projection appears to be reasonable, placing Scotland's coverage between that of Australia and Canada, and significantly above that of Ireland.

Table 5 11 Comparison of affordable broadband coverage as at 3Q 2005 [sources: Ovum (2006), SQW]

5.68 By applying the counterfactual coverage levels to our business take-up model, we can observe the acceleration in business broadband take-up resulting from the intervention ²². The chart below suggests that c. 11,000 more Scottish business sites were connected to broadband as at the end of 2005 than would have been the case otherwise - c. 10% of the total business connections at that time.

Figure 5 12 Projections of broadband take-up by businesses in Scotland, with and without intervention [source: SQW]

5.69 Finally, we have applied the counterfactual business take-up to our economic impact model. As illustrated in Table 5 12 and Figure 5 13 below, our model suggests that the economic impact of the Broadband for Scotland intervention will peak in the year 2008, with the annual market sector GVA being approximately £150 million higher in that year than it would have been otherwise (at 2000 prices). If we assume real market sector growth of 1.9% p.a., this difference would represent c. 0.23% of total market sector GVA in the year 2008.

Table 5 12 Projected annual GVA impact of affordable broadband for Scotland's market sector - with and without intervention, £ millions (2000 prices) [source: SQW]

Figure 5 13 Projected annual GVA impact of affordable broadband for Scotland's market sector - with and without intervention, £ million (2000 prices) [source: SQW]

5.70 Over the period 2003-2015, the Present Value ²³ of the above differences in market sector GVA is c. £500 million.