The diagnosis and treatment of disparities in United Kingdom regional economic performance: a critique

The effectiveness of these arrangements in addressing low levels of regional productivity is to be assessed in part by applying benchmarking indicators to monitor regional performance in respect of the five drivers of growth (DTI, 2006a). More ambitious efforts to identify the specific impacts of RDA activities have been eschewed (DTI, 2006b). The regional benchmarking of productivity indicators has been supplemented by Treasury-led Public Service Agreements agreed at national level to provide a framework for monitoring regional performance against agreed targets (NAO, 2003). This has turned the focus of attention onto the indicators chosen for this purpose and their fitness for purpose.

Central to this arrangement are the indicators for measuring productivity. The initial Treasury report on regional productivity justified the emphasis given to the role of productivity in New Labour’s regional policy by concluding that “on average, productivity differentials account for around 60% of regional GDP per capita differentials” (H. M. Treasury, 2001: para 1.12). Although GDP per head is used to identify underperforming regions at an EU-level, its validity as a specific indicator of spatial productivity gaps has been questioned. Drawing on expert evidence, a House of Commons report into the Blair administration’s approach to regional policy concluded that output per resident was “not fit for purpose” (House of Commons, 2003: 3) as a robust measure of regional economic performance.

One reason given for this verdict was the inadequacy of regional statistics on output. The committee found that regional GVA measures, which exclude taxes and subsidies and so offer a clearer picture of local performance than GDP, suffered from significant data limitations, providing only “approximate estimates of the trend rate of growth of GVA per head” (House of Commons, 2003: 17). Another concern was the Treasury’s contestable claim that regional variance in GVA per head was predominantly attributable to spatial differences in productivity. As demonstrated below, more careful analysis of improved regional and sub-regional data reveals that this measure includes an amalgam of other factors that can outweigh the influence of productivity.

Following the critical House of Commons report, the Office for National Statistics (ONS) has devoted resources to improving its methodology for estimating sub-national data on economic output and to reducing the delays in its publication. A series of reports (Vincent, 2003; Camus, 2007; Haskel, 2007; Swadkin & Hastings, 2007a, 2007b & 2007c; Swadkin, Louca & Virdee, 2007) has analysed some of the new data, to provide a much clearer insight into the determinants of regional GVA per head. The overall conclusion to be drawn from this fresh evidence is that the Treasury has oversold the role of productivity in explaining UK regional and sub-regional variations in GVA. Unless adequate consideration is given to the impact of other factors contributing to the spatial variance of this indicator, both the performance targets set for RDAs and the means of assessing their effectiveness in fulfilling these targets will be compromised.

In one of its most recent publications, the Treasury rehearses the argument that within a well-integrated national economy with efficient factor markets, “[e]conomic convergence theory suggests that… firms and people will move so that per capita disparities between [regions] are reduced” (H. M. Treasury, 2007a: para 1.18). It then observes that “experience has shown that regional performance has not converged in practice. GVA per head growth rates for 1971-2001 show that both the size of regional differentials and the relative ranking of regions has not markedly altered over this period” (ibid.: para 1.19). The rest of the section of this report then provides an analysis of NEG explanations for the uneven distribution of disembodied technology to account for the UK’s non-convergent spatial variations in productivity. The overall weight of argument that productivity gaps perpetuate regional differentials in GVA growth is subject to only minor qualifications (ibid.: para. 1.26).

A careful inspection of the most recent ONS data published on UK regional GVA reveals a more ambiguous picture. A chart illustrating how the GVA per head in the 12 UK NUTS1 regions differs from the UK average over the period 1995-2004 shows that only two of these regions, London and the South-East, demonstrate a level of GVA per head above the UK average: in both cases the variance from the mean is increasing (Figure 2). Variance from the mean has also risen for the two worst-performing NUTS1 regions: the North-East and Wales, although the third-worst region, Northern Ireland shows little change in this respect. Overall, trends in GVA per head data for NUTS1 regions appear to support the Treasury’s contention that UK regional performance remains non-convergent: the coefficient of variation for this series has risen from 16.7% to 19.7% between 1995 and 2004.

Similar analysis of GVA per head spatial variance is now possible for new ONS datasets offering regional disaggregations for the same period at NUTS2 and NUTS3 levels. At the NUTS2 level, a number of regions outwith London and the South-East are shown to have GVA per head above the national average, including one in Scotland (the North-East, which includes Aberdeen) which enjoys a GVA per head only exceeded by Inner London. Three NUTS2 regions in London and the South-East (Outer London; Hampshire and the Isle of Wight; Kent) fall below the national average. Overall, while the dispersion around the mean for NUTS2 regions exceeds that for NUTS 1 regions, it shows a similar trend, with the coefficient of variation rising from 26.6% in 1995 to 30.7% in 2004.

The NUTS3 level of disaggregation involves 133 UK sub-regions, revealing a more complex pattern of spatial variance from the mean for GVA per head. Thirty-six of the NUTS£ regions (27%) are shown to exceed the UK mean in 2004, but the picture is dominated by an extra-ordinary outlier, Inner London West, for which GVA per head was 324% above the UK average. No other NUTS3 region approaches this level of divergence: the two with the next highest GVA per head are the City of Edinburgh and Berkshire (both with GVA 61% above the UK mean). The trend for the 1995-2004 time series is comparable to the non-convergent picture revealed for NUTS2 regions, with the coefficient of variation for this series rising from 35.4% to 38.5%.

The difficulty with using GVA per head data as an indicator of productivity variations between UK regions is that it contains other factors apart from output per hour differentials. To isolate the genuine elements of productivity variations within the data, the indicator needs to be decomposed into its constituent determinants for each level of regional disaggregation. The ONS has recently published the results (Swadkin & Hastings, 2007a & 2007c) of its application of an OECD algorithm designed for this purpose. This decomposes GVA per head at regional and sub-regional levels [1] into five elements:GVA [1]/P = GVA [2]/HW+ HW [3]/EW + EW [4]/LFW + LFW [5]/LFR + LFR [6]/P ;

where [2] is GVA per hour worked, an unambiguous measure of average labour productivity; [3] adjusts for variations in part-time employment between areas; [4] adjusts for variations in unemployment between areas; [5] adjusts for differences between the labour force actually working in an area and its resident workforce to take account of differences in levels of commuting between areas; and [6] adjusts for variations between areas in resident activity rates. Removing variations in GVA per head attributable to commuting, activity and unemployment rates and part-time working isolates the influence of spatial variations in GVA per hour worked, to provide what the ONS terms its “preferred indicator of productivity” (Swadkin & Hastings, 2007c: 50).

This algorithm can be used to decompose 2004 UK NUTS1 gross GVA per head differentials into their constituent elements (Table 1). As the ONS researchers observe of the results of this procedure: “when using GVA per hour worked, there are significantly fewer and smaller differences in regional economic performance than when making comparisons based on other indicators” (Swadkin & Hastings, 2007c: 50). Regional productivity variations based on GVA per hour work still diverge over the period 1995 to 2004, but the scale of such differentials is much reduced from that suggested by the use of gross GVA per head estimates. For 2004, the coefficient of variation for the preferred indicator of productivity falls by 47%, from 19.79 for gross GVA per head variations that include non-productivity elements, to 10.40 for variations attributable solely to GVA per hour worked. Earlier Treasury assertions (2001: para 1.12) that around 60% of regional variance in GVA per head can be attributed to spatial variations in productivity are shown to be an exaggeration.

The decompositions provided by the OECD algorithm reveal the extent of the contributions made by non-productivity variables to UK NUTS GVA per head regional differentials. As the data for 2004 indicate, differences in rates of unemployment between UK NUTS1 regions explain little of the gross variance in GVA per head. However, the coefficients of variation created by variations in part-time employment (2.72) and in labour force activity rates (4.49) are important, while those produced by variations between residential and workforce population caused by commuting (7.33) are second only to productivity as a source of variation in GVA per head across NUTS1 regions.

Although spatial differences in GVA per hour worked is the largest single overall explanatory factor for 2004 GVA per head variations, for individual NUTS1 regions such as the North East and Wales other factors match or exceed the contribution of average labour productivity. Even for London, labour productivity accounts for less than half (+23.4%) of the amount by which its GVA per head exceeds the national average (+53%). More worryingly in terms of the use of gross GVA per head as an indicator of regional variations in productivity, the East and East Midlands are the two NUTS1 regions with GVA per hour closest to the UK average, yet they appear to be ‘underperforming’ purely because of adverse commuting effects that pull their GVA per head below the UK average.

This anomaly is entirely due to a statistical quirk, stemming from the way UK regional GVA per head is calculated for UK regions. The value of output produced by the workforce employed within the region serves as the numerator, but the resident population of the region is used as the denominator to convert this into regional output per head. As the ONS acknowledges in using a workplace-based measure of GVA per head, where significant levels of commuting-to-work across regional boundaries occur, this procedure creates serious distortions. UK regions such as the East and East Midlands that experience net out-commuting of residents to workplaces in other regions find that their GVA per head is artificially deflated, by -6.6% and -7.0% respectively in 2004, because the calculation of their numerator is confined to the output generated by those working within their boundaries, whereas all their residents are included in the denominator. In effect, this official treatment of the data means that net out-commuting regions ‘export’ part of the GVA per head generated by their residents to those regions with net in-commuting of workers. In recipient ‘importing’ regions such as London, GVA per head is artificially inflated by including in the numerator the output generated by non-resident workers commuting from other regions, and then confining the denominator to their resident population. For 2004, this boosted GVA per head in London by 21.4%.

Use of the OECD decomposition algorithm by the ONS highlights an obvious source of non-convergence in regional GVA per head that is clearly unrelated to productivity variations: the increasing disparity in housing costs between London and other parts of the country. This stimulates long distance commuting from other regions amongst London workers forced to reside where they can afford to purchase a home. Building more affordable homes in London or relocating economic activity to areas outside London would dampen the growth in such long-distance commuting. In the process this would also reduce regional differentials in GVA per head, but it would be incorrect to attribute any reduction in the growth of long-distance commuting to the success of government policy in addressing the ‘productivity gap’ between regions.

The recent release of new data series has enabled the ONS to apply similar decomposition procedures to UK NUTS3 sub-regional data. As yet, the ONS still lacks a compatible ‘hours worked’ series to allow the final stage of decomposition from GVA per filled job to GVA per hour worked to be applied at sub-regional level. This means that productivity has to be calculated in terms of GVA per job filled, making no allowance for sub-regional variations in part-time employment. With this qualification to the analysis, Table 2 indicates the ONS decomposition of the ten UK NUTS3 sub-regions with the highest GVA per head in 2004; while Table 3 applies the same analysis to the ten UK NUTS3 sub-regions that had the lowest GVA per head in the same year.

As might be expected, when Inner London West’s 324% apparent ‘over-performance’ as measured by GVA per head is decomposed into its four constituent elements, in-commuting and the resulting ‘import’ of GVA from non-London residents (+220.4%) accounts for by far the largest part of this. Netting out the other factors in this way leaves this part of London with a more realistic measure of productivity (+93.5%) that is just under double the UK average. A similar adjustment procedure reveals that Inner London East enjoyed the second highest level of average labour productivity in 2004 (33.2% above the UK average), followed by Berkshire (+32.1%) and Swindon (+26.8%).

As most of the top-ten sub-regions in Table 2 demonstrate, intra-regional disaggregation of data reveals a standard functional specialisation of place across the country as a whole, with sub-regions forming their own city-region economic nodes. Rather than serving as a good indicator of spatial productivity gaps, reliance on raw GVA per head data produces anomalies in the pattern of spatial variations. Because of the way the data are treated to calculate spatial GVA per head, accessible rural sub-regions containing relatively low levels of economic activity ‘export’ a significant part of their GVA per head via net out-commuting of part of their residential workforce, boosting this measure in the recipient sub-regional cities. These anomalies generate misleading information which can sometimes mask serious sub-regional productivity weaknesses within these nodes.

For example, Glasgow City’s GVA per head just makes the 2004 UK top ten sub-regions. However, this is largely due to the ‘export’ of GVA per head by commuters from sub-regions such as East Dunbartonshire, West Dunbartonshire, Helensburgh and Lomond (EDWDHL). Adjusting the Glasgow City data for such effects by deducting 47.4% of its GVA per head to reflect this ‘import’ of GVA per head from non-resident workers leaves the city -4.4% below the UK average for labour productivity. Belfast masks an even lower level of labour productivity (-13.8% below the UK average) thanks to the very large contribution ‘imported’ workers make via commuting (+95.2%), which leave its GVA per head 48% above the UK average, allowing the city to rank sixth highest in all 133 NUTS3 sub-regions.

Unsurprisingly, the ten sub-regions with the lowest GVA per head illustrated in Table 3 are all rural. Adjustments to isolate labour productivity from other influence again reveal some major anomalies in GVA per head ranking. Despite having a GVA per head -38% below the UK average, after allowing for out-commuting (-34.6%), the EDWDHL sub-region demonstrates a level of GVA per filled job (-4.2%) which brings it much closer to the 2004 UK average. By contrast, the low rankings in terms of GVA per filled job for three of the other bottom ten GVA per head NUTS3 sub-regions, Caithness & Sutherland and Ross & Cromarty (-38.9%), Lochaber, Skye & Lochalsh, Argyll & the Islands (-33.9%), and Torbay (-29.2%) are reasonably well approximated by their gross GVA per head data.

The decomposition of GVA per head for the two remote Scottish sub-regions reveals rates of labour force participation and unemployment above or equal to the UK average together with significant levels of net out-commuting, suggesting that their poor productivity levels reflect their peripherality and the lack of adequate investment in infrastructure to address this. In between these extremes, productivity measured in terms of GVA per filled job is well below the UK average for the North of Northern Ireland, Wirral, the Isle of Wight and the three Welsh NUTS3 sub-regions. This is compounded by adverse activity rates, combined with adverse commuting rates (except for the North of Northern Ireland), and adverse employment rates (except for South West Wales), which leave them in the bottom ten as measured by gross GVA per head.

6. Conclusions

UK regional policy under New Labour is no longer primarily concerned with redistributing the fruits of national economic performance to the areas that have not managed to harvest a reasonable crop of their own. Instead, the term ‘new regionalism’ articulates New Labour’s desire to empower communities to take responsibility for raising their own performance, while its regional policy places an onus on all regions to make an effective contribution towards the task of boosting overall national economic growth. In the process, so the argument runs, regions which successfully address the underlying causes of their own under-performing economies through supply-side reforms can give their communities a more sustainable improvement in their living standards than traditional demand-side spatial Keynesianism could achieve by fostering dependency on central government’s benevolence.

Under this approach central government retains firm control over setting the policy agenda for economic development, but accords its newly-fashioned system of regional governance primary responsibility for its implementation, because “it is devolution which is seen to offer the most appropriate framework for making progress on the five drivers of productivity, by allowing flexible interventions that are responsive to local needs”, Keep et al, 2006: 548). Charging regions with the task of fuelling the UK’s engine of growth effectively moves regional policy to the centre of New Labour’s overall strategy for the national economy. Critics of this major change of emphasis, such as Lovering (1999) and Fothergill (2005), argue that it has been driven largely by policy initiatives, with the theoretical analysis (e.g. H. M. Treasury, 2005; 2006b; 2007a; 2007b) belatedly appearing retrospectively as a way of justifying rather than informing strategy. As a result, there continues to be confusion between policy ends and means. Rather than investigating the accuracy of the diagnosis proffered, “much of the debate is about matters of detail, not about the substance of underlying assumptions” (Keep et al, 2006: 540).

As the previous section of this paper has attempted to demonstrate, when exposed to close scrutiny the evidence supporting this policy framework is shown to be less convincing than its advocates maintain. Nevertheless, discounting the overblown claims made for its explanatory power at a regional level, some substance still remains to the notion that productivity is a major long-run determinant of economic performance, and that spatial variations in productivity contribute unnecessarily to the impoverishment of parts of the UK and retard the overall growth of the UK economy. Does it follow from accepting this point that RDAs and other regional actors provide the best way of addressing the issue?

The difficulty in accepting this argument is twofold. Firstly, there are serious practical difficulties in operationalising a strategy which devolves responsibility for implementing centrally determined targets whilst simultaneously attempting to make regional governance responsive to regional needs through the adoption of locally-agreed RESs. In its report on RDAs, the National Audit Office identified the problem the agencies encountered in reconciling these conflicting objectives and simultaneously satisfying the requirements of their national and regional clients. Observing that there was no regional engagement in setting central government targets, its report concluded that “[s]takeholders report a sense of ownership of Regional Economic Strategies but not of national targets” (NAO, 2003: 7). Similarly, at national level “[d]epartments cannot easily see how Agencies’ work aligns with national priorities, because the Agencies’ long-term targets link only broadly to PSA targets… and there is no systematic monitoring of progress towards them” (ibid.).

Some efforts have recently been made to improve this situation, but even if New Labour’s regional strategy were to ameliorate the market failures limiting the effectiveness of regional productivity drivers, it is unclear from the Treasury model how this would ensure that productivity gaps narrowed. This leads to the second and central problem in accepting the logic of the current strategy underpinning New Labour’s regional policy. Although the Treasury publications on this topic outline various mechanisms by which productivity may be boosted by attention to one or more of its drivers, none of these is linked to a transmission mechanism ensuring that attainment of higher productivity will reduce regional differentials in GVA per hour worked. As things currently stand, the application of regional policy may prove effective in boosting the UK’s overall level of productivity and raising its national growth rate, but do little to rectify its regional disparities in performance. Indeed, as Martin & Sunley (1996) recognise, there are plausible analytical reasons for suggesting that a successful application of the current regional strategy could exacerbate such disparities.

The empirical evidence suggests that regional disparities in economic performance are slowly widening. Do the theoretical arguments presented for adopting the current regional strategy provide any coherent grounds for believing that such a situation can be turned around, and that lagging regions can catch up with the leaders, simply by ensuring that factor markets work more efficiently, that competition is effective, that innovation and enterprise are boosted and that investment in skills, infrastructure and capital equipment is buoyant? In other words, where is the transmission mechanism that ensures that under-performing regions will benefit to a greater extent than more successful regions from the pursuit of this approach? Must we assume that the current regional differentials are largely attributable to ignorance of best practice, and that concerted efforts to disseminate such knowledge will open up new opportunities for lagging regions that well-organised parts of the country have already exploited?

There is a curious hiatus in all such official rehearsals of the standard model used to justify New Labour’s focus on spatial productivity gaps (e.g. H. M. Treasury 2000; 2001; 2005; 2006a; 2007a). Firstly, we are told that neo-classical growth theory postulates the convergence of economic performance. Then we are provided with the arguments as to why this does not happen in reality, centred on aspects of disembodied technology, human capital and the ‘new economic geography’. Evidence is produced to demonstrate how these factors can explain part of the uneven economic performance of different parts of the UK. Up to this point, the analysis of government policy retains its theoretical coherence. There is then a jump in logic, which consists of assuming that if factor markets are made more flexible and all the elements of the NEG, human capital and disembodied technology which contribute to uneven development are allowed to function in a less spatially-constrained and more open economy, this would reduce spatial disparities in performance and promote regional convergence.

Implicit in such an assumption rests an unacknowledged transmission mechanism that views cumulative divergence as a function of inefficiency in factor markets, and that considers more flexible factor markets as contributing to economic convergence. Yet it is perfectly possible to argue the opposite, and to envisage the effects of government policy in boosting the five drivers of productivity as augmenting fissiparous pressures on growth and exacerbating regional differentials. As one perceptive commentator has observed, “[w]orking at their best, market forces are… just as likely to trigger a spiral of cumulative decline as a process of spontaneous revival” (Fothergill, 2005: 664).

In the initial chapters of The Wealth of Nations, Adam Smith observed that specialisation was limited by the size of the market and that the living a person could gain from the division of labour in a city the size of Edinburgh would be impossible in the confines of the Highlands. Nothing has changed to modify the truth of this observation. Indeed, in one of its reports (H. M. Treasury, 2005) the Treasury makes an equivalent claim for the City of London as the world’s financial centre, with its workers able to reap the benefits (and, more recently, costs) of global specialisation and division of labour.

Are we to suppose that assiduous efforts on the part of RDAs will succeed in identifying and exploiting global markets for the base activities of each of the UK’s regions, which will boost local labour markets to the same extent as London? Assuming this to be unlikely, what would be the more probable effect of removing market imperfections that hinder the free movement of capital, labour and technology across the country and beyond? It is surely not improbable to contend that enveloping every part of the UK into a global market for its resources and opening each region up to the full force of competition would further polarise spatial economic performance rather than reduce this.

Let us take as a test case the role of human capital in accounting for spatial variations in economic performance. Suppose, as the Leitch Report (H. M. Treasury, 2006c) urges, that special efforts are devoted to a radical overhaul of UK educational standards to bring them up to world class levels, with parts of the country with below-average educational attainments brought into parity. Would this promote a burst of productive energy that allowed the enterprises of lagging regions to exploit the enhanced skills of their workforces and boost their productivity, allowing productivity gaps to shrink not only between different parts of the country but between the UK and our nearest competitors, as Leitch contends?

Keep et al (2006) point out that the Leitch Report includes a case study that enables this conclusion to be tested: Scotland. Its long-standing devolved education system generates indicators of labour quality and human capital well-above the UK average and comparable with the best in the world. Yet, as Table 1 indicates, the commitment of high levels of expenditure on education and training its workforce has not sufficed to remove Scotland’s productivity shortfall. Instead, this has enhanced the mobility of Scottish workers, with many of its well-educated graduates migrate to regions offering better career prospects.

The Treasury response to this fact is to argue that regions should pursue measures to create better jobs to retain more skilled workers. Current policies have ruled out the redirection of businesses and jobs from elsewhere in the UK, and foreign direct investment is no longer seen as providing a central part of the solution, so the Treasury must assume that all of the UK’s regions possess the inherent potential to become global economic actors to rival London. Some of the country’s city-region nodes can undoubtedly respond effectively to globalisation and provide job opportunities sufficient to attract and retain a ‘world class’ workforce, but the remote peripheral areas identified in Table 3 have no prospect of becoming part of such a successful node.

These areas appear consigned by New Labour’s regional policy to becoming increasingly ostracised from the economic mainstream, if measures to boost productivity and increase the efficiency of UK factor markets simply result in greater out-migration of resources and an increase in cumulative economic divergence as the more successful parts of the country succeed in capturing even more of its most productive assets. Should this prove one of the principal outcomes of the implementation of current UK regional policy, it would seem that what was a policy agenda initially driven by concerns for spatial equity is now predicated on a desire for spatial efficiency.

Acknowledgement:

An earlier version of this paper was presented at a research seminar at the Centre for Local Government, University of Technology Sydney, on 29 April 2008.