DYNAMIC BENCHMARKING OF THE WELSH LABOUR MARKET
Melanie K. Jones
November 2002
Welsh Economy Labour Market Evaluation and Research Centre (WELMERC)
Department of Economics
University of Wales Swansea
Swansea SA2 8PP
Email address:
Abstract
This paper represents an initial investigation into benchmarking the performance of the Welsh labour market. It focuses on how labour market comparison can be simplified, using a composite measure of performance, whilst capturing the multi-dimensional nature of labour market performance. The issues that surround the development of a composite measure are discussed, including the selection of domains, indicators, an appropriate benchmark and weights. The index is applied to Wales between 1997 and 2002 at the national, unitary authority and EU Objective area level.
Key words: benchmarking, regional labour markets
Acknowledgements
The author would like to thank Richard Jones and Peter Sloane for their useful comments and suggestions. Financial support from the European Social Fund is gratefully acknowledged.
Dynamic Benchmarking of the Welsh Labour Market
1. Introduction
The evaluation of labour market performance is essentially comparative, assessed over space (countries, regions and urban areas) and time (previous decade, year and month). Performance is also multi-dimensional and labour markets may perform well in some domains but poorly in others. Thus, a single indicator can only provide a limited, one-sided view of labour market performance (see Watson, 2000 for a critique of using only the unemployment rate as a measure of performance). However, the use of multiple indicators can create ambiguity, since each indicator may rank labour markets differently. Without assessing the relative value of each measure it is not possible to identify one labour market performance that dominates another. The well-established use of composite indicators for assessing differences in deprivation and well-being (for an international example see Osberg and Sharpe, 2002 and for a Welsh application see Higgs and White, 1997) has more recently been applied to simplify the monitoring of labour market performance. Composite labour market indices have received most attention in European labour market research (see Tronti, 1998), where the focus has been on developing effective labour market benchmarks, from which the impact of national labour market policies on labour market convergence can be assessed.
This paper represents an initial investigation into the development of a composite measure of labour market performance for Wales. In particular, its application is intended to concentrate on divisions at the unitary authority level and between the Objective One and Three areas. There are several reasons why this application of benchmarking is appropriate.
Economists have become more aware of the spatial divisions within regions. Within Wales, Morris and Wilkinson (1989 and 1993) identify an East-West divide. Areas to the East, where there is greater access to larger centres of population, exhibit stronger economic performance. For example, in 1998, GDP per head was 98% of the UK average in East Wales, in contrast to 70% in West Wales. To evaluate differences in the degree of deprivation, the National Assembly has developed the Welsh Index of Multiple Deprivation (WIMD), which compares deprivation at the level of electoral divisions (EDiv) within Wales at a single point in time. The extent of the differences identified has become a matter of policy concern; A Winning Wales (2001), the economic strategy for Wales, sets targets to reduce intra-regional differences.
Labour market indicators also signal differences in performance at the unitary authority level. A range of indicators, using both the Unitary Authorities and EU Objective Areas are presented in Jones et al. (2002), each indicating the presence of labour market differences within Wales. For example, high unemployment areas are clustered in West Wales and the South Wales Valleys. Indeed, in September 2002, the unemployment rate in Blaenau Gwent (7.1%) was more than three times that in Powys (2.0%) (see Minford and Stoney, 2001 for a discussion of North Wales). Over time, these divisions have grown as structural change has had an unequal spatial impact; areas previously reliant on heavy industry (such as the South Wales Valleys) have suffered a disproportionate number of job losses. In contrast, economic growth has centred along main infra-structural developments such as the M4, which has contributed to the growth in Cardiff.
As a result of these disparities, there is now a difference in level of assistance being given to regions within Wales. Objective One funding has been allocated to West Wales and the Valleys (2000-2006) and this provides an opportunity to compare the labour market performance of the Objective One area with the remaining areas in Wales.[1] The index is constructed over the period 1997-2002 and, thus, the performance differential between these areas can be compared before, and after, the introduction of the Objective One scheme. In addition, the first results from the Welsh Assembly sponsored boost of the Labour Force Survey (LFS) have been released (November 2002). The increased sample size in Wales means the analysis of sub-regional labour market disparities previously constrained by data availability is now possible. This important step will increase future research and attention at the finer spatial level.
This paper is structured as follows. Previous applications of benchmarking to labour market performance are reviewed briefly; a general discussion of benchmarking is avoided since it has been considered extensively elsewhere (see Schutz et al., 1998). The structure of the composite index that is used to monitor the performance of the Welsh labour market is outlined, including a discussion of the domains used, namely employment, income and human capital and the benchmark chosen to assess relative performance. The results of an application of the index to Wales, at the national, Objective funding and unitary authority levels, are discussed.
2. Labour Market Benchmarking
The form of a benchmark depends crucially on which labour markets and what labour market features are being compared. Indices can be constructed to monitor performance of a single labour market over time. For example, Watson (2000) considers the Australian labour market from 1988-1999. Alternatively, Storrie and Bjurek (2000) compare labour markets over countries in the EU and Osberg et al. (2002) use spatial and temporal comparisons when analysing the national labour markets of the US and Canada.
Definitions of performance vary considerably between studies. Mosley and Mayer (1998) follow targets set out by EU labour market guidelines, whereas Watson (2000) considers the ‘health’ of the labour market using measures of the quantity and quality of employment. In contrast, Osberg et al. (2002) focus on the labour market well-being of workers, which is assessed using the domains of returns from work, accumulation, equality and security.
Several alternative techniques have been applied to monitor labour market performance, including both quantitative and graphical performance measures. The most common composite measure is an index, in which an index number can be compared to an optimal value.[2] Examples include Watson (2000), where the index runs from 1-10 and the optimal value is justified theoretically, and Osberg et al. (2002), where the index has a base of 1.0 in the initial period. Attention has focused on the radar chart approach, which allows graphical representation of a range of indicators. Schutz et al. (1998) and Mosley and Mayer (1998) use this technique. Indicators are standardised and represented by values from 0-1, where performance is assessed relative to the best observation (country) for each indicator or a theoretical ideal. A quantitative measure, the surface area measure of performance, ‘SMOP’, can be calculated using the area of the polygon. Schmid et al. (1999) discuss both the strengths and limitations of this technique.
All the above studies, however, share the notion that labour market performance can be assessed using outputs of the labour market production process. Theoretically, this has been questioned. Anxo and Storrie (1997) develop the concept of an efficiency frontier, where labour market performance is assessed using both outputs and inputs to the production process (in this case the input is spending on labour market policy). Tronti (1998) discusses an application of this process where the output refers to the efficiency of matching, as illustrated by a Beveridge (U-V) curve, and this is compared to the percentage of GNP spent on labour market policy. Labour markets performing on the frontier can be identified and the distance from the frontier is used as a measure of the inefficiency of other markets. Moreover, this technique seems to solve a fundamental problem of all multiple indices- the choice of weighting (discussed in section 3).
The subsequent use of the ‘efficiency frontier’ method has been confined to solving the weighting problem for a range of output measures. Storrie and Bjurek (2000) construct a performance frontier on the basis of employment and unemployment output indicators. The performance frontier is plotted using the best performance for each indicator across countries. Observations are then compared to the frontier using the ratio X/Y, where Y is the distance of the observation from the origin and X is the distance from the origin to the frontier (using the same slope); thus observations on the frontier have a value of one.[3] Weights are determined by the location of the observation, relative to the frontier. Thus, countries are compared to those on the frontier that resemble their indicator mix and, at high levels of any indicator, the relative weight for that indicator is low. This application has been criticised by both McCarthy (2000) and Eichhorst and Profit (2000), since the point of reference for each labour market is not the outcome of an optimisation process and thus the term ‘efficiency frontier’ is misleading. As yet, the concept of benchmarking labour market efficiency has had limited empirical investigation.
Schmid et al. (1999) discuss an alternative, the employment systems approach, which considers the set of institutions that determine employment. This approach is intended to provide a broader framework for benchmarking, with greater theoretical content. The wider economic environment is taken into account and concepts such as efficiency can be considered. For example, GDP is decomposed into efficiency measures of employment and quantity employment measures.
Composite Welsh Indicators
Applications of composite indices in Wales focus on deprivation. The most recent detailed measure is the WIMD. This index includes the domains of income, employment, health and disability, education, skills and training, housing and geographical access to services. The WIMD compares deprivation at a more disaggregate spatial scale, in electoral divisions, though for a single period (2000).[4] This has updated the previous comparisons of local deprivation, that focus on 1991 Census data such as the Townsend index, the Welsh Office index, DoE Index of Local Conditions and the Breadline Britain Index discussed by Higgs and White (1997).
In contrast, the Welsh Labour Market Index (WLMI) has a more specific focus on labour market performance and is constructed to enable comparison both over time and between unitary authority areas. Similar to previous national labour market indices, the WLMI focuses on performance in terms of labour market outputs and aims to identify performance gaps within the domains, which will enable ‘learning by comparison’ to promote convergence.
3. Development of the Composite Labour Market Performance Index
Domain and indicator selection
As has been suggested, labour market performance can be assessed using numerous indicators, but, in this case, the choice is constrained by data availability at the local level over time.[5] The data used need to be published regularly and be consistent over time. Thus, potential information in the form of intermittent surveys could not be included. Each quantitative indicator is appropriate to capture the main features of the included domain. Economists may disagree over desirable labour market characteristics[6] and the choice of domains and indicators is somewhat normative. However, all three domains of the chosen index are central in any labour market assessment and the goals are justified below. The index avoids indicators that have been questioned in previous work, such as casual employment, full-time/part-time balances and measures of gender equality in the labour market (see, for instance, CEDA Bulletin 2000).
One concern, when constructing the index, is excessive correlation between indicators. Some correlation is unavoidable and results from the natural inter-linkages between the components of labour market performance. However, choosing the most appropriate indicator can reduce unnecessary correlation. For example, the proportion of unemployed classified as long-term unemployed, rather than the long-term unemployment rate (formed as a proportion of the workforce), is included, since the latter is highly correlated with the unemployment rate and would effectively increase the weight of unemployment in the index.
The proposed index is constructed over three domains, employment, income and human capital. The indicators in each of these domains are outlined below.
Employment Index (EI)
In line with the National Assembly targets for increasing employment set out in ‘A Winning Wales’, employment is a fundamental indicator of labour market success. Three indicators are included in this domain; the unemployment rate, rates of economic inactivity and proportions in long-term unemployment.
Unemployment (URI)
The unemployment rate measures the proportion of the working population that is seeking, but unable to find, employment. Although some unemployment will be evident in any dynamic labour market,[7] high unemployment may suggest the local economy is unable to generate enough suitable employment.[8] Unemployment is associated with economic and social costs. The value of output forgone represents the most serious economic cost of unemployment, but unemployment also has an important impact on government finances, increasing benefit payments and reducing taxation revenue. Bivand (2001) estimates an annual treasury cost of £10,000 per person. Studies also find unemployment to cause a range of social costs such as crime (Carmichael and Ward, 2001 and Burdett, Lagos and Wright 1999), ill-health (Warr 1987, Clark and Oswald, 1994) and anti-social behaviour (MacDonald and Pudney, 2000), which are not included in the above estimate.
The claimant count unemployment rate is used due to greater availability at the local level.[9] However, unemployment claimant count measures are thought to under-estimate the true number unemployed (see Gregg, 1994 and MacKay, 1999). This is because the claimant count depends on the rules for benefit entitlement, which have been tightened over time. The International Labour Organisation (ILO) unemployment measure would have been preferred but is suppressed by the LFS when the sample sizes are small. A greater critique of unemployment measures has emerged (see Beatty et al., 2000 and Beatty and Fothergill, 2002, amongst others). It is argued that unemployment has become hidden due to the increased numbers on government training schemes, in early retirement and, most importantly, claiming sickness benefits. Combining several measures in the employment index, therefore, provides a more accurate picture of the level of employment than the use of the published unemployment rate.